Академический Документы
Профессиональный Документы
Культура Документы
Second edition
Olaf Ziemann Jrgen Krauser
Peter E. Zamzow Werner Daum
POF Handbook
Optical Short Range Transmission Systems
Second edition
123
Professor Dr.-Ing. Olaf Ziemann Dipl.-Ing. Peter E. Zamzow
Georg-Simon-Ohm-Fachhochschule Nrnberg R & D Cable Systems
Wassertorstr. 10 Erlen-Str. 5b
90489 Nrnberg 44795 Bochum
Germany Germany
olaf.ziemann@pofac.fh-nuernberg.de peter.e.zamzow@t-online.de
DOI 10.1007/978-3-540-76629-2
Library of Congress Control Number: 2007943247
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microlm or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,
1965, in its current version, and permission for use must always be obtained from Springer. Violations
are liable for prosecution under the German Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specic statement, that such names are exempt from the relevant protective laws
and regulations and therefore free for general use.
987654321
springer.com
Preface
In almost all areas of daily life the demands on the communications infrastructure
increased dramatically. With regard to the volume of data that has to be trans-
mitted, no matter whether we are dealing with public or private networks, indus-
trial fields or automobiles, needs will continue to rise. Consequently, the demands
on the bandwidth of communications systems will continue to increase since more
and more video data streams with high picture quality (IP TV), in addition to tele-
phone and data circuits, will be transmitted. More data streams are coming about
through the connection of an increasing number of Wi-Fi hotspots with high capa-
cities. All these services require a basic infrastructure with high capacities which
only optic technologies can offer. The expansion of the DSL network will bring
glass fibers closer to the end customer and will generate a demand for easy-to-
install, efficient and favorably priced cabling solutions in buildings. In such a case
polymer optical fibers are a veritable alternative.
After POF demonstrated its performance capabilities in industrial use and in
automotive engineering, nothing more stands in the way for employing these
optical solutions within buildings. The use of inexpensive visible LEDs, simple
plug connectors and insensitive cables and lines, allows for favorably priced sys-
tems which any private user can even install himself whenever required. In con-
trast to Wi-Fi or Powerline, POF is immune to interference and always guarantees
high capacity in point-to-point connections. In combination with new electronic
solutions and coding scheems it is possible today to bridge distances of over
100 m at a data rate of 100 Mbit/s using standard step index POF. Thus the tech-
nical basis is laid for large-scale use in building cabling. It now remains the task of
the component manufacturers to make such systems available to the market as an
economical alternative.
The main emphasis of polymer fiber applications for data transmission lies in
Japan as well as in Germany, Spain and Italy. Germany plays a leading role in
many of these applications.
The Polymer Optical Fiber Application Center (POF-AC) at the Nuremberg
University of Applied Sciences has developed over the last few years into a Euro-
pean Competence Center for POF. The POF AC and Leoni have cooperated close-
ly and successfully for years.
For many years, the Leoni AG has been one of the leading manufacturers of
POF and fiber glass cables for use in mobile networks (cars, trains, airplanes), in
automation technology and in sensor technology.
VI Preface
Leoni has also recently gone into the manufacture of glass fibers and the requi-
site preforms. Therein lies a particular main focus on the production of multimode
and special fibers as described in this book.
The areas of application range from building networks to medical technical
applications used in spectrometric systems. Furthermore, Leonis developers have
presented at conferences held over the last few years new solutions for optical
pressure-sensitive sensors based on special plastic fibers.
In all these areas of activity Leoni has aimed at combining its experience with
optical waveguides from the field of telecommunications with the knowledge and
the solutions from the market for special and POF fibers, thus making available
new practical solutions to users in the most diverse fields.
In 1997, Andreas Weinert, working today with colleagues at Leoni Fiber
Optics, published one of the first comprehensive POF books.
The book POF - Optical Polymer Fibers for Data Communication published
in 2001 provided an overview of POF technologies and in the meantime is out of
print. This new edition is now a handbook which, in addition to describing POF
and thick core glass fibers, can also be used for short-distance communication.
This specialized reference book is intended to help users and developers to
obtain information themselves quickly and comprehensively about the state-of-
the-art development of POF fibers and to become acquainted with their perfor-
mance capabilities. In a summarized form it conveys a number of experimental
results along with new trends in development and is a valuable adviser for develo-
pers of POF systems. The physical fundamentals as well as practical applications
are presented in a simple and understandable manner.
In the past few years polymer optical fibers (POF) and their applications have
continued to develop at a dizzying pace. This was the decisive factor in 2005 to
decide on completely revising the book, POF - Optical Polymer Fibers for Data
Communication, which appeared in German in 2001 and in English in 2002.
Before you now lie the results of two years of work - almost double the number of
pages of the First Edition.
One essential reason for the new edition is the diverse results which have been
obtained at the Polymer Optical Fiber Application Center at the Nuremberg Uni-
versity of Applied Sciences (POF-AC) since its inception in 2001. The scientific
director of the POF-AC has written the majority of the sections in this book. Dr.
Christian-Alexander Bunge from the Technical University of Berlin has contri-
buted two sections (Microstructured Polymer Fibers and the Simulation of Optical
Fibers).
The organization and layout of the new book are essentially based on the
following considerations:
All parts of the first edition have been taken over in order to provide the many
newcomers in the field with the opportunity to completely understand the
contents, without having to buy the first volume - which in the meantime is
out of print.
In addition to the optical polymer fibers many details on other thick-core
fibers, e.g. glass fiber bundles and plastic-coated glass fiber, have been added.
Many of these fibers not only have the same applications, but also similar
characteristics and place similar demands on measurement techniques.
Whereas the first edition mostly summarized the results from technical lite-
rature, the new edition now presents the POF-ACs own measurement results
on practically all fibers. Consequently, this new book presents and documents
the first five years of our institute.
The individual chapters correspond to the subject matter of the first edition.
The organization and sequence, however, have been adapted to the changed
points of emphasis. For example, the waveguides are now dealt with in a
separate chapter. The chapters on Fibers (No. 2) and on Systems (No. 6) are
among the predominantly new sections. Together these two parts form the
core contents of the book and document the progress of technology. Unfortu-
nately, however, these new parts will also become obsolete most quickly
since many new solutions will be found in the next few years.
The English edition is completely identical with the German one, including
all page numbers. No new content has been included.
VIII Editor's Preface
At the time when the first edition of this book appeared POF applications were
still exotic. Only in the fields of automation and lighting engineering had this
medium already been established. In the meantime, millions of vehicles drive
throughout Europe with polymer fiber on-board networks and the next generation
is right in front of the door. Many telecommunications companies are working on
solutions for transmitting ever higher bit rates via POF within apartments. The
Deutsche Telekom, for example, offers its customers a complete set for fast Ether-
net. Other large-scale applications for POF in multimedia applications will be
introduced into the market in the near future.
The authors are therefore very optimistic that this book will accompany the de-
velopment of polymer fibers from a niche existence to an important basis for data
and communications technology. In addition to use in telecommunications, sensor
technology and multi parallel data connections above all promise to be wide and
interesting fields for their use. A separate section is devoted to each of these areas
in Chapter 8. POF are not necessarily a rival of the established technologies such
as data transmission on symmetrical copper wires or radio. Different sections of
the book show how the diverse technologies can be combined in an optimal
fashion in order to achieve the best solutions technically and economically.
We have done our best to represent the scientific results and the products
available on the market as completely and as impartially as possible. Nevertheless,
we are aware of the fact that this goal can only approximately be attained. Should
a manufacturer or an institute mentioned in the book not feel being sufficiently
represented - this was not intended. The POF-AC gladly offers all interested
parties support in gaining access to the growing POF community. The POF-AC
offers scientific activities such as the ITG Sub Committee 5.4.1 Polymer Optical
Fibers as well as technical information such as the POF-Atlas as a German
POF product catalog.
As editor of the Second Edition I would like to express my thanks to all
colleagues at the POF-AC Nuremberg, the Nuremberg University of Applied
Sciences and not least my family for their support during the last two years and
their giving up so many hours that I actually should have spent with them. A
special thank to Prof. Economides (Berlin), who is the translator, and the Prof.s
Poisel and Hartl for their assistance during the correction phase.
I wish all readers much pleasure when reading this volume. The aim and inten-
tion of this book is to provide you with some help, support, information and food
for thought in your work. Please excuse the unavoidable errors and mistakes and
do feel free to pass on your reservations, criticisms and ideas to us.
Olaf Ziemann
Scientific Director of the POF-AC Nrnberg
Chairman of the ITG Sub Committee 5.4.1 Polymer Optical Fibers
Member of the International Cooperative of Polymer Optical Fibers
as the responsible editor of the second edition, November 2007
Content
2. Optical Fibers p. 37
2.1 Fundamentals of Optical Fibers p. 37
2.1.1 Refractive Index Profiles p. 37
2.1.2 Numerical Aperture p. 39
2.1.3 Ray Trajectory in Optical Fibers p. 40
2.1.4 Modes in Optical Fibers p. 42
X Content
7. Standards p. 561
7.1 Standards for Polymer and Glass Fibers p. 562
7.1.1 Polymer Fibers p. 562
7.1.2 Plastic Clad Glass Fibers p. 564
7.1.3 Fibers in General p. 565
7.2 Application Standards p. 566
7.2.1 ATM Forum (Asynchronous Transfer Mode) p. 566
7.2.2 IEEE 1394b p. 569
7.2.3 SERCOS (SErial Realtime COmmunication System) p. 572
7.2.4 Profibus p. 573
7.2.5 INTERBUS p. 574
7.2.6 Industrial Ethernet over POF p. 575
7.2.7 D2B (Domestic Digital Bus) p. 578
7.2.8 MOST (Media Oriented System Transport) p. 580
7.2.9 IDB 1394 p. 582
7.2.10 EN 50173 p. 583
7.3 Standards for Measurement Techniques p. 587
7.3.1 The VDE / VDI Guideline 5570 p. 588
References p. 823
Translator P. 874
Biographies p. 885
List of Abbreviations and Symbols
Symbol Explanation
B Bandwidth (generally)
BAM Federal Institute for Material Research and Testing
BB Bromobenzene
BBP Benzyl n-Butyl-Phtalate
BER Bit Error Ratio
BK Broadbanf cable (Breitbandkabel)
BPSK Binary Phase Shift Keying
BR Bit Rate
BzMA Benzyl Methacrylate
c Velocity of light
cm Velocity of light in a medium
cv Velocity of light in vacuum (2.99792458 108 m/s)
C Constant value (generally)
Cmn Coupling coefficient between the modes m and n
CPD Capacity of the photo diode
CAN Controller Area Network
CCD Charged Coupled Device
CCP Customer Convenience Port
CD Compact Disk
CDC Compact Disk Changer
CDM Code Division Multiplex
CDMA Code Division Multiple Access
CMT Corrugated Metallic Tube
CNR Carrier to Noise Ratio
CSO Composite Second Order
CTB Composite Triple Beat
CYTOP Cyclic Transparent Optical Polymer (Asahi Glass Comp.)
d Fiber diameter
d Diameter of the cable lay unit (Chapter 4.2)
d Reciprocal grating constant
d 1, d 2 Various diameters
dGM(4) Field penetration depth, dependent on the angle of incident 4
dm Cladding thickness
dmin, dmax Minimum and maximum diameter
dray Beam diameter
D Wire diameter
D Distance generally
D Diffusion constant
D Dispersion constant
DA Diameter of the stranding basket
DK Insertion loss
Dm Average diameter of the cable lay up layer
Drec Reciprocal dispersion
D2B Digital Domestic Bus (serial bus for automotives)
XXIV Abbreviations and Symbols
e/o Electro/optical
E Receiver (Empfnger)
EEl Electrical field of the modes
ECOC European Conference on Optical Communication
EL Effective Laser Launch
ELED Edge emitting LED
EN European standard
EMD Equilibrium Mode Distribution
EOF Elastomer Optical Fiber
ETFE Tefzel
EVA Ethylen-Vinylacetat-Copolymere
g Index coefficient
g(t) Pulse response
GI Graded Index
GI-MPOF Graded Index profile multimode MPOF
GI-PCS Graded Index Plastic Clad Silica
GOF Glass Optical Fiber
GRIN Graded Index (continuously index ....)
I Current generally
Iph Photo current
IRMS Noise current (root mean square)
Ith Threshold current
IDB Intelligent Data Bus
IGPT Interfacial Gel Polymerization Technique
IR Infrared
ISDN Integrated Services Digital Network
ISM Industrial, Scientific, and Medical Band
ITG Informationstechnische Gesellschaft
(Information Technology Society)
k Boltzmanns constant
Kl (v) Bessel function
kr Radial component of the propagation vector
KF Correction factor
KIST Kwangju Instuitute of Science and Technology
KPCF Korea POF Communication Forum
XXVI Abbreviations and Symbols
l Peripheral order
L Length
L1, L2 Lengths of different optical paths
Lc Coupling length
LAN Local Area Network
LD Laser Diode
LED Light Emitting Diode
Low-NA Reduced Numerical Aperture
LWL Optical waveguide (in German: Lichtwellenleiter)
m Order of refraction
M Material dispersion parameter
M Highest group number
M ('z) Mode coupling matrix
M1, M2 Various monomers
MC Multi Core
MC-GOF Multi Core Glass Optical Fiber
MCVD Modified Chemical Vapor Deposition
MFC Mode Field Converter
MGDM Mode Group Division Multiplex
MIMO Multiple Input - Multiple Output
MMA Methylmethacrylat
MM-GOF Multimode Glass Optical Fiber
MOST Media Oriented System Transport (serial bus in automotives)
MPOF Microstructured POF
MP3 Compression method for music
MP-P Multipoint to Point
MP-MP Multipoint to Multipoint
MPEG Motion Picture Expert Group (data compression standard)
MQW Multi Quantum Well
MSI Multi Step Index
MUX Multiplexer
n Refractive index
n Number of layer (chapter 2.8.3)
n0 Refractive index of air (approx. 1)
n1 Rotational speed of stranding basket (chapter 2.8.3)
n2 Rotational direction and speed of the capstan gear (chap. 2.8.3)
n1, n2, n3 Refractive index in various media
nair Refractive index of air
ncladding Refractive index of the cladding
ncore Refractive index of the core
ncore, max Maximum refractive index of the core in GI fibers
nPMMA Refractive index of PMMA
N Number of guided modes
NA Numerical Aperture
Abbreviations and Symbols XXVII
o/e Optical/electrical
OIIC Optical Interconnected Integrated Circuits
OTDR Optical Time Domain Reflectometer
OVAL Optical Video/Audio-Link
p Impulse
P Profile dispersion
P Power generally
P0 Output power
P0x, P1x Power for measurements of connector losses
Peff Effective power
Pelectr, Pel Electrical power
PL Power at the length L
PL1, PL2 Power at fiber outputs
Popt Optical power
Pout Output power
Pr Backscattered power
Preceiv Received power
P(f) Power at the frequency f
PA, PA-6 Polyamide, Polyamide 6
PAM Phase Amplitude Modulation
PAVNET Plastic Fiber and VCSEL Network
PC Personal Computer
PC Polycarbonate
PC(AF) Partially fluorinated polycarbonate
PCS Plastic Clad Silica
PE Polyethylene
PE-FRNC Polyethylene flame-retardant/halogenated
PE HD Polyethylene (high density)
PE LD; MD Polyethylene (low density; medium density)
PFA Tetrafluoroethylen-Perfluoroalkylvinyl-Ether
PFM Preform method
PFM Pulse Frequency Modulation
PF-POF Perfluorinated POF
PhMA Phenyl-Methacrylate
pin-PD Photo diode with p-i-n-semiconductor structure
PLC Power Line Communication
P-LED Polymer LED
PLL Phase Locked Loop
PNA Phone Network Association
XXVIII Abbreviations and Symbols
PMMA Polymethylmethacrylate
PMMA-d8 Complete deuterinated PMMA
PMT Photo Multiplier Tube
POF Polymer Optical Fiber
POF-AC Polymer Optical Fiber Application Center
at the University of Applied Sciences Nrnberg
POF-ALL Paving the Optical Future with Affordable Lightning-Fast
Links (EU project: www.ist-pof-all.org)
POFTO POF Trade Organization
PP Polypropylene
P-MP Point to Multipoint
P-P Point to Point
PS Polystyrole
PRBS Pseudorandom Bit Sequence
PSK Phase Shift Keying
PTC Positive Temperature Coefficient
PTFE Polytetrafluoroethylene
PUR Polyurethan (thermoplastic)
PVC, PVC 90 Polyvinylchloride, Polyvinylchloride 90C
PVC flame ret. Polyvinylchloride flame retardant
r Radius generally
rk Radius, which is not remained by helical rays (Fig. 2.7)
R, Responsivity
R Electrical resistance
R Bending radius
R Radius vector
RH Relative Humidity
R Radius of MC fibers
RC-LED Resonant Cavity LED
RIE Reactive Ion Etching
RML Restricted Mode Launch
s Pitch length
s Axial distance of fibers
sH Produced pitch length
S Transmitter (German: Sender)
S Backscattering coefficient
S Safety coefficient
Sin, Sout Modulation signal at input and output
SC Strain Compressed
SCM Subcarrier Multiplex
SDM Space Division Multiplex
Abbreviations and Symbols XXIX
t Time
t1, t2, t3 Different propagation times
ta, tb Efficiency parameter (in Tab. 3.3)
tf Fall time
ti Length of the launched impulse
tin, tout Pulse width at input and output
tr Rise time
tA Aging time
tL Life time (durability)
T Transmitter
T Temperature generally
Tmin, Tmax Minimum and maximum temperature
TG Glass transition temperature
TS Reference temperature
T-DSL Telekom ADSL
TDM Time Division Multiplex
TDMA Time Division Multiple Access
TTP Time Triggered Protocol
v Group velocity
v Normalized propagation constant
vm Draw velocity
v, x, y, z Variables for the calculation of SNR (chapter 1.3.3)
V Fiber parameter for the calculation of the mode number
XXX Abbreviations and Symbols
Figure 1.1 shows an overview of the electromagnetic spectrum. The zone of opti-
cal waves includes the ultraviolet, visible, and infrared ranges.
frequency [Hz]
1019 1018 1017 1016 1015 1014 1013 1012 1011 1010 109 108 107 106 105 104 103 102 10 1
10-11 10-10 10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 1 10 102 103 104 105 106 107 108
wavelength [m]
The interesting ranges as far as optical transmission is concerned are the near
infrared range between 850 nm and 1,600 nm for SiO2 fibers (glass optical fiber,
GOF) and the visible range between 520 nm and 650 nm for polymer optical
fibers (POF) because these show the least attenuation.
A detailed overview of the optical range is offered in Fig. 1.2; the white line
qualitatively represents the attenuation behavior of PMMA-POF.
wavelength [nm]
300 400 500 800 1000 2000
P G
O O
F F
5 4 3 2 1 energy [eV]
UV range visible range IR range
Fig. 1.2: UV-, IR- and visible range of the electromagnetic spectrum (POF: Polymer
Optical Fiber, GOF: Glass Optical Fiber)
When light propagates in a medium such as a polymer, the speed of this light is
reduced. The ratio of the vacuum light speed cv to the speed in the medium cm is
referred to as the medium's refractive index n, which is expressed as follows:
cv
n
cm
Apart from the speed, the wavelength O of light passing through a medium is
also altered, while the frequency f and thus the energy W remain constant. In
Fig. 1.3 we see a light ray entering the optically denser medium at an angle 4 and
being refracted at an angle D to the axis of incidence. A part of the light is reflec-
ted. Refraction is expressed with:
sin 4 n2
.
sin D n1
being transferred to the other medium when angle D reaches a critical value;
instead, it is completely reflected. For the boundary case of total reflection, i.e.
when 4 = 90q, the following applies:
n1
sin D T
n2
n1 optically
4 4 thin 4
n1 < n2
DT
optically
n2 dense
D D D
An optical fiber consists of a highly transparent core having a refractive index ncore
and a surrounding transparent cladding having a refractive index ncladding. To
ensure that a light ray that has entered the fiber can be guided along it, the
following must hold true: ncore > ncladding (Fig. 1.4), so that below a certain angle
4max total reflection takes place at the interface between the core and the cladding.
The surrounding medium is air with the refractive index n0 | 1.
ncladding cladding
n0
ncore J
Dmax core
4 max
refractive index n
Rays that strike the end face of the fiber at an angle greater than 4max are no
longer completely reflected at the core/cladding interface; instead they are partly
refracted into the cladding so that they are no longer completely available for
4 1.1 Light Propagation in Optical Fibers and Waveguides
transmitting a signal. The following example illustrates how even small diffe-
rences here can have great effects: with a core refractive index of 1.56 and a clad-
ding refractive index of 1.49 the critical angle of total reflection is 72.77. Thus,
light rays with a maximum angle of max = 17.23 with respect to the fiber axis,
can propagate.
Should the propagation angle exceed this value by only 0.001, then the
reflection coefficient is reduced from 100% to 95%. With this angle 310 reflec-
tions per meter will result in a fiber with a diameter of 1 mm. The remaining light
output power would then be 0.95310 = 1.2 10-7 which corresponds to a loss of
69 dB.
The shape of the waveguide can vary greatly as the three examples in Fig. 1.5
demonstrate. On the left is a singlemode glass fiber which is almost exclusively
used today in the field of telecommunications. A planar waveguide can be seen in
the middle and on the right a semiconductor laser in cross-section with which an
optical waveguide is also formed.
: 10 m 5 u 5 m 2 u 0.5 m
Fig. 1.5: Examples for optical waveguides
Should the waveguide have very small dimensions in the light wave length
ranges, then the ray optics description is not sufficient. As can be read in standard
works, e.g. [Vog02], the number of possible propagation angles (modes) dimi-
nishes with decreasing diameter. An extreme case thereby is the singlemode wave-
guide which will be introduced in the next section.
The different types of optical fibers are described in detail in this chapter; an over-
view of the standards can be found in Chap. 7.2. The two following pictures show
an overview of the different refractive index profiles. It can easily be seen that not
only the index profiles, but also the refractive index differences - which determine
the numerical aperture - and the core diameter vary considerably.
The fibers with the greatest core diameter used in data transmission can be seen
in Fig. 1.6. Standard SI-POF has an approx. diameter of 1 mm at a NA of 0.50.
GI-POF with this diameter, but with a somewhat smaller NA, has also been
available for a short time ([Yoo04]).
6 1.1 Light Propagation in Optical Fibers and Waveguides
0.00
The index profiles of different glass and polymer fibers can be seen in the next
fig. 1.7. A hybrid is the so-called PCS - polymer clad silica - i.e. a silica glass fiber
with a polymer cladding.
The singlemode glass fibers have the smallest core diameters. For use in the
1300 nm to 1600 nm range these fibers only have a core diameter of about 10 m.
Special fibers, e.g. for erbium-doped fiber amplifiers or for fibers with non-linear
properties, can even lie in the range of only 2 m for the core diameter. These
fibers are not the subject of this book. We can recommend [Vog02] as a work with
an excellent overview for this area.
0.000
There are no general international guidelines for the designation of optical fibers.
Due to the enormous variations in the different parameters it is hardly possible to
give all fibers clear-cut designations since these would otherwise be much too
long. The following list of parameters could be used for providing names.
For the communications engineer there is no doubt that, apart from spectral
attenuation, the most important parameter of an optical fiber is its bandwidth. In
waveguides there are usually different optical paths possible (with the exception
of the very thin singlemode waveguides). The different lengths of travel along
these different light paths lead to different time delays for an optical pulse, as
illustrated schematically in Fig. 1.8.
t
input pulse output pulse
1 0 0 1 1 1 0 0 1 1 0 1 0 0 1 0
a)
fiber b)
length
c)
d)
e)
Fig. 1.9: Influence of mode dispersion to data transmission
An optical signal is launched into the fiber and is switched on and off at the
respective bit rate (curve a). With the increase in pulse broadening, the bit edges
get more and more fuzzy (signal sequence with downwards increasing trans-
mission length). As long as the amount of broadening is clearly less than the bit
time, the signal will remain easily identifiable (curves b and c). If the width of the
edges exceeds the bit time, the signal can no longer be detected (curves d and e).
The process of pulse broadening is called dispersion. The difference between
different light paths described here is called mode dispersion (any possible con-
dition of the propagation of light in a waveguide is called a mode). Apart from
mode dispersion there is also the phenomenon of chromatic dispersion (different
time delays for different wavelengths) and polarization mode dispersion; however,
we will not be dealing with these two phenomena at this stage.
The second important quantity which determines signal quality is the signal-to-
noise ratio (SNR). In POF systems the optical receiver alone is almost always res-
ponsible for the noise. Under certain circumstances you have to pay attention to
mode distribution noise within thin multimode glass fibers. In modern singlemode
glass fiber systems there are even many more sources of noise, e.g. fiber ampli-
fiers.
The following sections provide a short insight into the fundamentals of analog
and digital transmission methods, especially in regard to the various sources of
interference. The effects which are important for short-range communication are
particularly elucidated.
10 1.2 Digital and Analog Signal Transmission
U(t)
Here, analog has two meanings. First, the signal is measured at any random
time t. Furthermore, U(t) can take on any value. When a signal is digitized, two
things usually happen. While the signal is being sampled, the values are not read
in continuously but rather at discrete points (Fig. 1.11).
U(t)
sampling points
t
The second step is that the voltage U must not take any random value but only
particular or discrete ones (quantization, Fig. 1.12).
U(t)
sampling points
quantization
steps
Figure 1.12 shows that the values no longer lie exactly on the actual curve but
always at the next quantization level.
The digitalization of a signal always distorts the original one. Initially, the
range of the recorded frequencies is limited by the choice of the sampling rate
(sampling points per second). According to the Sampling Theorem, only signals
whose upper limit frequency ful is equal or smaller than the half of the sampling
rate fs (ful d fs/2) are fully transformed into the digital signal. Figure 1.13 graphi-
cally illustrates this problem.
U(t) U(t)
t t
In the left figure, the sampling points are spaced sufficiently close to each
other. In the right figure, the signal also changes very quickly between the samp-
ling points (higher frequencies are present). The original signal cannot be recon-
structed from the points that are too far away from each other. Quantization also
distorts the signal. The difference between the actual value and the quantization
level can be interpreted as added noise (Fig. 1.14).
12 1.2 Digital and Analog Signal Transmission
The signal of a CD player is a good example of a digital signal. The human ear
can detect frequencies up to 15 kHz to 18 kHz. Music is stored on a CD with
44,200 values per second. Thus, signal frequencies up to 22.1 kHz can be recor-
ded. Each of these sampling values is divided into 65,536 amplitude steps (216).
The original continuous signal is thus broken into 44,200 numbers per second, for
example,
23,546; 22,125; 19,714; 13,120 etc.
The errors that occur as a result of this quantization are negligible. If the avail-
able levels are equally distributed to positive and negative voltages, for example to
the range between +1 V and -1 V, the deviation of the real value to the next quan-
tization level can be at most 15 V. This is a difference of approximately 96 dB.
This is the equivalent of the difference between a whisper and the sound of a loud
airplane propeller 5 m away.
In the world of digital signal processing, the numbers are represented with the
symbols 1 and 0. 65,536 values can be represented by 16 binary characters
(16 Bit). As a binary number, the signal above would then look like this:
0101101111111010,0101011001101101,0100110100000010,0011001101000000
When the signal is transmitted, the commas are, of course, omitted. Both sym-
bols are characterized by various signal states, for example, -1 V for the 0 and
+1 V for the 1 or also light off for the 0 and light on for the 1.
Figure 1.15 illustrates the difference between the original analog signal and the
binary signal generated by means of digitalization.
t
U(t) digital signal
t
The figure shows that the digital signal changes much faster than the analog
signal. This is easier to see if you consider that music at a maximum frequency of
approximately 20 kHz must transfer 44,20 16 = 707,200 bit/s. Why digital
signal processing still has many advantages over analog processing is explained in
the next paragraph.
U(t)
fg f fg f fg f
Even though some of the sources of interference have been somewhat exagge-
rated, they are intended to show the problems involved in transmitting analog
signals. Each element involved in the transmission can distort the original signal.
These forms of interference can only be eliminated in exceptional cases and are
also relevant to digital transmission, as Fig. 1.17 illustrates.
14 1.2 Digital and Analog Signal Transmission
U(t)
fg f fg f fg f
input distortion attenuation noise output
signal bandwidth limitation ext. perturbations bandwidth limitation signal
The signal behind the receiver appears to be strongly distorted; but this is where
the digital trick sets in. The receiver knows that the signal can only have one of
two levels and that the signal was transferred with a specific bit rate. This know-
ledge is then used to reconstruct the signal free of error.
The signal is at first filtered to eliminate as much noise as possible. A decision
threshold is then defined. For binary signals, this is the border between 0 and
1. At the sampling points that correspond precisely to the bit raster, the signal is
compared with the threshold and then reconstructed. This procedure is shown
schematically in Fig. 1.18.
U(t)
010011
t
received filtering decide sampling reconstructed
signal threshold signal
clock recovery
Fig. 1.18: Signal reconstruction in digital transmission systems
Although the signal was clearly distorted, the complete reconstruction of the
original bit sequence is possible. This is what users recognize as CD quality.
The question still remains as to how the analog signal, i.e. the music, is restored. A
digital analog converter is used for this purpose. In the example shown, the con-
verter uses 16 bits in the 65,536 intervals between -1 V and +1 V. The signal is
then subsequently filtered to eliminate the resulting harmonic waves and then is
ready for use (for example, to be fed into a loudspeaker). This step is not needed
for communication between the digital devices.
1.2 Digital and Analog Signal Transmission 15
0.4
0.2
0.0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
voltage [V]
As you can see in Fig. 1.19, the levels +1 and -1 are the most likely ones but
there are also other levels. In the case at hand, the decision threshold is 0 V. It
appears that it is still possible to clearly differentiate between the two symbols
despites the noise. However, the Gaussian curve never drops completely to 0. This
is shown in Fig. 1.20 in which the same curve is scaled logarithmically.
16 1.2 Digital and Analog Signal Transmission
10 -1
Symbol 0 Symbol 1
10 -2 with noise with noise
10 -3
10 -4
optimal
-5
decision
10 threshold
10 -6
10 -7 area of bit
errors
voltage [V]
10 -8
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
Fig. 1.20: Generation of bit errors in digital transmission systems due to noise
Here you can see that the curves for 0 and 1 overlap. This means that at
times a 0 can be distorted to such a degree that it is detected as a 1 or vice
versa. The hatched area represents exactly the bit error provided that the decision
threshold is actually 0.
Integrating this area and putting them into relationship with the integrals of the
probabilities for 0 and 1, yields the bit error probability (Bit Error Ratio
BER). In this case the result would be 10-7 . Returning to the CD example, this
would mean one error every 10,000,000 bits or, on the average, one error every
14 s. For normal use, this would not pose a major problem. However for data
transmission, this would be entirely inadequate. If saved as a file, this book would
consist of about 109 bits. At this error rate, 100 bit would be faulty. At best, this
would result in incorrect characters or errors in the illustrations. A more likely
scenario would be, however, that many of the errors would cause the system to
crash and make the file unusable. Considering all the time we authors have
invested in this documentation, this would be anything but a welcome course of
action. Hence, data connections should be considerably more reliable (for
example, BER < 10-15). The characteristics of noise anticipate the conclusion that
error-free data transmission is nothing more than an illusion. But this is where we
can call on statistics for help. If the voltage levels are increased to 2 V, the error
probability drops to approximately 10-25 for the same noise level. For our CD
signal, this would mean one error in 9 billions years, i.e., this book could be trans-
mitted several trillion times without error.
1.2 Digital and Analog Signal Transmission 17
However, there are other ways to reduce the error probability. Particular coding
schemes make it possible to detect individual errors on the receiver end (FEC:
Forward Error Correction). This is achieved by adding so-called control bits to
the signal flow. The bit rate is usually increased by only a few percent. On the
receiver end, practically all errors can be corrected. These procedures are often
used in mobile communications where, for example, the less than ideal trans-
mission channels cause an inordinate amount of errors.
Laser Noise:
Laser diodes and LEDs are normally very low noise and stable sources. Clean
driving is especially important for signal quality. Edge emitting laser diodes in
particular can show high noise if light is reflected back from the transmission link.
This can hardly be avoided since about 4% of the light is already reflected at the
very first frontal area of the fiber. All reflections which appear within the cohe-
rence length are disruptive. These can be on occasion many kilometers when using
high quality laser diodes. In order to avoid fluctuations in laser performance
caused by reflections, anti reflection coatings, optical insulators and special low-
reflecting plug-in connections are used.
The LEDs used in most POF systems have a coherence length of a few m and
are hardly influenced by reflections. Conventional LDs or VCSELs can be dis-
turbed by reflections. One advantage of using multimode fibers is their greater
diameter compared with the emitting surface of the laser. Even if much light is
reflected, only a small fraction reaches the active surface of the laser (Fig. 1.21) so
that the effect is negligible.
reflected light
PCS
core = 200 m
VCSEL
active = 10 m
iD2 2 e ID B
ID is dark current and lies in the nA range with normal pin-photodiodes so that
the additional noise is also negligible.
Receiver Noise:
The most important source of noise for the systems under consideration in this
book is receiver noise. In principle, every optical receiver can be described, at
least in a rough comparison, as a combination of a photodiode, an ohmic input
resistor and an amplifier (transistor or operational amplifier, Fig. 1.22).
POF
amplifier
input
resistance
The photo current is converted into a voltage at the input resistor which then
continues to be amplified. The greater the resistance, the higher the signal voltage.
On the other hand, however, every resistor generates a thermal noise according to
the following equations.
2 2 4 k T B
uth 4 k T B R or: ith
R
f3 dB 1 2 S R C
and a minimum bandwidth of the receiver - which corresponds to one half the
bit rate - the maximum input resistance may amount to:
1
R
bit rate S C
With a diode capacitance of 0.5 pF a R of about 500 , as indicated above,
results. This calculation is only a very rough approximation, but it does show the
principle problem involved.
Diodes intended for use in singlemode fiber systems only have to be a little bit
larger than the core diameter of the fiber. A typical size for photodiodes is from
30 m to 50 m. Their capacitance only lies at some 10 pF, but they do allow
large input resistances with high data rates and thus have good sensitivity. For
thick polymer fibers and PCS, however, photodiodes with very much larger sur-
face will be needed. Their capacitance lies at some nF or some tenths nF. With
high data rates the input resistance must correspondingly be reduced and thus the
sensitivity. This is the only indirect influence of the fiber diameter on the possible
bit rate with thick optical fibers.
20 1.2 Digital and Analog Signal Transmission
Modal Noise:
A special kind of noise only appears in multimode fibers: mode distribution noise.
In multimode fibers light propagates in different specific modes, whereby each
one possesses its own distribution of power over the fiber cross-section. The kind
of power distribution between the modes changes with tiny changes in the exterior
conditions, e.g. temperature, wavelength of the transmitter or also vibrations of the
fiber, but the overall power remains constant. Figure 1.23 shows examples of
power distribution of modes (multimode GOF/POF at 650 nm, laser excitation).
Fig. 1.23 Examples for energy distribution of modes in multimode fibers (left: 50 m
GI-GOF, right: 1 mm SI-POF with much higher mode number)
opt. power
lost power
fiber with
fluctuating cross section of
power the second fiber
distribution time
The intensity of the modal noise depends on the number of the power maxima
on the cross-section (in the order of magnitude of the number of modes) and the
differences in power. For 50 m multimode glass fibers the mode distribution
noise is typically only a good 20 dB below the overall transmitted level ([Vog02]).
This is insignificant for digital transmission, but makes analog transmission im-
possible.
In order to avoid modal noise, transmission must take place either with single-
mode or with very many modes since this effect is inversely proportional (statisti-
cal effect) to the root of the number of modes. Polymer fibers with several million
modes are therefore of interest for analog data transmission and are much better
suited than GI glass fibers.
Whenever digital and analog signals are transmitted, there is virtually always an
electrical voltage U(t) present at the beginning and at the end of the transmission
path. To cover the distance, other physical parameters can also be used. The chan-
ging of these parameters is referred to as modulation. As we will see, when we
want to transmit signals with light, there are a number of problems to consider. In
general, the desired parameters can be modulated analog, i.e. with varying
strength, or with only a small number of discrete steps. Below we will show some
examples of binary, digital modulation procedures. We shall start with the electri-
cal transmission procedure depicted in Fig. 1.25:
U(t) U(t)
t t
transmitter receiver
On the transmitter end, the signal is the voltage difference between two con-
ductors (for example, the wires of a twisted data cable or the core and shielding of
a coaxial cable). If we neglect such factors as attenuation and band limitation, the
signal is tapped again directly at the output end of the cable. This method is, of
course, the easiest but, experts know, also the one most susceptible to interference.
22 1.2 Digital and Analog Signal Transmission
t t receiver
source for
carrier
frequency
The carrier wave can be described with three parameters: amplitude, frequency,
and phase. All three parameters can be used for modulation. For binary signals
these are:
ASK: Amplitude Shift Keying; Amplitude Modulation
FSK: Frequency Shift Keying; Frequency Modulation
PSK: Phase Shift Keying; Phase Modulation
The principle is shown in Fig. 1.27.
signal
The advantages of optical communications are undisputed and have been know for
some time. The reader will undoubtedly be familiar with many of them. With the
low attenuation characteristics of modern singlemode fiber optic cables, many
hundreds of kilometers can be bridged with high data rates. If fiber amplifiers are
used, transcontinental communication is even possible. Particularly systems with
polymer optical fibers are well suited for short distances due to their insensitivity
to interference.
Light is an electromagnetic wave with a particular frequency. At 500 nm, i.e.
green light, this is 6 1014 Hz. No electronic component is able to process this
frequency. Photodiodes only measure the optical power of a light signal. Further-
more, optical sources cannot maintain their frequencies nearly as accurately as
electric oscillators. Directly modulating the parameters frequency or phase (and
polarization in particular) of light is only possible with so-called heterodyne
receivers. This involves modulating an extremely frequency-stabilized laser on the
transmitter end and heterodyning it on the receiver end with a second laser that is
just as stabilized. A mixed frequency is produced at the photodiode (the difference
between the both laser frequencies), which can be further processed by the sub-
sequent electronic components. Heterodyne systems offer, at least in theory, the
best frequency economy of all optical systems. Nevertheless they have not gained
acceptance due to the many technical problems involved. For further information,
see [Fra88] and [Ziem95].
Therefore, the only parameter left for modulation is amplitude. A photodiode
measures the optical power that is converted into a proportional photo current.
Since the electrical power, measured at a resistor, is proportional to the square of
the current, the following relationship holds true:
Pelectr ~ Iph ~ Popt
Although the electric field of the transmitted light can take on positive and
negative values, the actual measured parameter is always positive. This is a
significant difference compared to electrical communication systems. Let us take a
simple binary signal transmission as an example. In the electrical stage, the bits
are switched between -1 V and +1 V at the transmitter. The decision threshold is
set to 0 V. The levels 2 mW and 0 mW should be selected for an optical signal.
The decision threshold should be at 1 mW (see Fig. 1.28).
U(t) Popt(t)
+1 V 2 mW
t
0V 1 mW
t
-1 V 0 mW
Initially, both systems appear similar. In the Fig. 1.29 below, additional attenu-
ation is inserted, for example by increasing the temperature or aging the trans-
mitter. The level drops to 40 %.
U(t) Popt(t)
+1 V 2 mW
t
0V 1 mW
t
-1 V 0 mW
Fig. 1.29: Electrical and optical, digital signal transmission with attenuation
In the electrical system, both symbol levels are equally decreased. If the noise
is not too large, it will still function flawlessly. In an optical system, the zero level
will, of course, remain unchanged while the 1 level falls below the threshold.
The system then no longer functions. This problem, of cause, can be solved. Ca-
pacitive couplings or decision threshold controls are used, or the decision level is
set initially so low that the threshold is always above the noise level of the 0
symbol (Fig. 1.30 to 1.32).
All of the methods discussed have their advantages and disadvantages which,
however, will not be elaborated further here. Various commercial systems for POF
use the third method.
Popt(t)
2 mW
1 mW
t
0 mW
Fig. 1.33: Optical signal transmission with a frequency modulated subcarrier
It should be noted that the optical power curve still represents a much higher
optical frequency. One advantage of the method shown is that the average optical
output power remains unchanged from the succession of the 0 and 1 symbols.
Laser diodes are particularly well-suited for this type of modulation.
modulation current
As described above, lasers are best operated with a bias current. In direct power
modulation, this would be slightly below the threshold. In subcarrier modulation,
the bias current is set higher than the laser threshold so that the laser is always
operated above the threshold current as shown in Fig. 1.34.
Thus the tools for transmitting analog and digital signals are now also available
in optical communications. Apart from modulation, coding is of great importance.
We would like to refer the reader to the relevant technical literature, as this would
go beyond the scope of this introduction.
The next section of the introduction deals with various network architectures. Here
too, attention will be drawn to the unique features of optical communications. The
first topic is that of Point-to-Point transmission and distributed systems. The
Point-To-Point system (P-P) is the simplest form of data communication, in-
volving only one transmitter and one receiver. The second case deals with connec-
ting several devices to each other. In this case, there can be one transmitter and
several receivers (Point-to-Multipoint, P-MP), several transmitters and one recei-
ver (MP-P) or also several transmitters and several receivers (MP-MP, see
Fig. 1.35).
P-P P - MP MP - P MP - MP
It should be noted that it is often necessary to send data in both directions (up-
stream and downstream). For example, a network can be P-MP in one direction
and MP-P in the other. When more than two stations are to be connected to each
other, several solutions are conceivable. These are discussed in the sections below.
First we will briefly discuss active and passive networks. Then a network structure
or topology must be selected, for example, a tree-shaped network or bus structure.
Finally, a multi-access method must be selected.
A P-MP case will serve as an example of the difference between an active and
passive network. Figure 1.36 illustrates both possible solutions for this connection.
1.3 Network Architectures 27
P - MP P - MP
passive active
Passive means in this case that the receivers are physically connected to the
same medium. As a result, every receiver receives the complete signal, even if it is
intended for other receivers. This is an ideal situation for distribution services such
as broadcast radio. With active networks, there is an interfacing device between
transmitter and receivers that has the function of distributing the signal to the
proper receivers. In principle, this interface functions as a switch, which is the
reason why they are called exactly that: switches. If you take a closer look at this
architecture, you can see that it actually consists of several P-P connections.
Since the aim of this book is only to describe transmission technology, all other
active networks can be regarded as some form of a Point-to-Point solution. You
can say that every architecture can be made active by inserting the right number of
switches at the branch points.
The functionality of the active points can be formed in many different ways.
The active points, for instance, could be used solely as an amplifier. This means
that the signals are forwarded without regard to their destinations. Address and
access control must be performed by other elements. A multiplex function means
that the signals are routed according to their destinations and, in a MP-P structure,
are combined. Some form of access control must be available also in this case to
prevent collisions. A complete switch also takes on the access control functions.
This can be made by rejecting non-processable data or by serving as a data buffer.
The network structure describes the topology of the data connections. It can refer
to the physical structure i.e. the way the cables are arranged as well as to the
logical structure, i.e. the flow of the data streams. Figure 1.37 illustrates the most
widely known structures.
All modern data networks are designed as active star-type structures. The best-
known examples are Switched Ethernet and the ATM network (Asynchronous
Transfer Mode). Tree networks play an important role for television distributing
networks. Here it is crucial that each receiver receives the complete transmission
signals. Data networks for residential areas, such as USB or IEEE 1394 function
28 1.3 Network Architectures
logically as tree-shaped networks; however, they are physically set up as P-P con-
nections. This means that several other devices can be connected to one device.
Each device then forwards the complete data.
In all passive networks, one problem becomes apparent, in particular when you
examine the direction to the central element. Several transmitters can access the
same medium. To prevent mutual blocking, this form of access must be con-
trolled. This will be the subject of the next section.
The goal of all multiple access methods is the administration of a shared channel
(shared medium). The various parameters of the channel which can be used are
time, frequency or amplitude.
P - MP MP - P
TDM TDMA
1 1
2 2
3 3
t
t
t
t t
1 1
2 2
3 3
buffer memory
t t
t t
t t
f1 P - MP, FDM
band pass 1
f2 band pass 2
f3 band pass 3
f1 MP - P, FDMA
band pass 1
f2 band pass 2
f3 band pass 3
The advantage over TDM is that selection of the band pass filters determines
the synchronization. If these are permanently built in, there is practically no inter-
ference from other stations. But unfortunately it is then no longer possible to
dynamically allocate the capacity. This is why the frequency channels are often
assigned by the central node temporally on demand.
1.3 Network Architectures 31
U(t)
50 :
15 k: 15 k: 15 k: 15 k: 15 k: 15 k:
The minimum loss of 3 dB (half of the optical power) is easy to see for the
X-coupler as for the Y-divider. Because of the reversibility of the light path, this
also applies to the Y-coupler. If the divider at the far right is used as a coupler, its
attenuation will also be at least 6 dB. With polymer fibers, additionally the losses
that are unavoidable for multimode fibers must be taken into account.
There is a trick, however, for coupling and splitting light waves without
losses for singlemode fibers. If the light segments to be separated or combined
differ in wavelength or polarization, the proper WDM couplers (or polarization
couplers) can be used. Otherwise WDM systems would not be feasible with over
100 channels. For polymer optical fibers, the appropriate WDM components are
relative complex, as indicated below.
The special features of optical communications are adverse for particular
multiplex methods. If, for example, you want to use time division multiplexing
with two transmitters and two receivers that are located at two different sites, they
must be arranged as shown in Fig. 1.44.
1.3 Network Architectures 33
transmitter 1 receiver 1
The TDMA method ensures collision-free use of the shared transmission link.
The optical components yield a minimal loss of 3 dB + 3 dB = 6 dB. If the signals
could be combined before the optical transmitter and split behind the receiver, this
loss could be prevented. This would, for example, also increase the range. When
designing an optical transmission system, careful consideration should always be
given to the question as to which functions can be implemented better optically
and which ones better electrically.
disturbance,
transmitter 1 transmitter 2
depressed
x dBm y dBm
by v dB
channel
attenuation z dB
receiver 1 receiver 2
in in
fiber line
out out
receiver 1 receiver 2
transceiver 1
transceiver 2 time
transport: 12 21 12 21
These types of solutions are very good, particularly for systems with low and
medium data rates and short distances, for which the POF is also used, and can be
implemented at low cost since the entire data processing is performed in integrated
circuits.
A second possibility is frequency division multiplex, by which the data of both
directions are modulated upon different carrier frequencies (Fig. 1.47).
1.3 Network Architectures 35
isolation filter
transport transport
from 1 from 2
to 2 to 1 frequency
source O1 source O2
Y-splitter Y-splitter
in in
This procedure has the big advantage that the full capacity of the fiber medium
is available for each direction. Continuous operation without additional delays is
possible. The NEXT suppression is performed by optical filters placed before the
receivers. Signal processing is not necessary. A disadvantage is that two different
transceivers are always required on one link. However, this is a question concer-
ning the system concept. WDM is particularly interesting for fast data transmis-
sion, such as IEEE 1394. WDM also provides efficient solutions for systems with
asymmetric data rates.
2. Optical Fibers
The term optical fibers indicates special forms of optical waveguides, the most im-
portant special features of which are:
rotationally symmetrical cross-section
flexible
can be produced in great lengths
The characteristics of optical fibers are determined by a multitude of possible
constructive details. For example, the material selected primarily determines the
attenuation and the thermal stability. On the other hand, the optical bandwidth, in
essence the transmission capacity, is determined by the refractive index profile.
This is most likely the reason why most optical fibers are named after their index
profile. All current variations will be presented in the following sections.
The properties of wave guiding through a fiber are governed largely by the profile
of the refractive index of the core and cladding. In a step index profile fiber the
refractive index is constant across the entire cross section of the core and cladding
(Fig. 2.1) while the light rays propagate along straight lines in the core and are
completely reflected at the core/cladding interface.
r
a
refractive
index n(r)
-a
The profile of the refractive index in the core and in the cladding is expressed as
follows:
n(r ) ncore for r d a
n(r ) ncladding for r ! a
ncore 2 ncladding 2
'
2 ncore 2
r
a
-a
refractive
index n(r)
Fig. 2.2: Principle of a fiber with a graded-index profile
Those rays propagating in the center travel a shorter distance, but because of
the higher refractive index there, they travel at a lower speed. On the other hand,
the smaller refractive index near the cladding causes the rays traveling there to
have a higher velocity, but they have a longer distance to travel. By choosing a
suitable profile exponent it is possible to compensate for these differences in tran-
sit time. For negligible chromatic dispersion the ideal profile exponent is 2. One
then speaks of a parabolic index profile.
2.1 Fundamentals of Optical Fibers 39
When light enters the fiber's input face at an angle 4max, it is refracted at an angle
Dmax (Fig. 2.3). Applying the law of refraction we have:
n0 sin4max ncore sinD max ncore sin(90 - J max )
n0 sin4max ncore cos J max
n0 sin4max
ncore 1 - sin2 J max , with ncladding ncore 2 sin2 J max
n0 sin4max
ncore 1 - ncladding ncore 2
The sine of the maximum incident-ray angle 4max is defined as the numerical
aperture AN (Fig. 2.3). The angle 4max is referred to as the acceptance angle, and
twice the acceptance angle is referred to as the aperture angle. Using the relative
refractive index difference ', the value for AN is obtained as:
n0
Jmax
4max
Dmax
Thus, the value of the numerical aperture (NA) is solely dependent on the diffe-
rence in the refractive indices of the core/cladding material.
Example: The refractive indices of a standard PMMA fiber are ncore = 1.49 and
ncladding = 1.40; we thus obtain AN = 0.50 and 4max = 30q.
Whereas the numerical aperture of the step-index profile fiber remains constant
over the entire core, the graded-index profile fiber exhibits a decreasing accep-
tance angle from the center of the core to the cladding (Fig. 2.4).
40 2.1 Fundamentals of Optical Fibers
Compared with other fiber types (Fig. 2.5), POF has the largest numerical
aperture and the largest core diameter. This is one of the most important advan-
tages of POF, since the connection technology that can be used for POF is more
economical to apply than that used for glass fibers.
100/ 140 m
0 mm 0.5 mm 1.0 mm
Fig. 2.5: Aperture angle and core diameter of glass fibers and polymer fibers
In the step index profile fiber, light propagates along a zigzag path, being totally
reflected at the core/cladding interface; in the graded-index profile fiber, light pro-
pagates on a sinusoidal trajectory that is created within the graded-index profile
through refraction. If the incident light rays lie within one and the same plane
through which the fiber axis runs, meridional rays are formed. In all other cases,
skew rays are formed. Figure 2.6 shows the projection onto the fiber's incident
face. Step and graded-index profile fibers show the same behavior. The speci-
fication of the numerical aperture always refers to the meridional rays.
2.1 Fundamentals of Optical Fibers 41
D
D
D
D
Skew rays form an angle of \ < 90q with the tangential plane at the
core/cladding interface (Fig. 2.7). They never cross the fiber axis and propagate
along screw-like paths. For step index profile fibers, the projection onto the cross-
sectional area resembles a polygonal line so that these rays do not cross a circle-
shaped area having a radius rk around of the axis.
rk \
In graded-index fibers with a parabolic profile, ellipses are formed in the pro-
jection (Fig. 2.8 left) that may under certain circumstances form circles; these rays
are called helical rays (Fig. 2.8 right). Their distance from the fiber axis is always
constant.
Fig. 2.8: Helical rays (left) and skew rays (right) in graded-index profile fibers
42 2.1 Fundamentals of Optical Fibers
electrical field
Whereas the zigzag paths would lead to intensity distributions within the ray-
optical model that would change depending on the length of the fibers, the wave
model provides a constant light-dark distribution that is independent of the length
across the waveguide's cross-section.
The number N of the guided modes is approximately described by:
1 g
N| V2
2 g2
where V = 2S a AN/O, a is the radius and g is the profile exponent (see also Sec-
tion 1.1.5).
For step-index profiles g o f. This results in a value of N | V for the
number of modes. For parabolic profiles g = 2 and thus N | V. A polymer op-
2.1 Fundamentals of Optical Fibers 43
tical fiber with AN = 0.5, a core radius of 0.5 mm and a wavelength of O = 650 nm
can carry 2.9 million modes. If the angle of total reflection is exceeded, radiation
modes are created and the light is radiated into the cladding. If the refractive index
of the cladding is higher than the surrounding medium (air, for example), cladding
modes may be formed. In the POF, the optical cladding is encased in an absorbing
jacket so that no cladding modes can form. In contrast to guided modes, it is not
possible to count radiation modes. They do not take part in signal transmission.
(Fig. 2.10 special conditions for POF are explained below). Higher modes propa-
gate under a larger angle, lower modes under a smaller one. Under certain circum-
stances skew rays may turn into so-called leaky waves, which, on the one hand,
are guided in the Z-direction and, on the other hand, transfer energy to the
cladding. Under certain conditions they can still be detected in POF even after
several 10s of meters. Hence, they can influence both the transmission process as
well as the measuring techniques used.
radiation mode
higher mode
lower mode
cladding mode
The following equation describes the relationship between the angles D, \ and
G in Fig. 2.11 ([Sny83]):
cos D sin G sin
D is the angle of the incident and reflected ray relative to the surface normal of
the tangential plane in P. \ describes the angle between the reflection plane and
the tangent plane, and G is the angle between the projection of the skew ray on the
cross-sectional plane and the direction of propagation (parallel to the fiber axis).
Figure 2.12 summarizes the various ray types according to the respective angles
derived from the above equation ([Bun99a]). For guided rays holds G < Gmax and D
> Dmax. The leaky waves are shown in the subsequent rectangle while the ray
modes are shown above the line D = Dmax. For meridional rays D = 90q - G because
\ = 90q, i.e. they lie on the blue line.
44 2.1 Fundamentals of Optical Fibers
D
D G
\
P
Fig. 2.11: Designation of the angles of a skew ray; the right diagram shows the angle G,
which is obtained by projecting the skew ray on to the cross-sectional plane
0 \ q
D[] inside this triangle there
10 are the radiation rays
20
30 inside this triangle there \ q
are the guided rays
40 \ q
50
meridional \ q
60 rays
70
Dmax
inside this rectangle there
80
Gmax are the leaky modes
G[]
90
0 10 20 30 40 50 60 70 80 90
Fig. 2.12: The different types of rays
Several chapters of this book discuss the special characteristics of light propa-
gation in POF. Here now, the processes that need to be considered will be looked
at as a whole. The function of fibers as a waveguide for passing on light by means
of total reflection at the core/cladding interface has already been discussed.
If the ray model were applied consistently, then a light ray launched into an
ideal fiber would always propagate at the same angle relative to the fiber axis.
With a divergent light source, the far field would always remain constant along the
length of the fiber. This would not be true for the near field, as Fig. 2.13 illus-
trates: depending on the course of the ray, different locations along the fiber would
2.1 Fundamentals of Optical Fibers 45
generate different near fields in the form of point structures. However, this contra-
dicts the results obtained through experiments: there a continuous distribution of
intensity is obtained, and from a certain length onwards the intensity does not
change at all. Although the ray model is very illustrative, its practical application
is limited as the example above shows. In order to be able to describe experi-
mental results it is thus necessary to move on to the mode concept. In this respect
it is important to keep in mind that many optical simulation programs work on the
basis of discrete light rays. In order to obtain truly realistic results, a sufficient
number of rays has to be simulated.
Fig. 2.13: Near fields under conditions of the ray model with only a few discrete light
paths (in practice very difficult to measure and visible only on very short
lengths)
2.1.5.1 Attenuation
The most important process encountered by light as it passes through a fiber is at-
tenuation. When passing through an optical fiber of the length L, the power of the
light decreases (Fig. 2.14). The following equation applies to the optical power:
PL P0 e DcL
where PL and P0 are the power of the light after passage through a fiber of
length L in km and at the front end of the fiber, respectively; D is the value of the
attenuation coefficient in km-1.
P0 PL
To make it easier to work with the numbers involved here, it is usual to express
attenuation logarithmically. Thus, the attenuation coefficient is expressed as D in
dB/km.
10 P
D log 0 4,343 Dc
L PL
0.1 1 10 100
power ratio PL/P0 [%]
Fig. 2.15: Conversion of the power ratio PL/P0 in % into the dB value
Very often there is not a clear differentiation in the technical literature between
attenuation per unit length D and attenuation factor a. One often speaks simply of
the attenuation of the fiber. The addition spectral refers to the wavelength
dependence. A mistake is avoided, however, when the unit is indicated. We still
have to mention that attenuation and attenuation per unit length are practically
always indicated as positive numbers.
2.1 Fundamentals of Optical Fibers 47
100,000
attenuation [dB/km]
30,000
10,000
3,000
1,000
300 theory
measured
100
30
wavelength [nm]
10
500 600 700 800 900 1000
Fig. 2.16: Attenuation spectrum of the PMMA-POF (theory and measured by [Hess04])
angles are involved. In 100 m of POF, a light ray of this type will travel 6 m
farther which results in an additional loss of more than 1 dB when the attenuation
level is 200 dB/km. At 1,000 dB/km for polycarbonate fiber, this would result in
an additional loss of 4 dB after 20 m of travel (less than 50% of the launched
power reach the fiber output).
The second, more significant cause for mode-dependent attenuation is the
attenuation resulting from the cladding material. Fluorinated polymers are used as
optical cladding for PMMA fibers; these claddings may have an attenuation of
several 10,000 dB/km [Paar92]. Locking more exactly on the propagation of a
plane wave at the interface, we find that, even if total reflection results, the electri-
cal field escapes into the optically thinner medium by a distance in the order of
magnitude of the wavelength. This process is also known as the Goos-Hnchen
Shift ([Bun99a]) and the model explains this as resulting from a shift of the reflec-
tion plane into the optically thinner medium. The reflected ray is hence slightly
displaced on the interface surface, as can be seen in Fig. 2.17. In this model, the
additional light path would be subjected to the higher attenuation of the cladding
material.
cladding
core
area of higher
attenuation
Although the light path in the cladding is only in the Pm range for each reflec-
tion, it still plays a significant role because of the much higher attenuation encoun-
tered there. This effect is particularly striking when the core diameters are reduced
in size. Theoretically speaking, attenuation and bandwidth should not be depen-
dent on the core diameter. Nevertheless, thin cores such as those used in multi-
core fibers have indeed considerably larger bandwidths [Tesh98], a slightly
increased attenuation and narrower far-field widths. These effects are explained
quite well in [Bun99b] and [Ziem99c].
This effect also occurs in glass fibers. Silica glass fibers with a polymer
cladding (PCS) have losses in the core below 10 dB/km (wavelength range from
650 nm to 1,300 nm), whereas the polymer cladding has an attenuation of several
100 to 1,000 dB/km.
Attenuation values of 180 dB/km for the core and 9,000 dB/km for the cladding
are indicated in [Ebb03] for step index profile glass-glass fibers (used in fiber
bundles). Reasonably priced conventional glasses - albeit much purer than in win-
dow glass - are used in these fibers and not silica glass.
In singlemode and graded-index profile silica fibers there are no mentionable
differences in attenuation between the core and the cladding since both consist of
2.1 Fundamentals of Optical Fibers 49
Si02. The germanium dopant in the core does not have any great influence. An
important consequence of the mode-dependent attenuation is, as will be discussed
later on, a significantly narrower far field after greater fiber lengths than one
would expect from the fiber NA.
scattering center
Many experimental results clearly indicate that mode coupling occurs predomi-
nately at the core/cladding interface (Fig. 2.19). This can be explained by the fact
that is it not possible to create an ideal surface in the sub-nanometer range when
very large polymer molecules are involved. Thus, mode coupling is also depen-
dent on the angle of propagation.
cladding
core scattering
center
ding to the laws of statistics, the differential delay (or more precisely, the standard
deviation) does not increase in a linear relationship to the length but approxi-
mately only proportional to the square root of the length. This applies to lengths in
excess of a characteristic coupling length, which for PMMA-POF is generally
several 10 m.
Mode coupling always results in additional attenuation. Whenever there are
changes in the light propagation, energy is coupled into those angle ranges in
which there is no longer any light guiding. The shorter the coupling length, the
larger the additional attenuation will be. If the observed behavior of the POF,
namely the filling up of the near field after a few 10 cm of fiber, could be ex-
plained exclusively because of the mode coupling, then additional attenuations in
the range of 1000 dB/km would result - which indeed does not occur.
Figure 2.20 shows an electron microscope picture of the core-cladding interface
layer (photo ZWL, 2003). The marked smooth part running from the top left to the
bottom right is the surface of the core with the cladding removed. At the top right
you can see the cracked core. The step is the 10 m thick optical cladding. Further
theoretical considerations on the problems of scattering can be found in [Kru06a]
and [Kru06b].
Fig. 2.20: Photo of the core-cladding interface of SI-POF taken by electron microscope
(ZWL Lauf)
Fig. 2.22: Far fields of different POF (product A/B at the top/bottom); left/right after
20 m/50 m of fiber, launch with collimated light (AN Launch < 0.016)
52 2.1 Fundamentals of Optical Fibers
400 D [dB/km]
fiber A
350 source I
source II
300 source III
source IV
250
200
150
100
lPOF [m]
50
1 2 5 10 20 50 100
Both diagrams show very clearly that the different launch conditions (source I
emits very widely, source IV nearly collimated) lead to extremely different attenu-
ation values. After some ten meters, however, the differences disappear for the
most part through mode coupling. Evidently, there are great differences among the
fiber types.
2.1 Fundamentals of Optical Fibers 53
D [dB/km]
400
fiber B
350
source I
source II
300
source III
source IV
250
200
150
100
lPOF [m]
50
1 2 5 10 20 50 100
Fig. 2.24: Attenuation of another SI-fiber at different launch conditions
The next two figures 2.25 and 2.26 show measurements of far field widths for a
POF and a PCS each with altered launch conditions. Once again it can clearly be
seen how the differences caused by the different coupling conditions are evened
out after some 10 to 100 m.
In the 200 m thick PCS it takes considerably longer to establish the equili-
brium mode distribution especially when the length is related to the fiber diameter.
The values of the NA (calculated from the 5% far field width) are represented for
lengths up to 500 m.
54 2.1 Fundamentals of Optical Fibers
0.40
measured NA
0.35
0.30 AN = 0.02
AN = 0.09
0.25 AN = 0.17
AN = 0.26
0.20 AN = 0.34
AN = 0.48
0.15
fiber length [m]
0.10
1 10 100 1000
Fig. 2.26: Excitation dependent far field width of a 200 m-PCS
500
characteristic at Lc = 100 m
400
parameter
deviation for parameter
300
short fibers deviation for
1, 2 und 3 u Lc
200
Fig. 2.27: Approximation of an optical parameter to the equilibrium value by mode coup-
ling (schematically)
2.1 Fundamentals of Optical Fibers 55
The significance of leaky modes has already been touched upon earlier. For the
sake of completeness, it should be noted here again that light rays that lie above
the critical angle of the total reflection do not entirely vanish but still contribute
significantly to light propagation even after several 10s of meters.
Not until we examine the interaction of attenuation, mode-dependent attenu-
ation, mode coupling and mode conversion and take leak modes into account, can
we establish a model for the light propagation of SI polymer fibers that can at least
qualitatively describe the experimentally observed behavior. In principle, the same
processes take place in GI-POF; however there are basic differences:
With GI-POF, there is no core/cladding transition to serve as an essential
cause for mode coupling, mode conversion, and mode-dependent attenuation.
Fluorinated GI-POF are used in wavelength ranges in which Rayleigh
scattering is less significant.
To form the index profile, various zones of the fiber, as seen from the axis,
are provided with varying concentrations of a dopant or a copolymer so that
the attenuation usually gets a gradient. This is probably the most significant
cause of mode-dependent attenuation in GI-POF.
Yabre and Zubia made comprehensive observations on mode propagation in
GI-POF [Yab00a], [Yab00b], [Arr99], [Arr00].
The problem of mode coupling and mode conversion is sure to be very inte-
resting for multi step index fibers. Bandwidths could result that are larger than
what is theoretically expected. Some different theoretical investigations were
made in cooperation between the POF-AC and the University of Bilbao (Spain).
More details will be given in the fiber simulation chapter.
As the example of the multi-core fibers shows, mode-dependent attenuation can
be used to exchange attenuation for bandwidth. Less attenuating cladding would
reduce the overall attenuation of the POF, but more than likely also reduce the
bandwidth (always assuming equilibrium mode distribution). The future will
decide which parameter is of greater significance for users. If the transmission
budget is sufficiently large, it would be possible to increase the bit rate though
multi-level coding or by electrically compensating the dispersion so that a reduc-
tion in attenuation is the minimum goal to be targeted in this field.
P( f ) P0 ( f ) e
- f 2 f02
where P(f) is the power of a random frequency f at the end of the measuring
path, P0(f) is the launched power and f0 is a constant that describes the bandwidth.
Figure 2.28 illustrates the process schematically.
a)
P0(f)
b) pulse response
c)
d)
e)
P(f)
time t
Curve 'a' shows the sine-modulated source optical signal (it must be noted that
optical power can only take positive values). Figure 'b' shows how a single pulse
approaching a Gaussian function after traveling through the fiber. This is a theore-
tical borderline case because the Gaussian function extends from -f to +f, but the
output pulse cannot begin before the input pulse has started. To measure the shape
of the complete output signal, the input signal can be split into a series of pulses,
as shown in Fig. 'c'. After traveling through the fiber, every pulse forms a Gaus-
sian function of the respective height (Fig. 'd'). These have to be brought together
again to achieve the result in curve 'e' (mathematically speaking, this is a convolu-
tion of the input pulse with the so-called pulse response of the transmission link).
2.1 Fundamentals of Optical Fibers 57
It is easy to see that the amplitude of the signal has decreased. Attenuation of
the light has not been taken into consideration.
A short light pulse is briefly broadened when it travels the length of a fiber
(Fig. 2.29) and this in turn reduces the transmission bandwidth.
optical optical
input output
power power
100 %
50 % 100 %
optical
50 %
fiber
time time
tin tout
If Gaussian-shaped pulses are assumed, the result of the pulse broadening 't is
the square root of the difference of the squares of the input and output pulse width
(FWHM full width at half maximum):
2 2
't t out t in
The consequence of this broadening is that the time gap between the bits
becomes smaller, that the pulses finally overlap and that the receiver can no longer
differentiate between the two. The transmission bandwidth is limited as the light
waveguide functions as a low-pass filter. The product of bandwidth and length
characterizes the transmission capacity of a fiber. [Gla97] applies to Gaussian-
shaped pulses:
0.44
B L | L
't
Pulse broadening is caused by mode dispersion and chromatic dispersion. For
multimode fibers it is necessary to consider the factors of material, modes and pro-
file dispersion (in graded index fibers). Waveguide dispersion additionally occurs
in singlemode fibers, whereas profile dispersion and mode dispersion do not.
All the kinds of dispersion appearing in optical fibers are summarized in
Fig. 2.30. The mechanisms dependent on the propagation paths are marked in
yellow, whereas the wavelength-dependent processes are marked in green.
58 2.1 Fundamentals of Optical Fibers
dispersion
In regard to the fibers and applications dealt with in this book only mode and
chromatic (material) dispersion play a role so that the following sections deal
solely with these two effects.
ncladding
Jmax
L2 a
2
Dmax
1
L1
ncore
The propagation times of the two different propagation paths are determined
purely geometrically for:
n
t1 L1 core
c
2
n L1 ncore 1 L1 ncore
t2 L 2 core
c c sin J max c ncladding
n ncore ncladding
' t mod t 2 t1 L1 core
c n cladding
L1 2 L1 ncore
AN | '
2 c ncladding c
Figure 2.32 shows the dependence of the bandwidth on the numerical aperture
with which the light is launched. The assumption is that the far field, i.e. the
angular distribution of the light in the fiber, will remain constant over the entire
length of the sample (no modal coupling or conversion). For a PMMA standard
fiber with an AN = 0.50, a differential delay of 't | 25 ns for 100 m is produced.
The transit time is proportional to the square of the NA. From the above-men-
tioned expression B | 0.44/'tmod, a value of 15 MHz results for the bandwidth.
500
200 fiber-
length:
100 10 m
50
25 m
20 50 m
75 m
10 100 m
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
numerical aperture
The critical angle Jmax of total reflection is determined by the ratio of both
refractive indices (example, 1.492 for the core and 1.456 for the cladding):
1.456
J = arcsin = arcsin 0.976 = 77.4q
1.492
(max. angle to axis : Dmax 12.6q)
D>q@ \ q
70
\ q
meridional \ q
75
rays
80 \ q
Gmax
85
guided
rays
90 G>q@
0 5 10 15 20
As measurements of the far field (that is the power as a function of the angle to
the fiber axis, measured in a sufficient large distance) of a POF shows, this is also
reflected in the greater power obtained with larger angles. If the power is ex-
pressed in relation to the solid angle element, a constant power density is found
because larger angles cover a correspondingly larger arc. This is shown schemati-
cally in Fig. 2.34.
0.6
0.4
0.2
The differential delay increases approximately by the square of the angle rela-
tive to the fiber axis. If a short pulse having a mode distribution that correspon-
ding to UMD is launched into the fiber input, an approximately rectangular pulse
is generated at the output of the length of which corresponds to the approximate
values shown above for the maximum differential delay. Figure 2.35 demonstrates
the precise results for an assumed attenuation-free standard NA POF for the pulse
form obtained after 10 m, 20 m, 50 m, and 100 m of ideal POF (from [Bun99a]).
62 2.1 Fundamentals of Optical Fibers
norm. signal
100%
80%
60%
10 m 20 m 50 m 100 m
40%
20%
time [ns]
0%
0 5 10 15 20 25 30 35
Real SI-POF provide considerably higher bandwidths. The main reason for this
is the presence of mode-dependent attenuation in conjunction with mode mixing,
as will be shown in the next chapter.
The differential delay 't increases proportionally to a particular length Lc
(coupling length); for longer lengths, the increase is sub-linear (Fig. 2.36). The
following holds true:
't v L for L L c
't v LN for L ! L c with N 1
whereby the exponent N must be determined for each fiber. It is typically bet-
ween 0.5 and 0.7. The coupling length Lc ranges between 30 m and 40 m for stan-
dard SI-POF.
2.5
pulse broadening [a.U.] W a l
2.0
1.5 Wal
in reality
1.0
0.5
Lc
length [m]
0.0
0 20 40 60 80 100 120 140 160 180 200
Fig. 2.36: Schematically representation of the pulse broadening reflecting mode coupling
effects
2.1 Fundamentals of Optical Fibers 63
The impulse response of a 50 m long standard POF can be seen in Fig. 2.37.
The half-value width of the impulse amounts to about 50 ns, i.e. only about 30%
of the expected value. Furthermore, it is noticeable that the rear pulse edge drops
more slowly. It is in this range that the higher modes lie which are attenuated very
greatly by the mode-dependent losses. The dropping off of the rising edge can be
explained by the effect of modal mixing.
0.7
U [V] theoretical
0.6 pulse shape
'W = 16 ns
0.5
0.4
0.3
'W = 5 ns
0.2
0.1
0.0
0 5 10 15 20 25 30 35 40
t [ns]
L1 ncore '2
't prof ,
c 2
in other words, a factor '/2-reduced broadening of the pulse as compared with
step index POF; for a typical graded-index POF this means a reduction by
approximately 2 orders of magnitude [Blu98]. Mode dispersion or profile disper-
sion can only be avoided by using singlemode fibers. As explained later on, due to
the combination with the chromatic dispersion, certain polymer fibers, have some
advantages as opposed to silica glass fibers.
64 2.1 Fundamentals of Optical Fibers
output pulse
't
length
'O
wavelength
The real influence of the chromatic dispersion from different polymer optical
fibers to the system bandwidth will be shown in the next chapter which will
contain detailed descriptions of the materials and fiber types.
2.2 Index Profiles and Types of Fibers 65
After the theoretical descriptions on the properties of optical fibers in the section
on the fundamentals of light propagation and the observations indicated above on
mode propagation and the essential characteristics of fibers this following section
will deal with concrete, available fibers. First, the different index profiles, as
briefly mentioned in 1.1.6, will be introduced using examples.
The next section shows the historical development especially in regard to the
different POF variants. Thereafter the important characteristics attenuation and
bandwidth will be shown in a series of experimental results.
Three parameters are basically responsible for the actual properties of optical
fibers. The core and cladding materials used determine the attenuation and chro-
matic dispersion. The refractive index profile determines the mode dispersion and
the core diameter is also responsible for the number of modes. Especially the core
material and the index profile are at least recognizable from the name of the fiber,
a designation method widely used in this book.
In the following section the historical development of the different polymer
fibers is summarized. The POFs are dealt with in regard to their index profiles.
Thereafter, different hybrid and glass fibers for short-range data transmission will
also be introduced. The following chapter deals especially with the bandwidth of
thick optical fibers since this characteristic is particularly important and also it
makes the greatest demands on measurement techniques.
As was the case with silica glass fibers, the first polymer optical fibers were pure
step index profile fibers (SI-POF). This means that a simple optical cladding sur-
rounds a homogenous core. For this reason a protective material is always in-
cluded in the cable. Figure 2.39 schematically represents the refractive index
curve.
As already shown above, the refractive index step determines the numerical
aperture (NA) and thus the acceptance angle. Some typical values are shown in
Table 2.1. The refractive index of the core was always taken as 1.5, whereas the
cladding has a correspondingly smaller refractive index. The last line is valid for
wave guiding against air (n = 1). Here an acceptance angle of 90 is valid since the
NA exceeds the value of 1.
ncore
ncladd
Table 2.1: Relationship between relative refractive index difference and numerical aper-
ture (core refractive index = 1.50)
A larger acceptance angle of the fiber simplifies the launching of light, e.g.
from a semi-conductor source. In addition, a high NA reduces the losses asso-
ciated with fiber bending, as schematically illustrated in Fig. 2.40.
launched light
rays
bend
rays, guided
radius
behind the
bend
Fig. 2.40: Loss at fiber bends
2.2 Index Profiles and Types of Fibers 67
Due to the effects of bending, the propagation direction of each individual ray
is changed relative to the axis of the fiber. In the case of multi-mode fibers, a part
of the rays is always extracted because the rays exceed the angle of total reflection
at the interface between core and cladding. For fibers with a large NA, the effect
of a change in angle for a certain amount of bending is not so significant so that
the bending losses diminish. Likewise, when coupling fibers to each other (at
connectors) the loss due to angle errors is less significant when there is a large
numerical aperture.
A disadvantage of fibers with a large NA is the greater difference in time delay
between the different light paths, and this in turn leads to a greater level of mode
dispersion. This limits the bandwidth. In addition, the loss at coupling points in-
creases if there is a gap between the abutting faces. Some advantages of larger or
smaller numerical apertures are listed in Table 2.2.
However, when it became necessary to replace copper cables with polymer optical
fiber to accomplish the transmission of ATM data rates of 155 Mbit/s (ATM:
asynchronous transfer mode) over a distance of 50 m, a higher bandwidth was
required for the POF. In the mid-nineties all three important manufacturers deve-
loped the so-called low-NA POF.
POF with a reduced numerical aperture (low-NA POF) feature a bandwidth
increased to approximately 100 MHz 100 m because the NA has been reduced to
approximately 0.30. The first low-NA POF was presented in 1995 by Mitsubishi
68 2.2 Index Profiles and Types of Fibers
Rayon ([Koi98]). Figure 2.41 shows that the fiber construction corresponds to the
standard POF, the distinction being that the refractive index difference is smaller
(approximately 2 %). Usually the same core material is used, but the cladding
material has a modified composition.
ncore
ncladding
Unfortunately, practical testing showed that although this fiber met the require-
ments of the ATM forum ([ATM96b]) with respect to bandwidth, it did not meet
the requirements with respect to bending sensitivity. These requirements specify
that for a 50 m long POF link the losses resulting from a maximum of ten 90
bends having a minimum bending radius of 25 mm should not exceed 0.5 dB. In
order to meet both these requirements at the same time it became necessary to find
a new structure.
The double-step index POF features two claddings around the core, each with a
decreasing refractive index (Fig. 2.42). In the case of straight installed links, light
guiding is achieved essentially through the total reflection at the interface surface
between the core and the inner cladding. This index difference results in an NA of
around 0.30, similar to the value of the original low-NA POF.
ncore
ncladding1
ncladding2
When fibers are bent, part of the light will no longer be guided by this inner
interface. However, it is possible to reflect back part of the decoupled light in the
direction of the core at the second interface between the inner and the outer
cladding. At further bends, this light can again be redirected so that it enters the
acceptance range of the inner cladding. The inner cladding has a significantly
higher attenuation than the core. Light propagating over long distances within the
inner cladding will be attenuated so strongly that it will no longer contribute to
pulse propagation. Over shorter links the light can propagate through the inner
cladding without resulting in too large a dispersion. A schematic illustration is
shown in Fig. 2.43.
rays behind
the bend
The first generation of DSI-POF primarily served the purpose of increasing the
bandwidth of 1 mm fibers from 40 MHz 100 m to 100 MHz 100 m with an un-
changed minimum bending radius of 25 mm. The respective applications are to be
found in LANs and home networks.
The fiber producers offer these fibers under the same type names as the original
real low-NA fibers. It has since become standard procedure to call the fibers
low-NA and to indicate DSI as the index profile.
Currently, another goal is being pursued: the bandwidth of standard POF is
sufficient for applications in vehicle networks, but the bending radius should be
reduced. Presently being discussed are POFs, the index steps of which correspond
to a NA of 0.50 or 0.65 respectively to the inner and outer cladding. The bending
radius can thus almost be halved.
70 2.2 Index Profiles and Types of Fibers
R R
dm
dm
r
N=5
N=1 n = 19
Fig. 2.44: Schematically arrangement of cores in a MC-POF
In the figure, R denotes the radius of the complete fiber (typically 0.5 mm) and
d the thickness of the optical cladding (e.g. 5 m). Let us assume first of all that
the individual cores are arranged in a hexagonal shape with N = 2z + 1 cores
positioned next to each other.
The next Fig. 2.45 shows how the arrangement for fibers is changed for
z = 1 to 5. While these sketches can give a clear definition of the number of fibers
that can be arranged within a circular shape, for smaller and smaller individual
cores the possibilities are more complex. The arrangement at the bottom right
shows one possible deviation. For the first five arrangements the number of indi-
vidual fibers is calculated as follows:
n = 3z2 + 3z + 1.
2.2 Index Profiles and Types of Fibers 71
In Table 2.3, the degree of coverage of the circle area is calculated for the cases
shown. First, the number of individual cores is calculated from z. The radius r
results from the overall radius of the fiber (here always 500 m). Parameter ta indi-
cates what percentage of the total circular area is covered by the individual circles
(for the hexagonal arrangement of an infinite number of circles a maximum of
90.69 % of the area can be covered). When calculating parameter tb, the fact that
part of the cross-section is lost to the optical claddings (all uniformly 5 m thick)
is taken into account.
Table 2.3: Core cross area degree of coverage for MC fibers (ideal)
z: N: n: r: ta: tb:
0 1 1 500 m 100.00 % 98.01 %
1 3 7 167 m 77.78 % 73.18 %
2 5 19 100 m 76.00 % 68.59 %
3 7 37 71.4 m 75.51 % 65.31 %
4 9 61 55.6 m 75.31 % 62.36 %
5 11 91 45.5 m 75.21 % 59.57 %
11 85 49.3 m 82.47 % 66.57 %
6 13 127 38.5 m 75.15 % 56.88 %
7 15 169 33.3 m 75.11 % 54.27 %
8 17 217 29.4 m 75.09 % 51.73 %
14 29 631 17.2 m 75.03 % 37.82 %
f - - - 90.69 % -
72 2.2 Index Profiles and Types of Fibers
Figure 2.46 shows the proportion of core area tb as depending on the number of
cores for four different thickness of the optical cladding.
60%
40%
20%
0% number of
1 7 19 37 61 91 127 169 217 single cores
Fig. 2.46: Proportion of core area for different cladding thickness
As can be expected, the proportion of the overall covered area decreases with
an increasing number of cores because the proportion of cladding area will be-
come larger and larger. A certain minimum thickness of cladding is necessary for
it to be able to fulfill its function and still be technologically feasible. The four
individual data points show the case of the optimized fiber arrangement with 85
individual cores in accordance with Fig. 2.45.
Given a minimum thickness of the optical cladding between 5 m and 10 m,
these considerations indicate that a maximum number of some 100 single cores
should be used, in which case the proportion of useable area will hardly exceed
70 %. It is easy to conclude that a smaller proportion of useable core area would
lead to an increase in the losses encountered when connecting transmitters to, and
fibers between each other.
Practical experience shows that a better utilization of the area can be achieved.
During the manufacturing process the fibers are placed together at higher tempe-
ratures which means that they change their shape and thus reduce the gaps bet-
ween the fibers. Apparently, the resulting deviations from the ideal round shape do
not play a significant role in light propagation (the causes for this are not yet
completely understood; some points worth discussing can be found in the chapter
on light propagation in POF). Figure 2.47 shows a schematic illustration of the
cross-section of a fiber with 37 cores, such as e.g. in [Tesh98]. Data of available
MC-POF and -GOF are grouped together later.
Figure 2.48 shows the refractive index profile of a MC-POF, shown as a cross-
section through the diameter of the fiber. The index steps correspond to those of a
standard POF.
ncore
ncladding
Since the bandwidth only depends on the NA for SI fibers, it should be possible
to measure values comparable to the standard POF. However, the fact is that the
measured values are actually significantly higher, which has been explained in the
chapter 2.1.5.2 discussing mode-selective attenuation mechanisms.
Glass fibers are also produced for use in many areas as fiber bundles. In
lighting technology fiber glass bundles with a large NA are widely spread. (The
lighting of the headlight outer ring at BMW via such a fiber bundle is well-
known.) In the meantime, such fibers are also available for data communication
([Lub04b]).
In the MC-POF, too, an increase in bandwidth was achieved by reducing the index
difference. Due to the smaller core diameters it was still possible to avoid an
increase in bending sensitivity.
Even better values were achieved with individual cores having a two-step
optical cladding such as illustrated in Fig. 2.49. The principle is the same as in the
double-step index POF with an individual core. In this case a bundle with single
cladding is completely surrounded by a second cladding material (sea/islands
structure).
74 2.2 Index Profiles and Types of Fibers
ncore
ncladding1
ncladding2
When using graded index profiles (GI) an even greater bandwidth becomes pos-
sible. In these profiles, the refractive index continually decreases (as a gradient),
starting from the fiber axis and moving outwards to the cladding. Of particular
interest are profiles that follow a power law (remember chapter 1.4.1).
g
distance to fiber axis
refractive index n=n 1- '
fiber axis core radius
ncore
ncladding
Due to the continually changing refractive index, the light rays in a GI fiber do
not propagate in a straight line but are constantly refracted towards the fiber axis.
Light rays that are launched at the center of the fiber and do not exceed a certain
angle are completely prevented from leaving the core area without any reflections
occurring at the interface surface. This behavior is illustrated schematically in
Fig. 2.51. The geometric path of the rays running on a parallel to the axis is still
significantly smaller than the path of rays that are launched at a greater angle.
However, as can be seen, the index is smaller in the regions distant from the
core. This means a greater propagation speed. In an ideal combination of para-
meters the different path lengths and different propagation speeds may cancel each
2.2 Index Profiles and Types of Fibers 75
other out completely so that mode dispersion disappears. In reality, this is only
possible in approximation. It is possible, however, to increase bandwidths by two
to three orders of magnitude compared with the SI fiber.
n n
Fig. 2.51: Comparison of step and graded index profile (see also chapter 2.1.1)
When considering not only the pure mode dispersion but also chromatic disper-
sion, i.e. the dependence of the refractive index on the wavelength and spectral
width of the source, an optimum index coefficient 'g' deviating from 2 is achieved.
This has been the subject of comprehensive investigations by the research group
around Prof. Koike ([Koi96a], [Koi96b], [Ish00], [Koi97a], [Koi96c], [Koi98] and
[Ish98]). In [Ish00] and [Koi00] the significance of this effect is particularly pro-
nounced (see also Chapter 2). Due to the smaller chromatic dispersion of fluori-
nated polymer compared with silica, the bandwidth of GI-POF theoretically achie-
vable is significantly higher than that of multi-mode GI silica glass fibers. In parti-
cular, this bandwidth can be realized over a significantly greater range of wave-
lengths. This makes the PF-GI-POF interesting for wavelength multiplex systems.
However, in this case the index profile must be maintained very accurately, a
requirement for which no technical solution has as yet been provided.
Another factor involved in the bandwidth of GI-POF is the high level of mode-
dependent attenuation ([Yab00a]) compared to silica glass fibers. In this case
modes with a large propagation angle are suppressed resulting in a greater band-
width. An example is the simulation that was carried out in [Yab00a]: the band-
width of a 200 m long PMMA-GI-POF increases from 1 GHz to over 4 GHz,
taking into account the attenuation of higher modes. This is also confirmed in
practical trials. Mode coupling is less significant for GI fibers than it is for SI
fibers since the reflections at the core-cladding interface do not occur.
ncore
ncladding
jacket core jacket
optical cladding optical cladding
In this case light rays do not propagate along continually curved paths as in the
GI-POF, but on multiple diffracted paths as demonstrated in Fig. 2.53. However,
given a sufficient number of steps, the difference to the ideal GI profile is relati-
vely small so that large bandwidths can nevertheless be achieved. MSI-POF were
presented in 1999 by a Russian institute (Tver near Moscow [Lev99]) and by
Mitsubishi (ESKA-MIU, see [Shi99]). In the meantime, other companies are pro-
ducing such fibers which are often called GI fibers. These GI and MSI fibers are
classified in the same class of standards, e.g. A4e.
A relatively new version of index profiles are fibers which have a gradient with a
slightly varying index above the core cross section, but do have an optical clad-
ding with a great index step as shown in Fig. 2.54 ([Sum00], [Sum03], [Ziem05f]
and [Ziem06i]).
ncore
ncladding
jacket core jacket
optical cladding optical cladding
Fig. 2.54: Structure of a semi-graded index profile fiber
2.2 Index Profiles and Types of Fibers 77
At first sight this variety of fiber has enormous advantages. Light which propa-
gates within the gradient is only subject to very little mode dispersion. If a ray of
light has a greater propagation angle, e.g. after being bent, then it continues to be
led to the core-cladding interface layer through total reflection. However, these
rays do have a very much higher mode dispersion. Figure 2.55 shows how light
spreads theoretically and what consequences this has for the pulse response.
input output
GI-modes
SI-modes
t
In principle, two different groups of modes can be seen in the picture. The
paths designated as GI modes do not touch the cladding and only show a very
slight difference in propagation times. The shares designated as SI modes are
completely reflected at the core-cladding interface layer. These light paths are also
bent in the core, but the light path, now very much longer, can no longer be com-
pensated for in the outer areas by the lower refractive index. With very high data
rates the second mode group is drawn out so widely that it is presented solely as a
kind of DC offset in the eye diagram. At the POF-AC a data rate of 1 Gbit/s was
transmitted over 500 m of a GI PCS fiber with a PRBS signal ([Vin05a]). Data
rates up to 3 Gbit/s could be attained with a small surface APD receiver ([Kos95]).
In order to do justice to the complex behavior of the semi-GI POF, corresponding
modulation formats should be selected.
Figures 2.56 through 2.58 again show all index profiles described in an overview.
Due to the wide range of possibilities offered in polymer chemistry further deve-
lopments are certainly to be expected. For example, multi-core graded fibers,
fibers with special cladding for a reduction of the losses at the core/cladding inter-
face or to increase the bandwidth or even multi-core fibers with different indivi-
dual cores are all conceivable. In the following figures POF variants are shown
with typical parameters.
78 2.2 Index Profiles and Types of Fibers
Single-core fibers with diameters between 125 m and 3 mm are available from
different manufacturers at a reasonable price and in robust quality. Most of the
polymer optical fibers used in practical applications are of these types.
MC-SI-POF MC-DSI-POF
e.g. 200 cores e.g. 37 cores
AN = 0.30 AN = 0.19
100 MHz100 m 400 MHz100 m
Fig. 2.57: POF with multiple cores and step index profile
GI-POF MSI-POF
AN = 0.20 AN = 0.30
2 GHz100 m 500 MHz100 m
Fig. 2.58: Polymer fibers with graded index and multi step index profile
Graded index as well as multi-step index profile POF are commercially avai-
lable today. Laboratory experiments and a series of practical installations in Japan
and Europe, (e.g. [Ms04]) show the great potential in regard to the bit rates
possible. Asahi Glass introduced them into the market around 2001. Lucent Tech-
nologies, later called OFS and trading under the name of Chromis Fiberoptics as
of 2004 ([Whi04], [Park05a]), also announced the possibility of producing large
amounts of GI POF in case of demand.
2.3 The Development of POF 79
In Europe, fibers by Nexans are manufactured in Lyon ([Gou04]). All three fi-
bers will consist of the fluorinated polymer material CYTOP. The core diameter
of the LucinaTM Fiber by Asahi Glass is 120 m with an AN = 0.28. A protective
cladding made from PMMA and measuring 500 m is placed around an area of
fluorinated polymer outside the core profile. The duplex cable has external dimen-
sions of approximately 3 by 5 mm. The lowest attenuation achieved to date is ap-
prox. 15 dB/km for a wavelength of 1,300 nm. The specified value is < 50 dB/km
for 700 nm - 1,300 nm.
There has also been significant progress in the manufacture of GI or MSI-POF
respectively on a PMMA basis (see Section 2.3.4).
The following sections will describe the polymer fibers presented so far, whereby
particular attention will be paid to the chronological sequence of the develop-
ments. Section 2.4 supplements these observations with some types of multimode
glass fibers which were not discussed in the first edition.
The first POF were manufactured by DuPont as early as the late sixties. Due to the
incomplete purification of the monomer materials used, attenuation was still in the
vicinity of 1,000 dB/km. During the seventies it became possible to reduce losses
nearly to the theoretical limit of approximately 125 dB/km at a wavelength of
650 nm. At that point in time glass fibers with losses significantly below 1 dB/km
at 1,300 nm/1,550 nm were already available in large quantities and at low prices.
Digital transmission systems with a high bit rate were then almost exclusively
used in telecommunications for long-range transmissions. The field of local com-
puter networks was dominated by copper cables (either twisted-pair or coaxial)
that were completely satisfactory for the typical data rates of up to 10 Mbit/s com-
monly used then. There was hardly any demand for an optical medium for high
data rates and small distances so that the development of the polymer optical fiber
was slowed down for many years. A significant indicator for this is the fact that at
the beginning of the nineties the company Hchst stopped manufacturing polymer
fibers altogether.
During the nineties, after data communication for long-haul transmission had
become completely digitalized, the development of digital systems for private
users was commenced on a massive scale. In many spheres of life we are being in-
creasingly confronted with digital end user equipment. The CD player has largely
replaced analog sound carriers (vinyl records and cassettes). The MP3 format is
leading to a revolution in music recording and distribution. The DVD (Digital
Video Disc) and large hard disk drives could lead to the replacement of the analog
video recorder within a few years. Even today more digital television programs
80 2.3 The Development of POF
are available than analog programs. Decoder boxes have become standardized
(MPEG2 format) and will be integrated into television sets in the future. More and
more households are using powerful PC and digital telephone connections
(ISDN). With offers such as T-DSL (ADSL technology provided by Deutsche
Telekom AG) as well as fast internet access via satellite or broadband digital ser-
vices on the broadband cable network, private users are being offered access to
additional digital applications even before the start of the new millennium.
Likewise, in the automotive field the step towards digitalization has long been
made. CD changers, navigation systems, distance-keeping radar and complex con-
trol functions are increasingly part of the standard equipment being provided in all
classes of vehicles. The development of electronic outside mirrors, fast network
connections even from within an automobile as well as automatic traffic guidance
systems will ensure a further increase in the range of digital applications for the
motor vehicle. All these examples demonstrate that completely new markets for
digital transmission systems are being developed for short-range applications.
Polymer optical fibers can meet many of these requirements to an optimum degree
and are therefore increasingly of interest.
A significant indicator for this development is the history of the International
Conference for Polymer Optical Fibers and Applications which has been taking
place annually since 1992 and represents the most significant scientific event in
this specialized field. Many of the developments described below were presented
for the first time at these conferences.
The SI-POF is the oldest variant of all polymer fibers. Its development goes back
to the beginning of the 1960s, i.e. in a period when silica glass fibers were being
developed. Today the SI-POF is by far the most common POF variant. In Table
2.4 data from different publications on this fiber type are summarized - without
claiming to be complete.
5,000
attenuation [dB/km]
2,000
1,000
500
200
100
wavelength [nm]
50
400 450 500 550 600 650 700 750 800
It was not until about 1980 that technology made possible the production of
POF which came relatively close to the theoretical attenuation minima. Initial
problems with the service life and with certain mechanical loads were quickly
solved with on-going developments. In Fig. 2.59 the spectral attenuation curves of
three SI-POFs are shown (data sheet information). All three fibers from Japanese
manufacturers are close together. The visible differences may possibly be due to
different methods of measurement.
Most manufacturers offer SI-POFs in different diameters. In [Zub01b] and
[Nuv04] the properties of these fibers are compared (Table 2.5).
For Toray fibers, the losses of fibers with different diameters are listed in the
data sheet and are shown in Fig. 2.60.
10000
attenuation [dB/km]
3000
1000
With a few exceptions the losses for all fiber diameters are similar. Some
reasons for the increase in attenuation with thinner fibers could be that either the
high attenuation of the optical cladding plays a greater role or that more stress is
exerted on the thin fiber during manufacture. A fiber with a mm core diameter
2.3 The Development of POF 83
has only one sixth the thermal capacitance. When the cladding and opaque jacket
are applied this fiber is necessarily warmer. The process temperatures during ma-
nufacture can indeed lie clearly above the glass transition temperature.
The youngest manufacturer of PMMA SI-POF is the Italian company Luceat.
Here fibers for diverse applications, mainly in mechanical engineering, are pro-
duced. The highest quality is still in the developmental stage. A comparison of the
measured values of Luceat fibers (POF-AC 2006, [Ziem06h]) with the values
from [Wei98], more or less the POF reference curve up until now, is shown in
Fig. 2.61.
500
attenuation [dB/km]
300
200
Luceat
100
80 [Wei98]
60
50
400 450 500 550 600 650 700
wavelength [nm]
In the area of 520 nm this fiber is even somewhat better that the data of the best
fibers so far. Thanks to the availability of reasonably priced and fast green LEDs
this advantage can be assessed very highly. As part of the European POF project
POF-ALL (see www.ist-pof-all.org) the transmission of a 10 Mbit/s data stream
was able to be demonstrated over 425 m (see System Chapter).
We have already discussed the principle idea of a double step index profile POF.
All three important Japanese manufacturers presented such fiber types around
1995. After the expectations that ATM would become the dominating network
technology in the home were not fulfilled, these fibers have more or less become
niche products today, albeit at relatively high prices. Today in many areas there is
a demand for data rates which require the use of these fibers instead of the normal
SI-POFs. Technically, DSI-POFs are on a comparable level and would hardly be
more expensive than SI-POFs when produced in high volumes.
84 2.3 The Development of POF
We would like to point out once again that the DSI-POFs are usually offered
now as before as low NA POF. In the first few years manufacturers did not pro-
vide any information at all about the double cladding structure. In [Eng98b] the
double cladding structure was proven quite early on the basis of measurements of
the far field and with optical microscopy. In Fig. 2.62 you can see the far field dis-
tributions for different fiber lengths measured with the inverse far field method at
the FH Gieen/Friedberg.
1.0
Popt
1m
0.8 10 m
50 m
0.6 O= 594 nm 90 m
0.4
0.2
0.0 4 []
-30 -20 -10 0 10 20 30
Fig. 2.62: Inverse far field measurement of a DSI-POF
You can clearly see that after short distances much light from the interface
layer between inner and outer cladding is still guided. After 50 m these shares
have disappeared and the angle distribution corresponds to a true low NA POF.
Figure 2.63 shows two microscope photos of DSI-POF (Univ. of Ulm). Both
optical claddings can be easily recognized.
At the 2003 POF Conference Mitsubishi was the first manufacturer to present
the actual structure. The effect of suppressing higher modes by high attenuation of
the inner cladding was also confirmed theoretically and experimentally. For
example, Asahi gives a value of 6000 dB/km at 650 nm for the losses in the inner
cladding.
2.3 The Development of POF 85
Since 1994, polymer fibers as multi-core fibers have been introduced, e.g. in
[Tesh98], [Mun94], [Asa97] and [Tesh98]. Table 2.7 shows a few parameters
from these publications.
The MC-POF features a noticeably reduced sensitivity to bending and only in-
significantly increased attenuation as well as a significantly increased bandwidth
compared to single core fibers, this being due to the possibility of smaller nume-
rical apertures. Whether these fibers can be produced at the same price is still an
open question. Should this be possible, data rates of 500 Mbit/s up to 1 Gbit/s over
50 m can easily be achieved in commercial applications. At the POF-AC a data
rate of over 1 Gbit/s over 100 m MC-POF has already been achieved.
At present, only Asahi chemical offers MC-POF for data communication while
other manufacturers offer this kind of fiber for lighting purposes or also as image
guiding fiber. The following photos show the cross-sections of the three, presently
available MC-POFs with 37, 217 and 631 cores (the 19 core variant is no longer
available).
86 2.3 The Development of POF
Fig. 2.64: Photo by microscope of MC-POF, 37, 217 respectively 631 cores
We do have to point out one special feature of these four MC POFs: the fibers
are tightly bound in the cable as opposed to the individual fibers in a fiber glass
bundle or other MC POFs used in lighting technology. The share of the core
surface is not only enlarged, but it is considerably easier to work the fibers. These
strands can be mounted like quite normal 1 mm SI-POFs.
The two enormous advantages of MC-POF, namely the high band width and
the low bending losses, have in the meantime been somewhat qualified since con-
siderably cheaper GI-POFs on a PMMA basis have become available. The latter
will be treated in the next paragraph.
2.3 The Development of POF 87
The greatest bandwidths of all fibers - with the exception of the singlemode fibers
- are shown by graded index profile fibers. They have been used extensively for
some time in the field of silica glass fibers and are a standard. In the USA, predo-
minately fibers with a core diameter of 62.5 m are used, whereas in Europe and
most other countries fibers with a core diameter of 50 m are used. This diameter
is nevertheless 5 to 6 times greater than with singlemode fibers whereby the plug
costs are greatly reduced and the coupling of lasers is also easier. The bandwidth-
length product (BLP) of these multimode glass fibers lies in the range of 200 to
500 MHz km. For the transmission of 10 Gbit/s a new fiber specification with a
BLP of 2,000 MHz km at a wavelength of 850 nm is even being developed (for
example, see [Oeh02] and [Geo01]).
The advantages of the large core diameter and high bandwidth would be an
optimal combination with POFs. Furthermore, numerous problems with the core-
cladding interface area would cease to exist with GI fibers since the light guiding
would take place exclusively in the core. Glass GI fibers are produced by applying
many layers of a SiO2-GeO2 mixture with different compositions to a quartz glass
pipe. Finally, the fiber is drawn (several 100 km) out of such a preform. Unfor-
tunately, this is not possible with POFs. The different methods and combinations
of materials with which attempts have been made to produce GI-POF will be des-
cribed further on. Since GI fibers are difficult to produce - as we shall describe
later on - a series of multi step index profile POFs have been introduced. These
MSI-POFs also offer high bandwidth depending on the number of steps. For now,
the optical characteristics are summarized here.
Table 2.9 shows an overview of the values for PMMA-based GI, MC and MSI
fibers. To the best knowledge of the author, all PMMA-GI-POF published to date
are produced by doping, whereas only MSI-POF are produced in a co-polymeri-
zation process.
Table 2.9: Published data of PMMA-GI-, MSI- and MC-POF (IGPT: interfacial gel poly-
merization technique; PFM: preform method)
Ref. Year Producer Material core Attenuation at O NA Remarks
m dB/km nm
[Koe98] 1998 1 11 8 100 m
[Koi95] 1982 Keio Univ. MMAco VPAc - 1070 670 - first GI-POF
[Koi96c] 1990 Keio Univ. PMMA - - - - 670 nm: 300 MHzkm
[Koi95] 1990 Keio Univ. MMA co VB - 130 650 -
[Koi90] 1990 Keio Univ. MMA-VB - 134 652 - IGPT, 260 MHz1 km
[Koi90] 1990 Keio Univ. MMA-VPAc - 143 652 - IGPT,125 MHz 1km
[Koi92] 1992 Keio Univ. PMMA 200-1500 113 650 - IGPT, 1,000 MHzkm
[Koi92] 1992 Keio Univ. PMMA 200-1500 90 570 -
[Non94] 1994 Sumitomo PMMA 400 160 650 0.26 'n=0.014, 8GHz50m
[Shi95] 1995 BOF PMMA 600 300 650 0.19 3 GHz100 m
88 2.3 The Development of POF
1.2
Intensity [a.u.]
1.0
1m 66 m
0.8
0.4
0.2
0.0
t [ps]
-0.2
0 200 400 600 800 1000 1200 1400 1600
Since 2004, a new GI-POF on a PMMA basis has been available on the market.
The OM-Giga (see [Rich04] and [Yoo04]) has a core diameter of 900 m or
675 m respectively and a nearly parabolic profile. It is produced through poly-
merization of several layers, although the steps are almost completely smoothed
through thermal treatment. According to the data sheets available in the Internet
the fibers have the following parameters (Table 2.10).
The fact that this fiber possesses thermal stability comparable to a standard
POF, different from GI-POF with doping, must be rated as a particularly great
step. Even after 5,000 hours of operation at 80C no change in the bandwidth
could be determined. The cross-section of a 1 mm OM-Giga is shown in Fig. 2.66
(microscope photograph shown in wrong colors). The approx. 10 index steps can
still be seen quite well.
Fig. 2.66: Cross section of an OM-Giga (POF-AC) and a MSI (Tver, [Ald05])
In Fig. 2.67 the change in the refractive index profile of a doped PMMA
GI-POF is shown after accelerated aging (122 hours at +109C, from [Bly98a] and
[Bly98b]). You can see quite well that the index profile is still parabolic at the
beginning of the aging process. The share of the dopants is the greatest in the
center of the fiber which is why the glass transition temperature has sunk the most.
2.3 The Development of POF 91
2000
1500
1000
time [h]
0
0 400 800 1200 1600 2000 2400
Fig. 2.68: Long-term behavior of OM-Giga
92 2.3 The Development of POF
The frequency range of the network analyzer extended to 1.3 GHz. The values
represented were determined through extrapolation and thus burdened with a rela-
tively large error. A clear deviation from the parabolic index profile would in any
event have caused a very strong decrease in the bandwidth.
The stable bandwidth proves that co-polymerization is obviously a suitable
means to produce thermally stable and thus long-life PMMA GI-POF.
A comparison of the measured attenuation of ESKA-MIU and OM-Giga is
shown in Fig. 2.69. The attenuation of the OM Giga is somewhat higher at 650 nm
than that of the Mitsubishi fiber and also of the SI-POF. However, it clearly shows
the greatest bandwidth.
1000
attenuation [dB/km]
800
600
OM-Giga
400
217 dB/km
300
200
ESKA-MIU 161 dB/km
100
400 450 500 550 600 650 700
wavelength [nm]
Fig. 2.69: Spectral attenuation of ESKA-MIU and OM-Giga
Fig. 2.70: PMMA GI-POF index profile (left: [Kim03], right: [Luv03])
Thanks to their simple production and great robustness silica glass fibers with
polymer cladding have been used for a long time. Figure 2.71 shows the principle
structure. A core (typically with a diameter of 200 m) of homogeneous SiO2 is
surrounded by a high-strength, transparent polymer with smaller refractive indices
(about 15 m thick).
500 m
inner jacket
2.3 mm
outer jacket 200 m
SiO2-core
230 m
polymer
Fig. 2.71: Structure of a 200 m PCS
Production is so easy because the core is drawn from a quartz glass cylinder.
The polymer cladding is applied by extrusion after it has cooled off. First of all, all
glass fibers are extremely sensitive to water and must be protected by a plastic
coating as thick as possible. Furthermore, pure glass fibers do not have a great
94 2.4 Glass Fibers for Short-Range Data Transmission
mechanical load capability. The polymer cladding gives the fibers the capacity to
bear extreme loads. The jacketed fiber can thus hardly be shattered. Pure glass-
glass fibers (glass core with an optical glass cladding) are always surrounded by
similar protective layers, e.g. acrylates which, however, do not have any optical
function.
Because of its refractive index and attenuation the polymer cladding determines
to a great extent the optical parameters of the PCS. In short wavelength ranges the
attenuation nearly corresponds to pure SiO2 fibers. Above approx. 1,000 nm the
losses in the polymers are so high that the effective PCS attenuation also rises ra-
pidly. Silica glass can endure temperatures up to 1,000C, but not the polymer
cladding. Consequently, the primary coating material determines the thermal and
chemical characteristics. Most PCSs available in the market have been specified
for an application temperature of +70C. Some more recent types have been di-
mensioned for use in automobile networks for temperatures up to +125C. Infor-
mation on such PCSs can be found for example in [Hub03] and [Sch03]. Fig 2.72
has been taken from the latter work. You can clearly recognize how strongly the
attenuation spectra of different PCSs can depend on the cladding materials
selected.
10,000
attenuation [dB/km]
1000
100
diff. PCS
10
theoretical limit
0.1
200 400 600 800 1,000 1,200 1,400 1,600 1,800
wavelength [nm]
Fig. 2.72: Attenuation of different 200 m PCS according to [Sch03]
Just as with glass-glass fibers the absence of water plays an important role for
PCS for keeping losses low especially in the long-wave ranges. So-called all-silica
fibers in which the optical cladding consists of silica glass are used at high tem-
peratures. These fibers are also employed for the transmission of very high light
power (working with lasers) since it is very important that no light is absorbed at
the core-cladding interface layer.
2.4 Glass Fibers for Short-Range Data Transmission 95
Table 2.11 lists some of the representative types taken from a number of diffe-
rent PCS variants which differ in cladding material, core diameter and NA (data
from [Hub03] and [OFS02]).
system with
1 mm POF
system with
Popt [dBm] 200 m PCS
-4 -6 -8 -10 -12 -14 -16 -18 -20 -22 -24 -26 -28 -30
LED-power range (launched into the fiber)
receiver sensitivity range
allowed path loss (with margin)
Fig. 2.73: Link power budget for POF and PCS
96 2.4 Glass Fibers for Short-Range Data Transmission
Fig. 2.73 shows power budgets for both possibilities, each with the same trans-
mitters and receivers (system for 125 Mbit/s).
The result for PCS is a permissible fiber attenuation of at least 11 dB - taking
the system margin into consideration - thanks to the greater input power. In this
way at least 20 m of POF can be bridged. The guaranteed loss for PCSs is only
7 dB. However, at least 100 m of fiber can be bridged, limited here due to the
bandwidth.
100
70
40 recommended recommended
application application
area with POF area with PCS
20
Fig. 2.74: System parameters of the HP-system with POF and PCS (according to [HP01])
The bandwidth for PCS indicated in the data sheets has to be viewed with a
certain degree of skepticism. Measurements conducted at the POF-AC show that
all PCSs investigated with An = 0.37 at full launch have a BLP in the range of
5-7 MHz km. This lies clearly below the specified data of 10-20 MHz km. This
is not a contradiction, however, since none of the manufacturers as a precaution
provided any information about the measurement conditions. One reason may be
that the PCS was developed for relatively low data rates (10 Mbit/s and less). The
fiber bandwidth therefore did not play any role whatsoever while the POF was
also designed from the very beginning for higher data rates. Diverse information
and publications on bandwidth exist for the different polymer fibers as summa-
rized in Chapter 2.5. The most recent draft for the standardization of PCS is
viewed by the IEC as having a bandwidth of 5 MHz km for fibers with a NA of
0.40 0.04.
A specific problem with PCS in the past was that the temperature coefficients
of glass and plastic did indeed deviate considerably from one another. In the case
of some fibers this resulted in a refractive index difference - and NA, too - which
dropped to zero at low temperatures. This effect is shown in Fig. 2.75 taken from
[Dug88].
Far field distributions are represented in the picture after 2 m of fiber at diffe-
rent temperatures. They were measured with laser stimulation at altered angles. In
this case the optical cladding was a silicone plastic. Modern PCSs no longer show
this effect.
2.4 Glass Fibers for Short-Range Data Transmission 97
rel. power
1.0
0.9 +40C
0.8 -2C
-31C
0.7 -51C
0.6 -65C
-72C
0.5 -92C
0.4 -98C
0.3
0.2
0.1
T []
0.0
-5 0 5 10 15 20 25 30 35
Fig. 2.75: Temperature dependence of PCS-NA, presented as far field
Up until some time ago this class of fibers was only available as a product from
the manufacturer Sumitomo ([Sum03]). Except for the gradients introduced this
fiber corresponds to conventional PCS. The index variation is attained by adding
germanium which is also usual for silica glass. Even with normal 50 m GI fibers
the germanium share represents a considerable cost factor. The semi-GI PCS,
however, has a 16-fold cross-section. This type of fiber is still extremely expen-
sive. It is still open how far the price can drop when manufacturing greater
lengths. In the meantime, OFS has appeared as a second manufacturer
([Ziem06i]).
100
80 spectral attenuation [dB/km]
60
40 Sumitomo
30
20 OFS
10
8
6
4 wavelength [nm]
450 500 550 600 650 700 750 800 850 900 950 1000
Fig. 2.76: Spectral attenuation of the semi-GI-PCS
98 2.4 Glass Fibers for Short-Range Data Transmission
Figures 2.76 and 2.77 show the attenuation curve and the pulse response of the
semi-GI-PCS based on measurements made at the POF-AC. The following table
gives the parameters from the data sheet - the bending radius and the operating
temperature are not specified. The bandwidth and maximum data rate measure-
ments are dealt with in the corresponding sections.
0.2
t [ns]
0.0
0 10 20 30
Fig. 2.77: Pulse response of Semi-GI-PCS
the cable, then they can also serve as cross-section converters, e.g. monochro-
mators. The end surfaces are usually prepared: the bundle is glued in the plug and
then polished. Figure 2.78 shows an example.
The transmission of such a bundle is shown in Fig. 2.79 (acc. to [Ori01]). The
greatest part of the 100% missing share is determined by the only about 60% part
of the core surfaces and the Fresnel losses. The numerical aperture of the bundle
shown is 0.22, the length is about 1 m, and the single fiber diameter is 200 m.
70%
transmission
60%
50%
40%
30%
20%
10%
wavelength [nm]
0%
200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000
1,000
D[dB/km]
500 POF
200 MC-GOF
100
O [nm]
50
500 550 500 550 600 650 700 750 800 850
Fig. 2.80: Spectral attenuation of glass fiber bundles and POF
An entirely new application for such glass fiber bundles has come about with
the ever increasing use of optical networks in vehicles. The previous systems are
specified with 1 mm POF. Two parameters especially limit the use: the tempera-
ture range is limited to a maximum of +85 and the relatively large bending radius.
Both limitations can be reduced considerably with glass fiber bundles, whereby
the usual optical characteristics are for the most part retained so that the identical
active components can be used. Table 2.13 from [Lub04b] compares the parame-
ters of a glass fiber bundle (MC-GOF) with those of a POF for vehicle networks.
The construction of the plug is especially problematical. The usual method of
cementing and polishing takes too much time for mass production and results in
the core surface having too low a share with correspondingly high losses with the
plug connections.
Megomat TS AG, working together with Schott, has developed a new kind of
assembly procedure ([War03]). The actual fiber bundle has a diameter of 1.2 mm.
The plug has a metal ferrule with a corresponding opening. During production the
fiber bundle is heated to such an temperature that the glass can be compressed.
The fibers are pressed closely together when crimped so that the diameter of the
bundle is lowered to 1 mm.
2.4 Glass Fibers for Short-Range Data Transmission 101
The core share of the plug end face then amounts to about 85%. After the
crimping the bundle is broken off and polished. Figure 2.81 shows a photo of the
plug end face.
Since the bundle consists of about 400 individual fibers this irregular defor-
mation of individual fibers does not play any role overall. Since the deformations
only arise over a few millimeters there is no significant additional attenuation.
In conclusion, Fig. 2.84 shows an x-ray photo of the bundle within the cable.
The individual fibers have to move freely within the cable. When there is a tight
bend the change in length is distributed on the inside and outside for a long stretch
so that the fibers are only subject to a slight load. That is why the bundle can take
bending radii of only a few mm.
0.5
bandwidth f3dB
frequency f
0.0
0.5
f [a.U.] f f
0.0
0 1 2 0 1 2 0 1 2
A high pass filter is used for compensation. In the case of low frequencies, the
signal is attenuated - in the case of higher frequencies the signal is passed through
without attenuation. The resulting function has a significantly higher bandwidth;
however, due to the overall existing level of attenuation, a higher level of signal is
necessary.
In addition, the type of signal involved (digital or analogue) is also of signifi-
cance and finally the required system reserves must be considered. The following
general relationship can be used as a rule of thumb for digital systems:
maximum bit rate [Mbit/s] = 2 u bandwidth [MHz].
We intend here to look at bandwidth as a function of fiber characteristics. For
this reason, the effect of chromatic dispersion will be initially neglected because it
is directly proportionally dependent on the spectral width of the source.
In this section we will show experimental investigations on the bandwidth of
SI-POF fibers. After explaining the measurement procedures, we will show to
what extent bandwidth is particularly dependent on the launching conditions.
P( f ) P0 e
f 2 / f02
As can be easily demonstrated, the amplitude of a Gaussian low pass filter
(P(f) = P0 exp (f/f0)) for f = 1.17741 f0 has dropped to half the value that
applies for f = 0. When using a spectrum analyzer to measure the frequency
response of a fiber link, it is necessary to determine the electrical 6 dB width
because the photodiode will convert the optical power proportionally into a
current. Therefore the following applies:
2.5 Bandwidth of Optical Fibers 105
2
Pel Popt
-3
electr. power [dBm]
measured
-6
15 m SI
-9 650 nm
239 MHz
NA: 0.34
-12
-15 Gauian
approx.
-18
1 10 100 f [MHz] 1,000
Fig. 2.87: Bandwidth measurement at a SI-POF
Due to the limited dynamics of the measurement system, the frequency res-
ponse can be measured only up to a certain distance. In this case a measurement of
up to 200 MHz was easily possible. The 3 dB bandwidth is found simply by deter-
mining the point at which the electrically measured transmission function has
dropped by 6 dB, here approximately 150 MHz.
Apart from the values actually measured, an approximation with a Gaussian
low pass function has been entered into the figure. By determining the frequency
f0 it is then possible to determine the bandwidth even when the measurement is not
possible because of the limited dynamics or bandwidth of the measuring system.
Figure 2.88 shows the measured transmission functions for a SI-POF and a
DSI-POF of 50 m length each. The optical 3 dB bandwidth for an SI-POF is
approximately 67 MHz, corresponding to a bandwidth-length product of
33 MHz 100 m, with the NA of the fiber being 0.52. It follows that the measured
value is substantially greater than had been theoretically expected (approximately
14 MHz 100 m, see Fig. 2.31). For DSI-POF (AN = 0.30) the measured value is
130 MHz, corresponding to 65 MHz 100 m, with the theoretical value being
42 MHz 100 m.
The measurement was carried out with a 520 nm LED. The LED had a wide
emission angle so that approximate equilibrium mode distribution can be assumed.
106 2.5 Bandwidth of Optical Fibers
0
rel. power [dB] 50 m
DSI-POF
-3
-6
50 m
St.-NA-POF
-9
frequency [MHz]
-12
1 10 100 1,000
Fig. 2.88: Bandwidth measurement for SI-POF and DSI-POF
When measuring bandwidth, a two to four ranging factor of deviation from the
theoretical value of an ideal SI fiber can be generally expected, even when wor-
king in an EMD condition. The reason for this is the combination of mode depen-
dent attenuation and mode coupling described in Chapter 1. As a result of the con-
tinuous energy exchange that takes place between the faster and slower modes, the
delay does not rise in proportion to the length. The increased attenuation of those
beams having a particularly large propagation angle - many reflections at the
cladding - has the additional effect of reducing the pulse width.
Figure 2.89 shows the bandwidth measurement of a standard NA-POF for 3
different wavelengths for samples between 20 m and 100 m in length.
1000
bandwidth [MHz] bandwidth [MHz]
100
525 nm 525 nm
590 nm 590 nm
30
650 nm 650 nm
fiber length [m] fiber length [m]
10
10 20 50 100 10 20 50 100
Once again, the measurements in Fig. 2.89 were carried out with an LED
having an emission characteristic near to EMD (see [Gor98] and [Rit98]). The
figure reveals 2 significant items of information:
The bandwidth of the POF does not decrease in proportion to the length-1; its
decrease is less than proportional.
The bandwidth of the POF is nearly identical for the 3 attenuation windows.
20
fiber length [m]
10
10 20 50 100
Fig. 2.90: Bandwidth measurement according to [Tak91]
108 2.5 Bandwidth of Optical Fibers
1,000
bandwidth [MHz]
receiver detection
angle range
300
AN Det = 0.22
AN Det > 0.65
100
theory
30
whereby tmod is the modal pulse propagation and B z is the product of bandwidth
and length. Parameter C is a free selectable constant which depends on the coup-
ling conditions. The speed of light is c. In the formula AN , FF is not the fiber para-
meter indicated, but the value measured depending on length.
For a sample length of 10 m, the difference between the measured bandwidth
for launching the light with AN = 0.10 and AN = 0.65 is more than one order of
magnitude. For lengths up to 100 m this factor decreases to 2.
2.5 Bandwidth of Optical Fibers 109
10,000 launching
bandwidth [MHz]
conditions
3,000
AN Launch = 0.10
1,000 AN Launch = 0.65
theory based on
300 far field width
100
When launching light with a small NA, the bandwidth drops disproportionately,
from approximately 80 MHz km to approximately 16 MHz km. This suggests
an increasing filling out of modal field. By comparison, when launching with a
large NA, the bandwidth is reduced somewhat more slowly than the length, from
approx. 4 MHz km to approx. 5 MHz km. This is due to the effect of mode
coupling and mode related attenuation.
The bandwidth values determined by means of the far field width correlate very
well with the results of the bandwidth measurements made by pulse propagation.
This suggests that mode dependent attenuation and mode conversion are the deter-
mining processes because they affect the bandwidth by changing the mode distri-
bution. In contrast, if mode coupling were more pronounced, the bandwidth would
also change without affecting the far field. However, any estimated quantification
based on these measured results alone would be questionable. In [Rit93] measured
results for the bandwidth of standard NA-POF at launching conditions of
AN = 0.10 and AN = 0.65 (Fig. 2.93) are also shown.
5,000 launching
bandwidth [MHz]
2,000 conditions
AN Launch = 0.10
1,000
AN Launch = 0.65
500
200
100
50
fiber length [m]
20
10 30 100 300 1,000
Fig. 2.93: Measured bandwidth of a SI-POF according to [Rit93]
110 2.5 Bandwidth of Optical Fibers
Here too, the measured bandwidth for short lengths (20 m) differs by more than
an order of magnitude. For large lengths the difference is reduced corresponding-
ly. The authors calculate the bandwidth based on their own theory that follows the
concept of the diffusion model. Instead of investigating separate modes, this
model investigates modal groups that differ in their 2 angles of propagation (radial
and azimuthally).
The coupling between the modes is described by a diffusion constant that only
takes into account the energy transfer in neighboring mode groups. The model
also takes into account mode dependent attenuation.
In this work the remaining deviation between theory and measured values is
explained by means of the mechanism of mode coupling. In variance to the model,
this is a factor that is not independent of the angle. Simulations provide good
results if elongated scattering centers of 37 m length and 2.5 m diameter are
assumed in the fiber with random distribution and orientation along the axis of the
fiber (caused by the drawing process), as shown schematically in Fig. 2.94.
scattering centers
Figure 2.96 shows further experimental results for the bandwidth of polymer
optical fibers [Kar92]. In each case collimated light or light with an angle adapted
to the fiber's NA (UMD) was launched into the POF. As was the case in the results
previously shown, very large differences result for short lengths of fibers. The
parameter shown in the figure here is the product of bandwidth and length.
Apart from the effect of the launching NA, [Kar92] also investigates whether
the bandwidth depends on the size of the launched beam. In fact, for UMD laun-
ching, a larger bandwidth was found as well as a smaller light spot, compared with
complete illumination of the fiber cross-section; however, the differences are not
as pronounced as when the launch angle is changed. For collimated light the
relationship is reversed.
Because all processes described up to this point are only dependent on the
angle, it seems surprising to find that the size of the launching spot has an effect
on the measured bandwidth of SI fibers. However, when considering the fact that
mode conversion can cause deviations in location and deviations in angle after just
a short length of the specimen (see schematic in Fig. 2.97), the result becomes
understandable [Kar92].
bending
deviation in location
deviation in angle
1,000
theory
300
100
length [m]
30
1 3 10 30 100 300
Fig. 2.98: Measured bandwidth of different SI-POF according to [Poi00]
B3dB [MHz]
5,000 NAlaunch:
2,000 0.64
0.48
1,000
0.33
500 0.19
0.09
200
0.05
100
50
20 length [m]
5 10 20 50 100
Fig. 2.100: Bandwidth measurement of a 1 mm SI-PMMA-POF
114 2.5 Bandwidth of Optical Fibers
For a 1 mm PMMA POF (Toray PFU CD1000, see also [Ziem04a]) 3 dB band-
widths for lengths between 5 m and 100 m were measured. The coupling angle
was changed for NA values between 0.05 and 0.65 with the unit described above.
For short fiber lengths the bandwidths measured differ by almost a magnitude
which demonstrates once again the importance of correct measurement conditions
for correctly indicating the bandwidth values. After a 100 m test length there still
is a factor of two between the values measured. The curves for under filled launch
(small NA) fall more steeply than with length caused by a predominance of mode
mixing. For overfilled launch (large NA) the curves run flatter. Here the mode-
dependent attenuation dominates. The next figure shows the results with a 1 mm
POF made of modified PMMA (Toray PHKS CD1001). The fiber is specified
with a NA of 0.54.
B3 dB [MHz] NAlaunch:
3,000
AN = 0.05
AN = 0.09
1,000
AN = 0.19
300 AN = 0.33
AN = 0.48
100 AN = 0.64
30
5 10 20 50 100 length [m]
Fig. 2.101: Bandwidth measurement of a 1 mm SI-mod. PMMA-POF
Since the losses of this fiber lie at about 300 dB/km at 650 nm, test lengths of
only up to 50 m could be measured. Incidentally, the measurement results are
similar to a large degree to the results of the PMMA POF.
3.000 NAlaunch:
B3 dB [MHz]
AN = 0,05
1.000 AN = 0,09
AN = 0,19
300 AN = 0,33
AN = 0,48
100 AN = 0,64
length [m]
30
1 2 5 10 20
Fig. 2.102: Bandwidth measurement of a 1 mm SI-PC-POF
2.5 Bandwidth of Optical Fibers 115
The third fiber tested is the polycarbonate POF FH4001 from Mitsubishi. The
NA of the fiber lies at 0.75, the attenuation amounts to 650 nm at about
800 dB/km, whereby the maximum measurement length remains limited to 20 m.
Surprisingly, the bandwidth differences between the three types of fiber are
only very slight although there were clear differences in the NA. One explanation
for this could be the greater effects for mode mixing and above all for the mode-
dependent attenuation which occurred in the fibers made of modified PMMA and
polycarbonate. Figures 2.103 and 2.104 illustrate the far fields of the three fibers
in comparison (cf. [Bun02a]).
B3 dB [MHz]
3.000
PC
1.000 PHKS
PMMA
300
NALaunch = 0.33
100
length [m]
30
2 5 10 20 50 100
1000
power [a.U.]
mod. PMMA
800
PC
600
PMMA
400
200
T []
0
-40 -30 -20 -10 0 10 20 30 40
Fig. 2.104: Comparison of the farfields of different SI-POF
Fig. 2.105: Comparison of the far fields of different SI-POF (3-d representations)
10,000
B3 dB [MHz] NALaunch
5,000 0.05
0.10
2,000
0.19
1,000 0.34
0.47
500
0.65
200
100
50
1 2 5 10 20 50 100
fiber length [m]
Fig. 2.106: Bandwidth measurements of a 1 mm SI-POF (Luceat, HQ)
Both fibers essentially show comparable results. Since the fibers also have very
similar attenuation values they can be used in almost all the same applications.
The advantages of the thinner fibers are primarily the smaller space needed, an im-
portant point with multiple cables, and the smaller bending radius. The argument
that the fibers with a smaller core diameter would enable higher bit rates or better
receiver sensitivity because of the smaller photodiodes has for the most part since
been dropped because of technical developments.
2.5 Bandwidth of Optical Fibers 117
10,000
B3 dB [MHz] NALaunch
5,000
0.05
2,000 0.10
0.19
1,000 0.34
500 0.47
0.65
200
100
50
1 2 5 10 20 50 100
fiber length [m]
launching with magnified light spot: launching with mode field converter:
- medium fibers obtain small angles only - all fibers obtain around the same
- outer fibers obtain large angles optical power and rays of all angles
2,000
B3 dB [MHz] launch NA:
PMC 1000
AN = 0.09
37 cores
AN = 0.19
1,000 AN = 0.33
AN = 0.48
AN = 0.64
500
200
length [m]
100
20 30 40 60 80 100
2,000
B3 dB [MHz] MCS 1000 launch NA:
217 cores AN = 0.09
AN = 0.19
1,000
AN = 0.33
AN = 0.48
500 AN = 0.64
200
100
20 30 40 60 80 100
length [m]
B3 dB [MHz]
10,000
NALaunch
0.05
5,000
0.65
2,000
1,000
500
200
0.1 0.3 1 3 10 30 100
fiber length [m]
Fig. 2.111: Bandwidth measurements of a MC-POF (measurements on a single fiber
sample, cut-back method)
Multi step index fibers have already been introduced by different manufac-
turers. However, they are not yet ready to go into mass production. The youngest
product so far is the ESKA MIU from Mitsubishi-Rayon, a fiber with three diffe-
rent layers. Using a sample length of 100 m of this fiber, a bandwidth of almost
300 MHz was ascertained. Figure 2.112 shows the frequency response.
120 2.5 Bandwidth of Optical Fibers
3
rel. level [dB]
0 NA 0.10
NA 0.34
-3 NA 0.64
-6
-9
-12
-15
-18
-21
-24
10 20 50 100 200 500 1000
frequency [MHz]
-1
-2
-3
50 m OM-Giga
-4 OLD = 650 nm
ANLaunch = 0.34
f3 dB opt. = 1,504 MHz
-5
frequency [MHz]
-6
10 20 50 100 200 500 1,000
In order also to be able to measure bandwidths of several GHz with thick core
fibers, an optical oscilloscope is a practicable device, whereby the widening of a
short laser pulse (about 120 ps) is measured. In [Lwin06] the results for the
OM-Giga are shown compared with the microstructured POF (with effective
graded index profile).
350
pulse broadening [ps]
300
Optimedia 1,000 m
250
200
MPOF: 500 m
150
length [m]
100
15 25 35 45 55 65
The next illustration shows the frequency response for a PF-GI-POF at the
wavelengths 650 nm and 850 nm together with the fitted Gaussian functions.
-2
-3
300 m PF-GI-POF
-4 OLD = 650 nm/850 nm
ANLaunch = 0.10
-5
-6
0 200 400 600 800 1000 1200
frequency [MHz]
The 3 dB bandwidths are around 1,600 MHz for both fibers. The bandwidth-
length product is at about 500 MHz km, somewhat in the range of conventional
multimode graded index glass fibers (cf. further results in [Bach01]).
2
rel. electr. level [dB]
0
-2
20 m fiber
-4
at 650 nm
-6
launch NA:
-8
AN = 0.10
-10 AN = 0.34
AN = 0.46
-12
AN = 0.60
-14 AN = 0.64
f [MHz]
-16
10 20 50 100 200 500 1000
B3 dB [MHz]
3,000 1 mm MC-GOF
375 cores
NAfiber: 0.50
O = 650nm
1,000
launch NA:
AN = 0.64
300
AN = 0.34
AN = 0.10
length [m]
100
2 5 10 20 50
Finally, the bandwidth for different lengths was determined using a 650 nm
laser. In order to be able to make measurements relatively independently of mode,
a 1 m long SI-POF was used as an adapter fiber at both the transmitter and the
receiver. Figure 2.118 shows the results.
124 2.5 Bandwidth of Optical Fibers
1000
B3 dB, opt. [MHz] excitation by laser
NAlaunch | 0.30
O = 650 nm
500
200
length [m]
100
10 20 30 40 50 60
Fig. 2.118: Bandwidth of a MC-GOF excited by a laser source
This type of fiber is suitable for the transmission of data rates in the Gbit/s
range over lengths of 10 m to 20 m.
Another glass fiber version which has gained increasing attention is the PCS,
i.e. silica glass fibers with a polymer cladding. The typical NA lies around 0.37.
However, there are versions available with a NA up to 0.48. Accordingly, the
bandwidth of PCS should lie in the range of DSI-POF. At the POF-AC predo-
minantly fibers with a core diameter of 200 m - the most commonly used value -
were measured. In Fig. 2.119 the length and launch-dependent results for a typical
PCS are represented. The fiber, 200/230 m with a 500 m primary coating, was
laid out for this measurement as a loose bundle with a diameter of about 30 cm
(see also [Ziem04a]).
B3 dB [MHz]
2000
launch NA:
1000 0.02 0.26
0.09 0.34
500
0.17 0.46
200
100
50 200 m PCS
loose bundle
length [m]
20
10 20 50 100 200 500
Fig. 2.119: Bandwidth of a 200 m PCS
This fiber was specified with a bandwidth of 100 MHz 100 m. This value can
be achieved for an under filled launch. For a full launch, however, you can only
attain about 60 MHz 100 m. The differences between the different launch condi-
2.5 Bandwidth of Optical Fibers 125
tions hardly decreases with fiber lengths up to 250 m. Mode mixing hardly occurs
with this measurement. The measurement was repeated for the same type of fiber,
whereby the fiber was wound around a spool. The results are shown in Fig. 2.120.
2000
bandwidth [MHz] launch NA:
1000 0.02 0.26
500 0.09 0.34
0.17 0.46
200
100
50 200 m PCS
fiber on a spool
length [m]
20
10 20 50 100 200 500
Fig. 2.120: Bandwidth of a 200 m PCS
The results pretty much agree for short fiber lengths. For longer lengths, how-
ever, the differences roughly disappear between the different launch conditions for
the rolled up PCS. This can only be explained by a recognizable increase in the
mode mixing. The bandwidths dependent on the coupling NA are compared for
250 m long samples in Fig. 2.121.
90
Bopt, 3 dB [MHz] 250 m PCS
80
70
60
50
40 fiber on a spool
30
20 loose bundle
10
launch NA
0
0.00 0.10 0.20 0.30 0.40 0.50
Fig. 2.121: Bandwidth comparison of 250 m PCS
126 2.5 Bandwidth of Optical Fibers
BL [MHz km]
40
30
20
15
10
8
4
0.0 0.1 0.2 0.3 0.4 0.5
launch NA
Fig. 2.122: Bandwidth dependence on launch conditions for 5 different PCS types
For fast Ethernet (125 Mbit/s) the bandwidths entirely suffice to bridge distan-
ces of up to 1 km. The first range limitations (maximum of 275 m at 850 nm emit-
ters and 62.5 m fiber) arise with Gigabit-Ethernet so that a new class of fibers
(OM2) has been defined which generally guarantees a transmission range of
550 m.
In the worst case a data rate of 10 Gbit/s could be transmitted on OM1 fibers
over about only 30 m. OM2 fibers are also limited to about 80 m. In order to be
able to transmit high data rates, three different procedures have been suggested:
Splitting the data rate into 4 2.5 Gbit/s which are then transmitted by
WDM on a fiber.
Emitter with so-called Restricted Mode Launch (RML) or Effective Laser
Launch (EL) respectively, whereby the power is coupled if possible within
the annulus with a diameter of between 4.5 m and 19 m. Moreover, the
NA of the emitter may not be too large.
Use of the new OM3 fiber class which has been optimized for the employ-
ment of 850 nm VCSEL.
An overview of the specified characteristics of the different GI-GOFs is
presented in Table 2.14. Specific products can on occasion clearly surpass these
parameters.
OM3
Class Unit OM1 OM2 OM3
550m
Fast Gigabit 10Gbit 10Gbit
typical applications
Ethernet Ethernet Ethernet Ethernet
core- [m] 50/62.5 50/62.5 50 50
D at 850 nm [dB/km] 3.5 3.5 3.5 3.0
D at 1.300 nm [dB/km] 1.5 1.5 1.5 1.0
BW 850 nm (OFL) [MHzkm] 200 500 1,500 3,500
BW 1.300 nm (OFL) [MHzkm] 500 500 500 500
BW 850 nm (LD) [MHzkm] n.d. n.d. 2,000 4,700
(OFL: Overfilled Launch)
Other less customary fiber types are, for example, GI-GOF with a core dia-
meter of 100 m and a cladding diameter of 140 m. Fig. 2.123 shows the fre-
quency response of a 500 m long sample with three different launch conditions. At
200 MHz km the results lie in the range of the fiber specifications.
The last fiber presented here is the semi-GI-PCS described above. The
measurement conditions become extremely more noticeable here so that the
measurement results shown may not be conclusively representative.
128 2.5 Bandwidth of Optical Fibers
1
rel. electr. level [dB]
0
-1
500 m fiber
-2 at 650 nm
-3
-4
-5 launch NA:
AN = 0.10
-6
AN = 0.34
-7 AN = 0.64
-8
1 10 100 f [MHz] 1000
Fig. 2.124 first shows the frequency response with a 500 m long sample for 6
different launch conditions measured at a wavelength of 650 nm.
-4
-6 500 m
Semi-GI-PCS
-8
-10
-12
1 3 10 30 100 300
frequency [MHz]
3000 B3 dB [MHz]
launch NA:
AN = 0.02
1000 AN = 0.09
AN = 0.17
AN = 0.26
300
AN = 0.34
AN = 0.46
100
length [m]
30
10 20 50 100 200 500 1000
Fig. 2.125: Bandwidth measurement of Semi-GI-PCS
What is striking is the low dependence of the bandwidth on the launch con-
ditions with longer sample lengths. Evidently, there is a significant exchange of
energy between the SI and GI modes in the fiber. The specified bandwidth value
could only be determined in short fiber lengths with under filled launch.
Bandwidth measurements on semi-GI PCS have also been published by
[Aiba04] and [Aiba05], whereby a method was used in which a light pulse circu-
lates in a 100 m long ring and passes an acousto-optic modulator after every pass.
The numerical aperture of the coupling optics amounts to only 0.25 and SI modes
are for the most part suppressed. The results for the frequency response, deter-
mined by Fourier transformation, are shown in Fig. 2.126.
-2 1st circulation
-4 10th circulation
-6
-8
f [GHz]
-10
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Fig. 2.126: Frequency responses of a Semi-GI-PCS according to [Aiba04]
The bandwidths thus determined are shown in Fig. 2.127. The values lie higher
by a factor of ten than the values measured with full launch on long fibers. This
130 2.5 Bandwidth of Optical Fibers
1.0
0.8
0.6
0.4
length [m]
0.3
100 200 400 600 1000
Fig. 2.127: Bandwidth of a Semi-GI-PCS according to [Aiba04]
Semi-GI fibers have large bandwidths, above all over short lengths and when
coupling into small angles.
The bandwidth of individual fibers - not yet placed in cables - under labora-
tory conditions can depend to a great extent on the external conditions, depen-
ding on the degree of induced mode coupling.
A comparison between POF and PCS is particularly interesting since both can
be used alternatively in many applications. The length-dependent bandwidths of
both types of fiber with full launch are illustrated in Fig. 2.128.
300
100
length [m]
30
3 10 30 100
Fig. 2.128: Bandwidth comparison of POF (fiber-NA: 0.50) and PCS (NA: 0.37)
Theoretically, the PCS should show about 50% greater bandwidth because of
its smaller NA - which has just about been confirmed by measurements. Both
measurement curves run approximately parallel which suggests similar magni-
tudes in mode-dependent processes. The angle-dependent attenuation of a typical
PCS fiber is illustrated in Fig. 2.129.
400
attenuation [dB/km]
300 50 m
200
100 m
100
11 dB/km
T []
0
-30 -20 -10 0 10 20 30
Fig. 2.130: Mode dependent loss of a Semi-GI-PCS (at 650 nm)
PCS does indeed show very large mode-dependent attenuation, the intensity of
which is comparable to POF. This explains the similar behavior even if the core
material itself has a very much lower attenuation.
A schematic comparison of typical bandwidth values for the different multi-
mode fibers described above are illustrated in Fig. 2.131. The values, as already
mentioned several times, can clearly deviate for specific products or under diffe-
rent measurement conditions.
PC-POF : 1000 m
MC-GOF : 1000 m
St.-NA-POF : 1000 m
200 m PCS : 200 m
DSI-POF : 1000 m
SI-MC-POF : 1000 m
Semi-GI-PCS : 200 m
DSI-MC-POF : 1000 m
MSI-POF : 750 m
OM-Giga : 900 m
GI-GOF OM1 : 62.5 m
GI-GOF OM2 : 50 m
PF-GI-POF : 120 m
GI-GOF OM3 : 50 m
OM3 mit LD : 50 m
1 10 100 1,000 10,000
bandwidth [MHzkm]
Fig. 2.131: Bandwidth comparison of different optical fibers (typical values)
2.5 Bandwidth of Optical Fibers 133
The bandwidths of the fibers presented vary over more than 3 magnitudes. If
singlemode fiber is used, however, then nowadays there is practically no longer
any bandwidth limit. Mode dispersion no longer arises. Chromatic and polariza-
tion mode dispersion can be compensated for as one likes. The significance of
chromatic dispersion will be discussed in the next section.
In all optical media we can observe the effect that the speed of propagation of
light of different wavelengths differs. When we differentiate the propagation con-
stants according to wavelength, we obtain the so-called chromatic dispersion,
usually expressed in ps/nmkm. This constant indicates by how much a signal's
delay will vary with the wavelength. In the typical application range of optical
fibers this value is negative which means that with increasing wavelength the
delay becomes smaller (corresponding to greater speed). Figure 2.132 shows the
chromatic dispersion for silica glass, PMMA and a typical fluorinated polymer
(according to [Koi97a]).
200
dispersion [ps/(nmkm)]
0
-200
-400
-600 PF-Polymer
-800 silica glass
PMMA
-1,000
-1,200
400 500 600 700 800 900 1,000 1,200 1,400 1,600
wavelength [nm]
Fig. 2.132: Dispersion of different materials
Typical semiconductor sources feature certain spectral widths that range from
some 10 nm for LED up to a few MHz for lasers (corresponding to some 10-5 nm).
In addition, there is the fact that when a light source is modulated there is always a
spectral broadening that cannot be less than a certain theoretical limit. This effect
only plays a role, however, with spectral singlemode lasers and with very high
data rates.
Figure 2.133 shows a schematic illustration of the effect of chromatic disper-
sion on a light pulse that has a given spectral width. A pulse with a certain spec-
134 2.5 Bandwidth of Optical Fibers
trum of the width 'O is launched into the fiber. After passing through the fiber
(length L) and experiencing a certain amount of dispersion D, the pulse has the
width 'W = D L 'O, whereby the shorter wave components arrive first. (cf.
Fig. 2.38 as well).
spectral shape
of the source
'O
O
't = DL'O
POF
length L
t t
input pulse output pulse
broadening by time
For silica singlemode fibers, the value for chromatic dispersion at 17 ps/nmkm
lies within the range of the smallest fiber attenuation at 1,550 nm wavelength.
Today, DFB-laser diodes are predominantly used for long-distance systems, the
spectral width of which is a maximum of a few MHz. What matters here essen-
tially is the broadening effect that is brought about by the data itself. In this case,
1 nm corresponds to approximately 125 GHz of spectral width. This means that
for a data rate of 10 Gbit/s a spectrum in the range of one-tenth nm is generated.
Where the permissible bit broadening is 0.05 ns, the fiber link may have a length
of approximately 30 km. For 2.5 Gbit/s this value rapidly increases to approxima-
tely 500 km due to the narrower spectrum and the greater pulse broadening per-
mitted. Conventional 2.5 Gbit/s systems can operate without specific actions
against dispersion. However, all systems that have many inline fiber amplifiers or
higher bit rates require devices to counteract chromatic dispersion. The most
common method today is the use of dispersion compensating fibers with strong
negative dispersion. Since these fibers utilize waveguide dispersion they can only
be produced as singlemode fibers.
The situation is significantly different for POF. The chromatic dispersion of
PMMA-POF with over 300 ps/nmkm at 650 nm wavelength is over 20 times lar-
ger than of silica fibers at 1,550 nm wavelength. For POF it is also usual to use
LED with a typical spectral width of 20 nm to 40 nm and not lasers that have just
a few tenths of a nanometer of spectral width. On the other hand, there are the
typically short distances of POF systems and the moderate bit rates. Table 2.15
lists some examples for the effect of chromatic dispersion in POF systems.
2.5 Bandwidth of Optical Fibers 135
The first three examples are based on LED for transmitting data rates up to
155 Mbit/s over a maximum length of 100 m. Even in the unfavorable case of
using green LED, pulse broadening is less than one the bit length so that there is
only a small effect on the system. In the fourth example, the intention is to trans-
mit an IEEE1394 S400 data stream (with 500 Mbit/s physical data rate) over a
distance of 70 m using a green LED. Here pulse broadening is nearly in the same
range as the bit length. When this deteriorating effect due to mode dispersion is
added, one can see that this system can only work with considerable additional
efforts. It may, for example, be possible to partially provide electrical compen-
sation, whereby higher optical receiving power is required. When using data rates
from Gbit/s to 1 Gbit/s, the use of spectrally narrower sources becomes neces-
sary. These primarily include RC-LED and VCSEL (see Chapter 4), and for even
higher requirements DFB laser diodes. In most cases this selection is required
anyway due to the limited modulation bandwidth of LED.
Fluorinated graded index profile polymer fibers feature significantly reduced
chromatic dispersion compared with PMMA-POF. These fibers are designed for
use in Gbit/s systems operating at spectral ranges between 800 nm and 1,300 nm.
It is for these demands only that laser diodes can be considered, not least due to
the smaller core diameters, the spectral width of which is a few nanometers at
most. The last row shows that in such a case chromatic dispersion can be neglec-
ted even for a transmission length of a few 100 m.
laser diodes are used where the emitting angle is significantly smaller than the
angle of acceptance of a SI-POF. The use of solid-state lasers or gas lasers, the
exact wavelength of which is often required for measuring purposes, is even more
problematical. These lasers emit collimated light so that only a small proportion of
the POF modes can be excited. When using glow lamps or discharge lamps, opti-
cal devices are used to collimate the light to the fiber. For this reason it is difficult
to find lenses that actually work with consistent efficiency in the given acceptance
range. All this has the effect in a concrete experiment of increasing the deviations
of the actual bandwidth in comparison with the theoretical limit value. This is a
very undesirable effect when attempting to define characteristics by making
measurements of this kind, as shown in Chapter 7. However, for high bit-rate data
transmission this situation can in practice also be beneficially exploited as shown
by the following examples.
Figure 2.134 demonstrates the most important methods for increasing the
bandwidth of a POF.
Launching light at a small angle as well as detecting just a selected angle range
has the effect of restricting the modes involved in signal transmission and thus
reducing pulse broadening. It is possible to electrically compensate for the resul-
ting low pass behavior, both before as well as after the POF link. To date the most
significant increases in bandwidth for a POF system have been described in
[Bat92] (see also [Bat96a] and [Yas93]). The following components were utilized:
Launch with a small AN = 0.11, thereby exciting only a few modes with only
small differences in delay.
Pre-distortion of the LD excitation signal (peaking); high pass (33 pF [[51 :).
Detection with low NA (modes with large delay differences are blanked out).
Dispersion compensation behind the receiver; high pass (8 pF [[200 :).
It was possible to transmit at more than 500 Mbit/s across a distance of 100 m
of standard NA-POF (see also chapter 6). However, all these measures are usually
at the expense of a reduced power budget, as summarized in Table 2.16.
2.5 Bandwidth of Optical Fibers 137
It follows that the use of such methods is of particular interest in systems that
have adequate power reserves. POF attenuation across very short distances is
hardly of importance; on the other hand, the use of high data rates is of interest in
various applications. Chapter 6 will describe experiments for transmitting Gbit/s
over distances of 10 m to 100 m conducted by T-Nova GmbH, the University of
Ulm, Daimler Chrysler, the Fraunhofer Institute for Integrated Circuits Nuremberg
and the POF-AC Nrnberg.
Figure 2.25 shows theoretical considerations with respect to the POF bandwidth
at different launching angles (Gaussian shaped far field with 3 dB width calculated
relative to fiber NA) according to [Bun99a]. With short lengths and small launch
NAs the light remains concentrated in areas with small propagation angles. The
small differences in propagation time result in large bandwidths. After approx.
100 m of fiber equilibrium mode distribution is just about reached through mode
mixing, and the influence of the launch conditions gradually disappears. This be-
havior corresponds to a great degree to the measurement results described above.
1,000
B3 dB [MHz]
rel. launch NA
500 (NAfiber = 1)
0.5 1.5
200 0.7 1.7
1.0 2.0
100 1.2
50
20
length [m]
10
10 20 50 100 200 500
In the experiment, the data rate was 1,200 Mbit/s with NRZ coding. The twin-
stage pre-distortion filter dampens the signal by 12 dB in the low frequency range
so that the higher frequencies can create a stronger modulation. For the pulses this
means steeper edges and overshoot at the beginning and end, hence the term
peaking, as shown in Fig. 2.137.
nations of the different systems. The diagram also shows the theoretical limits for
the bandwidth of standard NA-POF and DSI-POF (assuming NRZ coding and bit
rate = 2 u 3-dB bandwidth).
POF-AC
1,000
POF-AC
UNI Ulm
POF-AC
500
Bates 93
Daimler
Chrysler IEEE 1394 Bates 92
200
ATM 155
100BaseFX Kaiser 92
100
10 20 50 100 200
POF-length [m]
Fig. 2.138: Bit rates of different POF systems (status 2003)
It is easily discernible that a number of systems with standard NA-POF are sig-
nificantly above the theoretical limits. Particularly for greater lengths the potential
for exceeding the limit is considerable. As shown in the next section, the practical
application presents some problems such as the bending behavior. It is generally
true that extreme dispersion compensation must be adaptive in its execution. That
means that above all the limit frequencies of the high passes must be adapted very
precisely to the frequency response of the link. If the frequency response changes,
the result will be too much or too little compensation so that the pulses become
distorted. Such a change may, for example, occur as a result of different lengths of
cable; however even a bend in the fiber may have the same effect. In commercial
systems it is desirable to avoid having to use automatic adaptations, such as is
necessary, for example, in 1000BaseT-systems, or having to provide specific
receivers for different cable lengths. One practical solution is to adjust the com-
pensation in such a way that there is just-tolerable over-compensation for short
lengths of fiber; based on this level of compensation, the next step is to select the
maximum fiber length for which this compensation is still just sufficient. A sche-
matic illustration is shown in Fig. 2.139.
140 2.5 Bandwidth of Optical Fibers
f f f f
compensation filter
(fixed)
f
resulting frequency response
f f f f
overcompen- optimized undercompen-
sation compensation sation
A proposal for increasing the bandwidth by direct interference with the optical
path is described in [Kal99]. By using a mode filter immediately after the trans-
mitter, the light angle range in the fiber is reduced as shown schematically in
Fig. 2.140. With this method it was possible to achieve an improvement in band-
width by 53% and 89% respectively for two standard NA-POF provided by
Mitsubishi and Toray. The losses of the mode mixer are approximately 2.5 dB
which is perfectly acceptable in many applications.
AN | 0.43 AN | 0.29
POF
source receiver
mode filter
Fig. 2.140: Increased bandwidth with a mode filter ([Kal99])
In general, optical transmission systems are set up in such a way that the system
bandwidth amounts to at least 50% of the bit rate with NRZ transmission. Thus,
500 MHz are needed to transmit 1000 Mbit/s. This means that the eye is complete-
ly open with ideally adapted filtering. In other words, the transition from the zero
symbol to the one symbol and vice versa takes place within the bit duration. If the
system bandwidth is smaller than half of the bit rate, then the symbol transition
needs more time resulting in a reduction of the vertical eye opening. This effect
must either be compensated for through adapted filtering or the reduced eye
opening is compensated for by a correspondingly higher receiving level. The
deterioration of the signal-to-noise ratio at the receiver through the bandwidth
limitation is called penalty (measured in dB). The relationship between signal-to-
noise ratio, receiving level and penalty is shown in Fig. 2.141.
U U
U1 U2
t t
system without noise and with system without noise and with
sufficiend bandwidth - the eye limited bandwidth - the eye in
open completely closed partially
penalty: 20log(U2/U1)
U U
UN
US
t t
system with noise and with system with noise and with
sufficient bandwidth - the limited bandwidth - the
eye is open completely eye is more closed
SNR = 20log(US/UN) SNR is decreased by penalty
In the example shown 820 Mbit/s were transmitted over 100 m of DSI-POF.
Although the eye was almost completely closed, an error free transmission was
possible. In a real system, however, certain detection would be relatively difficult
since the sampling moment and the decision threshold have to be re-adjusted very
exactly. Furthermore, there are no margin whatsoever for fluctuations in the laser
power or bending losses.
25 penalty [dB]
20 simulated
with fiber
15
10
0
0.10 0.20 0.30 0.40 0.60 0.80 1.00
bandwidth/bit rate [MHz/Mbit/s]
Fig. 2.143: Effect of system bandwidth on the penalty
2.6 Bending Properties of Optical fibers 143
The simulated values were determined by calculating the penalties with the aid
of PSpice analyses. A Gaussian-shaped filter was used as a low-pass system. The
measuring points were determined on a 20 m long standard POF with different bit
rates. The penalty was estimated from the eye diagram. The measured values
tallied greatly with the simulation down to 25% of the system bandwidth, e.g. a
transmission of 1 Gbit/s with a system bandwidth of 250 MHz. With higher bit
rates the penalty increases more quickly than in the simulation. One main reason is
that the frequency response only corresponds to a certain degree to idealized
Gaussian behavior. It hardly makes sense to use practical systems with more than
a 10 dB penalty.
The results show that an exact relationship does not have to necessarily exist
between the maximum bit rate and the fiber bandwidth. Furthermore, even with
bandwidth-limited systems relatively high data rates can be achieved under labora-
tory conditions if enough emitting power is available.
The essential parameters which determine the bending sensitivity of a fiber are the
diameter and the numerical aperture. The larger the NA, the narrower the permis-
sible bending radii may be in relation to the fiber diameter. Figures 2.144 and
2.145 show the losses for bends of different commercially available fibers accor-
ding to information in the data sheets ([Tor96a] and [Asa97]).
The Fig. 2.144 shows the bending losses of two different SI-POFs with some-
what different NAs. You can clearly see that larger NAs reduce the bending
losses.
fiber
4.0
PFU-CD-1001
AN = 0.46
3.0
PGU-CD-1001
AN = 0.50
2.0
1.0
0,0
0 5 10 15 20 25 30 35 40
bend radius [mm]
Fig. 2.144: Loss for 360 bend according to [Tor96a]
Figure 2.145 shows losses resulting from bends in a standard NA-POF, a low-
NA-POF and a multi-core fiber (see Chapter 2.3).
The low-NA-POF shows significantly larger losses compared to a standard
NA-POF. Due to the smaller individual core diameters, the bending sensitivity of
the multi-core fiber is comparable with that of the standard NA-POF despite the
smaller NA.
If many bends directly follow each other, attenuation does not increase propor-
tionally with the number of bends because there is less and less energy present in
the higher mode groups. Figure 2.146 shows a measurement of the bending losses
for different POF according [Hen99].
2.6 Bending Properties of Optical fibers 145
6
bend losses [dB] fiber
5 TC 1000 (AN = 0.485)
10.0
loss [dB]
5.00 NC 1000 (Low-NA)
no longer available at
2.00 the market
1.00
0.50 AC 1000 (DSI) PFU 1000 (St.-NA)
The measurements were taken at 650 nm with LED-launch and a mode mixer.
The bending radius was 32 mm and the bends were located at the beginning of a
50 m sample length.
PFU 1000 is a standard NA-POF, while MH 4000 and AC 1000 are double-step
index POF. Their losses are approximately identical and up to 10 windings are
significantly below 1.0 dB. By comparison, the low-NA-POF NC 1000 is in the
range of 10 dB, which is too much for deployment in practical applications. The
ATM forum stipulates an admissible bending radius of 25 mm and at this radius
the attenuation was already above the range of measurement. Meanwhile, DSI-
POF offer significantly improved bending characteristics at comparable NA.
Figures 2.147 and 2.148 demonstrate the losses over the inverse bending radius
and the number of windings for a (genuine) low-NA-POF (NC 1000) and a stan-
dard NA-POF [Hen99].
146 2.6 Bending Properties of Optical fibers
12
loss [dB] 12 mm
10 10 turns
8 turns
8
15 mm
6 turns
6
4 turns
4
21 mm 2 turns
26 mm
2 32 mm
39 mm
0
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
inverse bend radius [mm-1]
16
loss [dB] 28 mm 10 turns
14
12 8 turns
32 mm
10
6 turns
8
4 turns
6
4 39 mm
2 turns
50 mm
2
0
0.020 0.022 0.024 0.026 0.028 0.03 0.032 0.034 0.036
inverse bend radius [mm-1]
For graded index POF slightly different conditions apply for bending sensitivity
compared with step index profile fibers. Here it is not the total reflection at the
core-cladding interface but the continuous bending in the index profile that is res-
ponsible for light guiding. In addition, there is a fundamentally different distri-
bution in the near and far field. Figure 2.149 shows a measurement for GI-POF
according to [Ish95].
2
1
0.5
bend radius [mm]
0.2
0 10 15 20 25 30 35 40 45 50 55
Fig. 2.149: Loss of two GI-POF ([Ish95]) for a 90 bend
Due to the different dopants used, the two samples with a core diameter of
0.5 mm each have a different NA, which has a very significant effect on the ben-
ding losses. Despite the smaller core diameter the losses for a 25 mm bend are still
significantly higher than the values for a SI-POF or a DSI-POF. Here, too, a
reduction in the core diameters leads to lower bending losses.
[Aru05] describes how the bending losses in PMMA GI POF can be signifi-
cantly reduced. In addition to an optimized index profile an additional PVDF layer
(polyvinylidenfluoride) was applied to the core with parabolic profile resulting in
a semi-GI-POF which combines high bandwidth with low bending losses. The
losses of a 90 bend are shown later in Fig. 2.205 compared with a conventional
PMMA GI-POF. (The sample length was 100 m.) Even with a bending radius of
5 mm there was no measurable increase in attenuation. The different methods for
reducing bending losses in PMMA GI-POF and PF-GI-POF are described in
greater detail in Section 2.8 on fiber production. Examples of measurements are
also shown.
However, bends do not only contribute to additional losses, but also have an effect
on bandwidth because certain mode groups are selectively attenuated. This effect
is exploited in mode filters and mixers.
148 2.6 Bending Properties of Optical fibers
Figure 2.150 (according to [Rit93]) shows what the effect of a 720 bend at the
beginning of a 50 m long POF link has on the measured bandwidth. In this case
the light is launched with AN = 0.10.
bandwidth [MHz]
160
POF:
140
250 m
120 500 m
750 m
100 1000 m
80
60
Due to the low launch NA, the bandwidth is relatively large (80 MHz 100 m).
In the case of tight bending radii at the beginning of the fiber there is mode mixing
so that the bandwidth is significantly reduced sometimes. This effect is naturally
more pronounced for smaller diameters. In the illustration selected here above the
inverse relative bending radius, relative to the core diameter, the effect of the core
diameter should disappear. It seams to be, that the effect described above of the
larger bandwidth for thinner fibers is already dominant here due to the more mode
dependent processes.
Comprehensive investigations of the effect of bends on the bandwidth of POF
links were presented in [Mar00]. The test fiber consisted of a 100 m long standard
NA-POF; 360 bends were inserted at the beginning of the link, after 25 m, after
50 m, after 75 m or at the end of the link. The source consisted of a 655 nm laser
diode, the NA of which could be adapted through different optics from 0.10 to
0.65. The bandwidth and the attenuation of the overall link were measured without
bends and with bending radii of 6.4 mm, 11.1 mm and 13.8 mm. The results are
shown in Fig. 2.151.
When light is launched into the fiber using a large NA, the original bandwidth
of approximately 33 MHz can be increased significantly. However, large improve-
ments with small bending radii occur at the expense of large additional losses. The
biggest gain in bandwidth is obtained with a bend in the middle because this
means that many modes of the first 50 m are filtered out and EMD is not com-
pletely regained in the remaining 50 m. The changes in attenuation are largely
independent of the length since the mode field is well filled out everywhere.
2.6 Bending Properties of Optical fibers 149
When light is launched into the fiber using smaller NA, the relative gain in
bandwidth compared to the original - approximately - 60 MHz is not as big.
Therefore the optimum position for the bend is clearly nearer to the end since the
mode field must first be filled. Again, tight radii have more effect. The additional
attenuation increases significantly when the bends are moved to the end, since at
the beginning of the fiber there are hardly any higher mode families in existence.
These results also confirm clearly for the existing assumptions with respect to
mode propagation in a coupling length of some 10 m.
A very simple method to decrease bending radii is to reduce the core diameter
while otherwise retaining identical parameters. If you wish to maintain the
advantage of the simple handling of ready-made thick fibers, then there is the
possibility of fiber bundles or multicore fibers respectively.
Fig. 2.152 and 2.153 show the measured bending losses, each with a bend of
360 in the middle of the sample, with UMD launch and measured with an inte-
grating sphere. A 10 m long fiber was used for the MC-GOF. The range of the
bending radii lay between 2 mm and 100 mm. The bending attenuation measured
lies below 0.1 dB.
150 2.6 Bending Properties of Optical fibers
0.06
bending loss [dB]
0.05
0.04
0.03
0.02
one bend
0.01 by 360
0.00
inverse bending radius [mm-1]
The bending losses of the MC-POF were measured on a 100 m long sample in
order to guarantee as much mode equilibrium as possible. The bend (360) was
made in the middle of the fiber length. Due to the different relations between
mode coupling and absorption the EMD conditions for 520 nm and 650 nm only
differ slightly. That is the reason for the somewhat different bending losses.
0.10
In many areas the 200 m PCS is used because it permits smaller bending radii.
Fig. 2.154 illustrates quite graphically that the same physical characteristics are
also valid for these fibers. Here the bending losses are given versus the relative
bending radius in relation to the fiber diameter. The numbers in brackets indicate
the bending radius in millimeters for the PCS. Both fibers thus have in relative
terms an identical bending sensitivity.
2.6 Bending Properties of Optical fibers 151
3.0 1 mm POF
2.5
2.0
1.5
1.0
0.5 bending radius [u Kern]
0.0
0 5 10 15 20 25 30 35 40 45 50
(0) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Thin POF could be used as an alternative to PCS in many areas when tight
bending radii are indeed required, but the attenuation and the temperature range of
the POF are satisfactory. A comparison between a 250 m SI-POF and a 200 m
PCS, measured at 650 nm with full launch for 5 m long samples, is illustrated in
Fig. 2.155.
10.0
bending loss [dB]
3.0
0.3
0.1
200 m PCS
0.03
bending radius [mm]
0.01
0 1 2 3 4 5 6 7 8 9 10
Fig. 2.155: Bending losses of small diameter POF and PCS in comparison
The somewhat thicker POF also has somewhat higher bending losses. A tenth
of a dB is attained for the POF at a bending radius of 8 mm and 6 mm for PCS.
The bending losses of three different SI-POFs with different NAs are compared
in Fig. 2.156. The lowest losses are shown by the 300 m thick POF with a high
152 2.6 Bending Properties of Optical fibers
NA. The 250 m and 500 m thick POFs have almost identical bending attenu-
ations. It is thus indicative that the NA is by far the most important factor for the
bending losses. Consequently, you should always choose fibers with the largest
possible NA for particularly tight radii, unless you decide to go back to multicore
fibers. In addition, the latter have the advantage of offering an even greater
bandwidth.
10.00
bending losses [dB] 250 m POF (AN = 0.63)
500 m POF (AN = 0.50)
1.00 300 m POF (AN = 0.63)
0.10
0.01
3.0
D [dB] 7.5 u r
7.5 u r
7.3 u r
1.0 8.0 u r
0.3
0.1
250 m 750 m
1000 m
0.03
500 m
0.01 r [mm]
0.3 1.0 3.0 10.0 30.0
Fig. 2.157: Comparison of bending losses of various SI-POF
2.6 Bending Properties of Optical fibers 153
Bending radii are drawn in the picture with which a bend (360) results in
exactly 1 dB additional attenuation. With the four fibers with their 250 m to
1000 m core diameters this is the case each with a seven-fold to eight-fold fiber
radius, i.e. a bending radius between 0.9 mm and 4 mm. As a comparison, the
bending losses of a 125 m SI-POF ([Witt04]) are shown in Fig. 2.158.
1.0
0.5
0.2
bending radius [mm]
0.1
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Optimedia has made available samples of a thinner PMMA GI-POF. The ben-
ding losses of this fiber with overfilled launch (LED) and a launch with a laser
(AN = 0.10) are shown in Fig. 2.159. Both measurements were carried out at
650 nm with a 5 m long fiber.
0.5
laser launch
0.2
0.1
0 20 40 60 80 100
bending radius [mm]
Finally, some results from a project work [Bau06] are shown. First, Fig. 2.160
compares the bending losses of three Toray fibers with different core diameters:
500 m, 750 m and 1,000 m. The NA of the three fibers is the same. As expec-
ted, the bending radius for a given attenuation is reduced nearly proportional to the
fiber diameter. Only with very thin fibers does the effect of stronger mode-depen-
dent attenuation make itself noticeable.
10
bending losses [dB] fiber type: Toray PFU
AN = 0.47
measured with 650 nm LED
1 bend 360, 10 m fiber
1
0.1 1,000 m
750 m
500 m
0.01
10
bending losses [dB] fiber type: St.-NA
measured with 650 nm LED
1 bend 360, 10 m fiber
1
0.1
0.01
0 10 20 30 40 50
bending radius [mm]
Fig. 2.161: Bending losses of various standard-NA-POF
2.7 Materials for POF 155
2.7.1 PMMA
The material most frequently used for polymer fibers is the thermoplastics PMMA
(Polymethylmethacrylate), better known as Plexiglas. Figure 2.162 shows the
structure of the monomer and its polymer chain.
MMA PMMA
H
H C H CH3 CH3 CH3 CH3
H
C C C CH2 C CH2 C CH2 C CH2
C H C C C C
O H O
O C OCH3 O OCH3 O OCH3 O OCH3
H H
As can be seen in the illustration, each MMA monomer has a total of eight C-H
bonds. The vibrations of this compound, or more precisely its harmonic waves are
a main cause for the losses encountered in PMMA polymer fibers. The attenuation
resulting from absorption at the respective wavelength is shown in [Mur96] and
[Koi96c] (see Fig. 2.163 and table 2.18). In particular the harmonic waves at
627 nm (6th harmonic wave) and 736 nm (5th harmonic wave) essentially deter-
mine the level of attenuation within the application range of PMMA-POF because
these are not narrow absorption lines but relatively wide bands. Further causes for
attenuation will be discussed in the chapter titled Characteristics.
108
attenuation [dB/km]
106 molecule
104 C-H
102 C-D
C-F
100
C - Cl
10-2
10-4
10-6
10-8
500 1000 1500 2000
wavelength [nm]
Quite early in the history of this technology, the idea came up to reduce the ab-
sorption losses of polymer fibers by using different materials in which less or no
C-H bonds were present. However, it is not easy to eliminate these; instead, the
hydrogen atoms are replaced by other atoms of the 7th main group. A heavier core
will result in a lower vibration frequency, thus moving the attenuation bands to a
larger wavelength. The illustration shows the attenuation bands for deuterium
(heavy hydrogen with the atomic weight 2), fluorine (atomic weight 19) and
chlorine (atomic weight 35 or 37, see also [Bau94]). Generally, the materials for
polymer fibers can be divided into three groups:
compounds containing hydrogen
compounds with partial substitution of hydrogen
compounds with complete substitution of hydrogen
2.7 Materials for POF 157
Fibers with high resistance to heat are especially needed for use in certain areas of
automotive engineering (engine compartment) and automation technology. In the
passenger compartment of a vehicle a maximum of +85C will arise. PMMA-POF
can easily be used with such temperatures. In the area near the center console or
under the roof temperatures can also go up to over +100C and near the engine to
+125C. Summaries of the data published so far and of comprehensive investi-
gations at the POF-AC Nrnberg can also be found in [Poi03a] and [Poi03b]. On
the whole the following methods for increasing the resistance to heat of polymer
fibers have been presented:
Cross-linking of PMMA: cross-linking between polymer chains can be
generated by chemical effects or by UV irradiation which results in a rise of
Tg. At the same time, however, the scattering and the mechanical charac-
teristics become worse.
Polycarbonate: PC has a considerably greater Tg compared with PMMA and
is likewise transparent. Fibers made of this material have been produced on a
large scale. PC fibers, however, age relatively quickly in combination with
humidity.
Elastomers: fibers made of this material could be used up to +170C and
show very low attenuation. So far, they have only been produced as labora-
tory samples.
Alternative polymers: a series of other polymers such as cyclical polyolefins
have Tg up to +200C.
When determining the thermal stability, a maximum increase in the kilometric
attenuation is established over a maximum period of aging. In case the aging pro-
cedures are thermally activated, then the permissible operating period decreases
almost logarithmically to the temperature. An example of the behavior of a stan-
158 2.7 Materials for POF
dard PMMA-POF can be seen in Fig. 2.164 (measurements were made at the
POF-AC). The increase in losses is represented here vs. the temperature. With an
approx. 10 K increase in temperature the speed of ageing increases about one
order of magnitude.
1000
increase of
attenuation
coefficient
dB/(km1000 h)
100
520 nm
590 nm
10
650 nm
temperature [C]
1
70 75 80 85 90 95
Fig. 2.164: PMMA-POF ageing
4.000
attenuation [dB/km]
2.000
1.000
500
Tver-POF1
H-POF
200
[Tan94a]
PHKS wavelength [nm]
100
400 500 600 700 800
Fig. 2.165: Attenuation of cross-linked PMMA-POF
On the whole it is true for this type of fiber that a higher degree of cross-linking
leads to higher application temperatures, whereby the scattering is also greater so
that the losses increase. A short piece of a Tver fiber sample exposed to red light
is shown in Fig. 2.166. The high degree of scattering leads to a clearly visible
lateral emission.
2,000
1,000
4
PC(AF)
2 PC-A
0
105 115 125 135 145 155
temperature [C]
Fig. 2.168: Temperature resistance of PC ([Hatt98])
5,000
attenuation [dB/km]
PC(AF)
2,000 [Hatt98]
D-POF
[Irie94]
1,000
PC-POF
[Irie94]
500
wavelength [nm]
300
400 500 600 700 800 900
Fig. 2.169: Data by Furukawa 1994-1998 (Polycarbonate)
120%
relative transmission at 92C / 95 % RH
wavelength: 650 nm;
100%
sample length: 10 m
SI-mod. PMMA
80%
60%
40% SI-PMMA
20% SI-PC
10,000
attenuation [dB/km]
[Ish92]
5,000 HPOF-S
HPOF-Sb
[Suk94]
[Zei03]
2,000
1,000
400450 500 550 600 650 700 750 800 850 900
wavelength [nm]
The attenuation curves of different EOFs (elastomer optical fiber) are compared
in Fig. 2.171. Particulars of the following fibers are compared:
2.7 Materials for POF 163
1000
500
500 550 600 650 700 750 800 850
wavelength [nm]
Fig. 2.172: PC-POF in comparison with silicone-POF
H H H H
C C C C
H R
x
R R y
Fig. 2.174 shows an aging experiment at +130C with different fibers described
above. The most suitable ones at these temperatures were evidently the EOF and
the PC-POF.
10,000
attenuation [dB/km]
5,000
T = +130C
TVER 2002
2,000
1,000 FH 4001
500 PHKS
200 HPOF-S
measuring time [hours]
100
0 100 200 300 400 500 600 700
Fig. 2.174: Ageing of various POF at high temperatures
The PC-POF from Mitsubishi (FH4001) only shows a moderate increase while
the two POFs made of cross-linked PMMA aged more quickly. The EOF even
gets better during the measurement period. Particularly noticeable is the clear drop
in attenuation after 15 hours. This was the point at which the temperature was
raised in the climate test chamber. It was noticeable that the bandwidth of the EOF
had dramatically diminished after this treatment. The combination of both events
provides the explanation that the adhesion of the cladding onto the core was
clearly improved by the high temperature so that even higher modes can now be
guided.
Another candidate for the production of polymer optical fibers is polystyrene (PS),
the molecular structure of which is shown in Fig. 2.175 ([Ram99]).
H H H H
C C C C
C H C H
H C C H H C C H
H C C H C C
C H H
C
H H n
To date, PS-POF have been manufactured e.g. by Toray (first PS-POF 1972),
NTT (1982) and CIS in Tver (1993). The initial fibers had an attenuation of over
1,000 dB/km; later on it was possible to reduce this to 114 dB/km at 670 nm
([Koi95]). The NA of these fibers which can be used at temperatures up to 70C is
0.56, i.e. a little higher than that for the standard PMMA-POF. Figure 2.216 shows
the attenuation behavior of a PS-POF ([Ram99], red curve and [Zub001b]).
1000
attenuation [dB/km]
800
600
400
200
[Zub01b]
wavelength [nm]
100
500 550 600 650 700 750 800 850
The refractive index of PS is n = 1.59 so that it is possible to use PMMA for the
optical cladding (n = 1.49), as is possible for PC (n = 1.58). The glass transition
temperature of PS is approx. 100C and therefore approx. 5 K lower than that of
PMMA. Hitherto there has been no reason to replace the PMMA-POF by PS so
that this material is not of any practical significance.
where which protons (normal hydrogen nuclei) slowly replace the deuterium so
that the absorption losses will increase again.
Although it is possible to solve the problem with a watertight coating of the
fiber (including all connections), this would defeat the object of obtaining a parti-
cularly low priced cable system.
10,000
attenuation [dB/km]
5,000
2,000
1,000
500 [Koi95]
[Ish92a]
200
[Koi96b]
100 [Koi96d]
50 [Mur96]
20
wavelength [nm]
10
500 600 700 800 900 1000 1100 1200 1300 1400
Fig. 2.177: Loss spectra of GI-POF (deuterated, 1996)
In the past few years work has once again been conducted in Japan on the
production of deuterated POF. GI fibers exclusively have been investigated - see
[Kon02], [Kon03] and [Kon04]. The attenuation of these fibers from [Kon04] is
compared in Fig. 2.178 with the values from 1995 and those of a PMMA-POF.
1500
attenuation [dB/km]
1000 PMMA
500
d8-PMMA 2002
0 1995
450 550 650 750 850
wavelength [nm]
10.000
attenuation [dB/km]
5.000
2.000 PMMA
PMMA
1.000
500
200
100 d8-POF
50
20
wavelength [nm]
10
500 550 600 650 700 750 800
Fig. 2.179: Attenuation of deuterated POF ([Kon04])
Since 2003, Fujifilm has been announcing the development of a new fiber
Lumistar in the versions I, V and X. According to their own statements this is:
the first POF with a large diameter which is able to transmit over 1 Gbit/s. This
is somewhat exaggerated, of course, since PMMA GI-POF and MC-POF have
been able to do this for many years.
power [dB]
0.0
-3.0
-4.0
-5.0
frequency [GHz]
-6.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Fig. 2.181: Refractive index profile of the Lumistar GI-POF after ageing (+90C)
Parameters for the Lumistar fibers are mentioned in different sources. Accor-
ding to this information a particularly low-attenuation polymer is used. Since the
company works closely with Keio University, where until 2004 there were reports
on the development of deuterated fibers with very similar parameters, we must
assume that we are dealing here with d8 PMMA-POF.
1000
attenuation [dB/km]
500
PMMA-d8 core
PMMA-cladding
200
100 PMMA-d8
complete
wavelength [nm]
50
500 550 600 650 700 750 800
The bandwidth of the fibers was determined through pulse broadening in the
time domain, a SI-POF was used as a mode mixer. The fiber with the PMMA d8
core and the PMMA cladding attains 1.2 GHz 300 m (overfilled launch). This
version does indeed have a somewhat higher attenuation, but also has a higher
bandwidth due to the index dip at the core-cladding interface.
In October 2004, Fujifilm introduced a DVI transmission system on the basis of
the Lumistar fiber. Using a 850 nm VCSEL a data rate of 10.3 Gbit/s over 40 m
could be transmitted (eye diagram in Fig. 2.183).
To what extent this fiber is actually available on the market cannot yet be
assessed since there are no channels of distribution yet in Europe. Even the actual
production costs are still unknown.
The use of fluorine instead of deuterium is indeed more complicated, but does
promise even lower attenuation values and above all long-life fibers. The
following section describes the development of these fibers.
2.7 Materials for POF 173
The atomic mass of fluorine is many times greater than that of hydrogen so that
the absorption bands are moved significantly further into the infra-red zone. The
theoretical minimum values are less than 0.2 dB/km ([Mur96]), i.e. comparable to
silica fibers in the wavelength range of about 1,500 nm. Figure 2.184 compares
the attenuation values theoretically possible for fluorinated polymers with those
achieved for singlemode glass fibers.
silica glass
PF-polymer
10
The process of doping involves inserting small molecules between the long
chains of the actual core material which increases the refractive index. What is
important is that the dopants do not diffuse out of the polymer material too easily
and do not show too strong absorption in the desired wavelength range. The
doping process always lowers the glass transition temperature. It is therefore
desirable to insert a molecule that accomplishes the required change in the refrac-
tive index even at small concentrations (a few percent).
In co-polymerization one uses chains composed of different monomers. The
ratio of monomers determines the refractive index. In this case it is important that
the sequence should be irregular - no long chains of one monomer are formed -
since otherwise the losses due to scattering increase considerably. This means that
the bonding force of monomers amongst each other must not be greater than the
bonding force to the respective other monomer. Of course, both monomers must
have sufficient transparency. Figures 2.185 and 2.186 show a schematic illustra-
tion of the principles.
monomer
dopant
monomer A
monomer B
CYTOP
CF2=CF-O-CF2-CF2-CF=CF2
momomer polymer
It was possible to reduce the attenuation of fibers step by step from initially
over 50 dB/km to 30 dB/km and finally to less than 10 dB/km at a wavelength of
1,300 nm, as shown in the data for different PF-GI-POF in Table 2.25.
Different attenuation spectra of GI-POF are compared in Fig. 2.189. The years
indicate the history of the development of this technology. Estimates in [Mur96]
suggest that attenuation for CYTOP will be less than 1 dB/km, bearing in mind
that the need for a GI profile will have a negative effect on this value.
176 2.7 Materials for POF
500
1995
200 1996
100
50
1998
20
2000 wavelength [nm]
10
500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600
100
attenuation [dB/km]
60
40
20
10 AGC
OFS
6
600 700 800 900 1,000 1,100 1,200 1,300 1,400
wavelength [nm]
Apart from the materials used in the fiber core, the material used for the jacket is
also important. It has a significant contributing effect on thermal resistance. In
addition, the jacket determines the mechanical properties of the cable such as re-
sistance to compressive load and tensile strength as well as flexibility. Tables 2.26
through 2.30 list different possible materials with some of their characteristics.
The use of PVC, PE or PA as typical jacket materials for applications within
buildings allows for maximum temperatures ranging from 70C up to 90C. The
materials in the last two rows (trade names are Teflon FEP or Teflon PTFE) can
be used at significantly higher temperatures.
178 2.7 Materials for POF
The processes for producing POF have been continuously improved in the last few
years. The fundamental methods have indeed always remained the same, but vari-
ous details have been improved. A very comprehensive treatment of POF produc-
tion and its history can be found in [Nal04]. Many fine points concerning the
materials can also be found in [Har99].
As opposed to the production of glass fibers there is a number of unusual
features with POF. First of all, the polymer chemistry involved, in part very com-
plicated and with its occasional safety aspects, has to be mastered. On the other
hand the process temperatures are very much lower - almost always below
+200C.
The demands on POF production can be sub-divided into four areas:
The core material must be produced uniformly without any impurities, air
bubbles, etc. and with a correct distribution of the molecular masses.
The fiber must be drawn or extruded exactly.
For SI fibers a suitable cladding material with low refractive index and an
attenuation not too high must be found and applied. In doing so, one must
guarantee that the interface is sufficiently smooth and that the cladding has a
good wringing fit.
For graded index fibers a copolymer or a dopant must be found in order to be
able to vary - usually increase - the refractive index. A suitable process is
needed in order to distribute this material over the core cross-section so that
you have a parabolic refractive index profile.
There are other steps, of course, such as the application of additional protective
layers, the production of duplex or ribbon cables and quality control.
Today glass fibers are produced in two different ways. The typically 125 m thin
fibers for telecommunication applications are produced - up to more than 1000 km
- from a preform. Light guiding fibers are drawn directly from molten glass.
Even with polymer fibers one differentiates between continuous methods,
spinning or extruding, and the drawing out of the preform.
In the preform method a cylinder is produced that already has the index profile
of core and cladding while having a much larger diameter. During the drawing
process, the diameter is reduced until the desired size has been reached
(Fig. 2.191, see e.g. [Wei98]).
Ideally, the index profile should be maintained during this process but at a
proportionally reduced scale. The length of the fiber per preform is determined as
follows:
Length of fiber = preform length (preform diameter/fiber diameter)2
2.8 Fiber and Cable Production 181
This method is applied generally for glass fibers. Automated processes are then
applied to make several 100 km of fiber out of each preform, as the following
example shows:
Length of glass fiber = 2 m preform (5 cm preform diameter/125m)2 = 320 km
It is easy to see that the large core diameter of common POF is not favorable
for this process since only a few km of fiber can be produced from each preform,
for example:
Length of POF = 1 m preform (5 cm preform diameter/1 mm)2 = 2.5 km
Drawing speeds for glass fibers today can attain 10 m/s; with POF about 0.2 to
0.5 m/s.
oven
take up drum
diameter control
unit
In addition to being able to draw the complete fiber out of the preform there is
also the possibility of producing the core as a polymer cylinder and then applying
the cladding by extrusion or enameling. The advantage here is that the polymeri-
zation of the core material can proceed under very much better controlled
conditions.
This process is used with PCS. A silica glass core is drawn out to 200 m - or
to other thicknesses as well - and is then surrounded by a polymer cladding,
typically 15 m thick. Understandably, the glass and the polymer have to be pro-
cessed using different procedures.
Other versions are discontinuous production in which polymerization first takes
place in the reactor and then the resulting block is extruded at low temperature, a
so-called batch extrusion.
182 2.8 Fiber and Cable Production
N2
vacuum-
pump
reactor
mixer
heater
monomer initiator cooler
polymerization
controller
cladding
polymer
POF with cladding
The monomer, the initior and the polymerization controller are first distilled by
a vacuum pump. After the polymerization is finished, nitrogen pushes the poly-
mers through the nozzle and the cladding is then immediately applied.
In addition, Mitsubishi has developed a method with which the polymerization,
described in [Nal04], can take place photochemically.
Figure 2.193 from [Hess04] shows such a method. The core and cladding
materials are pushed through a nozzle by a pump and a mixer. The cross-linking
then takes place with a UV lamp. This process could prove to be quite suitable,
especially for heat-resistant POF.
mixture
spinning
nozzle
cladding core take up
material material drum
UV-light for
crosslinking
POLYMER
filler heated
vessel
diameter
core control
extruder fiber
conveyor
pump
cladding
extruder
monomer, initiator,
polymerization controller
reactor
cladding extruder
heating
fiber
pump
extruder
take up drum
In order to guarantee the optimal functioning of graded index and multi-step index
fibers, the best index profile possible should be realized. The developmental goal
of the past few years has been to attain as much as possible with minimum effort
and to continuously produce GI fibers.
A number of different processes for the manufacture of graded profiles are
described in the technical literature:
Interfacial gel polymerization technique
Centrifuging
Photo-chemical reactions
Extrusion of many layers
In most of these techniques the principle is to initially create a preform of up to
50 mm diameter and then to subsequently draw this preform down to the desired
fiber size. Some of these methods are described below.
that has been typically heated to 80C. This results in a layer of gel and accele-
rates polymerization. The smaller molecule M1 can more easily diffuse into this
layer of gel so that the concentration of M2 increases more and more towards the
middle. The index profile is thus formed in accordance with the resulting concen-
tration gradient. For manufacturing a PMMA-GI-POF, [Koi92] proposes that
MMA (M1) be supplemented with monomers VB, VPAc, BzA, PhMA and BzMA.
The material that was finally used is BzA because its reactivity is comparable with
that of MMA. The 15 mm - 22 mm thick preform is then drawn at temperatures
between 190C and 280C to produce fibers ranging from of 0.2 mm - 1.5 mm in
diameter. Figure 2.196 illustrates the principle (see also [Ish95]).
[Koi95] describes this method in more detail. The PMMA tube is produced by
rotating a glass reactor at 3,000 min-1 at 70C that is partially filled with MMA.
The polymerization process for the core takes place at a speed of 50 min-1 and a
temperature of 95C and requires approximately 24 hours to complete. [Ish95]
describes the production of a PMMA GI-POF with DPS as dopants. For traditional
materials such as BB or BBP, one obtains fibers with a NA of 0.17 - 0.21, whereas
with DPS a NA of 0.29 is possible. The greater NA improves the bending charac-
teristics and makes the launching of light easier.
186 2.8 Fiber and Cable Production
The production of the preform is carried out in two steps. Once the monomer
mixture has been filled into a tube, the GI profile is formed at room temperature.
Then the temperature is increased so that polymerization takes place. Rotation
continues during this process. Then the fiber is drawn from this preform.
In this process the rotation speeds must be up to 50,000 min-1. Even for a pre-
form with 10 mm diameter the centrifugal acceleration (a = M2r) already equals
14,000 times the acceleration due to gravity. At the University of Eindhoven an
ultra centrifuge operating at 50,000 min-1 has been constructed for preforms up to
50 mm in diameter which produce a centrifugal acceleration of 70,000 g. In the
first trials, GI cylinders were produced from PTFPMA and MMA. The process for
forming the GI profile took 24 hours. This was followed by a period of 12 hours
during which the polymerization process was carried out at 60C to 80C. The
refractive index difference achieved was approximately 0.009. No research reports
have as yet been published on the production of fibers from such preforms.
50
40
30
20
10
0
0 100 200 300 400 500
distance to the fiber axis [m]
O2
porous preform
burner
ceramic or
SiCl4 rail graphite rod
gas mixing
controller
GeCl4
You can see quite well under a microscope that the fiber is built up of many
layers. Nevertheless, the index profile is almost ideally parabolic and does not
show any steps - see Fig. 2.201 acc. to [Park06a]. An attenuation spectrum of the
OM-Giga, 1 mm GI-POF (data provided by the distributor Fiberfin) is shown in
Fig. 2.202. At 650 nm the losses are below 200 dB/km.
2000
1000
500
200
wavelength [nm]
100
400 500 600 700 800 900
Fig. 2.202: Attenuation spectrum of a PMMA GI-POF made by Optimedia (Fiberfin)
dopant
diffusion
coextrusion
head GI-POF indexprofil
protective heated tube index
layer difference
extruder
coextru-
sion head
diameter
capstan
control
-100 -50 0 50 100
radius (m)
to the take up drum
The insert shows the final index profile with an approximately parabolic curve.
The manufacturer indicates the bandwidth-length product of the fiber as being
400 MHz km.
0.8 1.340
1.335
0.6 1.330
x [m]
1.325
0.4 0 20 40 60 80 100 120 140 160 180 200
0.2
bending radius [mm]
0.0
0 10 20 30 40 50 60
Fig. 2.204: Reduction of the bending losses due to a Semi-GI profile ([Sato05])
This method can also be employed for PMMA-GI fibers. The results for a
1 mm thick fiber are presented in [Aru05]. The attainable bending radius drops to
below 5 mm with an additional PVDF cladding (polyvinylidene fluoride,
n = 1.42). The bandwidth-length product of the fiber is 1,500 MHz 100 m and re-
mains quite constant up to 10 mm. It only drops under full launch and with a
5 mm bending radius to 500 MHz 100 m. The attenuation at a 90 bend is com-
pared to a conventional PMMA GI-POF in Fig. 2.205.
192 2.8 Fiber and Cable Production
2.0
bend loss [dB]
PVDF clad GI-POF
NA of the GI core region = 0.17
1.5
PMMA based GI-POF
NA of the GI core = 0.21
1.0
0.5
0.0
bend radius [mm]
-0.5
0 10 20 30 40 50 f
In addition to the extra cladding layer a so-called W-profile for GI fibers has
also been developed. Here the goal is to improve the attainable bandwidth.
Measurements on PMMA GI-POF with this W-profile and different index expo-
nents are presented in [Tak05b]. The W-profile is characterized by a very steep
index drop directly at the core-cladding interface. Figure 2.206 shows the index
curve.
5.0
3 dB bandwidth [GHz100 m]
3.0 W-shaped POF
2.0
1.0
0.5 GI-POF
0.2
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
profile index exponent g
PF-GI-POF with optimized index profiles are presented in [Ebi05]. Their band-
width attain that of MM-GOF and in the short-wave range even surpasses it (Table
2.32). The high bandwidth is attained through the approximately ideal index
coefficients of 2.05, i.e. in combination with the low chromatic dispersion of the
material.
Bandwidth
wavelength 650 nm 780 nm 850 nm
PF GI-POF 8.39 GHz 8.50 GHz 9.54 GHz
SiO2-GI-GOF 5.27 GHz 7.34 GHz 9.31 GHz
Figure 2.208 shows the best attenuation values over time for some of the fibers
listed above. PMMA fibers (SI and GI) reached their theoretically maximum
possibilities in the mid-80s. Since then, other index profiles (MSI, MC, DSI) have
also reached this order of magnitude (approx. 130 dB/km at 650 nm and 80 dB/km
at 570 nm). Any differences in measured values and specifications are more likely
to result from different measuring conditions than from differences in quality.
The PF fibers have been continually improved, at least as far as the laboratory
results are concerned. The best values were attained in 2003 with about 8 dB,
almost one magnitude still above the theoretical limits. In the past three years no
further progress has been made with the attenuation. On the other hand, there has
been some success in attaining a high launch-independent bandwidth with opti-
mized refractive index profiles and in reducing the bending sensitivity.
194 2.8 Fiber and Cable Production
attenuation [dB/km]
1,000
500
GI-PMMA
200
100
50
SI-PMMA at 650 nm
20 SI-PMMA at 570 nm
SI-d8 at 680 nm
PF-GI at 1.300 nm
10
d8-GI at 688 nm
5
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
year
This chapter discusses the structure and properties of various cable structures with
POF wires. Different applications place different demands on the mechanical
shielding of the polymer optical fiber. SI-POF (Step Index Polymer Optical Fiber)
is a promising medium for relatively short transmission distances of 100 m. Poly-
mer plastics such as polymethylmethacrylate (PMMA) or polycarbonate (PC) are
used as the primary core material for manufacturing these fibers. Fluorinated poly-
mers, silicone or fluorinated PMMA materials are used as cladding material with a
reduced refractive index of ncladding ~ 1.42 as compared with the core material
ncore > 1.48 (Fig. 2.209).
Due to the large refractive index difference, numerical apertures of up to 0.50
are attained. Various manufacturer versions of optical fibers are shown in
Fig. 2.210, in which glass or plastic are combined for the core and cladding mate-
rial. The relatively thin glass fibers are mechanically fragile and must therefore be
protected by a multilayer cable construction. The POF is so flexible that a simple
jacketing of the optical cladding suffices as a cable construction.
2.8 Fiber and Cable Production 195
d 0.98 mm r ncladding
D 1.00 mm ncore D
ncore 1.492 d
ncladding 1.416 n
NA 0.47
Glass fibers with polymer optical cladding represent an intermediate step. They
also have a relatively simple construction (two-layer plastic coating around the
optical cladding). The large core diameter allows only step-index profiles.
980/ 1000 m
0 mm 0.5 mm 1.0 mm
10,000
attenuation [dB/km] attenuation
5,000 minimum
PC
2,000
1,000
500
PMMA
200
100
50
450 500 550 600 650 700 750 800 850
wavelength [nm]
Fig. 2.211: Attenuation spectrum of different POF made from PMMA or PC
Polymer optical fibers that are flexible and break-resistant can be produced
with a relatively large diameter (up to 1.5 mm or even more) and are thus easy to
handle and to install. The large core diameters in combination with the numerical
aperture make simple connection fittings and equipment possible with low de-
mands on precision.
flexibility
length of lay
Fig. 2.212: Schematic diagram of the relationship between the pitch length and the flexi-
bility of the stranding construction
outer sheath
metal band
inner coating
cladding
fiber core
2.2 mm 2.3 mm
Fig. 2.214: SI-POF duplex cable in a round cable and flat cable form
With these duplex cable constructions, care must be taken to ensure that the
strain-relief elements in the plugs or on the connectors are included in processing.
This is necessary because the temperature influence on the SI-POF wires is con-
structed in such a way that optimum temperature characteristics are ensured in the
temperature range from -40C through +80C.
10 cables group
26 mm
Fig. 2.215: SI-POF ribbon cable with traction and support elements
2.8 Fiber and Cable Production 199
The cross-sections of two POF ribbon cables from [Boc04] are shown in
Fig. 2.216. The individual fibers have each been extruded in a joint acrylic
cladding.
Fig. 2.216: Ribbon with four 500 m SI-POF (above) und eight 120 m/500 m GI-POF
(below, [Boc04])
For the OVAL project (see Chap. 6) of the POF-AC Nrnberg Nexans had pro-
duced 8-strand ribbon cables made of SI- and GI-POF each with a 500 m dia-
meter. The cross-section of a prototype with PMMA-GI-POF (Optimedia) is
shown in Fig. 2.217.
The spacing between the individual fibers deviates only slightly from 500 m.
Only in a vertical position great deviations do arise which can easily be avoided
by better guiding of the individual fibers in the extrusion tool.
In order to investigate the influence of the ribbon cable production on the
optical parameters, the spectral attenuation and the bandwidth were determined on
the SI-POF ribbon cables. The results are shown in Fig. 2.218 and 2.219.
200 2.8 Fiber and Cable Production
1000
attenuation [dB/km]
800
600
wavelength [nm]
100
450 500 550 600 650 700 750 800
Fig. 2.218: Single fiber attenuation in the ribbon cable
The attenuations of the 8 fibers agreed within the usual measurement error of
0.5 dB. There were also no significant deviations in the frequency response in
Fig. 2.219.
+5
rel. level [dB]
0
-5
-10
-15
-20
-25
-30
frequency [MHz]
-35
1 3 10 30 100 300 1000
Fig. 2.219: Frequency response of the fibers in the ribbon cable
36
FWHMeff [] high NA-fiber
NAlaunch = 0.10
32
28
24 fiber
ribbon cable
20 annealed
aged
16
lPOF [m]
12
0.1 0.3 1.0 3.0 10.0 30.0 100.0
isolation
2 ... 3 mm 3 ... 4 mm 7 ... 9 mm 4 ... 5 mm
Fig. 2.221: New design for POF with CMT as electrical conductor
copper wire
POF
980/1000 m
foil
support element
POF
980/ strain element
1000 m
copper wire
inner coating
inner coating
outer sheath
6.5 mm outer sheath
7.5 mm
In these cases, insulated copper wires and POF wires are processed either into a
group of four or as stranded layers with several stranding elements. The copper
wires are used with diameters of 0.5 mm to 1.5 mm. Thicker copper wires are pro-
cessed as braided wires, because the flexibility of the cable usually does not meet
customers requirements.
facturing process or for transport purposes or for winding the cables or lines on
production-machine reels or shipping reels or when sold in rings. The individual
SI-POF elements are twisted in a screw-like fashion around an imaginary center-
line. Twisting is necessary in order for the manufactured products to be flexible
and portable.
The advantage of twisting is that the stranding element is stretched and com-
pressed alternatively on the inner and outer side of a curved section (Fig. 2.224). If
the section in which a SI-POF stranding element is wrapped 360 around a twist
axis that is considerably smaller than the curved section, the strain and pressure in
a stranded construction are constant and it is possible to bend this SI-POF cable
without deformation.
Fig. 2.224: Comparison of cable constructions with short or long lay lengths in terms of the
bending characteristics
1
2 3 5
s 4
n1 d n2
DA
1. rotor s: pitch length d: diameter of the
2. stranding elements n1: rotational speed of stranding unit
3. stranding unit the stranding basket n2: rotational direction
4. capstan gear DA : diameter of the and speed of the
5. stranding axis stranding basket capstan gear
D A S n2 v m 1000
s [mm] s
n1 n1
ding construction is the result and increases the pitch length manufactured. The
geometric assignment is easy to see in the enclosed illustration (Fig. 2.226); the
manufacturing pitch length SH is calculated from it.
d
3
1 1: capstan gear
2: fiber loop
3: POF
2
DA
Fig. 2.226: Diagram for explaining the concept of 'manufacturing pitch length'
Lay Direction
The rotational direction of the stranding basket determines the lay direction. The
following distinction is made depending on the sense of direction of the helix:
Z-lay means a right-handed thread
S-lay means a left-handed thread (Fig. 2.227)
S Z S Z S
Fig. 2.228: Explanation of the lay direction schematically
Multiplication Factor
The helical SI-POF stranding element (Fig. 2.228) is longer in the stranded unit.
The stranding method always leads to an increase in material consumption. The
ratio of the laid length L of the SI-POF stranding element to the lay length of the
stranded unit results in the well-known multiplication factor f = L/s. The multi-
plication factor f is determined from the pitch length and the average diameter Dm
in the stranding layer.
The multiplication factor can be easily derived from the triangle shown in
Fig. 2.229.
2
L ( S Dm )2 s2 S Dm
L ( S Dm )2 s2 and f 1
s s s
with L: Laid length L = s/cos Z
f: Multiplication factor
Dm: Average diameter of the stranded layer
s: Pitch length of each stranded layer
For relatively large pitch lengths (Dm s), the calculation can be simplified as
follows:
f | 1 S Dm / s 2 /2
2.8 Fiber and Cable Production 207
L SDm
s
Dm Z
s D
Number of Strands
To characterize the bending properties of an SI-POF stranding element v, the
number of strands is formed from the quotient of the pitch length and the average
diameter Dm (v = s/Dm).
s: Pitch length of each stranded layer
Dm: average diameter of this stranded layer
v: Number of strands
Production developments in stranded cable constructions or SI-POF cable con-
structions have lead to the number of strands being v > 8. By using the number of
strands v, the multiplication factor f can be easily calculated.
2
S S2 v 2
f 1 | 1 S/2v
v v
Layer Structure
Standard SI-POF elements have a simple geometric shape but have an exact dia-
meter. This makes it easy to calculate SI-POF cables or SI-POF lines. An SI-POF
cable in its classic form, i.e. with a core element, has the same diameter as the SI-
POF element; it can be constructed in a circular fashion with 6 SI-POF elements in
the same layer. The cladding lines are in contact with each other. Two different
core layers have been adopted schematically in Fig. 2.230. The other layers are
calculated and shown. In Table 2.33 and Table 2.34, the number of elements and
the diameters have been compiled for a general case and for the case with
d = 2.3 mm respectively, whereby the variables have the following meaning:
n: Layer number
z: Number of elements per position
6z: Total number of the elements to the layer n
d: Diameter of the cable unit
Dm: average diameter of the unit
D: Diameter of the layer
208 2.8 Fiber and Cable Production
d = 2.3mm
Dm1
D1
Dm2 Dm2
D2 D2
Dm3 Dm3
D3 D3
Cable Materials
The specification profile for SI-POF cable or SI-POF lines in various fields of
applications such as in industry, in office environments or in the automotive field
place the highest demands on the material components.
Thermoplastic materials (polymers) are preferred that have been mounted to the
cable using an extrusion process. Excellent mechanical properties are needed so
that the values listed below are ensured when SI-POF cable or SI-POF lines are
installed.
2.8 Fiber and Cable Production 209
Abrasion
Repeating bending characteristics
Torsion
Acceleration
Hammer blow
Small bending radii
Especially in the automotive field, the material must be highly resistant to the
following properties:
Resistance to oil
Cooling lubricant resistance
Steam
Hot gases
The demand for materials that are temperature resistant comes from users.
These customers are in the automotive field, in industry or in the cable-installation
field for buildings. Special halogen-free material properties are desired in order to
provide on-site safety to customers and consumers alike.
Todays selection of modern plastic insulation and cladding mixtures, which in
part can be improved through various methods of crosslinking, should and must
protect the SI-POF cables or SI-POF lines in all types of applications.
In case of an accident, special plastic optical fiber cables are to have emergency
running properties. SI-POF hybrid cable constructions ensure this reliability to a
very high degree.
The mechanical properties of thermoplastic materials such as
Hardness
Density
Tensile strength
Elongation at break
Tensile stress value
Compression strain
Impact resistance
Electrical properties
can be found in the relevant data specifications of the standardized norms or the
data specifications of the chemical industry. Preferred plastic materials are:
Polyethylene
Polypropylene
Polyurethane
Cross-linked thermoplastics
The properties that have been improved by cross-linking are those of thermal
resistance and higher mechanical strength. In addition, the resistance to solvents
has also been increased, which can be seen by the fact that less swelling and
cracking occur for polymers with residual tensile stress.
210 2.8 Fiber and Cable Production
The essential physical properties of some of the important materials are listed
in section 3.3.6.
A very good alternative is a combination of plastic and metal, for example, with
the corrugated micro tube. Metal in the most varied constructions, whether as a
steel alloy, in aluminum or in copper keeps the SI-POF in an expanded tempera-
ture range protected against mechanical and thermal strain.
Possible applications for CMT cables will be discussed later in Chapter 8.1.1.7.
The manufacturing process for corrugated tubes is described below.
The UNIWEMA is used to weld copper, aluminum and steel strips or steel
alloys or alternative materials. The machine creates smooth and corrugated metal
tubing in an economical manner.
The tube welding process is continuous and fast. All weldable metals such as
copper, aluminum, steel and their alloys can be processed. The process can be
used for manufacturing small metal tubes for core diameters ranging from 1 mm
to 500 mm. Strip thickness of 0.05 mm to 4.0 mm are welded with a laser using
the WIG process. Neither burrs nor bulges are produced at the welding seam
(Fig. 2.233).
Due to the concentrated thermal effect of the welding source, the welding zone
is limited on the metal edges. The heat is quickly dissipated over the tube. Since
the welding zone is covered by a protective gas shield, the formation of an oxide
layer is prevented.
Laser Welding
The laser beam is monochromatic and coherent and can be easily focused. As a
result, a high power density can be achieved at the processing point - the V-seam
between the strip edges (Fig. 2.235).
2.8 Fiber and Cable Production 213
By applying the auxiliary gases argon and/or helium in such a way that the
beam power is absorbed in the capillaries, the coupling properties of the plasma
can be controlled. The actual welding joint is produced by the melt converging
behind the capillaries (Fig. 2.236).
metal vapour
laser induced
plasma welding zone
(fluid)
welding zone
(fluid) vapour
welding zone
(plasma) (solid)
welding zone
channel
(solid)
ND:YAG-laser laser
data
beam source control controller
acquisation
device
process computer
laser fiber
quotient
pyrometer
partially
transparent optic
O reflects
Nd:YAG
radiation
laser optics O transmits heat
radiation
beam-material
interaction zone detected heat
radiation
workpiece
feed direction
In addition to classic optical fibers which consist of a core and cladding there are
also microstructured fibers in which the wave guiding does not rely on a refractive
index profile, but on holes along the entire length of the fiber. Normally, wave
guiding in optical fibers is based on the effect of total reflection in the general
sense of the term. The core consists of a material with a higher refractive index
than the surrounding cladding material. In this fiber configuration special field dis-
tributions, so-called modes or eigenmodes, can be guided within the fiber. These
modes experience an effective refractive index of the fiber, which lies between the
maximum refractive index of the core and that of the cladding material.
In 1996, J. Knight et. al. demonstrated a new kind of optical fiber, the wave
guide characteristics of which were no longer based on a rotation-symmetrical re-
fractive index. This created a variety of completely new possibilities and novel
functions ([Kni96] and [Kni97]). These fibers now only consist of one material,
usually silica glass, and have a structure of the cross-section with air holes. The
holes in this structure are as a rule considerably smaller than the wavelength of
light so that they do not act like objects on which light is reflected or scattered.
Instead they change the refraction characteristics of the material.
The material is changed in such a way that it acquires new kinds of charac-
teristics. Relatively simple and specific characteristics can be created with these
fibers, e.g. for dispersion, dispersion slope, modal field radius and others.
For some years now microstructured fibers have also been made of polymer.
These fibers with low temperature processes can be produced on the basis of the
low melting point of polymers and other characteristics, thus resulting in possibi-
lities for new kinds of fiber geometries and also potentially new applications.
In the following section we would like to deal with the fundamental wave
guiding mechanisms. The different types of fibers and their specific characteristics
will be introduced and the methods for producing these different types of fibers
will then be shown. We would particularly like to take a close look at the diffe-
rences between microstructured fibers made of glass and polymers. Applications
which are possible with these fibers and are presently the subject of research will
then be introduced. Some of these applications can even be obtained commercially
now. Finally, the present state of development will be discussed and we will ven-
ture a prognosis as to where the limits for such fibers may lie in the future.
There are two fundamental mechanisms which exercise this influence: holes
either act as a kind of doping by changing the effective refractive index of the
material in average ([Gho99]) or they are put into a regular, grid-shaped arrange-
ment so that they act like a kind of meta-material ([Cre99]. Other materials which
have a greatly differing refractive index from that of the core material can also be
used). Such fibers can exhibit effects with a great degree of wavelength depen-
dence since such arrangements have similar characteristics as e.g. Bragg gratings,
in which the light at certain wavelengths can be constructive or destructive over-
lapped. The two-dimensional pendant to such a Bragg grating are the Bragg fibers
in which concentric areas with greatly differing refractive indices alternate at
regular intervals ([Yeh78]). Constructive overlapping waves can come about at
certain wavelengths thus resulting in wave guiding. At other wavelengths light is
not guided. One can therefore surmise that such fibers are capable of having
strong wavelength-dependent characteristics.
A new kind of wave guiding occurs in such fibers with regular structures. This
wave guiding is possible in cores made of air as opposed to those fibers based on
total internal reflection. For wave guiding with total internal reflection it is essen-
tial that the core material has an effective refractive index which is higher than
that of the cladding. This is not necessary with fibers having a photonic band
gap. Because of the regular structure within the fiber band structures are formed
analogous to electrical semiconductors in which certain energy states of light
waves are allowed and others are rejected resulting in light waves which can
remain within the material and others which cannot. When there are light waves
which have permissible energy states within the core area, but not in the cladding,
then the light must stay in the core and is guided through this band gap since they
cannot exit into the cladding.
being similar to normal step index glass fibers, whereby the fiber parameter de-
pends on the wavelength of the light ([Mor03a ] and [Mor05b]). The reason for
this is that the influence of the holes varies greatly depending on the wavelength
of the light which also depends on the relationship between the hole diameter and
the wavelength and whether light can resolve the holes.
nSiO2
1.4
1.2
1.1
n air
1.0
PBGF-mode
0 1 2 3 4 5 6 7
normalized frequency /O
The form of the energy bands, i.e. the energy areas, which correspond to the
permissible energy states is greatly dependent on the arrangement of the indivi-
dual holes. Even small deviations can lead to great changes in the energy bands so
that with this kind of fiber only slight tolerances are allowed in the arrangement of
the holes. Nevertheless, these fibers permit greater possibilities for structuring
([Arg06]). As a consequence, propagation characteristics such as dispersion, dis-
persion slope, effective area, etc. can have relatively large dimensions. Especially
for very narrow-band applications, e.g. sharp-edged filters, fibers with photonic
band gaps can be employed quite well. This is also true for high-performance
applications in which the linear characteristics of the hole core are used ([Lim03],
[Mat05b] and [Nie06]).
Fig. 2.240: Air-hole - MPOF with 220 m outer diameter/5 m hole distance, [Eij03a]
2.9 Microstructured Fibers 219
The rings can be produced in a variety of different ways. Refractive index pro-
files can be produced which have higher or lower refractive indices with specific
radii. Microstructured fibers, however, are also possible where the rings with dif-
ferent refractive indices are realized by hole structures. In this case rings with
holes are arranged at regular distances from the fiber axis which, because of the
effective index of this layer, acts like a layer with reduced refractive index.
Bragg fibers behave similarly to fibers with photonic band gap. They are also
based on the exact arrangement of the holes or the layers with different refractive
indices respectively. If the geometry is followed exactly very sharp-edged filters
can be produced or fibers which are very selective in regard to the wavelength.
ring-shaped hole structures are arranged around the core in order to reduce the
bending sensitivity of the fibers ([Guan04] and [Nak03b]). The outer structure acts
like an additional step in the refractive index profile which should hold part of the
output emitted in the bend in the cladding area. This measure is supposed to
increase the wave guidance without having to make compromises concerning the
propagation characteristics of the fiber.
Glass melting at low temperatures can also be extruded, whereby the glass is
either melted or liquefied. The ensuing viscous fluid can then be pressed through
specifically arranged nozzles which have the structure of the desired preform. This
preform can then be used to immediately draw the fiber or to make a preform.
This method of making preforms in effect allows the production of as many hole
geometries as one likes. In principle, round holes and any kind of arrangement can
be produced in this way. However, this production method is limited to glass with
a low melting point. Consequently, silica glass, for example, cannot be processed.
The production engineering of microstructured fibers has improved tremen-
dously in the past few years. Whereas the first fibers still had attenuations of seve-
ral 100 dB/km, today fibers based on an effective index with attenuations per unit
length can be produced below 0.3 dB/km at a wavelength of 1.55 m ([Taj03]).
Photonic band gap fibers permit attenuations per unit length up to 13 dB/km
([Smi03]).
Microstructured polymer fibers can also be extruded and then drawn into fibers.
The same limitations regarding geometry and production tolerances are valid for
them as for glass fibers.
Researchers at the University of Sydney ([Bar04c] and [Lar01b]) have
developed a particular kind of preform production in which a massive cylinder
made of polymer is structured using drills with different diameters. At present,
preforms up 65 mm in length can be structured with this method, otherwise the
drills would be too long. As many geometries as one may wish can be produced in
which both the arrangement and the hole diameter can be freely chosen. Present-
day production processes have hole diameters between 1 mm and 10 mm with
minimum spacing in between of about 100 m which then shrink to their original
size through drawing.
New kinds of process techniques can even produce elliptical holes which give
the fiber an intrinsic double refraction. Preforms can either be poured into molds
or around capillary tubes and then drawn into fibers ([Zha06]).
Other materials can be introduced into the fiber in addition to the holes. Fibers
with metal wires for the poling of the material have been demonstrated as well as
fibers with liquids in the capillary for controlling the propagation characteristics
and doping materials for changing the optical and electrical characteristics
([Cox03b] and [Cox06]).
After the first MPOF was introduced at the end of 2001 ([Lar01b]), the tech-
nology has continued to develop at an amazing pace. The fibers introduced back
then still had an attenuation of 30,000 dB/km. In the course of time the individual
process parameters have been continuously improved so that the attenuation could
be steadily reduced. The process parameters optimized include conditions when
drilling the preforms, rinsing and cleaning steps as well as drawing parameters.
The best microstructured polymer fibers today have an attenuation of 200 dB/km
and are thus not very far away from conventional polymer fibers which have an
attenuation of about 120 dB/km at a wavelength of 650 nm.
2.9 Microstructured Fibers 223
100
achieved attenuation [dB/m]
Sept. 2001
10
April 2005
0.1
0 5 10 15 20 25 30 35 40 45
months for the first publications
The direct cutting of the fiber with conventional cutting pliers can destroy the
fine structures because of these lateral forces.
224 2.9 Microstructured Fibers
Other processing methods such as hot plate or subsequent polishing have also
been investigated, but did not deliver any good results. The hot plate technique
leads to inclusions at the end surfaces so that the original geometry can no longer
be recognized. On the other hand, polishing leads to the deposition of rubbed off
shavings and their removal into the holes. A reproducible coupling is therefore not
possible since the influence of these inclusions or that of the deposited foreign
matter in the structures holes is not controllable.
Better processing characteristics are shown by those MPOF which are surroun-
ded by another, so-called buffer layer made of hard polyester. This layer absorbs a
large part of the mechanical forces when cutting and prevents the breaking of the
fine webs within the structure. Since such fibers consist almost exclusively of
polymer they can almost be worked on like polymer fibers. Figure 2.248 shows
the end surface of such an embedded fiber with a buffer layer. You can see that
the fiber is not embedded centrically which leads in practical use to a lateral mis-
alignment of the plugs and thus to plug losses and power redistribution. In the
future you can expect, however, that the dimensions of the fibers will become
greater and that the fibers can be better centered with new drawing techniques.
Fig. 2.248: End face of an embedded fiber with buffer layer; 100-fold magnification
2.9 Microstructured Fibers 225
No practical solution exists yet which can provide for good reproducibility and
a high degree of reliability. Processing methods still have to be found for both
practical and laboratory use which can meet the necessary requirements. In the
case of termination in the field the end faces must allow acceptable losses; in the
laboratory, preparation with high reproducibility is necessary. Both kinds of pre-
paration still have to be developed.
only one mode for V < 2.405, the first zero of the Bessel function of zeroth order.
If the wavelength selected is large enough then V will become small enough at
some point so that the fiber becomes singlemode. In microstructured fibers the
fiber parameter is not simply anti-proportional to the wavelength since the holes in
the cladding area act differently with large wavelengths than with small ones
leading to a wavelength-dependent numerical aperture so that fibers can be pro-
duced which are singlemode for all wavelengths ([Bir97], [Mor03b] and [Zag04]).
2.9.3.3 Birefringence
Since microstructured fibers are not rotation-symmetrical such as conventional fi-
bers with a refractive index profile, for example, they tend to be birefringent.
Typical hexagonal structures do not exhibit any birefringence. However, when this
symmetry is disrupted, e.g. through production tolerances, then these fibers are
birefringent.
This effect is used positively in some fibers, whereby the high birefringence
causes the fibers to retain their polarization ([Ort04]). In the case of very great
differences between the propagation constants of both polarizations they can then
only very weakly interact with each other and exchange power. When only one
polarization is launched into the fiber, then the power in this polarization is
retained and is propagated in this way to the end of the fiber.
gence occurs which can be created in a controlled and thermally stable manner, or
the holes are elliptical and not round which contributes to the birefringence
([Issa04b]). It is more difficult to control this kind of birefringence, but it does
allow complete freedom of fiber design because the arrangement of the holes and
their size can be freely chosen.
This technique can also be used for controlling the form of the mode field in
order to adapt it to other types of fibers and thus minimize coupling losses at the
connector. For example, Furukawa introduced such fibers at the ECOC in 2004
([Guan04]) the mode fields of which are adapted to standard singlemode fibers.
2.9.3.6 Filters
Microstructured fibers can show very strong wavelength-dependent effects. As
described above, the dispersion can be adapted to a wide area, but other wave-
length-dependent characteristics can be specifically designed, e.g. group velocity
or even the attenuation per unit length of the fiber.
Fibers with an effective refractive index permit the relatively simple adaptation
of the group velocity with which one can generate all-pass filters with specific
phase responses.
Fibers based on a photonic band gap can have very sharply delimited wave-
length ranges with which light is guided. Thus, filters with specific amplitude
response and sharp edges can be produced ([Vill03], [Kim05c], [Kim06d] and
[Sai05]).
Fig. 2.250: Double core-MPOF with 9.6 m spacing between the cores ([Eij03b])
2.9.3.9 Imaging
As we have seen above, microstructured fibers can be produced with more than
one core for parallel data transmission. If you continue to increase the number of
cores, you can use the same method to produce image guides in which every indi-
vidual core transmits a part of the image (a pixel). As mentioned above, the
arrangement of the holes stays the same and the cores along the fiber are retained.
Each individual pixel reaches the end of the fiber in its definite position so that the
image is retained ([Eij04c]).
230 2.9 Microstructured Fibers
Fig. 2.253: Cross section of a graded index profile multimode polymer fiber (GI-MPOF)
with 135 m core- and 520 m outer diameter ([Eij04d]) and of a MPOF
according to [Lwin06]
Figure 2.253 shows a multimode fiber in which the effective refractive index
continuously decreases with increasing distance to the fiber axis. If you take an
average of the entire circumference of the refractive index, then you have a para-
bolic refractive index profile in the radius direction. Measurements have shown
that these fibers have a similar propagation behavior as a conventional multimode
fiber. However, the differences lie in the detail. If you stimulate the GI-MPOF
with a small spot, for example, the fiber behaves differently, depending on
whether or not the light hits a hole or the core material; something that cannot
happen in conventional fibers. For this reason greater research and development in
measurement techniques and characterization are necessary before the GI-MPOF
is widely used in commercial applications.
3. Passive Components for Optical Fibers
Fig. 3.1: Copper data cable with separately shielded twisted pairs
Glass fibers have a core diameter between 10 m and 200 m. This requires
precise guides that are provided by metal, ceramics or high-grade plastics. Fur-
thermore, glass fibers cannot simply be cut. The face must either be precisely
broken by carving with a diamond blade or else polished after cutting.
Further advantages for POF result from the material itself. The surface of plas-
tics can be smoothed by both cutting and simple polishing. In addition, a thermal
smoothing of the surface is also possible for PMMA.
234 3.1 Connection Technology for Optical Fibers
In recent years, a wide range of connector types has been developed specifically
for various applications. These fall into the following categories:
Special plug-in connections for POF, e.g. V-pin, DNP
Plug-in connectors developed for glass fibers have been adapted for POF, e.g.
FSMA, ST
LAN plug-in connectors which have identical dimensions for both copper and
optical fibers (SC-RJ, RCC45)
Connectors for special standards (D2B, F07)
Connector systems without plugs (optical clamps)
Hybrid connector, generally a combination of copper lines and POF (e.g.
MOST)
Several connectors are available in metal or plastic constructions depending on
the requirements for mechanical stability. Most systems can still be obtained in
simplex or duplex versions.
At this point we would now like to define the term connector attenuation.
Strictly speaking, a connector does not have any defined losses, only a fiber-to-
fiber connection has losses. The light losses are caused by:
inexact alignment of the fibers to each other, whereby the parameters of the
fiber, the connector and the coupling could be responsible,
the fiber parameters not being adapted to each other, e.g. different numerical
apertures and
direct losses at the fiber end face through reflection, scattering and absorption
(see Fig. 3.2).
You can thus see immediately the great dilemma of the connector manufactu-
rers who are not at all responsible for an essential part of the losses. In most cases
you cannot even be sure that the connectors and couplers all come from the same
supplier. The greatest contradiction, however, lies in the specifications of the POF
itself. As will be shown later on, the IEC has approved wide ranges for the fiber
parameters, but still insists on including plug-in connections in the power rating
calculation. That would be like approving on the one hand surge impedance of 100
30 for a gigabit Ethernet on symmetric lines and demanding on the other a
reflection attenuation of at least 20 dB and a connector attenuation of < 0.1 dB.
For this reason attempts are being made to reduce the characterization of a con-
nector to the effects at the connector end face. This involves precision in posi-
tioning the fiber in the connector or in the coupler and the quality of the surface
treatment.
Another thing we have to mention is that the connector attenuation with all
multimode fibers depends on the mode distribution and thus on the launch con-
ditions and the measuring fiber length. In this respect there are no specified direc-
tives.
3.1 Connection Technology for Optical Fibers 235
3. mechanical tolerances
of fiber and connector
The choice of the end face preparation is of great importance (see [Moll00]). The
following procedures have proven themselves:
Cutting and polishing: The POF is roughly cut at the connector face and
subsequently sanded down to the face with sand paper. Using fine-grained
polishing paper, the surface quality can be further improved. For normal de-
mands it usually suffices treating the surface with 3 m polishing paper after
the first cut. If the connector attenuation should be minimal and for measuring
purposes we recommend polishing it with fiber grades of polishing paper,
first 10 m, then 3 m and finally 0.3 m.
Hot-plate: The POF is cut before the connector facet at a defined projection
length. Afterwards, the connector is pressed within a guide against a hot mir-
ror. The connector has a ring-shaped bulged opening on the face in which the
projected material is pressed. After the mirror has cooled down, the connector
is withdrawn.
236 3.1 Connection Technology for Optical Fibers
Cutting: Using a thin blade (usually a razor blade), the POF is cut in a guide
at a perpendicular angle. This method is often used for connections without a
special connector. The blades must not be used more than once at one point.
Laser cutting: The POF is cut vertically with a laser (for example, CO2
laser). This procedure is only feasible for ready-made cables.
Microtome cut: Extremely thin slices of a sample can be cut off with a
microtome, for example, for microscopy. If you cut off several thin slices of a
POF the remaining surface is then extremely smooth and even. This pro-
cedure is very expensive if you use a diamond blade and is consequently only
employed as a reference method or for special measurements.
POF-Press-Cut (PPC): [Moll00] and [Fei00] contain descriptions of how
cracking can be avoided in PMMAs when a suitable pressure is being applied
to perform the cutting. Thus surfaces are possible whose losses are near the
theoretical limit (Fresnel reflection), as shown in Fig. 3.3.
The following diagram shows the losses of a plug-in connection which are
counted back to one surface. Geometric influences have been eliminated through
optimized coupling.
0.20
0.10
0.00
conventional Hot plate POF press sanding/ microtom cut with
cot cut polishing diamond blade
finishing method
Fig. 3.3: Comparison of the connector losses for various procedures according to [Moll00]
The theoretical limit is 0.17 dB by Fresnel reflection. Simple cut POFs have a
loss of almost seven tenths dB. By means of PPC and multistep polishing almost
ideal end faces can be attained. The microtome cut here lies at only 0.07 dB above
the Fresnel limit. However, it is the most costly procedure and used only for
measurement purposes.
3.1 Connection Technology for Optical Fibers 237
2. removed coating
3. connector mounted
4. POF cutted
The advantages of this procedure lie in the very limited number of tools
needed:
a scissors or pliers
a stripper for the fiber cladding
polishing fixture
regular abrasive paper
polishing paper, in several grades if necessary
With a certain degree of practice you can make a usable plug in 30 seconds and
a very good one in two minutes.
238 3.1 Connection Technology for Optical Fibers
2. removed coating
3. connector mounted
5. surface pressed
to a hot plate
6. connector is ready
after cooling
Fig. 3.6: Irregularly broken POF due to the cutting process ([Fei01a])
Fig. 3.7: Cutted and hot plated POF with cracks in fluorescence light ([Fei01a])
Proceeding from these results, the POF press-cut procedure was developed at
the POF-AC in which the fiber is pushed to the right and left of the blade when
being cut.
blade
Fradial Fradial
compressed area
higher resistance
uncompressed area
lower resistance
Fradial Fradial
The crack is thus always forced to propagate evenly in front of the blade. The
result is a POF which has a good uniform cut over its entire cross-section (princi-
ple is in Fig. 3.8).
The cut with the PPC experimental set-up is compared with a conventional cut
in Fig. 3.9. In the meantime two manufacturers have turned this procedure into
commercial products. Figure 3.10 shows the cutting tool from the Rennsteig com-
pany based on this principle.
The fiber itself as well as a fiber already secured in a connector can be worked
on using this method. You can prevent the connector itself from being milled with
guides and limit stops. The advantage of this procedure is the high reproducibility.
An example of such a device is shown in Fig. 3.11.
In the first edition of this book a good half dozen families of connectors were des-
cribed. In the meantime, additional versions have come along. It is not possible to
review all products here. Many of the manufacturers have made data available in
an overview of the industry under www.pofatlas.de.
A number of connector systems for polymer optical fibers will be shown in the
following pictures. Each system has been optimized for specific applications.
There is no universal connector type as is the case with glass fibers or copper
cables.
is a duplex connector version available. These plugs and the active components
available for them are described in [HP06] and [HP03]. The emitter and receiver
versions HFBR-0507 and HFBR-15X7/25X7 are designed for data rates up to
155 Mbit/s for all these connector systems. The simplex and duplex connectors,
the inline coupling and an active component are illustrated in Figs. 3.12 - 3.14.
Fig. 3.12: V-pin-crimp connector system and coupling (above: crimped simplex connector
with coupling, below left: duplex connector, below right: connector with latch)
Later on the system was expanded with a crimpless version which included a
plastic part which locks in place when snap shut and fixes the fiber in position. If
these two parts are connected, then you automatically get a duplex connector
(Fig. 3.13). The fiber spacing naturally corresponds exactly to the spacing between
the two connected HFBR components. Moreover, there is a connector with a latch
so that the extraction force can be increased.
Fig. 3.14: V-Pin connector with latch, connector with an active component
Fig. 3.15: Fiberfin- connector system (presented at the International POF Conference2004)
The POF is cut off in front of the connector when being fixed in position and
cut/polished with the polishing fixture (see next section).
An identical connector can be used for 200-230 m PCS which, however, has
to be crimped twice. First the connector is attached to the 500 m primary coating
and then crimped onto the 2.2 mm outer sheath (Fig. 3.16).
The coupling for PCS only differs from that of POF through the metal piece
which reduces the tolerances. The active components can be used in part alter-
natively for POF or PCS.
244 3.1 Connection Technology for Optical Fibers
Figure 3.19 shows a front-view of the FSMA connector versions for prepara-
tion by polishing and the hot plate procedure. In the hot plate version a small slot,
which can also be slanted, is located around the fiber core hole. The core material
sticking out is pushed into this open space when heated (see Fig. 3.5). As we shall
see later, this causes a little additional attenuation.
3.1 Connection Technology for Optical Fibers 245
Fig. 3.19: FSMA-connector for polishing (left) / hotplate (right), front view
Fig. 3.20: FSMA-connector in plastic variants and FSMA-connector with kink protection
The duplex version of this connector is the F07 which is also used as a standard
connector in the ATM Forum (PN connector). There are different options with or
without latch. With the F05 and F07 the end face is almost always treated with hot
plate. Occasionally there are also versions for cutting and polishing.
SMI Connector
The SMI connector is a duplex connector for SI and GI-POF. It is already inclu-
ded in various committees and standards, not least in IEEE 1394. Hot plate as well
as cutting/polishing can be used for the surface preparation. A number of manu-
facturers offer transceivers.
SC-RJ
The SC-RJ connector for glass fiber applications is well known and is especially
popular in local networks. In any event it would not take much for many installers
to get used to it. The SC connector is a duplex connector with an interlocking
device. It is also available as a simplex version so that two simplex connectors can
be plugged into a duplex coupling. A connector, a coupling and a transceiver from
Reichle & De-Massari ([Chr05]) are illustrated in Fig. 3.33.
For glass fibers here are connector versions for singlemode fibers, green for
angled polished, and blue as physical contact, and beige for multimode fibers. The
connector is described in standard IEC 60873-14, parts 1 to 3. For POF the color
combination suggested is black/white. A simplex connector and another transcei-
ver prototype from [Dre05] are shown in Fig. 3.34.
EM-RJ
The EM-RJ introduced in [Neh06a] is a duplex connector which can be cut and
fitted on location with connectors and has housing dimensions of the RJ45 (acc. to
EN 60603-7). It is available in the safety classes IP20 to IP67 for home and indus-
trial applications. In addition to the connectors boxes, couplings and distribution
panels are also available. The ferrules are made of metal, but in the future should
also be available in plastic. It is also possible to insert 8 metal contacts into the
connector so that it can be used as a hybrid Ethernet connector. The connector and
a Fast Ethernet transceiver (Euromicron) are shown in Fig. 3.35.
The manufacturer FCI came up with another concept in which the connector
consists of a single piece of plastic. Gear teeth which can be directly crimped onto
the primary coating are located in the back. The end face is treated by hot plate.
Figure 3.37 shows some of these connectors.
A completely new connector system was developed by Tyco-AMP for the D2B
bus. A kind of transparent plastic cup filled with an index matching gel is placed
on the fiber. In this way surfaces which are not ideal can be compensated for.
Further details concerning the system are not publicly available.
The connector developed for MOST (see Chap. 8.1.1) has attained considerably
greater importance. It consists of a metal or plastic ferrule which is crimped or
spliced with a laser on the primary coating. The connector end face is milled with
a fast rotating saw blade. Two versions of this ferrule and a cutaway view are
shown in Fig. 3.38.
The ferrules and electrical contacts in different combinations are built into
hybrid plug-in connectors. There are a number of different versions for straight
and angled connectors, in-line couplings and sockets on the control units
(Fig. 3.39).
252 3.1 Connection Technology for Optical Fibers
ferrule
A number of different tools are available for assembling the various connector
types. Figure 3.42 shows the tool required for assembling a crimpless V-pin con-
nector (see Fig. 3.12).
A wire-end stripper is shown on the left side; it is used to remove the POF
jacket. The polishing receptacle can be seen at the right; it is used to hold the
connector for polishing the face.
The pliers in Fig. 3.43 is additional required for the V-pin version with crimp
ring. This is a combination tool for assembling connectors to POF and 200 m
PCS fibers.
Complete installations sets are available for a number of POF connectors, some
of which are illustrated here. Figure 3.44 shows the assembly and cutting locator
for a Lucina duplex connector (Asahi Glass) and the assembly of a SC-RJ
(Reichle & De-Massari).
A tool as shown in Fig. 3.45 can be used to assemble connectors with the hot-
plate procedure. The power supply is (not visible) at the left. It periodically sup-
plies power to the heated metal plate for a specified period of time. A red LED
indicates the heating period. Afterwards, the plate is cooled by means of ventila-
tion. The receptacle shown at the far left can be used to press several connectors
perpendicular against the plate with high precision. This is done initially when the
plate is cold. The connector must remain pressed against the mirror for the entire
heating period. When the green LED lights up, the surface has sufficiently
hardened so that the connector can be removed.
The receptacles in Figs. 3.46 and 3.47 serve to polish various connectors. They
guarantee that the face is always perpendicular to the polishing paper and the sand
paper. This also prevents excessive grinding on the connector itself.
Fig. 3.46: Polishing fixtures for simplex and duplex connectors (AMP)
Fig. 3.47: Polishing fixtures for ST connectors and FSMA connectors (at the right)
Figures 3.48 and 3.49 show a tool for stripping the POF and a crimping tool for
the TCP sleeve from FCI. The special feature of the stripper is that it can precisely
guide the fibers in such a way that damage to the optical cladding can be pre-
vented securely.
Fig. 3.48: Wire-end stripper tools for POF with wire guide and stop
256 3.1 Connection Technology for Optical Fibers
Figure 3.50 shows three simple cutting tools to cut POF with or without jacket
using a razor blade. The cut is relatively rough; the tool can be used once only for
each hole. These are devices for home installations that do not involve any
particularly high quality requirements.
Fig. 3.51: Machine for POF-cable fabrication, detail of the machine ([Mei02b])
3.1 Connection Technology for Optical Fibers 257
The automatic machine, a detail of the machine, the principle of attaching the
ferrules by crimping and the saw blade for treating the surface are illustrated in
Figs. 3.51 and 3.52 from [Mei01a] and [Mei02b].
Fig. 3.52: Ferrule crimping, saw for end face processing ([Mei02b])
Today there are a number of connectors for glass fibers which for the most part
have already been described in corresponding standards. As opposed to POF con-
nectors, reflection loss plays a very important role in glass fibers. A low return
loss can be attained through connectors with physical contact (PC, see Fig. 3.66)
or through beveled end faces (APC: angled physical contact, see Fig. 3.53).
The main difference however between the glass fiber connectors and the POF
connectors is the much greater assembly time and effort caused by the lower
tolerances of 5 m for MM-GOF and below 1 m for SM-GOF. Figure 3.54
illustrates a typical assembly procedure for a duplex connector.
258 3.1 Connection Technology for Optical Fibers
Spring ferrules are almost always used with glass fibers in order to guarantee
the physical contact with the end faces. Furthermore, glass fiber cables are usually
equipped with strain relief, e.g. Kevlar strands, which have to be separately
secured in order to protect the actual glass fibers from tensile load. On the other
hand good glass fiber connections also attain losses in the range of 0.1 dB to
0.5 dB.
Somewhere in the middle between POF and glass fibers lie PCS concerning the
complexity of the plug-in connectors. An overview of the available plug-in con-
nectors is provided by [Sch03]. The most important connectors with typical
values are listed in Table 3.1.
ST FC/PC SMA
glueing/polishing crimping/cutting crimping/cutting
Dtyp = 0.6 dB Dtyp = 0.6 dB Dtyp = 1.1 dB
FR = 40 lbs FR = 40 lbs FR = 40 lbs
Dtyp : typical connector loss, FR pull out force (1 lbs = 0.4536 kg)
3.2 Basis for Calculating Connector Losses 259
The following illustrations schematically show the different causes for connector
attenuation. In all cases, Uniform Mode Distribution (UMD) is assumed. This
means for a step-index profile fiber that both the near field as well as the far field
are constant within the acceptance range as shown in Fig. 3.55. In a realistic POF,
equilibrium mode distribution (EMD) is not identical with the UMD because
mode dependent attenuation must also be considered. However, the calculation is
initially made easier by assuming UMD.
-600 -400 -200 0 +200 +400 +600 -30 -20 -10 0 +10 +20 +30
distance to the fiber axis [m] angle to the fiber axis []
Fig. 3.55: Near field and far field under UMD conditions
The first process under consideration concerns the difference between the core
diameters of the POF deployed. In Fig. 3.56 below it is assumed that the light pro-
pagates from the left to the right. Apparently no loss arises where the output fiber
(on the right, shown in blue in the front view) is larger than the input fiber (on the
left, shown in yellow in the side view). However, if the output fiber is smaller,
there is a loss of part of the light (for a more detailed description see [Schw98]).
The difference in numerical aperture has a very similar effect. If the input fiber
has a smaller aperture, the light is completely guided by the output fiber. How-
ever, if the NA of the output fiber is smaller, there will be some losses as shown in
Fig. 3.57.
Tmax1 Tmax2
When calculating the losses under UMD conditions, one looks at the solid ang-
les of the far fields, which can be equated to the far field angle if the NA is not too
large.
Solid angle for input fiber: Tmax12S
Solid angle for output fiber: Tmax22S
Attenuation: D = 10 log (:1/:2) = 10 log (Tmax12/Tmax22)
Expressed by NA: D = 10 log (AN12/AN22)
For the worst case of the ATM Forum specification (AN = 0.35 or 0.30) this
results in the loss of 1.34 dB, whereas the ATM Forum specification permits only
0.8 dB for the sum of losses from core diameter and NA differences. This diffe-
rence can be explained by assuming EMD conditions where NA differences are
far less critical.
In EN 60793-2-40 the permissible values for the numerical apertures for the
fiber classes A4a to A4c lie between 0.35 and 0.65, i.e. the actual NA has not yet
been specified. For example, there is a theoretical loss of 5.38 dB with the coup-
ling of a POF with an AN = 0.65 and full launch on a fiber with AN = 0.35. Evi-
dently this relationship was unknown to the authors of this standard. The NAs of
3.2 Basis for Calculating Connector Losses 261
the fibers of important manufacturers are steady between 0.47 and 0.51 which
results in the possible coupling losses dropping to 0.71 dB. Under real conditions
the mode distributions of the fibers are even closer together so that coupling losses
are typically around some tenths of a dB. Figure 3.58 shows the theoretical losses
in dependence of the relationship of the NA of the input to the output fiber.
6
coupling loss [dB]
5
4
3
2
1
NAin/NAout
0
0.8 1.0 1.2 1.4 1.6 1.8 2.0
Fig. 3.58: Coupling loss dependent on NA-ratio
In the case of a lateral offset of fibers, the calculation of connector losses is just as
simple (Fig. 3.59). Where distance x is not too great (compared with diameter d),
the following applies:
Attenuation D = 10 log (A1/A2)
= 10 log [(dS/4)/(dS/4-dx)]
= 10 log [1/(1-4x/dS)]
5
loss [dB]
4
theory
POF-POF
3
GOF-GOF
2
1
lateral misalignment [mm]
0
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5
Fig. 3.60: Excess loss due to lateral misalignment
Another cause for attenuation at connectors is shown in Fig. 3.61. The surface of
the POF can be rough due to processing, for example, grinding. This changes the
light path and part of the power is lost through diffraction or scattering. The ATM
Forum specifies 0.1 dB of losses since the POF surface should be of good quality
due to hot plating.
Fig. 3.61: Connector attenuation due to roughness of the fiber's end faces
A specific problem of rough surfaces is that the light does not necessarily have
to be lost immediately at the coupling spot. Rather, a part of the light can be
converted into a large propagation angle. By means of mode-dependent attenu-
ation it can then be lost little by little in the following meters.
3.2 Basis for Calculating Connector Losses 263
Finally, Fig. 3.62 shows a deviation of the fiber axes as a possible cause for
attenuation. Here, the ATM Forum allows a deviation of max. 1 corresponding to
a loss of 0.1 dB.
Fig. 3.62: Connector attenuation caused by an angle between the fiber axes
This value is also relatively easy to calculate for small angles compared with
the acceptance angle. The far field comprises an angle-range :, which approxi-
mately equals : = Tmax2S, provided that the NA is not too large. For an angle H
(small against Tmax) between the axes, a solid angle area 2TmaxH is obscured. The
following applies:
Attenuation: D = 10 log [(Tmax2S)/(Tmax2S - 2TmaxH )]
= 10 log [1/(1- 2H/TmaxS)]
For an AN = 0.30 the acceptance angle of Tmax = 17 which results in an attenu-
ation of 0.16 dB, again slightly above the specified value. However, under EMD
conditions the additional loss to be expected is significantly smaller.
A measurement example from [Schw98b] is shown in Fig. 3.63. Two standard
POFs were tilted at angles up to 40 with the fiber spacing kept as low as possible.
10
coupling loss [dB]
9
8 St.-NA-POF
7
theory
6
520 nm
5 570 nm
4 650 nm
3
2
1
0
-40 -20 0 20 40
angular misalignment []
Fig. 3.63: Coupling loss because of angular misalignment of the fibers [Schw98b]
264 3.2 Basis for Calculating Connector Losses
The POFs were about a meter long. The losses measured are clearly lower than
the theoretical ones under UMD conditions. Under normal conditions there are
hardly any modes at the higher angles. Small angle differences at the couplings do
not immediately lead to initial losses. The measurement was conducted with three
different wavelengths. The diverse results can be explained by the different far
fields of the LEDs used. It also shows how important it can be to define the exact
measurement conditions when stating connector losses.
A detailed segment of the area tilted up to 10 is shown in Fig. 3.64. Here once
again two POFs and two glass fiber bundles were used. You can easily see that the
angle errors of 1 to 2 are perfectly tolerable for POF connectors. A diameter
tolerance of e.g. 100 m can be allowed with a 10 mm long connector holder.
0.4
0.2
0.0
-10 -8 -6 -4 -2 0 2 4 6 8 10
angular misalignment []
The value for the Fresnel losses results from the difference in refractive index for
air and PMMA. One can always assume that between POF connectors there is an
air gap between the cores which is substantially larger than the wavelength. Due to
the large mode number and the usually incoherent sources used, interference
effects can be neglected so that the reflection losses can be calculated in the same
way as for independent PMMA/air transitions (see Fig. 3.65).
nPMMA = 1.49
core
cladding
ferrule
F F
Fig. 3.66: Principle of the PC connector
Still, a pressure of 70,000 bar is reached with a pressure force of 5 N and the
contact surface of a 30 m wide diameter. With polymer fibers the area is larger
by a factor of 1000. Polymer is indeed somewhat softer, but much greater absolute
force is necessary for POF. This is problematical because you cannot apply the
forces over the soft cladding as you wish. You have to assume that there will
always be an air gap over the greatest part of the cross-section in POF connectors.
Another loss mechanism described here has not been covered in the ATM Forum
specification. In practice, the gap between the two fiber cores is often significant
for fiber attenuation. Figure 3.67 shows the mechanism for a distances.
266 3.2 Basis for Calculating Connector Losses
For d = 980 m and s = 200 m at AN = 0.30 the result would be 0.09 dB, as an
example. This effect is of particular significance in the case of hot-plate connec-
tors. With these connectors the excess material is pressed into grooves on the side.
This destroys the core's light guiding capability for a short distance. As an appro-
ximation, this effect can be numerically regarded as a gap equivalent (Fig. 3.68).
2
s
- 10 log 1 2 tan max
d
For AN = 0.30 we obtain Tmax = 19.4; since the groove is filled up, Tmax = 12.9
applies. At the stated values and a distance of 400 m the calculated loss is
1.98 dB. The two values are significantly different. In [FOP97] the assumption
was that the light will be evenly distributed on a circular disk expanding with the
max. angle. Even for small distances this assumption is not adequate. We propose
3.2 Basis for Calculating Connector Losses 267
here to use the following formula for small distances and UMD, assuming a linear
decreasing light power for an over radiated ring and an under radiated ring (see
Fig. 6.38).
Area of the input fiber: A1 = d2S/4
Spilled over area: A2 = dSsAN/n (to be weighted with )
Loss: D = 10 log ((A1+A2)/A2)
= 10 log (1 + (s AN)/(nd))
For the stated values this would be 0.33 dB which is near the value given in
[Wei98]. This approximation formula should be suitable for distances that are
significantly less than 1 mm.
s
intensity
distribution
Fig. 3.69: Proposal for calculating the attenuation caused by fiber separation
The actual losses for standard and DSI-POF are shown in Figs. 3.70 and 3.71.
They are clearly below the theoretical line for both fibers (with UMD assump-
tion). The reason for the deviation is again the strong suppression of the higher
order modes in the POF.
1.2
excess loss [dB] St.-NA-POF
1.0
theory
0.8 520 nm
570 nm
0.6 650 nm
0.4
0.2
0.4
0.2
154 m 46 m
For the above values the losses are 0.59 dB for the difference in diameter and
0.29 dB for the offset, now only entered x = 46 m. A similar ratio applies to the
relationship between angle error and NA difference. Nevertheless, under UMD
conditions the losses remain significantly larger than specified. We do not yet
have a comprehensive model for calculating EMD in SI and DSI-POF or even for
the connector attenuation. This is another open task for the standardizing bodies.
However, when assuming EMD instead of UMD, 2.0 dB is a realistic value. With
continuous improvement in POF technology, one could hope for an improved
situation due to reduction in fiber tolerances for the diameter and also for the NA.
It is not possible here to say anything about the likely specification of attenu-
ation in GI fibers. The data to be expected from the fibers are largely not yet
known. However, the diameter is likely to be between 100 m and 150 m with
an NA of around 0.20 so that the requirements for connector precision will be
more exacting by several orders of magnitude.
3.3 POF Couplers 269
Among the passive components for POF transmission systems, couplers play a
significant role that is borne out by the availability of a complete line of products.
Extensive descriptions of how to construct couplers can, for example, be found in
[FOP97] and [Wei98]. In recent years, coupler constructions have been introduced
in various studies such as: [Kal92], [Rog93], [Yuu92], [Woe93], [Yuu94], [Li96],
[Agu97], [Fau98], [Sug99], [Kob99], [Ern00], [Kaw00] and [Woe94].
The following diagrams show different principle possibilities for manufacturing
1 : X couplers (one input with several outputs). Figure 3.73 shows a face coupler
in which both output fibers are coupled with a butt-joint and directly to the input
fibers.
With polished couplers (Fig. 3.75), matched polished fibers are glued in such a
way that there are no protruding areas. Additional losses are created in particular
by sudden changes in the guided angle range. These must be minimized by provi-
ding sufficient flat polished angles. At any rate, a certain dependency of the addi-
tional attenuation on the launch conditions remains.
If a coupler has several ports, polished couplers become too complex. Even
planar mixing elements do not offer enough uniformity of attenuation to the out-
puts. Figure 3.76 illustrates a 1 : 7 coupler with cylindrical-mixing element. With
7, 19, 31 etc. output fibers, the excess loss is not that large.
To improve the uniformity of the attenuation, a curved element can be used in-
stead of a straight mixing cylinder, which then functions as a mode mixer
(Fig. 3.77 according to [Woe93a]).
Good results can be attained at a bend of 180 with a radius of several 10 mm.
The last example in Fig. 3.78 shows a coupler according to [Fau98] for which the
mixing element is a hollow funnel with a wall thickness corresponding to the fiber
thickness. Here too, the goal is - apart from the good uniformity - the reduction of
mode dependency.
Fig. 3.78: Principle of the coupler with mixing element in a conical form
Further possibilities for the coupler design are polished couplers in which
curved fibers are polished laterally or simply flat mixing elements. Melting coup-
lers are based on welding fibers to couplers that are drawn in the coupling zone to
adapt their diameters. Finally, couplers with reflecting elements are also conceiv-
able.
The need for POF couplers is relatively small since there are not yet any large-
scale applications. Up to now couplers have been used in the field of measurement
techniques and for special sensors. POF couplers for bidirectional transmission on
one fiber could find wide-ranging applications in the future.
Since POF couplers are relatively simple to produce, a number of manufac-
turers have attempted to enter the field over the years but none of them have been
able to become well established. At least you can get these components today, but
only if you order a certain minimum amount which is then made to order.
Figure 3.79 shows two examples of a Y-coupler commercially available at the
end of the 1990s. Another coupler in a 16 u 16 configuration is illustrated in
Fig. 3.80 (such components were conceived of for passive star networks in mobile
applications).
A very simple coupler for use in audio networks is sold by Hama in electronic
specialist stores (Fig. 3.83). Four fibers with a diameter of 500 m
are arranged in
the component and distributed over the two outputs. The connections correspond
to a F05 connector.
The typical additional losses of couplers manufactured in this way are below
2 dB including all interface losses. A frequency diagram of the measured additio-
nal losses is shown in Fig. 3.86.
24
no. of samples
20
16
12
8
4
0
0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5
excess loss [dB]
The Institute for Micro technology Mainz (IMM) introduced the production of
passive components for POF and PCS on the basis of waveguides in [Klo03] and
[Fre03]. The necessary moulds are produced by LIGA technology. The subsequent
production of a coupler by filling the waveguide structure with a UV-hardening
polymer is shown in Fig. 3.87.
The finished component with a coupled 1 mm POF is shown in Fig. 3.88. In the
experiment an average additional attenuation of 2.8 dB and a uniformity - in refe-
rence to the outputs - of better than 0.4 dB for the complete component was
attained.
3.3 POF Couplers 275
3.4.1 Filters
Optical filters have to fulfill numerous tasks in transmission systems and sensor
applications. They can, for example, serve the purpose of suppressing interfering
light or reducing near crosstalk in WDM systems. The different arrangements for
multiplexers and filters are described in detail in Chapter 6.
Fundamentally, all filters fall into two categories. With interference filters,
which include optical grating, dielectric multi-layer structures and interferometers,
a certain wavelength range can pass through while the rest is reflected. By means
of correspondingly complex structures you can attain almost any desired spectral
curve. Such components can also be used for multiplexers.
Suitable additions to dye filters absorb undesired light and the spectral curves
depend on the available dyes. These filters normally have worse parameters, but
are many times over simpler and more reasonably priced. They are well suited for
suppressing interfering light and have one great advantage in that they work
independently of the angle of incidence.
Table 3.2 summarizes some important characteristics.
There are not yet any special filters for POF or other thick fibers due to the very
small range of applications. Users have to rely for the most part on products which
are used for general measurement techniques. However, since PMMA POFs are
used in the visible spectral range there are numerous utilizable products from
optical applications such as photography. Section 6.3.7.3 reveals that dyes for ink-
jet printers are very well suited for suppressing NEXT in 520 nm/650 nm WDM
systems.
3.4 Filters and Attenuators for POF 277
3.4.2 Attenuators
Devices are needed for the variable attenuation of light in many areas, especially
in measurement techniques. A typical case is the measurement of the sensitivity of
receivers in which the bit error probability is measured with different light inten-
sities. You could simply vary the power of the transmitter, but then again other
parameters such as the spectrum and the modulation bandwidth would change,
too. That is why devices are used which can be inserted into the fiber link and
cause a variable or given change in the light intensity. The requirements for such
Variable Optical Attenuators (VOA) are:
low insertion loss
large adjustment range
high resolution
good reproducibility
wavelength independence
independence of the far field distribution in the fiber
Different versions of such VOAs have been developed for several fibers and
employed in devices. Two of the simplest procedures are indicated in Fig. 3.90.
The version with the axial offset is particularly well suited for setting very large
attenuation values. You can always carry out precision changes with relatively
great spacing between the fibers. The main disadvantage is that extreme mode
filtering occurs. Collimated light is practically coupled into the output fiber. The
measurement results are distorted enormously in systems in which the bandwidth
of the fiber plays a great role. For example, if you measure the sensitivity of a
wideband receiver, the VOA will generate a much greater bandwidth since all
high modes will simply be filtered away. Consequently, there would be an entirely
different bandwidth value with a real fiber.
The lateral offset also generates a certain mode filtering, but not quite as ex-
treme - at least with step index profile fibers. On the other hand it is not as well
suited for greater attenuation values. If both fibers have almost an offset of one
diameter, then even the smallest shifts will cause definite changes in the attenu-
ation. The theoretical characteristic curve for standard POF - under UMD condi-
tions - is shown in Fig. 3.91.
278 3.4 Filters and Attenuators for POF
35
attenuation [dB]
30 lateral
misalignment
25
20
15 axial
misalignment
10
5
misalignment [mm]
0
0.1 1 10 100
Fig. 3.91: Loss caused by axial or lateral misalignment of two fibers
POF POF
The choice of the neutral filter determines the angular and spectral dependence.
Both are usually relatively low in absorption filters. The smaller the spacing bet-
ween the two fibers, the smaller the mode filter effect (higher modes are decoup-
led more strongly at the gap). The size and stepping of the filter determine the
resolution of the adjusted attenuation and the adjustment range. The precision of
the mechanical design is responsible for the reproducibility. This universal prin-
ciple is widespread.
One possibility of also being able to use filters with angle dependence is the
expansion of the light path with lenses as illustrated in Fig. 3.93.
On the other hand this method has poor characteristics for standard POF in re-
gard to mode independence. For a NA of 0.50 you can hardly find suitable lenses
which can evenly display the entire angle range. This results in a high additional
attenuation for higher modes. In addition, there are a total of 8 interface layers
already resulting in 1.56 dB Fresnel losses. The surfaces would have to be anti
reflection coated for a low-loss set-up.
3.4 Filters and Attenuators for POF 279
grey filter
POF POF
lens lens
In order to minimize the effect of mode dependence and mode filtering, some
commercial devices have inserted mode mixers in front of and behind the filtering
element. This does indeed increase the basic attenuation by some dB, but it does
make the function much more reproducible. The disadvantage is on the one hand
that the stated attenuation values are only valid under approximate EMD condi-
tions and on the other that no systems can be investigated with which non-EMD
conditions are intentionally used.
Figure 3.94 shows an example of the far fields in front of and behind a VOA
(POFA-3 from Bauer Engineering) with varied coupling conditions.
It can easily be seen that there is always approximate modal equilibrium at the
output of the device regardless of the far field distribution at the input. This does
not represent any problem for most applications involving measurement tech-
niques; on the contrary it is rather advantageous.
However, there are also applications in which the light is supposed to be
attenuated without influencing the mode distribution in the fiber. The so-called
Mode Independent Variable Attenuator (MIVA, [Los04b]) was developed at the
POF-AC for just this reason. Figure 3.95 shows the principle set-up.
280 3.4 Filters and Attenuators for POF
optimized
mirror
POF
filter wheel
step drive
POF
Optimized mirrors are used here for the optical imaging instead of lenses.
These mirrors have the advantage of also being able to display very large angle
ranges with constant efficiency. The neutral filter is located in the beam area
which has been expanded to 1 cm to 2 cm. The second mirror images the input
fiber exactly onto the output fiber. Under ideal conditions you can even use the
device for entirely different fibers since the spot size and NA at the output always
correspond for the most part to the fiber at the input.
The output far fields for different attenuation adjustments with three different
launch NAs at the input of the MIVA are shown in Fig. 3.96. The distributions
remain almost unchanged.
0.8
0.6
0.4
0.2
0.0
-30 -10 10 30 -30 -10 10 30 -30 -10 10 30
T [] T [] T []
Fig. 3.96: Far field distributions at the output of the MIVA
The prototype of the MIVA and the commercial device available from Bauer
Engineering can be seen in Fig. 3.97 (see also www.pofatlas.de). An example of
the mirror design is shown in Fig. 3.98.
3.4 Filters and Attenuators for POF 281
We have pointed out in different sections of this book how important it is under
certain conditions to have as defined and reproducible a mode distribution as pos-
sible in the POF. This is especially true for almost all areas of optical measure-
ment techniques with POF and other fibers. In the simplest case you simply use a
launching fiber which is long enough. Unfortunately, you need over 100 m of
fiber under certain conditions before equilibrium mode distribution (EMD) occurs
in the POF. The additional attenuation which thereby arises makes the actual
measurement impossible.
A device is therefore needed which creates EMD conditions on a short piece of
fiber without having too much insertion loss. Such components are designated as
mode mixers or mode scramblers. In fact, they always represent a combination of
mode mixing and mode filtering which is why the designation mode converter is
factually more correct.
One of the widely used mode converters and the only standardized one so far is
the cylinder mixer according to JIS 6863 (Fig. 3.100) for 1 mm standard POF.
cylinder:
42 mm
distance:
3 mm 40 mm
5 mm
90 mm 70 mm
The POF is wound ten times in the shape of an 8 around both cylinders. First of
all, higher order modes are filtered away and secondly, the lower order modes
come in contact much more often with the core-cladding interface which leads to
new modes arising through conversion and scattering. Figure 3.101 shows how the
roughly collimated light after passing through a converter has a distribution very
close to EMD.
The parameters for cylinder diameter, spacing and number of turns have been
empirically determined. A comprehensive analysis on variations of these parame-
ters can be found in [Arr03b]. The authors find that experimentally 7 turns around
a cylinder diameter of 40 mm results in a better mode mixture at a lower insertion
attenuation than a turn with a 20 mm diameter. For 120 m PF-GI-POF they re-
commend a configuration with a 38 mm cylinder diameter, 120 mm cylinder
spacing and 6 turns at approx. 3 dB insertion loss.
3.5 Mode Mixers and Converters 283
Fig. 3.101: Far field distribution in front of the mode converter and behind
A very simple, but quite effective method for mode mixing is the meander
mixer (Fig. 3.102). With correspondingly varied bending radii it can also be used
for other fibers.
A comparable set-up was used in [Fus96] for reducing the far field width of a
wide emitting LED which resulted in an increase in the bandwidth.
NA | 0.43 NA | 0.29
POF
The next mode converter shown here was introduced in [Att96b] in which the
fiber was also laid in meander shape. However, the spacing and bending radii have
been calculated exactly in order to guarantee maximum mode mixing.
R=
10 mm
D = 25 mm d1 = 3.4 mm d1 = 2.2 mm
In the latter work the far field distributions of a 100 m long straight reference
fiber are compared with the measured far field distribution of the mode mixer and
the simulated far field distribution of the mixer. You can obtain very good equili-
brium mode distribution in this way.
1.0
Popt
0.8
0.6
Fig. 3.105: Far field distribution behind mode mixing according to [Att96b]
So far the method of operation of the mode converters has always been evalu-
ated on the basis of the measurement of the far field distribution. In fact, this dis-
tribution does not completely describe the mode distribution. It cannot differen-
tiate between meridional rays and helix rays. Different experiments have shown
that even after different mode converters EMD was not completely attained even
if the width of the far field would suggest that. Furthermore, it has been proven
that in particular the cylinder mixer for standard POF from various manufacturers
leads to quite different results. To a minor extent this is due to the differences in
the NA, but the different scattering parameters at the core-cladding interface are
mainly responsible. In order to attain EMD conditions as close as possible, one
should either work from the very beginning with wide emitting sources, i.e. LEDs,
or combine the mode converter with additional scattering means, e.g. with scatte-
ring foils.
3.6 Optical Slip Rings 285
bearing
tolerance distance
Polymer fibers would be used in such a connection when the attainable mecha-
nical deviations surpass some 10 m, i.e. resulting in unacceptably high coupling
losses with glass fibers and when at the same time no great demands are placed on
the data rate and the length of connection.
The task becomes somewhat more complicated when several fibers have to be
connected to a rotating coupling. This problem is solved using the so-called Dove
Prism (Fig. 3.107) which is positioned between the input and output fibers. Paral-
lel beam bundles are generated using collimators. Total reflection occurs at one of
the outer surfaces of the prism (see for example [Schi07] and [Sta05]).
pass 1 pass 3`
pass 2 pass 2`
pass 3 pass 1`
fiber-collimator-array
Fig. 3.107: Dove-prism in rotary joints
286 3.6 Optical Slip Rings
A special gear unit sees to it that the prism rotates at exactly half the speed of
the moving fiber. As the schematic shows in Fig. 3.108, the rotating fibers are dis-
played exactly on the standing fibers. Such components are available for POF,
MM-GOF or singlemode glass fibers.
Fig. 3.111: Optical 13-channel- rotary joint and collimator array (Schleifring/BLZ)
288 3.6 Optical Slip Rings
The optical axis is not always accessible for data transmission. A classic example
is computer tomography for which very large amounts of data have to be trans-
mitted from the rotating detector and there is only one external ring to do the job.
circular to the
path of fixed
the source detector
In order for the principle to function, an efficient and fast transmitter which can
be modulated must be available. Furthermore, the fluorescence dye has to react
quickly enough to the light, i.e. if possible have a life span in the ns range.
Figure 3.115 shows a typical fluorescence spectrum of one of the dyes used: on
the left the absorption efficiency, on the right the emitted spectrum.
Fig. 3.115: Excitation efficiency and emitted spectrum of the fluorescent POF ([Poi99b])
290 3.6 Optical Slip Rings
Both wavelength ranges lie in acceptable low areas for the attenuation of poly-
mer fibers. The fluorescence life span is shown in Fig. 3.116. The launch pulse
(red curve) is about 2 ns wide, while the emitted pulse has been widened to about
5 ns. The shortest life span for a dye with adequate efficiency lays at 1.9 ns which
means that data transmission rates up to 500 Mbit/s would be theoretically
possible.
Fig. 3.116: Excitation and emission (red) peak of the fluorescent POF ([Poi99b])
In another project at the POF-AC a solution for transmitting even higher bit rates
was developed. The project name GigaFOS stands for Gigabit - Fiber Optic Slip
ring. This project is supported by the Bavarian Economics Ministry as part of the
Microsystem Technology program. The goal is the development of rotary optical
connectors for data rates > 10 Gbit/s. These slip rings have to have a large, free
inner diameter since they are primarily intended for use in medical applications
(computer tomography). The POF-AC has been working together with the BLZ
and the Schleifring and Spinner companies on this project.
Fig. 3.117 shows one of the realized principles with which light is coupled into
a halved optic fiber by means of a prism coupler. When in use, this prism head
moves at several meters per second and glides at a distance of 0.1 m over the
optic fiber. The requirements here are comparable to those of DVD players of the
next generation such as Blue-ray. The fiber used is a PF-GI-POF. On the one hand
the high mode dispersion of the SI-POF can be eliminated and on the other long
wave lasers from communication technology can be used.
3.6 Optical Slip Rings 291
A demonstrator with a model about one-third the size was set up at the
POF-AC which is supposed to prove the function. The central element next to the
floating prism is the polymer fibers made of CYTOP with graded index profile
which are halved lengthwise. The principle and a photo of the model shown here
with visible light can be seen in Fig. 3.118.
The most recent development of the joint project between the POF-AC and the
Schleifring und Instrument-Making Co. is a transmission system based on a mirror
groove [Schl06] and [Schi07]. A laser beam is radiated at a very small angle into a
circular slot. Figure 3.119 shows the principle of light guiding by repeated reflec-
tions and the rotating head for coupling in the light.
Fig. 3.119: Principle of the reflecting groove slip ring and the movable head for light
coupling
Since the mirror groove has high reflectivity over a wide range, almost as many
wave lengths as you wish can be used. Should you wish to use telecommunication
components, i.e. lasers with 1.55 m, then very high data rates can be achieved.
At the same time the singlemode fiber connection of these components generates a
very good collimated beam which hardly increases along the extent of the ring.
The cross-section of the groove and the prototype of the entire system from
[Schl06] are shown in Figs. 3.120 and 3.121.
Fig. 3.120: Cross section of the reflecting groove and the complete transformer
3.6 Optical Slip Rings 293
Fig. 3.121: Complete rotating slip ring (with electrical contacts in the foreground)
In this chapter we will deal with active components, the most important compo-
nents for transmission systems. Except for a few special applications semiconduc-
tor components are employed.
First, we wish to talk about the theoretical basics, then summarize the most
important types and structures of emitters and receiving diodes. Finally, we will
describe the components available. Just as with fibers, technical development in
the field of optoelectronic components is proceeding at such a rapid pace that we
can only present a snapshot of developments here.
Other laser sources are used only to a small degree in optical free-space con-
nections or optical heterodyne systems. This is why the subject will not be further
discussed here. Because of their low modulation speed and their size, none of the
thermal light sources available today are suitable for optical communication.
Organic light sources may become interesting alternatives particularly because of
their potential for being manufactured inexpensively.
Semiconductors differ from metal and insulators through their band structure. A
valence band completely occupied with electrons in the ground state (0 K) is sepa-
rated from an empty conduction band by an energetic gap (the band gap) having a
width of WG.
If an electron is lifted into the conduction band through thermal activation, a
hole remains in the valence band (it is more efficient to fill the valence band with
holes and the conduction band with electrons by using an external electrical
source, i.e., through injection). Both particle types can recombine with each other,
whereby a photon having a frequency of f = WG/h is emitted, and the electron falls
back into the valence band. The band gap is dependent on the material and the
state of motion (impulse) of both particles. The impulse p of a particle can be ex-
pressed as p = h/O = h k/2S.
With direct semiconductors, the maximum of the valence band and the mini-
mum of the conduction band (this is where the charge carriers tend to concentrate)
are located directly on top of each other, i.e., at the same impulse value (Fig. 4.1,
left illustration). Thus, recombination can occur very efficiently by emission of
photons. With indirect semiconductors, the extreme values are found at different
k-values (Fig. 4.1, on the right).
During recombination, the electron must change its impulse, which is achieved
through interaction with a phonon. Since three particles must interact, the radia-
ting recombination is less probable and inefficient. Only a small number of the
well-known semiconductors possess a direct band structure. With mixed semicon-
ductors, i.e., a combination of more than two elements, there are often only certain
areas in which the material is direct.
Indirect semiconductors can also be used as light sources; nevertheless, they are
inefficient and slow and will not be taken into consideration here. Some green
LEDs, GaP for example, are produced from indirect semiconductors in which so-
called deep defects are embedded through which the light is emitted. Recently,
even silicon, the best-known indirect semiconductor was stimulated to emit light.
In addition to their extremely low power efficiency these emitters are as a rule still
too slow.
4.1 Emitters and Receivers 297
direct indirect
semiconductor semiconductor
W W conduction
band
- - - -
--- -
-- - - - - -
- -
photon phonon
WG photon
+ +
+ ++ +
+ +
k + +++ ++ + + k
valence
e.g. GaAs e.g. Si band
Eg [eV] O [nm]
GaAs InP InAs GaSb
2.5 0,.5
GaP
AlAs
2.0
A lS b 0.75
1.5 GaAs InP
1.0
1.0
1.5
GaSb
2.0
0.5 InAs 3.0
direct semicon.
5.0
indirect semicon.
lattice constant []
0.0
5.5 5.6 5.7 5.8 5.9 6.0 6.1
Fig. 4.2: Lattice constants and band gaps of various semiconductors
298 4.1 Emitters and Receivers
When current is injected, it is exactly this zone that forms an area in which
there is a large concentration of holes and electrons at the same time. Another
effect is that, through doping, conductivity is increased by a high order of magni-
tude in contrast to a pure semiconductor. Figure 4.2 illustrates the range in which
many optical semiconductors are direct. The colored strips mark the attenuation
windows of the PMMA-POF.
The material system (AlxGa1-x)yIn1-yP is of particular interest (band gap and
lattice constant are shown in Fig. 4.3). With 50% part indium the semiconductor is
lattice-matched to GaAs. Theoretically, the wavelength range of 525 nm to
656 nm can thus be covered, i.e. there where POF has its minima.
700
1.7
750
800
1.5 GaxIn1-xP
GaAs
850
900
InP
1.3 950
5.4 5.5 5.6 5.7 5.8 5.9
lattice constant []
AlP: Eg = 2.45 eV InP: Eg = 1.35 eV
A0 = 5.4510 A0 = 5.8686
GaP: Eg = 2.26 eV GaAs: Eg = 1.424 eV
A0 = 5.4512 A0 = 5.6533
Fig. 4.3: Lattice constants and band gap energies in the AlP/GaAs material system ([Li05])
Figure 4.4 shows the corresponding lattice constants and band gaps for the GaN
semiconductor system. In this case there is unfortunately no suitable, i.e. lattice-
matched, substrate material. Usually, SiC as conduction material or sapphire as
isolator is used. In both cases an interlayer has to be inserted in which the lattice
constants have to be adapted to the LED material. This layer is full of defects and
warping which are luckily electrically neutral unlike AIIIBv semiconductors. Only
in the past few years has there been a theoretical explanation for this phenomenon.
4.1 Emitters and Receivers 299
The figure shows another necessary requirement for the construction of a lumi-
nous component. The various layers are applied to a substrate through different
processes (epitaxy). The substrate is a carrier made of semiconductor material that
was cut from a mono-crystal block (with a typical thickness of 100 m to
300 m). On the one hand, this carrier makes it possible to handle the otherwise
extremely thin layer structure; on the other hand, it determines the crystal structure
of the fully grown semiconductor layers.
3.5 350
GaN
3.0 400
2.5 500
InN
2.0 SiC ZnO 600
1.5 800
3.0 3.1 3.2 3.3 3.4 3.5 3.6
lattice constant []
If the lattice constants of the substrate and the layers are perfectly identical then
no forces arise in the crystal. Should the lattice constants differ to only a slight
extent the layers applied are slightly deformed (strained).
This lattice misfit can usually amount to a few parts per thousand. If it becomes
larger, dislocations occur which can seriously influence the functioning of the
component.
300 4.1 Emitters and Receivers
The properties of the semiconductor source are defined mainly by three essential
design parameters. The choice of semiconductor material essentially determines
the emission wavelength, as shown above. Furthermore, the choice between a
direct semiconductor and indirect semiconductor is critical for efficiency and the
modulating characteristics. Thirdly, there are various possibilities for spatially
structuring the diode whereby the carrier density also can be influenced such as
the light path within the component and the decoupling of the emission. A number
of these structuring possibilities will be described below.
Semiconductor diodes consist of a layer sequence of various materials in order
to make light generation efficient. Thus, several requirements must be combined:
The light emitting layer should have a band gap that corresponds to the desi-
red wavelength.
This material as well as the material of the other layers must be adapted to the
substrate lattice.
By doping, a p-n junction must be created near the light-emitting layer. All
other layers must be good conductors.
It must be possible to manufacture all layers with an uniform process.
4.1 Emitters and Receivers 301
energy W
Wg1
Wg2
Wg1
n-doped p-doped
semiconductor semiconductor semiconductor
band gap W g1 with a smaller band gap W g1
band gap W g2
In laser diodes, it is not only the different band gap that is interesting, but also
the different refractive indices. The intermediate layer forms an optical waveguide
that also enhances efficiency. This effect does not play a role for surface emitters.
Other possibilities for a vertical structuring are layers for a separate optical
guide or the use of so-called quantum wells. Here, the layer thickness is reduced to
the point that the charge carriers can only move in a single plane. This radically
changes the band structure. Practice has shown SQW and MQW structures (single
and multi quantum well) to have a high degree of efficiency and very stable wave-
lengths. However, they require very costly processes and cannot be implemented
with just any material.
302 4.1 Emitters and Receivers
upper contact
cladding layers
active layer
substrate
bottom contact
If a conduction substrate is used, one contact can be attached to the bottom and
a second one to the top of layer package. LED are often glued to the layer package
face down on the metal carrier. Since the substrate is transparent for the light
emitted, it does not effect the radiation. The heat generated due to its function can
be dissipated better so that the potential optical power increases. The component is
finally placed into a transparent package, so that the semiconductor is not subjec-
ted to humidity and is also protected against mechanical impact. By creating a
lens, the emitting characteristics can be modified in a suitable manner.
sity, the stimulated emission is so strong that losses in the component are
exceeded. The last requirement is that a resonant cavity must be made. In the most
simple of cases, it is formed by two parallel cleaved semiconductor edges (Fabry-
Perot laser diode). Figure 4.8 illustrates the principle structure.
resonator mirrors
upper contact
active layer
substrate
bottom contact
upper contact
active layer
substrate
bottom contact
upper contact
upper mirror
active layer
bottom mirror
substrate
bottom contact
Fig. 4.10: Structure of a VCSEL
To limit the current, the area of the light emitting surface is practically always
restricted by a limiting device acting on the conduction area under the upper con-
tact (aperture diameter is typically some 10 m). Thus, the active volume is much
smaller than with conventional lasers. This results in threshold currents that are in
the range of a few mA but can also amount to 100 A. This also limits the output
power to a few mW. Efficiency is also as high as with the best laser diodes. For
data communication with POF, a laser power restriction for eye protection is in
effect that makes typical VCSEL power completely adequate. Advantages of
VSCEL technology are:
The change in wavelength with temperature comes to approx. 1/3 of the value
for LED.
The laser emits light perpendicular to the surface. This makes it easier to
couple the light to fibers and test the components on the wafer.
The threshold current is very low which makes the power consumption of the
transmitter very small.
The VCSEL emits light at a small emission angle that is nearly circularly
symmetrical and is thus ideal for coupling into the fibers.
The spectrum of a VCSEL is very narrow compared to an LED.
4.1 Emitters and Receivers 305
The biggest problem in producing VCSEL is the mirrors. At times, they must
reflect more than 99% of the light. To achieve this, various layers of semicon-
ductor materials are applied alternatively. For both mirrors, this can mean more
than 200 additional layers. Unfortunately, the selection of suitable semiconductors
for short-wave ranges is very limited. As will be shown in Chapter 4.2.5, VCSEL
may be the ideal source for POF systems.
In the meantime very powerful VCSELs are available in the 780 nm to 850 nm
range which also permit speeds up to 12 Gbit/s in addition to a temperature range
of over +125C and high efficiency. In the 650 nm range, however, things look a
bit different. Because of the diminished thermal conductivity - caused by the Al
share - the maximum application temperature drops. Commercial 650 nm
VCSELs can only be used up to about +45C. Recently at the University of
Stuttgart a CW laser operation at 70C was also achieved for red VCSELs.
upper contact
upper mirror
active layer
bottom mirror
substrate
bottom contact
Fig. 4.11: Structure of a RC-LED
n-AlGaAs
polyimide/
gold mirror
Si-substrate
Whereas the first NRC LEDs were realized in the wavelength range around
800 nm they can also be produced in the meantime in the visible range. Compara-
ble decoupling power efficiencies of > 50% can also be attained with other
methods. Slanted sides of the LED chip have proven to be very effective.
The following sections describe examples of transmitting diodes which have been
developed for POF systems or at least can be used for such. There is a conside-
rable difference between semiconductors for glass fibers and those for POF,
namely, the price. The great number of lasers available today for 850 nm to
1.55 m have for the most part been specifically developed for data communica-
tion. All other applications are in comparison relatively few. The number of pieces
was still relatively small at the beginning of development, but laser diodes some-
times cost several DM 10,000. Today the components are much cheaper, but the
number of pieces is also considerably higher.
The situation for POF components is entirely different. In the beginning POF
systems had to be able to compete in price with other large-scale technologies.
Therefore the price span for transmitting diodes often only amounted to a few
10 ct. The development costs for new components cannot be shifted to the first
product generation.
Most manufacturers only want to initiate the development of new components
when production figures amounting to several million pieces annually are guaran-
teed. As long as there are no optimized POF transmitters available the expansion
of the technology is being hampered. Fortunately, many POF applications can
accept component parameters in many areas so that LEDs or lasers can be used
from different large-scale applications. In the chapter on POF systems many expe-
riments were presented in which LEDs for lighting purposes or lasers for laser
pointers and barcode readers could be used. New transmitters were first developed
for MOST networks since their use on a large-scale was very probable. The
situation in Germany should change with the introductory use of POF in 2006.
The increasing demand for POF systems will initiate developments among the dif-
ferent manufacturers which will lead to falling prices and quickly improved para-
meters and consequently inspire greater use of POF.
Since approximately the middle of the 1980s commercial POF components on the
basis of red LEDs and SLEDs have been sold. In the first few years it was
primarily GaAlAs LEDs whose emission wavelengths lay in the 660 nm to
670 nm range. Through the use of quaternary semiconductors the wavelength was
gradually better adapted to the 650 nm minimum.
308 4.2 Transmitting Diodes for Data Communication
Ring-LED at 650 nm
An LED especially adapted for POF is described in [Dut95] and [Yam95]. The
emission wavelength of the double hetero structure LED is 655 nm and optimized
for the attenuation minimum of the POF. Figure 4.14 shows the LED structure.
contact metal
isolation SiO2
substrate n-GaAs
The special adaptation of the structure consists in the ring-like contacts (outer
diameter 65 m) instead of a contact attached to the center, as is common for
LED. This adapts the near field better to the fiber coupling. When directly coup-
ling with a 2 mm ball lens to a 1 mm POF, 35% of the output power can be
launched into the POF. The coupling efficiency can be increased to 70% by
attaching a specially mounted plastic lens. The power can thus reach 1.7 mW at
100 mA diode current. The spectral width is 25 nm; the modulation data rate up to
156 Mbit/s.
SLED at 650 nm
Super-luminescence diodes enable faster modulation and have a smaller emission
angle than conventional LEDs. They were used in different POF transmitters at
the end of the 1990s, for example, by Hewlett Packard and NEC (see also
Fig. 4.71).
MOST-LED
A low temperature dependence of output power is especially decisive for use in
mobile networks. A red LED with a particularly small change in the output power
(only 2 dB) in the range of 20C to 125C was introduced in [Baur02] (see
Fig. 4.15).
4.2 Transmitting Diodes for Data Communication 309
-0.5
-1.0
-1.5
power in the POF:
150 W at 10 mA
-2.0 (without lens)
switching time: 12 ns T [C]
-2.5
20 40 60 80 100 120
Fig. 4.15: Very temperature stable red LED for MOST
Laser diodes in the range around 650 nm have been produced for a number of
years in great quantities. The most important applications are in CD and DVD
drives, laser pointers and scanners. Truly great modulation bandwidths are not re-
quired here. Since the lasers do not have to work continuously in most appli-
cations, the specified life spans are not always suitable for applications in data
communication. The main disadvantage is usually the type of construction which
has not been optimized for coupling with fibers. In general, these lasers work with
AIIIBV semiconductors like GaAlAs or AlInGaP.
Popt. [mW]
9
90C 80C 100C 110C 120C
8
70C
7
6 50C
5 25C
4
3
2
1
laser current [mA]
0
0 10 20 30 40 50 60 70 80 90
Fig. 4.16: P-I-T-characteristics of a 650 nm laser
The active layer of the component described consists of an GaInP quantum well
(tension-strained) in AlInGaP barrier layers (pressure-strained) encased with
(Al0,7Ga0,3)0,5In0,5P cladding layers. For lasers with a 600 m long resonant cavity,
the maximum possible optical output power is 72 mW at a threshold current of
38 mA and a wall plug efficiency of 1 mW/mA. An optical power of 30 mW can
be achieved at up to 75C in CW operation. At an power of 30 mW and 50C tem-
perature, the service life is over 1,000 h.
The possible performance of red laser diodes today is demonstrated in [Ohy99].
The laser described there is designed for DVD applications whereby high output
power and a long service life are expected at a low price. The 655 nm laser is
based on AlInGaP. The active layer is formed through a pressure-tensioned MQW
on a disoriented GaAs substrate. The resonant cavity length is 500 m. At 80C
and a constant optical output power of 5 mW, a service life of 92,000 h can be
achieved. An power of 5 mW can be achieved up to +115C. At room tempe-
rature, the threshold current is 36 mA.
MQW-Laser at 650 nm
In [Oka98] various MQW lasers are described that are based on tension-strained
active layers made from GaInAsP/AlGaInP. Threshold currents between 4.5 mA
and 23.4 mA are attained at wavelengths between 654 nm and 659 nm with
different resonator lengths and mirror coatings. Power values of 30 mW for up to
+90C are possible.
A laser that was adapted just for POF is introduced in [Mor95]. The LD posses-
ses an active layer made of AlInGaP with MQW structure. The lateral current
confinement is achieved by etched wells. At a threshold current of 24 mA (room
temperature), the wavelength is 650 nm. The maximum operation temperature is
+80C. The maximum modulation bit rate is more than 4 Gbit/s.
4.2 Transmitting Diodes for Data Communication 311
For the Sanyo laser the curve for laser wavelength in dependence on the
temperature is shown in Fig. 4.17 (the information has been taken from the data
sheet). In addition, the attenuation of a PMMA POF at the corresponding wave-
lengths is entered in the diagram.
666
laser wavelength [nm]
664 DPOF = 234 dB/km
At 70C the wavelength changes up to 664 nm. Here the POF-attenuation has
already increased by 100 dB/km. The temperature coefficient of 0.18 nm/K is de-
termined by the material of the red laser. You therefore have to take care that in
312 4.2 Transmitting Diodes for Data Communication
the middle of the intended operating temperature range the emission wavelength
lies at exactly 650 nm, or as close as possible thereto.
The temperature-dependent power-current characteristics of one of the lasers
(from Union Optronics) is shown in Fig. 4.18. Between 10C and 70C the thres-
hold current almost doubles and the differential power efficiency becomes less.
9
Popt [mW] 10C 25C 40C 50C 60C 70C
8
7
6
5
4
3
2
1 ILD [mA]
0
0 10 20 30 40 50 60 70 80
Fig. 4.18: Optical power-current-characteristics of a 650 nm laser (Union Optronics)
The spectral width of the laser normally lies in the range of about 2 nm. Chro-
matic dispersion therefore plays no role with PMMA fibers when laser diodes are
used. Examples of the spectra are shown in Fig. 4.19. The lasers are spectrally
multimodal. In the diagram the modes have not been resolved since they only lie a
few tenths of a nanometer apart.
1.0
Prel.
0.9
0.8 5 mW
25C
0.7
0.6
0.5 vertical
0.4
0.3
0.2
horizontal
0.1 T []
0.0
-40 -30 -20 -10 0 10 20 30 40
Fig. 4.20: Typical far field of a 650 nm laser (Sanyo)
Laser diodes are supplied in diverse types of housing. The set-up in a TO-18
housing is quite common. The laser chip is placed in a sealed housing. A window
with a diameter of 1 mm is located at the top about 1 millimeter away from the
emitting surface (Fig. 4.21). At a far field width of 30 the light spot at the
window is still only 1 mm large and can thus be readily coupled into a POF.
The three pins are used for the power supply of the laser and for connecting a
monitor photodiode. Common cathodes, anodes or also series connections are
possible. From time to time a lens is used instead of a plane window which either
focuses the light beam or generates parallel light (for laser pointers).
314 4.2 Transmitting Diodes for Data Communication
A particularly rapid development in the last few years has been that of LEDs on a
GaN basis. The emission wavelength of pure GaN lies in the blue range. By ad-
mixing aluminum the emission wavelength can be enlarged to about 560 nm. As
opposed to conventional AIIIBv semiconductors there is no lattice matched sub-
strate material. Today GaN LEDs are produced on a sapphire or SiC substrate. By
using suitable coating procedures you can prevent any dislocations from arising
despite the large lattice mismatch in the active area. Generally, GaN LEDs can be
modulated fast - up to some 100 Mbit/s -, have high power efficiency and good
reliability.
1.0
rel. power -40C
0.9
-20C
0.8 NSPG525 0C
Nichia
0.7 525 nm 20C
0.6 at 20 mA 40C
60C
0.5
80C
0.4
0.3
0.2
0.1
0.0
460 480 500 520 540 560 580 600
wavelength [nm]
Fig. 4.22: Spectra of the green LED NSPG525
Another example is illustrated in Fig. 4.23 with NSPG510. The spectral width
of 35 nm for this type is somewhat smaller and the temperature dependence is
even lower. The advantages when using such an LED instead of red transmitters
can easily be seen. First of all, the attenuation of the POF is lower by almost a
factor of 2 and secondly the temperature coefficients for wavelength and output
power play almost no role.
1.2
rel. opt. power
1.0 temperature
-20C to +70C
0.8
LED current 20 mA
0.6
Nichia
0.4 NSPG510
0.2
wavelength [nm]
0.0
450 500 550 600 650
Fig. 4.23: Spectra of the green LED NSPG510
The relative change in the output power in dependence of the temperature for
the NSPG500 is shown in Fig. 4.24. The power only changes about 1.1 dB in the
-20 C to +70 C range which is typical for home applications.
0.0
-0.5
-1.0
temperature [C]
-1.5
-60 -40 -20 0 20 40 60 80 100
Fig. 4.24: Change of the optical power with the temperature for a GaN-LED
In the meantime, Nichia has also produced LED samples which have been opti-
mized for POF transmission. Essentially, the chip surface has been reduced in size
in order to maintain a lower diode capacitance.
316 4.2 Transmitting Diodes for Data Communication
1.0
0.9 Prel. Nichia LED
sample
0.8
0.7 T = 25C
0.6 ILED = 20 mA
0.5
0.4
0.3
0.2
0.1 O [nm]
0.0
400 450 500 550 600 650
Fig. 4.25: Emission spectrum of the fast green LED
The highest wavelengths which have been attained with GaN LEDs so far
amount to 562 nm, i.e. in the range of the absolute attenuation minimum for POF.
The power efficiency is reduced with higher wavelengths, but the LEDs tested still
emitted 1.9 mW at 50 mA.
Figures 4.27 and 4.28 show the temperature-dependent spectra of the LED at
20 mA and the change in LED power with the temperature at 25C - measured
directly and after 250 m of PMMA POF having taken the spectral filter effect into
account.
4.2 Transmitting Diodes for Data Communication 317
1.8
rel. power
1.6 ILED = 20 mA
temperature:
1.4
-20C to +70 C
1.2
1.0
0.8
0.6
0.4
0.2
wavelength [nm]
0.0
520 540 560 580 600 620
Fig. 4.27: Spectra of 560 nm LED (sample from Nichia)
Fig. 4.28: Change of the optical power due to the temperature (560 nm LED)
Blue LEDs in the range of 430 nm to 470 nm have also already been used for
POF systems. The attenuation of PMMA POF at 470 nm is indeed about
20 dB/km higher than at 520 nm, but the blue LEDs are considerably more effi-
cient and as a rule can be modulated more quickly. One main reason may lie in
their better conductivity. Furthermore, many blue LEDs show a clearly lower
spectral width so that the chromatic dispersion is of no influence.
The temperature-dependent spectra of a blue LED in the -20C to +70C range
(type SHR470 from Sander Electronics) is shown in Fig. 4.29. The temperature
coefficients for power and wavelength are even lower than with green LEDs, thus
ideal for POF applications.
In 2006, the POF-AC first succeeded in realizing an error-free transmission of
data at over 1 Gbit/s with a blue LED (DieMount). Once the GaN LED manufac-
turers turn to optimizing POF transmitters, then further increases in performance
are foreseeable.
318 4.2 Transmitting Diodes for Data Communication
1.1
rel. power
1.0 SHR470, 20 mA
T = -20C .. +70C
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1 O [nm]
0.0
420 440 460 480 500 520 540
n-contact
upper mirror
InGaN
quantum wells
p-contact
AlGaN/GaN
DBR
GaN
Sapphire
AR-coating
norm. power
1.0
0.9
0.8
0.7
0.6
0.5 -40C
0.4
+20C
0.3
0.2 +70C
0.1
0.0
470 480 490 500 510 520 530 540 550 560
wavelength [nm]
1.0 10C
PPOF [mW]
0.9
20C
0.8
30C
0.7
0.6 650 nm 40C
RC-LED
0.5 50C
60C
0.4
0.3
0.2
520 nm LED
0.1
I [mA]
0.0
0 10 20 30 40 50 60
Fig. 4.32: P-I-characteristics of a green LED compared with a RC-LED
Lasers with a wavelength in the green attenuation minimum of the POF would of
course be ideal transmitters. In the meantime, green laser pointers are part of the
product range of any hobby shop. Hidden inside these laser pointers, however, is
always an infrared laser source, the frequency of which has been doubled. Real
green lasers have so far not been technically employed on a large scale. Lasers
based on ZnSe have been successfully realized. The GaN material system has only
permitted blue lasers so far.
For POF systems efficient lasers at 520 nm and 560 nm would be suitable. Effi-
cient lasers with ZnSe have already been developed whose service life, however,
do not meet the practical requirements. Blue lasers on a GaN basis have been in-
troduced into the market and are especially used for mass storage systems (DVD).
Green lasers based on InGaN so far operate optically pumped. They definitely can
be expected to be electrically powered in the near future.
the layer surface are 10.5 and 13 respectively. The service life is not speci-
fied. The optical parameters would be ideal for use in POF systems.
Pd/Au contact
LiN lateral isolation
ZnSe/ZnTe MQW contact layer
ZnMgSSe cladding layer
ZnSSe waveguide layer
ZnCd0.25S0.07Se0.68 active layer
ZnSSe waveguide layer
ZnMgSSe cladding layer
n-GaAs substrate
Pd/Pt/Au contact
Extensive studies on the development of vertical laser diodes (VCSEL) in the red
and near infrared spectral range were conducted at the Universities of Ulm and
Stuttgart (for example [Ebe96], [Ebe98]). The best VCSEL are those currently
available in the spectral range of 800 nm to 1.000 nm with AlInGaAs quantum
wells. The most efficient components attain a 47% degree of conversion efficiency
(optical power relative to the electrical dissipation power) and 50 mW of optical
power or also threshold currents of 0.29 mA. In the short-wave range, 670 nm are
achieved. Intensive studies are underway with red and green VCSEL. In the long
wavelength range, 1,550 nm lasers are being developed, larger wavelengths (for
example, 6 m with IV-VI) semiconductors are also possible.
9
Popt. [a.U.]
8
40 mA
7
6 30 mA
5
20 mA
4
3 10 mA
2
1
O [nm]
0
620 630 640 650 660 670 680 690
Fig. 4.34: Spectra of a 655 nm RC-LED
Further details on this diode are provided in [Gui00a] and [Dum01]. The
sources developed at the Optoelectronics Research Center of Tampere University
of Technology can be modulated to 1 Gbit/s (see also Chap. 6). 622 Mbit/s and
400 Mbit/s were transmitted over 1 m or 10 m respectively of POF. The relation-
ship between the magnitude of the active zone, the optical power and the modula-
tion bandwidth is shown in Table 4.2.
RC-LEDs at 650 nm
In [Gray00] a new 650 nm RC-LED is described for use in POF systems. The best
RC-LED so far based on InGaP/AlGaInP reach an external quantum efficiency of
4.8%, and 0.5 mW of optical power at 5 mA diode current at an emission wave-
length of 660 nm.
The diode described has two mirrors with 32 periods (bottom) and 8 periods
(top) made of Al0.5Ga0.5As/AlAs with a distance of one wavelength (and thus
considerably fewer layer pairs than for the VCSEL described above). The active
layer is formed by three In0.5Ga0.5P quantum wells, (Al0.5Ga0.5)0.51In0.49P barriers
and (Al0.7Ga0.3)0.51In0.49P cladding layers (MQW structure). The shift to 650 nm is
achieved by straining the active layer. A mesa 400 m large was etched into the
layer. The wavelength lies between 647 nm and 649 nm, depending on the angle
of emission at a spectral width of 4 nm. At 1 mA current, the optical power is
0.1 mW, which corresponds to a quantum efficiency of 6%, with a diode voltage
of 1.7 V.
The principle of detuning of the mirror resonance against the emission wave-
length of the active layer is described in [Gray01], whereby both Bragg mirrors
have a somewhat too high wavelength at room temperature. When the temperature
is increased, the emission wavelength of the active layer runs so-to-speak into the
resonance. As a result the normal decrease in efficiency is compensated for by a
wide temperature range. Another effect is that the emitted wavelength varies with
the emission angle as shown in Fig. 4.35.
40
rel. power 25 2015
30 10
30 5
35 0
20 40
45
10 50
55
60
65 O [nm]
0
620 625 630 635 640 645 650 655 660
Fig. 4.35: Angle dependent spectra of a RC-LED with detuned mirrors ([Gray01])
324 4.2 Transmitting Diodes for Data Communication
Here the emitted spectra were measured angle dependent between 0 and 65.
At 0 the emission maximum lay at about 653 nm. On the other hand at 65 the
maximum lay at 632 nm. You have to keep in mind that the differences in angle in
the diode itself are smaller because of the light refraction when exiting the
material.
This effect can also be clearly seen in Fig. 4.36. Here the far field has been
measured at different wavelengths. The higher wavelengths are transmitted from
the center while the shorter wavelengths come out of the diode slanted.
Popt [a.u.]
0
4
-30 30
-60 60
2
652.5 nm
651.5 nm
650.5 nm
1 648.5 nm
646.5 nm
644.5 nm
642.5 nm
0
-90 90
Fig. 4.36: Wavelength dependent far field of the RC-LED with detuned mirrors
This diagram is particularly impressive because you can see that not only the
power and spectrum of the detuned RC-LED change at different temperatures, but
that the coupling power efficiency in a fiber can also change considerably and that
these changes can proceed spectrally quite differently. This effect can be used to
keep the power at the receiver largely temperature-independent, although it can
only function well with one type of fiber within a specific length: for example, in
vehicle networks in which the fibers are specified exactly and where there are no
fibers longer than 10 m.
30
Popt. [a.u.] -10C
0C
25
+10C
+20C
20
'O = 20 nm +30C
dO/dT = -0.082 nm/K +40C
15 +50C
+60C
10 +70C
+80C
5
O [nm]
0
600 610 620 630 640 650 660 670 680
1.2 +10C
PPOF [mW] +20C
1.0 +30C
+40C
0.8
+50C
+60C
0.6 +70C
+80C
0.4
0.2
I [mA]
0.0
0 10 20 30 40 50
Fig. 4.38: P-I-characteristics of the RC-LED
1.0
Prel OHF00147
0.9
0.8
0.7 25C
0.6 60C
0.5 90C
0.4
0.3
0.2
0.1 M []
0.0
-100 -80 -60 -40 -20 0 20 40 60 80 100
Fig. 4.39: Far field of a RC-LED at different temperaturs
The decrease in optical power with the temperature is shown in Fig. 4.40 accor-
ding to [Baur02]. The drop in power up to 85C (working range for MOST net-
works) is less than 1 dB, up to 125C it is 3 dB (at present the power rating for
MOST allows for up to 6 dB fluctuations in transmission power).
-0.5
-1.0
-1.5
-2.0
-2.5
-3.0
T [C]
-3.5
20 40 60 80 100 120
Fig. 4.40: Temperature dependence of an optimized RC-LED
As is shown in Fig. 4.41 from [Wir01b], the geometry of the electrical contacts
also plays an important role. In order to maintain high optical power, large chip
surfaces are used with large-area contacts which distribute the current (on the left)
optimally. Small diodes which have less capacitance make sense for data trans-
mission. Furthermore, the center of the emission area should not be covered by
electrical contacts in order to enable an optimum coupling of the light into the
fiber (on the right).
4.2 Transmitting Diodes for Data Communication 327
type A type B
high power data links
300 u 300 m : 80 m
VCSEL at 690 nm
[Saa00] describes a 690 nm VCSEL. Here 55 and 38 layer pairs are used as a
mirror. The mesas have a diameter of 34 m to 50 m at a current aperture dia-
meter of 4 m to 20 m. 1.3 mA of threshold current and max. 0.56 mW of power
at 5.6 mA (6.9% external efficiency) are attained. Laser operation is possible up to
+45C. The P-I characteristic curve of the VCSEL with a 10 m aperture diameter
is shown in Fig. 4.42. The greatest attainable power efficiency lays at 6.9% at
3.7 mA. This component is hardly suitable for POF since the attenuation amounts
to over 300 dB/km.
328 4.2 Transmitting Diodes for Data Communication
1.0
opt. power [mW]
5C
0.8
690 nm 15C
0.6
25C
0.4
35C
0.2
45C
0.0
0 1 2 3 4 5 6
current [mA]
VCSEL at 675 nm
Figure 4.43 shows the characteristic curves of another 675 nm VCSEL [Lam00b].
The threshold current here is even lower. At 25C an optical power of over 1 mW
is attained. However, laser operation can only be achieved up to around 50C.
1.8
opt. power [mW]
10C
1.6
1.4 20C
O = 675 nm
1.2
1.0 30C
0.8
0.6
40C
0.4
0.2 50C
0.0
0 1 2 3 4 5 6 7 8 9
current [mA]
Fig. 4.43: P-I-(T) characteristics of a 675 nm VCSEL according to [Lam00b]
VCSEL at 674 nm
Another VCSEL with an emission wavelength of 674 nm is described in
[Tyn00b]. Here laser operation has even been attained up to 75C (Fig. 4.44). For
a comparable type with an emission wavelength at 670 nm the maximum applica-
4.2 Transmitting Diodes for Data Communication 329
tion temperature still amounts to somewhat over 60C. The mirrors of the VCSEL
consist of 35 pairs Al0.95Ga0.05As/Al0.5Ga0.5AS (top) and 54 pairs (bottom). The
active zone consists of four quantum wells.
1.0
opt. power [mW]
VCSEL
674 nm
0.8
10C
0.6 20C
30C
0.4 40C
50C
0.2 60C
75C 70C
0.0
0 1 2 3 4 5 6 7 8 9 10
current [mA]
VCSEL at 670 nm
The temperature behavior of a 670 nm VCSEL (see [Tak99]) is shown in
Fig. 4.45. The threshold current lies at only 4 mA and is clearly below the typical
values of an edge emitter at 15 mA up to 60 mA. Laser operation is, however,
only possible up to about 50C. If you take the typical temperature rise into consi-
deration, for example, in a PC housing, such a component can hardly be used over
30C, in other words only in an air-conditioned environment.
1.0
opt. power [mW] 10C 20C
0C 30C
0.8
0.6
0.4 40C
0.2 50C
current [mA]
0.0
0 1 2 3 4 5 6 7 8 9 10
Fig. 4.45: P-I-(T) characteristics of a 670 nm VCSEL laser according to [Tak99b]
330 4.2 Transmitting Diodes for Data Communication
VCSEL at 665 nm
[Lam00b] describes the structure of arrays from 665 nm VCSEL for parallel data
communication with POF (see also [Lam00a]). The aim of the studies is the
capability to manufacture VCSEL with >1 mW between 0C and +50C. In the
component described, mirrors with 54 or 34 layer pairs are used. The active zone
consists of 4 tension-strained GaInP quantum wells and AlInGaP barriers. The
lateral current is confined with etched mesas (49 m diameter) with an oxidized
aperture (15 m).
The change in power of 2%/K is relatively high when heated. For a 10 m
aperture diameter, up to 2 mW is attained. Laser operation is possible up to
+60C. At 13 m aperture, the threshold current is 1.9 mA.
10 m
Fig. 4.46: VCSEL detail photo (Univ. of Stuttgart)
4.5
Popt [mW]
4.0 0C
3.5 cw operation
O = 670 nm 10C
3.0 7 m VCSEL
20C
2.5
30C
2.0
1.5 40C
1.0
60C 50C
0.5
70C I [mA]
0.0
0 2 4 6 8 10 12
Fig. 4.47: P-I-characteristics of a 670 nm VCSEL ([Schw03b])
12
11 Pmax (20C) [mW]
10
9
8
7
6
5
4
3
2
1
wavelength [nm]
0
630 640 650 660 670 680
Fig. 4.48: Maximum CW output power for red VCSEL at 20C
That the poor thermal conductivity is indeed responsible for the diminishing
efficiency is proven by the measurement of the P-I characteristic curve with a
small pulse duty factor. Laser activity of the 670 nm VCSEL is attained here up to
a temperature of +150C in the active zone (Fig. 4.49).
Usable 650 nm VCSELs would be the ideal source, especially for PMMA
GI-POF. The results just shown reduce the problem only to efficient heat dissi-
pation out of the active zone. A number of promising methods are currently under
development.
332 4.2 Transmitting Diodes for Data Communication
Popt [mW]
0.30
23C
pulse: 0.3 s (200 Hz)
0.25 43C 8.6 m VCSEL, 670 nm
0.20
0.15 144C
107C
0.10
0.05 154C
0.00
0 20 40 60 80 100 120 140
Fig. 4.49: P-I-characteristics for pulsed operation ([Schw03b])
The easiest method for reducing the temperature rise in the active zone during
laser operation is to minimize the size. However, the attainable output power
drops, too. The P-I characteristic curves of two 650 nm VCSELs with 20 m and
7 m large apertures are shown in Fig. 4.50. Whereas the output power drops to
about 40%, the maximum temperature for laser operation rises to +65 C.
0.8
Popt [mW]
0.7
aperture: 20 m aperture: 7 m
0.6 0 0C .. +65C
0.5
0.4
0.3
0.2
0.1
0.0
0 5 10 15 20 25 0 5 10 15
I [mA] I [mA]
Fig. 4.50: Comparison between two 650 nm VCSEL with different apertures
The authors have come to the conclusion that possible parameters for red
VCSELs are:
laser operation up to 110 C
up to 10 mW output power at room temperature
modulation bandwidths up to 10 GHz
4.2 Transmitting Diodes for Data Communication 333
Figure 4.51 summarizes the results achieved so far for maximum laser tempe-
rature and emission wavelength.
Tmax [C]
90
target
80
2000
70 2003
2006
2003
60 2000
50 1999 2000
2003 2000
40
center wavelength [nm]
30
645 650 655 660 665 670 675 680 685 690
Fig. 4.51: Overview of red VCSEL, presented up to now
VCSEL at 850 nm
VCSELs with emission wavelengths of 850 nm can be used for relatively high
ambient temperatures. A 850 nm VCSEL is presented for example in [Schn03]
which shows a laser emission up to +145C, the P-I characteristic curve is in
Fig. 4.52.
In this wavelength range VCSELs can even be used without any power control
over wide temperature ranges. The PMMA POF has losses of about 3 dB/m at
850 nm. The use of 850 nm sources nevertheless makes sense for example for
connections between computer components or in mass storage. Even at 780 nm
very good VCSELs are available. In this case PMMA fibers even permit lengths
up to 30 m.
334 4.2 Transmitting Diodes for Data Communication
VCSEL at 782 nm
In [Ueki99] a 782 nm VCSEL is introduced with a maximum output power of
3.4 mW at 10 mA. The threshold current is 0.61 mA (each at +20C). At +60C,
2 mW of optical output power is still attained. Longer wavelength VCSEL are
thus also suited for considerably higher temperatures.
The upper mirror of the VCSEL consists of 24 pairs, the lower of 40 pairs
Al0.3Ga0.7As/Al0.9Ga0.1As, with a distance of one wavelength. The active layer
contains three quantum wells made of Al0.12Ga0.88As/Al0.3Ga0.7As.
NRC-LED at 850 nm
To date (year 2000), non-resonant cavity LED have been described for wave-
lengths around 850 nm only. [Roo00] introduces an array of 850 nm NRC-LED
with an active diameter of 30 m and a distance of 100 m. This enables the
transmission of 1 Gbit/s across 10 cm image-guiding fibers (7 m single-fiber dia-
meter).
A series of studies on NRC-LED were published by Windisch. An efficiency of
up to 31% is cited in [Win99] for 870 nm NRC-LED. The roughened surface leads
to a more diffuse scattering and improvement in efficiency. A mirror at the rear
improves the efficiency. Further improvements of up to a 40% external quantum
efficiency for an 870 nm NRC-LED are described in [Win00a], [Win00b] and
[Win00c]. Various parameters were attained with various active layer thicknesses
(10 nm, 20 nm and 30 nm) and various diameters of the etched mesa diodes
(30 m and 45 m) as shown in Table 4.3.
4.2 Transmitting Diodes for Data Communication 335
When using a NRC-LED with 20% efficiency, 1,200 Mbit/s can be transferred.
A 100 m glass ball lens can be used to improve the coupling of the 30 m LED
on a POF. Thus the launch efficiency in a POF with AN = 0.50 attains a value of
50%. Figure 4.53 shows the relationship between the bit rate attained and efficien-
cy. A high efficiency is indeed achieved with larger active diameters, but the bit
rate is somewhat smaller. Nevertheless, 1.25 Gbit/s can be attained with NRC
LEDs with 20% external quantum efficiency.
10
0
0.4 0.8 1.2 1.6 2.0
maximum bit rate [Gbit/s]
NRC-LED at 650 nm
The first red NRC-LEDs were introduced in [Roo01]. The InGaP/AlInGaP LED at
650 nm attain an external quantum power efficiency of 31%, whereas the red
LEDs up till then had only attained a maximum of 12%. For example, output
power of 4 mA is attained with a current of 7 mA. The spectrum of NRC-LEDs is
illustrated in Fig. 4.54.
336 4.2 Transmitting Diodes for Data Communication
1.0
rel. opt. power
0.8
0.6
0.4
0.2
wavelength [nm]
0.0
600 620 640 660 680 700 720
A simple method for increasing the coupling-out power efficiency of a LED was
introduced in [Kra99], [Lew99] and [Hr03]. Normally, a semiconductor has a
refractive index of about 3.5. Total reflection already occurs above an angle of
17. If the LED chip is as usual cubical, then the angle does not change even with
reflection. Barely 3% of the emitted light can leave the LED through the upper
boundary layer. By means of a transparent substrate with mirror coating and
receptacle with a suitable funnel-shaped LED the light from all 6 boundary layers
can be used which can increase the power efficiency to a good 15%.
substrate
n-contact
n-GaP
The pyramid LEDs make use of this effect just like the NRC-LEDs, namely
changing the direction of light upon reflection. Intentionally slanted inner sides are
produced instead of a roughened-up surface. The principle is demonstrated in
Fig. 4.55 according to [Li05].
An external quantum power efficiency of 55% can be attained for red LEDs.
For blue LEDs 37% can be attained. Some examples of such pyramid LEDs are
shown in Fig. 4.56 from [Lew99] and [Hr03].
To create efficient data transmission with a particular POF, sources will always
first be sought whose emission wavelengths match the respective minimum
attenuation values. For PMMA fibers, these are the ranges near 520 nm, 570 nm
and 650 nm.
In Fig. 4.57, the external efficiencies (data from the current datasheets of
various manufacturers) are listed for available LED. The different materials are
marked separately.
100%
external quantum GaAlAs AlInGaP GaAs
efficiency GaAsP GaN InGaN
GaAlP GaP SiC
10%
1.00%
0.10%
wavelength [nm]
0.01%
420 440 460 480 500 520 540 560 580 600 620 640 660 680
Fig. 4.57: Efficiency of various LED material systems (from datasheet information)
338 4.4 Receivers
Until the GaN/InGaN technology had been developed, the efficiency of LED in
the direction of the short wavelength spectrum decreased markedly. Only SiC-
LED were available in the blue range; however, these were expensive and ineffi-
cient. In the mean time, very efficient LED have become available in the 370 nm
through 540 nm range. For 560 nm too, samples have been produced. In the next
few years, it is expected that the current gap will be closed into the red range so
that sources with an efficiency >10% will be available for the entire visible spec-
tral range.
Apart from efficiency, it is obvious that the possible modulation speed also
plays a decisive role. Figure 4.58 contains the quoted switching times for some
LED from Fig. 4.57.
300
100
30
10
wavelength [nm]
3
460 480 500 520 540 560 580 600 620 640 660
Fig. 4.58: Switching times of various LED material systems
Diodes that have a lower efficiency usually exhibit very low switching speeds.
This is particularly true of green LEDs in conventional technology. Diodes based
on GaN are very well suited for data transmission due to their high switching
speeds.
The various diode constructions such as lasers, VCSEL, RC-LED, or NRC-
LED have until now only been realized in the red spectral range. Only LED are
available in the yellow and green POF attenuation window. This situation will
probably change in the next few years.
4.4 Receivers
The smaller the light wavelength, the less the photo current per watt of optical
power emerges just because each photon possesses more energy. This may first
seem to be a paradox but you must keep in mind that a short-wavelength LED
with the same power efficiency emits more optical power than long-wavelength
diode. Consequently, a wavelength-independent relationship arises between the
driving current of the LED and the photo current of the photodiode.
0.8
responsivity [A/W] Kext. = 1.00
0.7
0.6
850 nm:
0.5 = 0.63 A/W
Kext. = 0.92
0.4 650 nm:
0.3 = 0.47 A/W
Kext. = 0.89
0.2 500 nm:
= 0.32 A/W
0.1 Kext. = 0.79
wavelength [nm]
0.0
400 500 600 700 800 900 1000 1100
Fig. 4.59: Responsivity of a Si-pin photodiode (Hamamatsu S6801)
340 4.4 Receivers
In the data sheets for photodiodes there is information about the responsivity ,
but seldom data about the quantum efficiency. This responsivity parameter des-
cribes the emerging photo current per light power in A/W or also in mA/mW. An
example of the length-dependent responsivity is shown in Fig. 4.59 (from Hama-
matsu data sheet).
At about 950 nm this diode has the highest responsivity which is typical for
silicon photodiodes, but even at 650 nm the quantum efficiency is close to 90%.
Above 1,000 nm the efficiency quickly falls back to zero since the energy of the
photons is too small.
In addition to the actual efficiency there is another important parameter, namely
the absorption length (depth of light penetration), which becomes larger and larger
at longer wavelengths. If it exceeds the thickness of the absorbing layer, then the
power efficiency drops because the light passes right through the photodiode. The
dependence of the absorption length on the wavelength for different semiconduc-
tor materials is shown in Fig. 4.60.
102
Si
101
GaAs
100
10-1
In0.7Ga0.3 In0.53Ga0.47As
Ge
As0.64P0.36
10-2
0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
O [m]
Fig. 4.60: Absorption length for different semiconductor materials
Different photodiode structures are employed in this technology. The three most
important versions are:
4.4 Receivers 341
Si-nitrid
metal contact
passivation
p-InP
i-InGaAs
absorbing
layer
n-InP
metal contact
The advantage of a pin-diode lies in its easy use. It only requires a bias voltage,
typically 5 V to 15 V. For an APD, the bias voltage has to be adjusted in such a
way that the optimum gain factor is attained. The necessary bias voltages can
reach some 100 V and are dependent on the temperature and the power (controls
are necessary). Furthermore, APDs are considerably more expensive. All commer-
cial POF systems work with pin-photodiodes.
The contact structure of an MSM photodiode is shown in Fig. 4.63. The metal
surfaces cause partial shading so that the efficiency is diminished. However, since
the capacity of the diode can be much smaller than with a PIN diode of the same
size, a greater transimpedance can then be used which improves the sensitivity.
342 4.4 Receivers
For very fast MSM PDs, i.e. up to 30 GHz, finger spacing in the area of 1 m is
necessary, whereby polarization dependencies do arise. MSM-PDs have not yet
been employed commercially for POF systems, but should be in use in a few
years.
A pin-photodiode, just like any other semiconductor diode, has a junction capacity
which depends on the reverse voltage applied. A typical example is shown in
Fig. 4.64. This capacity is much larger than with diodes for glass fiber systems - at
least above 3 pF with diodes up to a diameter of 800 m - because of the large
photodiode surfaces required.
7
capacity [pF]
6
Together with the input resistance of the following stage this diode capacity
forms a low-pass which as a rule limits the entire bandwidth of the receiver. In
order to get a good signal-to-noise ratio, the impedance of the receiver should be
as high as possible. A low diode capacity is therefore directly responsible for good
receiver sensitivity.
Consequently, for a long time it was supposed that the possible bit rate for
1 mm fibers was limited to a maximum of 150 Mbit/s. In the meantime, however,
different laboratories have attained up to 2,500 Mbit/s with pin-photodiodes ran-
ging in diameter from 600 m to 800 m (see Chap. 6). Even considerably higher
bit rates should also be possible with multi-level transmission, adaptive equali-
zation and multi-carrier transmission. Optimization of the fiber coupling to the
photodiode promises further improvements.
There is as a rule very little mention of the construction of receivers in the various
publications on POF systems. Many institutes use commercial receivers or at least
amplifiers. Low-noise transimpedance amplifiers are available for almost every bit
rate range. The greatest disadvantage is that these commercial components have
been designed for a capacity of only a few tenths pF. If you couple photodiodes
with considerably greater capacity, the bandwidth and sensitivity drop dramati-
cally. Details on receiver circuits, to the extent that there is information on them,
are listed in Chapter 6 on System Overview.
344 4.4 Receivers
An overview of the sensitivities attained so far with different bit rates as well as
for different fiber diameters and various wavelengths is illustrated in Fig. 4.65.
-8
-10 sensitivity [dBm] POF:
-12 1.0 mm
-14 0.5 mm
-16 0.2 mm
-18
wavelength
-20
-22 470 nm
-24 650 nm
-26
850 nm
-28
1300 nm
-30
-32 bit rate [Mbit/s]
-34
100 200 500 1,000 2,000 5,000 10,000
Fig. 4.65: Parameters of POF receivers up to now
The relationship between bit rate and sensitivity attained can clearly be seen. A
ten-fold increase in the bit rate costs about 15 dB sensitivity because of the greater
noise bandwidth as well as by the necessary reduction in the receiver resistance.
Data rates above 3 Gbit/s have only been attained so far with relatively thin fibers.
As is the case with transmitters, there are numerous technical starting points for
greatly improving the parameters of receivers which should be implemented when
large-scale use begins.
Putting together a nearly complete list of available types of diodes and receivers
would go far beyond the scope of this book which is why we have not included
such a compilation. This is the reason why we wish to present only two exemplary
photodiodes which have been used for POF systems for years by various
institutes.
Fig. 4.66: pin-diode SFH 250 by Infineon for 1.0 mm/2.2 mm POF (right with a housing
for clamping the POF with a screw)
4.4 Receivers 345
As early as 10 years ago Siemens - now Infineon - developed the type SFH 250
in Fig. 4.66. This Si-pin-PD has about a 1 mm diameter. The plastic housing has a
2.2 mm bore so that a sheathed 1 mm POF can be directly centered. The diode is
then ideally suited for systems without plug connectors.
The important parameters for this diode are summarized in Table 4.5 (data from
[Inf03]). Data rates up to 250 Mbit/s have been realized with different set-ups.
This diode has also been used successfully for analog signals.
Another readily used diode is the S5052 from Hamamatsu. The 800 m large
chip is encapsulated in a plastic housing with a 3 mm large lens which guarantees
an optimum coupling to 1 mm thick fibers. The diode was operated at data rates
up to 2,500 Mbit/s at the POF-AC. The important parameters are listed in
Table 4.4 [Ham01].
Receivers for polymer fibers are hardly obtainable as a single unit; they are
mostly sold as a system together with the corresponding transmitters. Data on the
components available have been summarized in the chapter after next.
346 4.4 Receivers
As already mentioned above, many receivers use a lens to improve the sensitivity.
If the photodiode is designed smaller than the fiber, you can have a higher trans-
impedance, but you lose more sensitivity because of the lost light. With the use of
an appropriate lens a large part of the light guided in the fiber can be concentrated
onto a smaller photodiode. Because of the POFs large NA this works up to a
reduction factor of 2 since anything above that would mean that the angle which
appears would be too large for the lens and for the photodiode. In Fig. 4.67 from
[Har01] the possible coupling power efficiency form a 1 mm standard POF (UMD
launch) is shown in different-sized photodiodes with spherical lenses.
70
coupling efficiency [%]
60
50 PD [m]
40 600
500
30
400
20 300
200
10
0
0 200 400 600 800 1000 1200
distance POF lens [m]
Fig. 4.67: Coupling efficiency POF-PD with spherical lenses according to [Har01]
Using a 500 m large photodiode, 50% of the light could be coupled in a pho-
todiode with a correctly positioned spherical lens. Even if you could now select a
four-fold greater load resistance at the receiver - because of the smaller capacity -
there would still not be any gain in the SNR:
loss through coupling: 3 dB
gain through greater transimpedance: 6 dB
loss through greater noise: 3 dB
On the other hand, higher data rates can generally be attained with the smaller
photodiode so that it is worth using.
Optical concentrators as non-imaging elements permit a more efficient coupling
from the fiber into the photodiode as described in [Poi04a]. There are versions
with mirrored surfaces as well as with wave guiding through total reflection. One
version from [Ueh02b], [Ueh03] and [Mat02b] is shown in Fig. 4.68.
4.5 Transceivers 347
Table 4.7 from [Poi4a] shows the theoretical power efficiency of different ver-
sions for the coupling of a 1 mm POF (UMD launch) to photodiodes of different
sizes as compared with direct butt joints.
The dielectric taper (reflection against air) attains the best power efficiency
since total reflection is more efficient than metallic reflection. Theoretically, 75%
of the light can be coupled (with a loss of 1.25 dB) in using a photodiode which is
only 400 m large (only 16% of the capacity). Without question future POF
systems will increasingly use these methods since the concentrators can be mass
produced very easily.
4.5 Transceivers
In the past five years there have been diverse developments particularly in the
field of POF transceivers. We were hardly able to present complete systems in the
first edition of this book. The components then were sold only by small producers
and for niche applications. Wiesemann & Theis have developed converters for
RS232 or 10BaseT to duplex-POF. Figure 4.69 shows both components. 594 nm
(T2P) or 650 nm LEDs are used as the sources (see also [Leh00]).
348 4.5 Transceivers
Fig. 4.69: RS232 (left) and 10 Mbit/s Ethernet transceiver (right) by W&T
Fig. 4.70: ATMF-compatible transceivers for 100BaseT and 155 Mbit/s ATM
(from left to right: Hewlett Packard, NEC, Univ. Ulm)
All three components are equipped with a 1 u 9-pin contact and can be plugged
into conventional PC cards or LAN components. Test networks were created for
this purpose at the University of Ulm ([Som98a]) and at Deutsche Telekom in
Berlin ([Lei98]). Table 4.8 summarizes the most important properties (according
to [HP01], [HP02], [NEC1] and the results of a research project with the
University of Ulm).
4.5 Transceivers 349
All three transceivers are based on the ATMF specification. Figure 4.71 shows
the temperature-dependent spectrum of the NL-2100 transceiver. It has virtually
all the properties of the HP transceiver. Neither HP nor NEC provides current
information on possible new product developments. This also applies to the trans-
ceiver provided by NEC as a prototype for data rates of up to 250 Mbit/s.
1.0
Popt rel. 24C
30C
0.8
40C
50C
0.6
60C
70C
0.4
0.2
0.0
590 600 610 620 630 640 650 660 670 680 690
wavelength [nm]
In the last few years Fast Ethernet (125 Mbit/s physical data rate) has increasingly
been used in automation instead of the conventional field bus systems. A number
of components have been developed for this field. In addition, there are products
which have been specifically developed for the home.
In the first half of 2006 alone there were half a dozen new manufacturers. That
is why the following summary can only be a short-lived snapshot and not a de-
tailed listing of parameters and company addresses. However, this information can
be found in the POF-Atlas (www.pofatlas.de).
DieMount is the only manufacturer which sells transceivers with blue and red
LEDs. A WDM version has also been developed for bi-directional transmission.
The simplex components achieve ranges of 70 m or 30 m at wavelengths of
470 nm or 645 nm respectively. As duplex versions each is capable of a trans-
mission length of over 100 m. Different versions of the Fast Ethernet transceivers
from DieMount are illustrated in Fig. 4.73. In the meantime these products are
also being sold under other brand names by large electronic companies. In addi-
tion to the external media converters presented here PC plug-in boards and comp-
lete switches are also sold.
Versions for Fast Ethernet with a red LED and a range of 70 m and for
10 Mbit/s Ethernet with 520 nm LED and up to 200 m transmission length are
available as media converters. The components can be seen in Fig. 4.75 ([Thi04]).
Fig. 4.75: Media converters for 10 und 100 Mbit/s from Ratioplast
Fig. 4.77: SC-RJ connector and Fast-Ethernet transceiver (left) and RCC-45 hybrid con-
nector with data cable and duplex POF from Reichle & DeMassari
Infineon expects strong growth in the POF market in the next few years for
which the introduction of VDSL and the increase in IPTV products are respon-
sible. (The estimated growth rate for IPTV-capable connections will be 92% annu-
ally.) Infineon gives a price of $12 USD for the complete electronic components
(POF transceiver and ADM6992SX Chip when mass produced (Fig. 4.79 from
[Inf06]).
Since October 2006 the system, now under the Siemens label, can be ordered
from the website of the Deutsche Telekom. Included in the set (Fig. 4.80) are
copper cables, power supplies, a cutting tool and 30 m of duplex POF (with 1 mm
core diameter and 1.5 mm jacket) in addition to the two media converters
(Fig. 4.81).
Fig. 4.80: Siemens Gigaset Optical LAN-Adapter, offered by the Deutsche Telekom (right:
media converter for Fast-Ethernet)
POF transceivers are also sold for other different interfaces in addition to the Fast
Ethernet components. Particularly in the Asian countries house networking is not
carried out in pure IP networks, but in accordance with the IEEE 1394 standard.
This standard permits real-time transmission with guaranteed data rates and maxi-
mum delay times in contrast to IP and has been designed for operation without a
main server. The components are still somewhat more expensive today because of
low-volume production. Whether Ethernet or IEEE 1394 is used plays no role
whatsoever for POF and the transceivers. Both systems can work up to 100 Mbit/s
and use 4B5B coding.
4.5.3.1 Comoss
The Taiwanese manufacturer Comoss sells a complete assortment of POF compo-
nents for IEEE 1394 systems on the basis of the SMI connector (Fig. 4.81). Media
converters as well as transceivers with red LEDs with a range up to 50 m and data
rates of S100 and S200 are available.
4.5 Transceivers 355
Fig. 4.82: IEEE 1394 S200 transceiver with SMI connector from Firecomms
The transceivers designed for use with MOST have been certified in the mean-
time and exceed, for example, the required power budget by 8 dB. The switching
times of 4.2 ns make it possible to use them up to 250 Mbit/s. The temperature
range can also be enlarged. A POF version without connectors is also available
from Firecomms (Fig. 4.83). Table 4.10 summarizes some parameters of the
Firecomms transceivers.
Fig. 4.83: IEEE 1394 transceiver OptoLock from Firecomms without connector
356 4.5 Transceivers
Table 4.10: Parameters of Firecomms POF products (typical values from data sheets)
Fig. 4.84: POF transceiver Hitachi DC9500 from 1999 (left), Fast Ethernet POF media
converter Mitsubishi OMCP-ETH100SA for S200 (middle and right)
Fig. 4.85: Media converters for 10/100 Mbit/s Ethernet and RS232 from Luceat
A big problem in present-day POF applications is the lack of devices with their
own POF interfaces, thus requiring the use of external media converters and
additional power supplies (also see Chap. 8).
Netopia is the first manufacturer to sell a DSL modem with built-in POF
interface (Fig. 4.88, transceiver from Firecomms, [OTS06c]).
As you can see with the losses stated in dB/cm, the transparency of these struc-
tures is not comparable to that of optical fibers. The main reasons for the very
much greater optical losses lie above all in the non-cylindrical waveguide geome-
try and in the poorer surface quality of the core-cladding-interface layer. If you
take into consideration, however, that only a few centimeters in length are needed
for typical applications, then the losses in the magnitude shown are justifiable.
PMMA at 1.3 m and 1.55 m is completely useless for optical fibers since the
attenuation is too high. However, it can be used as waveguide material since the
demands on the material here are not too great. Nevertheless, materials with lower
attenuation, especially at 1.55 m, are being searched for. As with fibers, an effec-
tive variation is the use of partially or completely fluorinated polymers. CYTOP
material also can be employed here. The positions of vibration bands for different
carbon compounds are shown in Fig. 5.1 (detailed description in [Gra99]). The
fluorinated compounds have the lowest vibration frequency - in the mid-infrared
range.
In addition to the relatively high attenuation, another inhibiting factor is the
lack of thermal endurance of the PMMA. For many polymer materials with high
Tg, the use as fiber materials fails because of the losses. However, it can make
sense to use them in waveguide structures.
5.2 Production of Polymer Waveguides 361
-C C-
-CH2-
-C N
C-Br
-CH3
-NH2
C=O
C=N
C-Cl
CH
-O H
C-C
-CN
-SH
C-S
C-F
=CH2
C-C
-CH3
C-N
C-O
-CH2-
2 3 4 5 6 7 8 9 10 20
wavelength [m]
This work describes, for example, the production of waveguides from aqueous
synthesized composite materials - combination of organic and inorganic compo-
nents (Fig. 5.2). The stamp used can be made of silicon or silica - produced by
anisotropic etching - or from nickel (formed galvanically).
removing of
the cover glass
deforming depositing of the
covering layer
waveguide
Fig. 5.2: Schematic sequence of waveguide production from aqueous synthesized com-
posite ([Gra99])
Table 5.4: Overview of the optical characteristics and production technology of passive
optical light waveguides ([Gra99])
UV-light mask
photoinitiator, monomer
cladding
substrate
cladding
waveguide
cladding
With this method a substrate is coated with a mixture of a monomer and a pho-
tosensitive material. By means of a mask a selective change in the refractive index
of the waveguide layer has brought about by inserting the initiators in the polymer
matrix. After exposure the remaining volatile molecules are removed from the un-
exposed areas through annealing. After another cladding layer is put on the wave-
guide is formed.
This method is used commercially for the production of multimode waveguides
under the brand name Polyguide from DuPont. PMMA can serve as a substrate.
The photoinitiators are, e.g. ketones and benzoins. The possible refractive index
changes lie at 0.001 to 0.010. According to [Keil96b] this method is used for
thermo-optical switches, whereby the waveguides show losses of 0.3 dB/cm (at
1.3 m) and 0.8 dB/cm (at 1.55 m).
Another method for producing waveguides mentioned in [Hen04] is selective
polymerization. In this case the waveguide is formed through photochemically
initiated polymerization. However, the unexposed parts are subsequently removed.
With ormocers, an inorganic-organic hybrid polymer of the Fraunhofer Gesell-
schaft, losses of 0.32 dB/cm (at 1.32 m) and 0.66 dB/cm (at 1.55 m) were
achieved.
On the other hand with photo bleaching the refractive index outside the later
waveguides is lowered through a photochemical reaction. The non-linear charac-
teristics of different dyes are used here. Waveguide attenuations of 0.8 dB/cm at
1.31 m were reached.
A method for the production of waveguides in a kind of printing process is
described in [Kal03b], whereby the polymer is melted and then applied directly
onto the carrier by means of an approximately 10 m thick micropipette. After-
wards the hardening is effected by means of UV light. A second polymer serves as
the upper cladding material. Single-mode waveguides can successfully be pro-
duced with a cross-section of about 16 u 0.8 m.
The production of waveguides by UV-induced refractive index changes is des-
cribed in [Bru06]. Through radiation with light in wavelengths of 200 nm -
260 nm 7.5 m wide waveguides are written into PMMA. At a wavelength of
1.55 m the waveguides achieve losses below 1 dB/cm. Different kinds of coup-
lers were also realized.
reduce coupling losses, the cross-section of the waveguides must either be fit to or
so-called tapers for transforming the mode field have to be used.
Not all production techniques are equally suitable for singlemode and multi-
mode waveguides. The casting methods in particular are more advantageous for
thicker structures.
The production of waveguides in a sensor application (interferometry) is des-
cribed in [Kor04], whereby a grooved structure is written into the glass through
laser ablatation (UV-excimer laser). The groove is subsequently filled with a
highly viscous polymer as the waveguide core. A second technique is based on the
UV-polymerization of thin layers, whereby the ridge waveguides produced are
then surrounded by a second material with a lower refractive index. The third
method investigated produced linear waveguides by modifying the refractive
index by means of a fs-pulse laser. Details of the waveguide structures are shown
in Fig. 5.4.
covering layer
core
d h
SiO2-substrate
w
The losses of both waveguides lie at 0.42 and 0.46 dB/cm respectively (at a
wavelength of 1.55 m).
Particularly temperature-stable polymer waveguides are also described in
[Kang02]. Fluorinated poly(arylene ether sulfide) (FPAESI) and fluorinated poly
(arylene ether sulfide fluorene) (FPAESF) serve as materials. For singlemode
waveguides with a cross-section of 6 u 7 m2 losses of 0.4 dB/cm at 1.55 m were
reached. Ageing over 1,000 h at +100C did not result in any increase in the
attenuation.
Another method for the production of polymer waveguides is described in
[Sum04]. Again a sheet of glass is used as a substrate. The core is made of PMMA
or SU-8 while NOA-88 is used for the cladding. The waveguide structures are
written with a proton beam (2 MeV, 2 pA proton current 1.875 1013 protons/cm2).
The production process is illustrated in Fig. 5.6 (waveguide cross-section is
5 u 5 m).
irradiation of SU-8 by 2
MeV-protons
chemical developing
and removing of the
non irradiated areas
SU-8 layer
glass substrate
glass covering
NOA 88
glass substrate
Fig. 5.6: Production of waveguide structures ([Sum04])
In addition to linear waveguides Y-splitters are also realized. The setup of such
a splitter is shown in Fig. 5.8. The spectral transmission for a 2 mm long wave-
guide is indicated in Fig. 5.9.
190 m
linear taper linear
cosinus region
linear S-bend
region
According to the information for the particle stream and the radiation perfor-
mance used (12 mill. particles/second) the writing of a coupler may have taken
several minutes. The power output of both arms lay in a ratio of 46 : 54.
transmission [%]
100
80
60
40
20
wavelength [nm]
0
400 600 800 1200 1400 1600
Fig. 5.9: Spectral attenuation of the cross-linked SU-8 (2 mm, [Sum04])
368 5.4 Multimode Waveguides
Multi-mode polymer waveguides, just like multimode fibers, have the disadvan-
tage of a limited bandwidth due to mode dispersion. The decisive factor is once
again the numerical aperture. However, since optic waveguides are typically used
in the length range below one meter, bit rates of many Gbit/s can be transmitted
without any problem. Measurements of the bandwidth are also hardly mentioned
in the technical literature since normally the values lie far outside the measure-
ment possibilities. In demonstrators bit rates up to about 10 Gbit/s have so far
been realized.
The main reason for the use of thick waveguides lies in the tolerances (see also
Section 5.6). When they are integrated into electrical printed circuit boards, con-
ventional automatic placement machines should be used for passive placement on
the optical components, whereby waveguide cross-sections of at least 50 u 50 m
are necessary.
Concepts for planar multimode waveguides, for example, were introduced in
[Schm00]. In this case TOPAS 6017, APECHAT 9371 and PMMI8817 were
used as materials. The waveguide cross-section is about 100 m u 250 m.
Nothing has yet been said about the attenuation.
While singlemode waveguides are primarily used as functional components in
classic glass fiber systems, among other things as couplers, filters, switches or also
as possible amplifiers, multimode waveguides serve above all as high-bit rate data
channels. Production techniques are primarily casting or the photochemical struc-
turing of thin layers. Fig. 5.10 from [Schr02] shows the most important production
methods for multimode polymer waveguides.
Fig. 5.10: Overview of the production methods for polymer waveguides ([Schr02],
[Sche05])
5.4 Multimode Waveguides 369
This work [Schr02] also describes the setup of a complete demonstrator as part
of the BMBF project EOCB (Electro-Optical Circuit Board). Up to 1.25 Gbit/s
were transmitted over each of 4 parallel channels with a spacing of 250 m
between each (50 m cross-section).
The cross-sections of polymer multimode waveguides are mostly square or tra-
pezoidal. In calculations made in the C-Lab Paderborn additional losses of this
geometry are calculated ([Bie02]). The advantage of trapezoidal-shaped structures
lies in the easier ability to cast and the resulting smaller roughness. For a wave-
guide of 30 cm in length additional losses of 0.38 dB/cm and 0.66 dB/cm at edge
angles of 5 and 10 respectively result from simulations. This can normally be
ignored in regard to losses because of the surface roughness.
The use of silicones for multimode waveguides is described in [Ney05]. The
high temperature stability of this group of materials simplifies above all the lami-
nating into printed circuit boards. At 850 nm losses are only 0.03 dB/cm - compa-
rable to the attenuation of a PMMA-POF of about 3000 dB/km at 850 nm.
Table 5.5 shows the parameters of the polymers for waveguides described thus far.
Table 5.5: Overview of polymers for waveguides in printed circuit boards ([Ney05])
0.10
attenuation [dB/cm] before 5 min/220C
2 h/180C 5 min/260C
0.08
0.06
0.04
0.02
0.00
1 2 3 4 5 6 7 8 9 10
waveguide number
Fig. 5.11: Temperature behavior of silicone waveguides (polyimide embedding)
depositing of
the core
polymere
preform
squeegee
curing
preform WG-cores
preform substrate
deforming
superstrate-carrier superstrate
grouting,
curing
superstrate polymer
A number of studies have been carried out for constructing components for sys-
tems with singlemode glass fibers. A few examples we wish to mention are the
works of the Heinrich Hertz Institute (HHI) in Berlin ([Keil96], [Keil97],
[Keil99], [Keil05]). Singlemode waveguides on a polymer basis have a cross-
section of about 10 u 10 m. One of their advantages is the fact that they can
easily be produced through casting. The great dependence of the refractive index
on temperature can be bothersome, but also advantageous for the production of
switches and the fact that tunable filters can be fabricated.
Figure 5.13 shows the schematic setup of a thermo-optical switch with polymer
waveguides. Heating electrodes are located over the core areas. The refractive
index can be changed in such a way through heating that the power only leaves
from one output because of the altered wave propagation.
heating electrodes
teflon layer
PMMA waveguide
cladding material
SiOx
Si carrier
heating electrodes
PMMA waveguide
In the HHI different switches have been developed and presented in the confi-
gurations 1 u 2, 2 u 2, 1 u 4 and 4 u 4 with crosstalk lying below -30 dB. The
attenuation of the waveguides is about 0.7 dB/cm so that the insertion losses of the
complete switches are very small. One advantage of this technology is the very
small necessary switching power of a few 10 mW. The switching speed lies in the
range of a few milliseconds.
The dependence of transmission of such a switch on the applied heat power is
shown in Fig. 5.14 from [Keil05].
372 5.5 Functional Components as Waveguides
transmission [dB]
0
-10
-20
-30
-40
direction
-50
P1 S1
P1 S2
-60
P2 S1
-70 P2 S2
Fig. 5.15: Four-fold thermo-optical switches made of polymer waveguides (HHI 2005)
5.5 Functional Components as Waveguides 373
5.5.2 Modulators
waveguides of the
interferometer
HV
The waveguides in the picture are each 2 m wide and 0.9 m thick with
spacing of 4 to 6 m. The losses of the waveguides lie at 0.43 - 1.22 dB/cm and
the uniformity of the splitter outputs at 0.23 - 1.30 dB.
The work previously cited ([Hen04]) also describes the production of couplers.
Fig. 5.18 shows such a splitter component. A best value of 0.9 dB/cm (at 1.55m)
was achieved as attenuation for the waveguides. The insertion loss of the y-coup-
ler amounts to about 6.5 dB which means an additional attenuation of about 1 dB
compared to the waveguide losses.
For some years now work has been carried on in DaimlerChryslers research cen-
ter on the development of computer backplanes in combination with electrical and
optical lines (see e.g. [Gut99], [Moi00a], [Moi00b], [Moi00c], [Rode97],
[Mon00], and [Kru00]. This work has been motivated by the fact that an ever
greater number of data connections in PCs can be found on wider and wider
busses. Optical solutions can prevent the problems of crosstalk considerably. Fi-
gure 5.20 shows the principal setup of a PC with an electrical - optical backplane.
A detailed illustration in the following Fig. 5.20 shows how the plug-in boards can
be coupled into the backplane without contact. Using lenses, a collimated beam is
generated which is then again focused onto the waveguide, whereby the distance
between the backplane and the board is mostly uncritical.
376 5.6 Waveguides as Interconnection Solutions
backplane with
combined terminals
optical
lines
electrical
lines
plug-in cards
In later works waveguides with a 200 u 200 m cross-section have been pre-
sented. Wavelengths between 650 and 850 nm had losses lower than 3 dB/m. The
additional attenuation is about 1 dB/cm for a bending radius of 15 mm. The nume-
rical aperture of the waveguide is 0.35. The usable temperature range lies between
-40C and +85C. For waveguides with a length of 55 cm the complete attenuation
only amounted to 2.5 dB. In addition to straight and bent waveguides, crossings
and couplers can also be realized so that complete optical networks can be set up.
By using VCSEL and MSM photodiodes (diameter of 300 m) the sensitivity can
be improved to -20 dBm at 2.5 Gbit/s. Figure 5.22 shows the design for a
56 u 1 Gbit/s connection [Moi00c].
Fig. 5.22: Example for the application of an optical backplane with 56 channels
The following Figures 5.23 to 5.25 demonstrate further details for realizing
optical backplanes. First of all Fig. 5.23 shows a carrier with different waveguides
(straight and as a 4 u 4 network).
Figure 5.24 shows the setup of a single transceiver with LD, PD and the lenses
for the coupling (on the left). On the right you can see the photos for combining
several transceivers as well as the individual lenses.
Fig. 5.25 depicts a bent polymer waveguide with a rectangular cross-section
(the laser cannot be seen). Such components are above all useful for the routing of
light.
378 5.6 Waveguides as Interconnection Solutions
The license to produce optical backplanes was granted to ERNI and Varioprint
as cooperation partners from 2002 on ([Ern02]).
Fig. 5.28: Optical polymer waveguides written by laser and produced photolithographi-
cally ([Gri06])
380 5.6 Waveguides as Interconnection Solutions
Other methods include pressing the waveguides into copper (0.1 dB/cm) and
the production method mentioned above by means of photolithographic processes
(0.1 dB/cm).
In addition to the waveguides themselves the coupling of active components
also plays an important role. A concept is presented in [Gri06] in which light is
coupled from the waveguides to the active components by a 45 mirror. The
connection can be made by using plugs. Reasonable demands on stability and tole-
rances can only be attained by using multimode waveguides.
Fig. 5.30: Eye diagram of the transmitter (12.5 Gbit/s), polymer waveguides ([Schm05])
Details of these investigations were also published in [Lenz 05] by the ETH
Zurich. Waveguides with a cross-section of up to 100 m are described. It was
pointed out in this work that one of the limiting factors for the integrated optics
was the small packing fraction which resulted from the minimal bending radii.
One solution could be photonic crystal structures which permit approximately
right-angled bends.
382 5.6 Waveguides as Interconnection Solutions
optical backplane
transmission [%]
90
80
70
60
50
40
30
20
10
0
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
wavelength [m]
Fig. 5.35: Transmission spectrum of a UV-cured epoxy resin ([Bau05])
The waveguides produced also do not reveal any clear change in transmission
after the printed circuit boards have been laminated and the soldering tests as well
as the temperature cycle tests (-40C/+125C, > 200 cycles) have been carried out.
On the contrary, there was even a decrease, Fig. 5.36.
384 5.6 Waveguides as Interconnection Solutions
0.20
0.15
0.10
0.05
0.00
1 2 3 4 5 6 7 8 9 10 11 12
waveguide number
Fig. 5.36: Waveguide attenuation after lamination/soldering and temperature load
([Bau05])
Fig. 5.37: Waveguides, 50 m lines, 200 m spacing; wafers with waveguide structure;
waveguides with smooth surface and vertical edges (from [Micr06])
A module for coupling the active modules to the optical printed circuit board is
shown in Fig. 5.38. Holes are drilled into the printed circuit board into which the
module is projected. The light is deflected by means of 45 mirrors.
In order to calculate the link power budget, one needs to know the range of fluctu-
ation of the source output power. The attenuation of the link must then be calcu-
lated. The two taken together indicate the range of the optical power arriving at
the receiver. If this exceeds the dynamic range of the receiver, the system will not
operate reliably.
The optical power of a semiconductor source is not constant. It is particularly
dependent on the temperature and it generally loses power over time. If one in-
tends to deploy a system within a certain temperature range, for example between
-20C and +70C, the output power fluctuations of the source must be considered
in this range. Figure 6.1 shows typical power current characteristics of LD and
LED, the most frequent sources used in POF systems, in relation to the tempe-
rature (schematic illustration).
Essentially, the LED shows a flattening of the characteristics curve, i.e. a reduc-
tion in efficiency. In the case of the LD, there is also the fact that the threshold
current Ith, the start of the laser operation, shifts to higher currents. Later on we
will demonstrate that this phenomenon requires very special precautions to facili-
tate using LD as transmitters.
The next step is to establish how the source is to be operated. In the case of the
LED, we assume that it is operated with a constant current. That means that we
can determine the change of optical power from the P-I(T) curve, see Fig. 6.2.
0.0 -3
0 10 20 30 40 50 -40 -20 0 20 40 60
current [mA] T [C]
Fig. 6.2: Typical P-I(T) characteristics curves for LED and determining the change in Popt
The figure is based on a reference power at 25C (0 dB). The vertical double-
pointed arrow represents the total possible power change within the permissible
temperature range. Typical values for the LED power variation between -20C and
+70C range from 1 dB up to 5 dB. In addition, there are up to 3 dB for changes in
output power due to the increasing aging of the source (see also [Sch00]).
Likewise, the sensitivity of receivers is not constant but depends on various fac-
tors. For example, one of these is temperature which particularly has an effect on
the noise of the amplifier as well as the wavelength of the light received, which in
turn is affected by the temperature of the source.
The proposed use of the receiver must be clear in order to evaluate it. If it is
used for the analogue transmission of a signal, a certain signal-to-noise ratio
(SNR) must be complied with as well as a certain degree of linearity which is
indicated by the distortions power. A general requirement of digital systems is that
a certain probability of bit errors (bit error ratio: BER) is not exceeded. Figure 6.3
shows the typical BER of a system in relation to the power at the receiver.
For example, in the case of data transmissions a BER = 10-9 is accepted. That
means that the sensitivity of the receiver shown is -32 dBm (0.63 W). However,
6.1 Link Power Budget of Optical Transmission Systems 389
the figure also shows a second characteristic of the receiver, the dynamic range.
When the power received is too large, it is possible that the amplifier may be over-
loaded. This will also lead to a deterioration of the signal, even exceeding the per-
mitted BER. In the example shown, this occurs at -12 dBm (63 W). This means
that the dynamic range of the receiver is 20 dB. It follows that it is important for
the system design to ensure that the optical power at the receiver always corres-
ponds at least to the sensitivity and that it also does not exceed the top end of the
dynamic range. Since transmission systems often have to work with very different
cable lengths, the latter requirement is not always easy to meet.
BER
10 -5
limited by noise limited by
10 -6 receiver overload
10 -7
sensitivity -32 dBm
10 -8
10 -9
dynamic range 20 dB
10 -10
10 -11
10 -12
-36 -34 -32 -30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8
Preceiver [dBm]
Of course, the simplest way for establishing the sensitivity of the receiver is
through measurement. The set-up for this test consists of a transmitter (of the best
possible quality), the transmission link, the receiver to be tested and a variable op-
tical attenuator (VOA). This attenuation element should allow the tunable setting
of a selected attenuation; otherwise it will be necessary to take a parallel measure-
ment of the received power via a coupler, as schematically shown in Fig. 6.4.
source
receiver
under test
variable attenuator
source receiver
under test
The second option for determining the receiver's sensitivity, which is particu-
larly important during the design phase, is to perform a theoretical calculation.
Assuming that there is sufficient bandwidth and linearity, the noise of the first
amplifier stage is usually the limiting factor. As a simple method, one can use the
current noise density at the amplifier input which is expressed as pA/Hz. The
calculation example below (Table 6.1) shows the necessary steps:
Up to this point we have established the necessary basic data for the transmitter
and the receiver for calculating the complete link power budget, as illustrated
schematically once again in Fig. 6.5. If the transmitter output power is within the
dynamic range (case a) in every case, the sum of losses on the fiber link may be
between 0 dB and the maximum value shown. If the maximum possible transmit-
ted output power is over the limit of the dynamic range (case b), a minimum link
attenuation, if necessary by providing an additional attenuator, has to be
guaranteed.
-32 -30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8
Preceiver [dBm]
When calculating the link power budget, the greatest attenuation must be placed
on the fiber link. Normally, the system provider has the least influence on this
part. In addition, this is the part where the environmental effects are greatest.
In this book we cover passive POF transmission systems. Although it is pos-
sible to conceive of POF amplifiers, these will not be of any practical significance
in the foreseeable future. This means that there will only be loss-making elements
between the transmitter and the receiver. Figure 6.6 shows an illustration of all
important elements.
3. fiber line
7. receiver
4. connectors
1. source
6. POF-receiver
2. source-POF- 5. passive elements coupling
coupling as couplers, filters
1. correct coupled
2. light
2. light, reflected at
1. the front surface
3. light, not falling on
POF
3. the front surface
source 4. light outside the
4. acceptance range
of the fiber
The most critical of all source parameters is the angle of emission or, more pre-
cisely, the far field, i.e. the emitted power in relation to the angle with the optical
axis. A standard NA POF has an acceptance angle of approximately 28. For a
DSI-POF this value is reduced to 17. For DSI-MC-POF or GI-POF it is only
11. However, the LED used for cost reasons in POF systems emit at a much
wider angle. To a certain degree it is possible to reduce the emission angle of a
LED by means of lenses. There are a number of LED available which have diffe-
rent emission angles. These are achieved through different designs of the LED
housing, which also has the function of acting as the lens. According to the laws
of optics, the product of image size and numerical aperture cannot be reduced.
This means that a reduction in the angle will result in an increase in the image of
the LED chip on the front face of the fiber. Typical LED chips are 200 m to
300 m in size. Thanks to a POF diameter of 1 mm there is some room here for
maneuvering, as is schematically illustrated in Figs. 6.8 and 6.9.
f f POF
LED
lens
Fig. 6.8: Imaging of the LED chip on to the POF with a 1:1 magnification
f f POF f f POF
LED LED
lens lens
Fig. 6.9: Imaging of the LED chip on to the POF with a reduction (left) and enlargement
(right) of the chip image on the POF
One can see that in the case of the reduction of the LED image, the angle range
of the rays increases. Conversely, in the case of a magnification, the angle range
becomes smaller. The illustration also shows a second effect too. Typical LED
emit at such a large angle that they can hardly be captured by normal lenses. That
means that the aperture of the lens determines how much light can be launched.
For example, the authors used plano-convex lenses for this purpose with a focal
length of 13 mm at 21.4 mm effective diameter. By arranging two lenses behind
each other it was possible to place the LED approximately at the focal point so
6.1 Link Power Budget of Optical Transmission Systems 393
that the useable lens NA was approximately 0.8. However, a much more efficient
coupling of the LED can be achieved if the chip is equipped with an appropriate
micro-lens fitted by the manufacturer, as shown in Fig. 6.10.
micro lens
POF
LED
Fig. 6.10: Projection of the LED chip on to the POF via a micro-lens
With direct butt coupling of a LED to the POF, the typical losses are in the
range of 10 dB to 12 dB. When the imaging is optimized, it is possible to reduce
these losses to within a range of 4 dB to 5 dB. Any improved values can only be
achieved where specially optimized components are used. Such components are,
for example, VCSEL or special LED, as will be described later on.
90
excess loss by cladding
80 attenuation [dB/km]
70
60
20
10
0 4 8 12 16 20 24 28
launch angle []
Fig. 6.11: Additional attenuation through cladding losses according to [Paar92]
However, it is generally true that when light is being launched with a small NA,
a significantly lower level of attenuation can be measured over the first few 10 m,
whereas when light is launched with a large NA, the level of attenuation is signi-
ficantly higher. Numerically this effect is expressed by an attenuation coefficient
for equilibrium mode distribution and by an additional value for the overall attenu-
ation deviation resulting from light launching using a different NA. However, the
exact determination of this value is still open and current standards only describe
it in basic terms. Attenuation data for POF quoted in data sheets have mostly been
established with collimated light and are therefore only of limited use in practical
application.
passive components). In the following section, the possible causes for attenuation
are described in detail. No practical splicing or bonding technology has as yet
been developed for PMMA-POF.
distributed coupler
central coupler
In the top arrangement, the loss through the couplers is at least 12 dB, i.e. 6 dB
through the coupling of the four arms and a further 6 dB through their subsequent
division. In the bottom arrangement, it is possible to reduce the loss to a total of
6 dB as shown in Fig. 6.13 with the detailed illustration of the coupler. While
there is in fact only one single fiber between coupling and division point in the top
arrangement, there actually are four fibers or a correspondingly wider wave guide
provided in the bottom arrangement.
396 6.1 Link Power Budget of Optical Transmission Systems
1/4 1/4
1/4 1/4
This short overview of coupler technology shows that one should not use gene-
ral formulas unless the topology applied is completely clear.
Filters are always useful in POF systems when wavelength multiplex is used. A
number of different solutions have already been presented in Chapter 3.4. The
large diameter of the POF makes it possible to fit filters directly between the end
faces of two connectors, as shown in Fig. 6.15. Here, the losses occur in the area
of the connector attenuation. The second option is the widening of the ray with the
help of a lens or mirror and placing the filter element into the parallel ray
(Fig. 6.16). Due to the large NA of POF, lenses are often not in a position to
efficiently capture the light so that losses of 5 dB are typical.
6.1 Link Power Budget of Optical Transmission Systems 397
filter
POF POF
lens
Fig. 6.16: Filter with beam expansion
Many of the processes described above in general terms have been studied in vari-
ous standards using exact quantities. Between 1996 and 1999, the ATM Forum
has established a specification for the transmission of 155 Mbit/s over 50 m of
PMMA-POF ([ATM96a], [ATM96b] and [ATM99]). In these documents the
different contributions to the link power budget are described in great detail so
that they can be used here as a very informative example.
Fig. 6.17: Link power budget according to the ATM Forum specification
The main part of the permissible changes is determined by the fluctuation of the
LED power caused by temperature. Figure 6.18 shows the spectrum of a 650 nm
LED, as used in components by Hewlett Packard (Agilent, Avago), in relation to
the temperature (according to [HP04]). The figure clearly shows the reduction in
emitted power at higher temperatures. Another effect is the shifting of the
emission wavelength towards greater values. As we will see later on, this process
also results in additional losses.
1.4
opt. power [a.U.] -40C
1.2
0C
1.0 +25C
0.8 +70C
0.6
+85C
0.4
0.2
0.0
610 620 630 640 650 660 670 680 690
wavelength [nm]
Fig. 6.18: Changes in the LED spectrum under the influence of temperature ([HP04])
approx 4.5 dB
1.0
GaAlAs LED
0.5
temperature [C]
-20 0 20 40 60 80
Fig. 6.19: Change in output power of a GaAlAs LED under the influence of temperature
changes according to [Nak97]
400 6.1 Link Power Budget of Optical Transmission Systems
[Sch99a] presents new LED for use in POF systems in automotive applica-
tions where temperature resistance is of particular importance. The LED shown
here (DH-MQW: Double Heterostructure Multi Quantum Well) feature less than
1 dB of output power change between -20C and +70C.
The amount of power reduction due to aging of the LED is determined prima-
rily by the operating temperature and the operating current. Halving the current or
reducing the operating temperature by 10 K can increase the service life by appro-
ximately one order of magnitude. The degree of production tolerance is pro-
portional to the care and effort afforded in manufacture. It may be necessary to
select the appropriate components.
10,000
attenuation [dB/km]
5,000 Toray St.-NA 1995 PCU-CD1002
Mitsubishi St.-NA 1995 Eska Extra
2,000 Asahi Low-NA 1996 NC-1000
Toray DSI 1997 PMU-CD1002
1,000 Mitsubishi DSI 1997 Eska Mega
Asahi St.-NA 1996 TC-1000
500
200
100
wavelength [nm]
50
350 400 450 500 550 600 650 700 750 800 850 900
Fig. 6.20: Attenuation spectra of different 1 mm PMMA-POF
6.1 Link Power Budget of Optical Transmission Systems 401
At a wavelength of 650 nm, all fibers meet the ATM Forum requirements with
an attenuation of less than 156 dB/km, as is shown in greater detail in the illustra-
tion of different fibers in Fig. 6.21.
Below 550 nm, clear differences can be recognized between the different types
of fiber. There are hardly any measurements of fiber attenuation under conditions
of equilibrium mode distribution. Manufacturers usually specify fibers with colli-
mated light or excitation with a small NA. Detailed studies on the attenuation in
polymer optical fibers, particularly in regard to the launching conditions, are des-
cribed in [Kell98], [Pfl99], [Hen99], [Pei00a] and [Pei00b].
400
POF attenuation [dB/km]
350
GH 4001
300 SH 4001
MH 4001
250 ATM-Forum MH 4002
specification PMU-1001
200 TC-1000
GH 4001
150
100
630 635 640 645 650 655 660 665 670 675 680
wavelength [nm]
Table 6.2: Steps for the calculation of the effective POF attenuation with LED
Parameter Formula / Calculation Unit
normalized LED spectrum PLED(O) 1/nm
LED power f 1
P0 = PLED (O)dO = 1
O 0
1.0 500
rel. power [a.u.] 0m POF-loss
0.9 [dB/km] 450
0.8 400
0.7 10 m 350
0.6 300
0.5 20 m 250
0.4 30 m 200
0.3 150
0.2 40 m 100
0.1 50
50 m
0.0 0
620 630 640 650 660 670 680 690
wavelength [nm]
Fig. 6.22: Deformation of the spectrum of a 660 nm LED through POF attenuation
Figure 6.23 shows the filter effect for lengths up to 200 m of POF in logarith-
mic scaling. For this purpose a measured LED spectrum as well as the measured
spectral POF attenuation were used to mathematically determine the spectra after
passing through different fiber lengths. Here too, one can clearly recognize the
shift towards 650 nm as well as the narrowing of the spectrum.
6.1 Link Power Budget of Optical Transmission Systems 403
0
rel. power [dB]
-10 0 m POF
10 m POF
-20 20 m POF
-30 50 m POF
-40 75 m POF
Due to the filter effect, the effective attenuation of the fiber is no longer propor-
tional to its length. A relatively wide LED spectrum at the beginning of a POF
experiences a relatively high averaged attenuation since a great degree of power is
placed in spectral areas with high POF attenuation. After some distance the spec-
trum has become narrower and its center of gravity is close to the attenuation
minimum so that the effective attenuation decreases. This indicates the difficulties
experienced when measuring POF attenuation with LED sources as well as with
correct specification of the effective attenuation.
The ATM Forum specification allows sources with a spectral width of max.
40 nm and a center wavelength of 650 r 10 nm. No statements are made about the
shape of the LED spectrum. However, most LED can be approximated quite well
to Gaussian shaped spectra.
In order to calculate the effective additional loss, we will use the Gaussian LED
spectra and the POF attenuation curve according to [Wei98] (losses at 650 nm:
132 dB/km). Figure 6.24 shows the results of the calculation for a spectral width
of 40 nm and deviations from the center wavelengths 650 nm up to 20 nm, i.e. for
center wavelengths of 630 nm, 640 nm, 650 nm, 660 nm and 670 nm (calculations
as explained in Table 6.2).
8.0
eff. excess loss [dB]
7.0 630 nm
6.0 640 nm
5.0 650 nm
660 nm
4.0
670 nm
3.0
2.0 ATM-Forum
specification
1.0
POF length [m]
0.0
0 10 20 30 40 50 60 70 80 90 100
Fig. 6.24: Additional effective losses of a 40 nm wide LED source
404 6.1 Link Power Budget of Optical Transmission Systems
The limit value of the ATM Forum for a maximum of 50 m has been entered.
The specification allows 3.4 dB of additional loss. The diagram shows 3.61 dB as
the max. value for the curves with 640 nm to 660 nm center wavelength, whereby,
however, a value of 132 dB/km is assumed for 650 nm, which means only 6.6 dB
for 50 m instead of the 7.8 dB stated in the specification. One can see very clearly
the flattening of the curve due to the filter effect, i.e. the convergence of the
effective attenuation to the value at the attenuation minimum.
Figure 6.25 shows comparable results for a source which again is of Gaussian
shape, this time with only 20 nm full width at half maximum and center wave-
lengths between 630 nm and 670 nm.
10.0
eff. excess loss [dB]
9.0
8.0 630 nm
640 nm
7.0
650 nm
6.0 660 nm
5.0 670 nm
4.0
3.0
2.0
1.0 POF length [m]
0.0
0 10 20 30 40 50 60 70 80 90 100
Fig. 6.25: Additional effective losses of a 20 nm wide LED source
In this case, the value stays significantly below the limit set by the ATM
Forum. For sources between 640 nm and 660 nm center wavelength, the narrower
spectrum is associated with smaller additional losses. However, for sources with
630 nm and 670 nm wavelength the additional attenuation is significantly larger
compared with those sources having a width of 40 nm. Lastly, Fig. 6.26 shows the
additional losses for monochromatic sources, for example, singlemode lasers.
As will be readily appreciated, there is no filter effect in this case, since only
the kilometric attenuation of the source wavelength is of importance. For 660 nm
the additional loss is 3.35 dB again, i.e. near the limit of the specification. For
greater deviations from the wavelength of the attenuation minimum, the additional
losses rise rapidly in accordance with the attenuation characteristics of the POF.
This leads us to the three basic statements below:
Wide sources create additional attenuation since a large part of the spectrum
is located in spectral areas with high POF attenuation.
Deviations of the source center wavelength from the attenuation minimum
lead to additional losses due to increasing POF losses.
If a source is significantly outside the attenuation minimum, a wide spectrum
limits the additional losses up to a certain degree, since parts of the spectrum
lie in low-loss areas. In the case of very long links, the additional losses in-
crease less than proportionally due to the filter effect.
The last point also provides an explanation of why the effects of center
wavelength and spectral width cannot be established separately but must always
be considered in their (non-linear) relation to each other.
ray optics but must be due to losses at the core/cladding interface. Up to the limit
of the acceptance range, the additional losses increase to some 100 dB/km, i.e. a
multiple of the basic attenuation. The measurement at 50 m length reveals signifi-
cantly lower attenuations. This can be explained by the fact that part of the input
power launched at the wide angle is shifted along the fiber path to smaller propa-
gation angles due to mode conversion. Therefore, the values no longer reflect the
conditions of exactly one angle of propagation but instead an average of the
modes generated along the path.
200 50 m
100
0
-30 -20 -10 0 10 20 30
propagation angle []
Figure 6.28 shows the comparable results for the measurement with a double-
step index profile POF which basically shows the same behavior. In contrast to the
standard POF, however, high additional attenuations in the range between 20
and 30 can be seen. These angle ranges carry light which is no longer guided by
the interface between core and inside cladding but is still completely reflected at
the interface between the inner and outer cladding. The significantly higher attenu-
ation of the inner cladding compared to that of the core causes the high additional
loss. It is precisely this effect that ensures that after some 10 m there is only light
present in the acceptance range of the inner cladding. In this case too, mode con-
version leads to a reduction in the additional losses after some length of the POF.
The measurement of angle-related additional attenuation according to the
method developed by Klein facilitates very swift and explanatory insight into the
function of mode-related attenuation and mode conversion which are jointly res-
ponsible for the mode filter effect. One can only hope that this method will be
adopted into the specifications for measuring POF.
6.1 Link Power Budget of Optical Transmission Systems 407
600
excess loss [dB/km]
10 m
500
O= 594 nm
400
300
50 m
200
100
0
-30 -20 -10 0 10 20 30
propagation angle []
Fig. 6.28: Angle-related additional attenuation for a DSI-POF according to [Kle00]
As was shown in Chapter 3, the protruding material, which arises with the hot
plate process, acts as an axial distance of some tenths of a millimeter which results
in additional attenuation of typically 0.4 dB. It is not clear why the ATM Forum
specification does not take this value into account despite the recommendation for
the hot plate process. Table 6.4 summarizes once again the losses considered here.
Table 6.4: Overview over the connector losses under UMD conditions
If the losses due to roughness and external mechanical losses are assessed right
away, then the overall result is almost a doubling of the attenuations with the
UMD conditions accepted here compared with the ATM Forum specification.
0
bending loss [dB]
-1
-2
PFU-UD1001 St.-NA
one 360 bend
-3
PFU-UD1001 St.-NA
15 bends by 90
-4
-5
bending radius [mm]
-6
0 10 20 30 40
If the bends are distributed over the length of the fiber, it is possible that mode
conversion occurs between them. The first bend leads to the extraction of certain
modes which exceed the acceptance range. If another bend follows immediately,
the modal field has already adapted, thus usually lowering the attenuation. If the
bend occurs at the same distance, the mode distribution has already changed
towards equilibrium mode distribution. In addition, a subsequent bend usually
bends in another direction.
Long storage at high temperature and relative humidity leads to a steady and
irreversible aging of the fiber leading to a typical additional attenuation of
60 dB/km for each 1,000 h of aging (for +85C/95% RH, see also
[Ziem00b]). Above a certain threshold, increased aging sets in which will des-
troy the internal structure of the fiber. For +70C this time span is over
20 years.
Overall the aging effects for polymer optical fibers have not been systemati-
cally investigated to any great extent. It is quite probable that the current specifica-
tions will have to be adjusted since, under certain conditions, the additional attenu-
ations calculated are exceeded under the specified conditions. On the other hand, it
does not make much sense to stipulate a relative humidity of 95% at +75C for the
total maximum link length in buildings since this type of climate would make a
normal building at least uninhabitable. Any short term exposures or the effect on
short sections can easily be managed by current POF.
We have now covered all aspects of the link power budget of the ATM Forum
specification. It shows that many of the individually discussed mechanisms re-
quire relatively complex calculations. Overall one cannot completely follow bud-
get considerations. In particular, several questions remain open in regard to the
definition of additional loss due to spectral source characteristics, connectors and
climatic effects.
In the next chapter, we wish to show what improvements can be achieved in the
link power budget through different components. The biggest potential is repre-
sented by the selection of source wavelength and the type of transmitter.
Initially, it would appear that the selection of source wavelengths for POF systems
is relatively easy, being determined by the minimum attenuation. However, closer
study reveals that a multitude of criteria must be considered, the most important of
which are listed below:
low POF attenuation for the LED wavelength
low effective additional loss, taking into account the spectrum
low drift of the wavelength with temperature
low spectral width
high optical power
small emitting surface area
small angle of emission
long service life
good efficiency
large modulation bandwidth
low price
good availability, preferably from several manufacturers
large operation temperature range
low temperature coefficients for wavelength and power
6.1 Link Power Budget of Optical Transmission Systems 411
Further points can be added to this list. Many of the parameters lead to com-
pletely different solutions so that the final selection always represents a compro-
mise. The best spectral characteristics, the largest modulation bandwidth and the
highest power are offered by laser diodes. However, these are not always cheap
and often have large temperature coefficients. Red LED are cheap and widely
available but have unfavorable spectral characteristics. Green LED or red VCSEL
have better parameters but are not yet widely available. In the following sections,
we will be looking initially at LED as possible transmitters for POF systems. Sub-
sequently, we will carry out a theoretical comparison of different semiconductor
structures.
5,000
POF attenuation [dB/km]
2,000
1,000
500
200
100
wavelength [nm]
50
400 450 500 550 600 650 700 750 800
Fig. 6.30: POF attenuation spectrum according to [Wei98] for calculating the effective
attenuation (range 700 nm to 800 nm added according to [LC95])
412 6.1 Link Power Budget of Optical Transmission Systems
Although these LED have a relatively high power for their wavelength, they do
not in every case represent the maximum achievable values. One essential cause
for this is the fact that most manufacturers of visible LED do not state the optical
output power. A typical specification covers the illumination and the angle of
emission. If the far field and spectrum are not known, these data only allow a limi-
ted conclusion regarding the optical power. Often, these data are also inaccurate
and have been determined using different measuring conditions. Another factor is
that many LED vary a great deal in their intensity from unit to unit so that only
typical values are stated.
As a first step, Fig. 6.31 shows the relation of optical power to the ambient tem-
perature, integrated across the whole spectrum. The change between -20C and
+70C is represented as an average value in dB/K. All LED show a decreasing
output power for an increase in temperature. However, the size of the coefficient
'Popt/'T varies a great deal. For the LED investigated, the values range from
-0.002 dB/K to -0.052 dB/K. GaN LED ranging from 470 nm through 560 nm
show the least change in optical power in response to temperature. The 593 nm
LED shows the strongest reduction in power.
Sander 430 nm
Sander 450 nm
RS 470 nm
Sander 470 nm
Sander 500 nm
Conrad 525 nm
Nichia 525 nm
Nichia 560 nm
RS 563 nm
RS 583 nm
RS 593 nm
Farnell 609 nm
Farnell 621 nm
Conrad 625 nm
Conrad 640 nm
RS 650 nm
RS 660 nm
'Popt/'T [dB/K]
RS 700 nm
0.00 -0.01 -0.02 -0.03 -0.04 -0.05 -0.06
All sources with values near the absolute minimum of 520 nm and 570 nm cre-
ate additional losses. The sources with short wavelengths at 430 nm and 450 nm
are on the downward slope of the first minimum and have a somewhat smaller
excess loss due to the filter effect. The relatively large increase in the effective
attenuation of the LED with 583 nm and 593 nm occurs because of the shift in the
direction of significantly higher attenuations at the 620 nm maximum. Figure 6.33
shows the results for LED above 600 nm.
Fig. 6.33: Effect of the spectrum on the additional attenuation for a POF of 50 m length
-5.8
-6.6 RS 583 nm
-20 -10 0 10 20 30 40 50 60 70
temperature [C]
Fig. 6.34: Effective attenuation of 50 m POF with different LED
The classic 660 nm LED still meet the requirements of the ATM Forum, albeit
with a high sensitivity to changes in temperature, which is caused in particular by
the LED wavelength running out of the attenuation minimum when heating up.
The LED with 650 nm wavelength from RS shows even worse results due to a
relatively wide spectrum and a big reduction in power with increasing tempera-
ture. The 640 nm LED by Conrad benefits from the tendency of POF attenuation
to counteract a shift in the wavelength apart from the fact that the power is less
sensitive to temperature. When heating up, the spectrum runs into the attenuation
minimum which means that the effective attenuation over the whole range is
significantly lower.
The maximum values of the LED at 621 nm and 625 nm lie practically exactly
on the attenuation peak. Nevertheless, their effective attenuation is only a little bit
higher than that of the 650 nm LED. The spectra are so wide that they reach into
the areas on both sides of the peak with lower attenuation. The shift towards the
red window is here compensated for by the reduction of power due to heating up
in an almost ideal way. For links with fixed installation it would be possible to
deploy a receiver with very low dynamic. However, the 609 nm LED shifts to-
wards the attenuation maximum during heating up which is indicated by the big
reduction of power.
Some LED in the green and yellow window are compared again in Fig. 6.36.
The 593 nm LED by RS has the lowest effective attenuation of all LED at low
temperatures, however, during heating up the peak wavelength clearly moves
towards an increase in attenuation up to 620 nm. In addition, there is a high tem-
perature coefficient of the output power.
-2.5
effective POF loss [dB/50 m]
-3.0
-3.5
-4.0
Nichia 525 nm
-4.5
Nichia 560 nm
-5.0 RS 563 nm
-5.5
-6.0
-6.5 RS 583 nm
-7.0
-7.5
-8.0
-8.5
-9.0 RS 593 nm
-20 -10 0 10 20 30 40 50 60 70
temperature [C]
Fig. 6.36: Effective attenuation of 50 m POF with different LED
We can conclude that selecting an LED only according to the parameters center
wavelength and spectral width does not always lead to optimum results. Instead,
one needs to determine the effective attenuation together with the attenuation
curve of the POF deployed. GaN LED in the green and yellow range generally
provide an advantage here, in the case they offer an adequate absolute power.
The numerical values established here depend on the actual LED type and the
POF type. Different types can lead to significantly different results especially in
the area of steep edges within the attenuation spectrum. However, the basic ten-
dency should remain the same. Figure 6.37 shows that for all ranges of visible
light there are capable LED available (spectra have been normalized to reflect an
equal maximum power).
418 6.1 Link Power Budget of Optical Transmission Systems
Prel
400 420 440 460 480 500 520 540 560 580 600 620 640 660 680 700 720
wavelength [nm]
Fig. 6.37: Overview of the spectra of different LED
With the optimized red LED a gain of more than 2 dB is obtained in the link
power budget in addition to the amount for the decrease in fluctuation of the
output power. Green LED have a low attenuation as well as the advantage of a
very flat attenuation curve so that the spectral parameters only play a minor role.
6.1 Link Power Budget of Optical Transmission Systems 421
Figure 6.40 summarizes the link power budget for the effective POF attenu-
ation of the four LED under consideration.
spectral
590 nm-LED parameters
665 nm-LED
0 2 4 6 8 10 12 14 16
contribution to the power budget [dB]
In addition, the spectral width of laser diodes is much smaller than that of LED.
The typical width of multi-mode laser diodes is just a few nanometers and there-
fore hardly of any significance for the effective attenuation. DFB laser diodes
(with distributed feedback in the active zone) emit light in single modes with a
width that can be completely neglected in POF systems.
Somewhat more significant is the temperature-related drift of the wavelength.
As an example, we wish to take a look here at an AlInGaP based laser from
Toshiba [Tos98]. Between -10C and +60C the wavelength increases in a practi-
cally linear fashion from 664.5 nm to 677.0 nm which corresponds to a coefficient
of 0.18 nm/K as is typical for Fabry-Perot laser diodes. For a POF length of 100 m
this corresponds to an increase in attenuation of 5 dB (from 238 dB/km to
288 dB/km).
When assuming this coefficient for the overall POF wavelength range for the
lasers that are of interest here, the emission wavelength drifts by r8 nm at 25C
within the range of -20C to +70C. Figure 6.41 shows the changes in the effec-
tive attenuation in a 100 m long POF link for an ideal laser for a value of 25C.
For a laser at 650 nm the change is no less than 3.4 dB and 4.9 dB for -20C or
+70C respectively. Components of the wavelength range around 520 nm or
560 nm would be very well suited again. DFB lasers are much less susceptible to
the effect of temperature; for POF connections, however, they are too expensive
by several orders of magnitude.
10
8 loss change relative
to 25C [dB/100 m]
6
4
2
0
-2
-4 change for -25C
-6
change for +70C
-8
-10
450 500 550 600 650 700
center wavelength at 25C [nm]
Fig. 6.41: Change in POF attenuation due to a drift in the laser wavelength
In the meantime there are new different LED versions which have strong
asymmetrical spectra. This definitely has significant consequences for the effec-
tive attenuation. Tests were conducted in a joint project with Agilent to determine
what the actual losses on a 10 m long fiber link could be in dependence of diffe-
rent spectral parameters. Peak wavelengths between 630 nm and 685 nm are per-
mitted in the MOST specification whereby the maximum spectral width may only
be 30 nm. The attenuation of 10 m of POF for LEDs with peak wavelengths (OPeak)
and spectral widths ('O) in the permitted range are shown in Fig. 6.43. In each
case a Gaussian-shaped spectrum was used.
2.8
FWHM: 30 nm
2.3 Gaussian
FWHM: 20 nm shaped
spectrum
1.8 FWHM: 12 nm
FWHM: 4 nm
1.3
630 640 650 660 670 680 690
peak wavelength [nm]
Exactly 3.3 dB were indeed attained with (OPeak) = 685 nm und 'O = 30 nm. At
630 nm the loss is still clearly smaller. The authors of this standard obviously
assumed that the greatest attenuations will always arise at maximum spectral
width. This is true for the long-wave side of the spectrum, but not for the short-
wavelength. An LED with OPeak = 630 nm and 'O = 4 nm does indeed meet the
MOST specification, but leads to 0.3 dB too much attenuation.
This very theoretical case should not come about, however, since LEDs always
have wider spectra. Furthermore, the short wavelengths only come about at low
temperatures with which the efficiency of the LED increases.
A much more realistic situation is when the LED spectrum is clearly asymme-
trical. This could be the case with an S-LED. A spectrum was tried in Fig. 6.44
which consists of two differently steep Gaussian edges with a width ratio of 1:7.
In the worst case an attenuation of 5 dB can arise (1.7 dB over the assumed
maximum attenuation), although the LED formally corresponds to the specifica-
tions. The reason is that the greatest part of the light energy lies above the peak
wavelength where the POF attenuation is the greatest.
6.1 Link Power Budget of Optical Transmission Systems 425
4.8
eff. loss [dB/10 m]
4.3
3.8
MOST limit: 3.3 dB
3.3
The effective attenuation was calculated overall with 7 different spectra. The
parts of the parameter field from OPeak and 'O in which the attenuation for at least
one spectrum lies above 3.3 dB are marked in Fig. 6.45. A significant part of the
range permitted by MOST is thereby covered.
60
FWHM [nm] loss >3.3 dB
50 for 10 m POF
40
30
20 specified
area for
10 MOST-LED
Opeak [nm]
0
600 620 640 660 680 700 720
Fig. 6.45: Calculated areas for LED parameters giving >3.3 dB loss
If you wish to avoid either having to significantly limit the permissible para-
meter range or to allow more attenuation, then you have to describe the LEDs by
means of parameters in which possible asymmetries in the spectrum have less con-
sequences. The solution recommended, and in the meantime accepted by the
MOST consortium, is to describe in the future LEDs through their spectral center
of gravity wavelength and the effective width (50% width of the equivalent
Gaussian distribution):
O central
Pi Oi and 'O eff 2.355 V 2.355
Pi (Oi - Ocentral )2
Pi Pi
426 6.1 Link Power Budget of Optical Transmission Systems
Two examples of real LED spectra are shown in Fig. 6.46. In addition, the
Gaussian curves which result from OCentral and 'Oeff are included. Determining this
data is simply more practicable for LED manufacturers.
0.4
0.2
0.0
580 600 620 640 660 680 700 620 630 640 650 660 670 680
O [nm] O [nm]
Fig. 6.46: Comparison of the old and new parameters for 2 LED
If you take these two parameters as a basis then almost the entire range speci-
fied by MOST remains with losses below 3.3 dB. There could even be a much lar-
ger range permitted (Fig. 6.47). This change is practically irrelevant for conven-
tional LEDs with symmetrical spectra. The calculability of POF losses could be
considerably improved especially for RC LEDs with unusual spectra.
30
20
specified area
10 for MOST-LED
0
600 620 640 660 680 700 720
Ocentral [nm]
Fig. 6.47: Calculated areas for the new LED parameters giving >3.3 dB loss
The effects become even more dramatic for greater lengths. In the future, other
standards will also have to make use of the new parameters introduced by MOST.
6.2 Examples of Link Power Budgets 427
The following is a list of some practical examples for calculating link power bud-
gets. In Fig. 6.48 we show again the calculation for the link power budget in the
ATM Forum specification for 155 Mbit/s over 50 m at 650 nm wavelength.
Fig. 6.48: ATM Forum link power budget for 155 Mbit/s
Apart from fluctuations in the LED power, connector losses, and attenuation of
the POF with respect to spectral source characteristics, the modedependen loss and
the effect of climate and bends have been taken into account.
22
LED power variation
20 2
mode dependent loss
POF loss at 650 nm
18 4
connector loss
source spectral width and drift of
6 the center wavelength
16
influence of temperature and rel.
humidity
14 8 fiber bends
system margin
12 10
Fig. 6.49: Link power budget of a 100 m POF link with 520 nm LED
[Ziem98a] and [Ziem98b] propose using the first optical POF attenuation win-
dow for transmitting 155 Mbit/s. Due to the reduced attenuation at 520 nm and
428 6.2 Examples of Link Power Budgets
A link power budget is also calculated for using POF in the standard IEEE 1394b
(Fire-Wire or i.link, Fig. 6.50). Here, the maximum used data rate is 200 Mbit/s.
Due to the 8B10B coding (NRZ), the physical data rate on the POF is 250 Mbit/s
and correspondingly higher for the further steps in the hierarchy. The sensitivity of
the broadband receiver is correspondingly worse so that the overall link power
budget is only 19 dB compared to 23 dB of the ATM Forum. No fixed length is
specified in IEEE 1394. Instead, the standard allows a maximum link attenuation
of 9.1 dB corresponding to the values of the ATMF specification without connec-
tor but taking into account climatic factors and mode dependent loss. If connectors
are used, again taking a maximum loss of 2.0 dB into consideration, the per-
missible length of the link is reduced correspondingly. A maximum of three con-
nectors is possible which reduces the maximum length to 27 m (42 m for one con-
nector and 34 m for two connectors). There is no margin provided in the system.
18 0
LED power variations
2
POF loss at 650 nm with
16 mode dependent loss,
influence of temperature and
humidity and connector loss
4 (optional up to three)
source spectral width and
14 drift of the center wavelength
fiber bends
6
12
8
10
The effect of the spectral source parameter for the 50 m fiber length is entered
at 3.4 dB and the values for the other lengths are reduced correspondingly, for
42 m: 2.9 dB, for 34 m: 2.3 dB and for 27 m: 1.6 dB. The calculation is based on
0.182 dB/m and 2.0 dB loss per connector and 12.5 dB overall loss.
Currently, there are receivers on the market which are significantly better, also
at a data rate of 250 Mbit/s. Therefore it would appear to make sense to revise the
specification, at the latest when short wavelength POF windows can be considered
or when deploying new source types (RC-LED, VCSEL).
Further specifications of the link power budget exist for the bus systems used in
the automotive field, i.e. D2B (Domestic Digital Bus) and MOST (Media Oriented
System Transport). Figure 6.51 shows the link power budget for the D2B specifi-
cation according to [Pet98]. The guaranteed LED power is -15 dBm. A maximum
LED power has not been stated in the literature. When taking the typical tempe-
rature susceptibility and a specified temperature range of -40C to +85C into
consideration, the value would probably be around -6 dBm.
11
1 POF loss at 650 nm
10
(max. 8 m, 0.4 dB/m)
LED-POF coupling
9 2 POF-PD coupling
connector loss
(optional one)
8 3 system margin
7 4
6 5
The data rate for D2B is 5.65 Mbit/s, of which 4.2 Mbit/s are the payload. Due
to bi-phase coding, the corresponding physical data rate on the POF is 11.3 Mbit/s.
The architecture of D2B forms an active ring. This means that each component
is equipped with one receiver and one transmitter. The components are switched
in a ring so that each element has to pass on the signals for all following equip-
ment.
The specification does not take into account mode dependent loss, climatic
effects or the effect of the spectral source parameters. On the other hand, the value
430 6.2 Examples of Link Power Budgets
for the POF attenuation is a conservative one at 400 dB/km so that even conside-
rable aging or a drift in the source wavelength to 670 nm can be tolerated. For a
maximum link length of 8 m, these losses are not really critical. Their specifi-
cation provides a large margin of 5.0 dB in the system. This probably reflects the
intention to make appropriate allowance for the newness of the technology for the
automotive field. Another factor is that the conditions for the cable installation
during vehicle manufacture are much more difficult compared to the installation
of cables in a building network, for example. Subsequent replacement of a defec-
tive cable is also expensive. The sensitivity of the receiver is specified as
-26 dBm. Considering the low data rate (11.3 Mbit/s), this value is also rather con-
servative. Overall, the link power budget shows that a large part of risk of the
cable installation and the design of the electrical interfaces has been moved to the
POF link. However, this can be well accommodated by the link. It is likely that the
existing reserves in the next generation will probably be used to increase the bit
rates as is already heralded by the introduction of the MOST Standard. At
approximately 21.2 Mbit/s of use data rate, the physical bit rate is approximately
50 Mbit/s due to the RZ coding.
Figure 6.52 shows the link power budget for MOST (Media Oriented System
Transport, for example, see [Tei00], [Pan99] and [Pan00]). The budget comprises
approximately 23 dB. For the attenuation of the POF link a value of 16.5 dB is
taken into consideration which includes losses at additional connections, in case
prefabricated pigtails are used in installation. Nevertheless, a large value has been
assumed considering the short lengths in a vehicle, for example compared to the
value of 13 dB of link attenuation assumed in the ATM Forum specification for a
link of 50 m length. The high value for attenuation takes into account the rough
conditions for POF during installation and practical use in the vehicle as well as
the necessary margin for long service life and the reliability requirements of the
systems.
22
2 LED power variations
20 influence of temperature
4 and current supply to the
electronics
18 LED aging
6 LED pigtail coupling
POF loss at 650 nm
16 (includes 4 dB for POF
8
coupling to transceivers)
14 10 pigtail to PD coupling
12
Finally, Fig. 6.53 shows an additional link power budget based on the author's
own proposal for transmitting ISDN signals via POF ([Ziem00c], [Ziem00d] and
[Ziem00e]).
Fig. 6.53: Link power budget for 250 m POF-ISDN with LED of 560 nm
The sources used here are 560 nm GaN-LED which are available as samples
from Nichia. The maximum power launched into the POF is assumed to be
-3 dBm. For a guaranteed receiver sensitivity of -48 dBm, the link power budget
amounts to 45 dB.
Due to the excellent temperature stability of GaN LED, only 3 dB have to be
considered to for the fluctuations in power. 1.0 dB has been taken into account for
the mode dependent loss. For a wavelength drift of 10 nm (GaN LED feature a
much lower drift) and a maximum of 40 nm spectral width, a fiber attenuation of
80 dB/km is obtained corresponding to 20 dB for the maximum transmission
length of 250 m assumed here. Following our own tests for climate exposure,
20 dB/km were entered for the effects of aging due to temperature and 20 dB/km
for the effect of water absorption (5 dB for 250 m, see also [Ziem00b]). Allowing
for 3 permissible connectors there remains a margin of 3.0 dB in the system which
would cover, for example, bends in the fiber link. This proposal is new in the field
of POF applications since it is the intention here to cover a large distance with a
relatively small data rate (192 kbit/s). This makes the use of the first two POF
windows mandatory.
detail. Below we will demonstrate link power budgets for possible systems with
asymmetric bit rates.
If in the coming years the very high transmission rates under consideration here
are achieved in vehicle networks, a highly asymmetrical requirement must be
assumed. Conversely, this means that only small data rates will be achieved in
backward direction. The transmission of different data streams via just one fiber as
opposed to a solution with duplex fiber can be solved particularly effectively with
WDM (see also [Ziem97b]). Two different concepts can be considered here:
5 dB 1 mm SI-POF 2 dB
WDM-filter WDM-filter
Si-PD Si-PD
100 Mbit/s, -26 dBm 1 Gbit/s, -16 dBm
4 dB 1 mm SI-POF 4 dB
Si-PD WDM-filter Si-PD
WDM-filter
100 Mbit/s 1 Gbit/s
-28 dBm -16 dBm
Fig. 6.55: Concept for WDM with symmetrical couplers
6.2 Examples of Link Power Budgets 433
Figure 6.56 shows a diagram for the complete link power budget calculations of
both systems. To date this set-up has not been implemented in practice. At the
time of this concepts, there was still a problem of obtaining sufficiently fast recei-
vers for POF of 1 mm. These systems should be considered as a theoretical con-
cept design.
asymmetrical coupler:
source powers receiver sensitivities
770 nm LD
520 nm LED
0 -2 -4 -6 -8 -10 -12 -14 -16 -18 -20 -22 -24 -26 -28
770 nm LED
650 nm LD
0 -2 -4 -6 -8 -10 -12 -14 -16 -18 -20 -22 -24 -26 -28
optical power [dBm]
Fig. 6.56: Link power budget for asymmetrical data rate WDM connections
434 6.3 Overview of POF Systems
medium is rated by the manufacturer at 300 dB/km. The source was a laser diode
NDL 3200 with a max. of 3 mW at 670 nm wavelength that was originally deve-
loped for barcode lasers. When modulated and launched with a collimated beam,
the maximum power in the POF was still 0 dBm.
A pin FET transistor arrangement was used as a receiver. At a diode capacity of
3 pF, the bandwidth of the receiver was 75 MHz (0.6 times the bit rate). The cal-
culated sensitivity was -31.4 dBm at BER = 10-9. A value of -28.5 dBm was
measured (with a 80 MHz preamplifier). This resulted in a coupling loss of 3.7 dB
at the receiver. With measured losses of 276 dB/km, a maximum transmission link
length of 90 m could be achieved with the power budget of 28.5 dBm. The system
is shown schematically in Fig. 6.57.
Fiber: St.-NA-POF EH 4001, 300 dB/km (Mitsubishi)
Length: 1 m/ 25 m/50 m/90 m
Bit rate: 125 Mbit/s
Transmitter: LD 670 nm; NDL 3200, 3 mW
Receiver: pin-FET-transistor-combination
Reference: [Pri92]
Company: Kennedy & Donkin Systems Control Ltd.
670 nm LD 1 m, 25 m, 50 m, 90 m
NDL 3200 Si-PD
980/1000 m SI-POF pin-FET
3 mW ESKA EXTRA EH 4001
125 Mbit/s 75 MHz
276 dB/km | 670 nm
Fig. 6.57: Transmission system according to [Pri92]
In [Kit92] the Mitsubishi POF ESKA Premier was used which was newly
developed at the time. At a NA of 0.51 and at losses of 135 dB/km at a wave-
length of 650 nm, this fiber material can be used at temperatures up to +85C.
With collimated light, the attenuation of 65 dB/km at 570 nm and 124 dB/km at
650 nm was measured.
Using the TOLD 9410 Toshiba laser diode and a receiver of its own, it was also
possible to transmit 125 Mbit/s over 100 m, since the launched power was corres-
pondingly higher (Fig. 6.58). Moreover, a system with a yellow LED was realized
(see below).
Fiber type: SI-POF, Mitsubishi ESKA Premier
Length: 100 m
Bit rate: 125 Mbit/s
Transmitter: 650 nm LD, Toshiba TOLD 9410
Receiver: own construction
Reference: [Kit92]
Company: Mitsubishi
6.3 Overview of POF Systems 437
In [Fuk93] the use of various wavelengths is also described for POF trans-
mission systems. At first an InGaAsP LED at 670 nm was used with which an
average power of -12 dBm could be coupled into the fiber. 100 Mbit/s were trans-
mitted over 30 m SI-POF.
Fiber type: SI-POF
Length: 30 m
Bit rate: 100 Mbit/s
Transmitter: InGaAsP-LED, 670 nm, -12 dBm
Receiver: commercial Toshiba receiver, -22 dBm sensitivity
Reference: [Fuk93]
Company: Toshiba
NEC started using the newly developed low NA-POF for transmissions of
156 Mbit/s over 100 m using 650 nm LED (for example, [Koi97a]) as early as
1995. In [Kob97] a new type of LED is used for such systems that makes a coup-
ling efficiency of 70% possible with simple plastics lenses through improved
emission characteristics.
Later a transceiver in a 1 u 9 pin type of construction was offered as a
commercial product [NL2100]. It was followed by the NL2110 of the S200 type
with a data rate of 250 Mbit/s over a maximum of 70 m. Today there is no longer
any known work being done by NEC in the POF field.
Fiber type: Low-NA-POF
Length: 100 m
Bit rate: 156 Mbit/s
Transmitter: 650 nm LED, improved coupling efficiency by an optimized
emission characteristic
References: [Koi97a], [Kob97]
Company: NEC
Sony introduced a transceiver for IEEE 1394 in 1997 ([Sak97]). The SI-POF
used had an attenuation of 160 dB/km and 130 MHz 100 m bandwidth at
650 nm. The silicon-pin diode was attached to a transimpedance amplifier, making
it possible to achieve a sensitivity of -25 dBm at a BER = 10-10 (125 Mbit/s). At
200 Mbit/s, a transmission length of 70 m was achieved.
Fiber type: DSI-POF, 160 dB/km, 130 MHz 100 m
Length: 70 m
Bit rate: 200 Mbit/s
Transmitter: 650 nm LD
Receiver: Si-pin-PD, -25 dBm at 125 Mbit/s, transimpedance amplifier
Reference: [Sak97]
Company: Sony
A test network was set up at the University of Ulm for office cabling with
SI-POF and later with DSI-POF. The first results were presented in 1998
([Som98a]). A total of 33 POF connections were installed with lengths ranging
between 5 m and 63 m (total installed cable length was 1,400 m). The bit rate was
125 Mbit/s.
A 100BaseFX(POF) switch served as the central node. The switch and the PC
cards were re-equipped with various commercially available POF transceivers in
1 u 9-pin form. Hewlett Packard HFBR 5527, NEC NL 2100 and self equipped
receiver were used. For the fiber, Asahi AC-1000W, Mitsubishi MH 4002F and
Toray PMU-CD 1002-22E according to the ATM forum specification were used.
The measured effective attenuation with the transceivers used was from
196 dB/km to 205 dB/km.
Fiber type: SI-POF, DSI-POF, Asahi AC-1000W, Mitsubishi MH 4002F,
Toray PMU CD 1002-22E
Length: 5 m to 63 m
6.3 Overview of POF Systems 439
652 nm LD Si-PD
Fig. 6.61: High bit rate data transmission over SI-POF
The Mitsubishi ESKA EXTRA EH4001 was used as the fiber medium. It has
139 dB/km of attenuation at 652 nm. A Philips laser diode CQL82 with a wave-
6.3 Overview of POF Systems 441
length of 652 nm served as the light source. The laser was operated at 290 K
(17C) with 36 mA of bias current. To increase the bit rate, a first order high-pass
filter was pre-connected as the peaking filter. With the help of input optics,
2.7 mWp-p of power was achieved at launch of NA = 0.11. During modulation, the
average power was -1.7 dBm (0.68 mW); with the peaking filter, the average
power fell to -6.7 dBm (0.21 mW).
An AEG-Telefunken BPW89 photodiode with 4.9 pF capacity at 20 V of re-
verse voltage was used as a receiver. The responsivity is 0.4 A/W at 650 nm (76%
external efficiency). The coupling to the POF is done with a ball lens. A second
high-pass filter was connected behind the receiver as a compensation filter for the
mode dispersion. The receiver achieved -22.1 dBm sensitivity at BER = 10-9. As a
result, a data rate of 265 Mbit/s was achieved.
With an improved Hamamatsu S4752 photodiode with 1.6 pF capacity at 10 V
reverse voltage and a diameter of 600 Pm of the active surface, the same sensiti-
vity was achieved without lens and compensator.
As is already known, the theoretical bandwidth of the standard NA-POF is
approximately 40 MHz 100 m. Consequently, the high data rates of the experi-
ments in question were not possible. Nevertheless, this value only applies to light
propagation in the equilibrium mode distribution. Figure 6.62 shows which
methods can be used to increase the bandwidth of the transmission system.
input output
Besides the compensation of the band limitations through transmitter, fiber me-
dium and receiver with suitable high-pass filters, it is particularly helpful to reduce
the number of the modes involved in data transmission and thereby the pulse
broadening. The following methods can be used individually or in combination
(values according to Bates):
Launch with small NA = 0.11. Only a few modes with small mode delay will
be excited.
Predistortion of the LD drive signal (peaking) with high-pass filter
(33 pF 51 :).
Output with small NA (modes with large mode delay are hidden).
Dispersion compensation behind the receiver with high-pass filter
(8 pF 20 :).
442 6.3 Overview of POF Systems
To the author's knowledge, the transmission of 531 Mbit/s over 100 m of SI-
POF is the fastest system in terms of the bit rate length product so far. In [Bat96a]
the theoretical limit was even estimated to be 1 Gbit/s over 100 m (which is
clearly proven in the meantime). The practical application of such systems is,
however, beset with difficulties since the filters must be dimensioned very
precisely and the specific parameters must be adapted to each individual link.
Through bendings and plug-in connections, the mode distribution is changed for
POF in such a way that the gain in bandwidth through selective launching and
detection lost at least in part.
Fiber type: SI-POF, Eska Extra EH4001, 139 dB/km @ 650 nm
Length: 100 m
Bit rate: 265 Mbit/s, 531 Mbit/s
Transmitter: Philips LD CQL82, 652 nm. -1.7 dBm
Receiver: AEG-Telefunken, BPW89, 0.4 A/W
Hamamatsu S7452, : 600 m
(comparable with the present product: S7482)
References: [Bat92], [Wal93], [Yas93], [Kuch94], [Bat96b]
Company: University of Essex
In cooperation with IBM and Keio University ([Kuch94]) a system was
developed in 1994 in which a surface emitting laser with a wavelength of 670 nm
was used as a transmitter. At a launch NA of 0.11, the averaged power in the fiber
was -10 dBm. A 400 m diameter Si-pin PD Hamamatsu S4753 served as the
receiver that was coupled to a GRIN lens. The sensitivity was -23.3 dBm at
1 Gbit/s (BER = 1.5 10-9).
A data rate of 531 Mbit/s was transmitted over 30 m of PMMA SI-POF. The
used POF was an INFOLITE F120 (Hoechst Celanese) with a core diameter of
500 m. The attenuation of this fiber is 130 dB/km at 650 nm or 300 dB/km at
670 nm. For 100 m of POF, the limit for the data rate was 300 Mbit/s for the
selected test setup.
Fiber type: 500 m SI-POF, Infolite F120 (Hoechst)
130 dB/km (650 nm); 300 dB/km (670 nm)
Length: 30 m
Bit rate: 531 Mbit/s
300 Mbit/s over 100 m
Transmitter: 670 nm VCSEL; -10 dBm
Receiver: 400 m PD, Hamamatsu S4753,
-23.3 dBm at 1 Gbit/s
Reference: [Kuch94]
Company: IBM, Keio University
The Swedish company Mitel has been working on the development of a 650 nm
VCSEL for use in POF systems for some time now. The resonant cavity LED
(RC-LED) can be regarded as the preliminary stage of the VCSEL and already has
a number of advantages such as the small radiation angle and the small wave-
length drift during changes in temperature.
6.3 Overview of POF Systems 443
The use of 650 nm RC-LED for the transmission of 250 Mbit/s over 30 m
SI-POF was demonstrated in [Stre98] and [Stre98b]. A BER of 3 10-10 was
achieved in the test. The RC-LED had a DBR mirror made of AlGaAs and an
active zone with 4 quantum wells of GaInP and barriers of AlGaInP. The cladding
layers consist of AlGaInP. The emitting area has a diameter of 84 m. At a center
wavelength of 660 nm, the RC-LED has a spectral width of 3 nm and emits 3 mW
optical power at 50 mA. The maximum power is 4.2 mW at 120 mA. The maxi-
mum external quantum efficiency (QEext) is about 3.2 %. The differential resis-
tance is specified by the authors to be 3 :.
A SI-POF with 980 m core diameters and AN = 0.48 was used. The effective
attenuation at 650 nm is 180 dB/km. The source was coupled to the POF directly
without any optics. At a DC current of 60 mA, the launched power was -2.2 dBm
(0.6 mW). A Si-pin photodiode (Tek P6701A) served as the receiver. The probabi-
lity of errors was estimated from the measuring the Q-factor (6.2). The data trans-
mission rate is limited by mode dispersion in the POF (fiber rise time = 2.85 ns,
corresponding to 44 MHz 100 m). It was possible to transmit 512 Mbit/s over a
1 m POF link.
Fiber type: SI-POF; AN = 0.48; 44 MHz 100 m
Length: 30 m; 1 m
Bit rate: 250 Mbit/s; 512 Mbit/s
Transmitter: 660 nm RC-LED
'O = 3 nm, 4.2 mW at 120 mA
Receiver: Si-pin photodiode (Tek P6701A)
References: [Stre98a], [Stre98b]
Company: Mitel
In one recent study [Schu01a] commercially available RC-LEDs for use with
lengths of 50 m to 100 m were investigated. Except for the use of a DSI-POF no
other measures which increased the bandwidth were employed. With 50 m of
DSI-POF the SNR loss (penalty) through mode dispersion at 500 Mbit/s still
amounts to 7 dB. This system permits the error-free transmission of 250 Mbit/s
and 125 Mbit/s with a standard POF over 50 m or 100 m respectively.
Fiber type: PMMA DSI-POF, AN = 0.30
Length: 50 m
Bit rate: 500 Mbit/s
125 Mbit/s over 100 m
and 250 Mbit/s over 50 m St.-SI-POF
Transmitter: 650 nm RC-LED
Receiver: Si-PD (tr, tf < 1 ns)
pre amplifier Infineon FOA 1061
sensitivity: -11.25 dBm
(7 dB penalty caused by the mode dispersion)
Reference: [Schu01a]
Company: Infineon Technologies
444 6.3 Overview of POF Systems
The last system presented in this section was set up at the DaimlerChrysler
Research Center in Ulm [Scha01]. It permitted the transmission of 500 Mbit/s
over 30 m of standard POF with commercial components in the receiver.
Fiber type: PMMA SI-POF
Length: 30 m
Bit rate: 500 Mbit/s
Transmitter: 650 nm LD
Receiver: pin-H125G-010 by OSI Fibercomm (400 m),
MAX3761 pre amplifier
-11.4 dBm sensitivity (butt coupling)
Reference: [Scha01]
Company: Daimler Chrysler research center Ulm
30 m PMMA-POF
650 nm LD AN = 0.50 Si-PD
500 Mbit/s 400 m
Fig. 6.64: Transmission of 500 Mbit/s over SI-POF with laser diode
The reaching of ever greater bit rates, on SI-POF, too, had a formative influ-
ence in the following years. The basis for this was less the development of better
components than the increased interest in fast systems over short distances.
SI-POF, AN = 0.48
655 nm RC-LED
622/1,000 Mbit/s 1m
400 Mbit/s Si-PD
10 m
Fig. 6.65: Data transmission with a 655 nm RC-LED
446 6.3 Overview of POF Systems
The transmission of larger data rates for use in the automotive field is examined
in [Scha00] (Fig. 6.67). Various PMMA SI-POF from Hchst, Toray and Siemens
(with Mitsubishi core, developed especially for MOST) were tested. With a
670 nm VCSEL and a fiber-coupled power of 0.32 mW a data rate of 500 Mbit/s
was transmitted over 10 m. In further tests a 650 nm laser with 50 m glass fiber
output (AN = 0.20) was used. A Tektronix converter P6701A with 850 MHz band-
width served as the receiver. 400 Mbit/s were attained over 10 m and 20 m Hchst
fiber medium (300 dB/km at 650 nm) and also 600 Mbit/s at 30 m. In all cases, the
eye diagram was measured and not the BER (Fig. 6.68).
50 m GOF
650 nm LD
400 Mbit/s
600 Mbit/s receiver
P6701A
850 MHz
PMMA-SI-POF
670 nm VCSEL Hchst (300 dB/km), 10 m, 20 m
500 Mbit/s Siemens, Toray (150 dB/km), 20 m, 30 m
0.32 mW
Fig. 6.67: Transmission experiments with different POF according to [Scha00]
10 m 980/1000 m SI-POF
160 dB/km at 657 nm
V-pin F-SMA
BIAS Peaking 657 nm PD broadband
2 stages LD S5052 receiver
24
20
16
12
8
Giehmann t [ns]
4
0.00 0.42 0.83 1.25 1.67
Fig. 6.70: Eye diagram at 1,200 Mbit/s
6.3 Overview of POF Systems 449
Different experiments with Gbit/s transmission using POF were also conducted
at the Fraunhofer Institute for Integrated Circuits (IIS) in Nuremberg. A commer-
cial laser diode for DVD applications was used as the transmitter. A relatively
small photodiode (330 m active diameter) was used, resulting in a coupling loss
of about 10 dB. A passive filter for dispersion compensation and a limiter ampli-
fier were employed for improving the signal. Different varieties of fibers with
lengths of up to 50 m were used.
Fiber type: SI-POF, DSI-POF, MC-POF
Length: 15 m to 50 m
Bit rate: 500 Mbit/s, 800 Mbit/s, 1,000 Mbit/s
Transmitter: 650 nm DVD-LD
Receiver: 330 m Si-pin-PD
Reference: [Jun04d]
Company: Fraunhofer IIS
The different tests result in the transmission of:
1,000 Mbit/s over 15 m SI-POF direct
1,000 Mbit/s over 20 m SI-POF with equalizing filter
500 Mbit/s over 50 m SI-POF
500 Mbit/s over 50 m DSI-POF
1,000 Mbit/s over 50 m MC-POF
(estimated theoretically,
not realized due to the limited received power)
In Figures 6.71 and 6.72 the measuring setup and an eye diagram for a S800
signal (effectively 1,000 Mbit/s) are illustrated. As can be seen, noise represents
the dominating system limitation. A definite improvement in the system could be
achieved with an optimized photodiode coupling, e.g. with a CPC.
Bit Error
Detector
(BERT)
TP1 TP2
Laser
Photodiode Dispersion
Driver &
& TIA Filter
Laser 20 m POF
Fig. 6.71: Measurement setup of the IIS for IEEE 1394-S800 transmission
450 6.3 Overview of POF Systems
The next two illustrations show an example of the S800 bit sequence after the
limiter and for the media converter already set up.
PD
S5052
high pass high pass
filter filter
SMA-
MMIC- MMIC- MMIC- MMIC- plug
amplifier amplifier amplifier amplifier
GALI-51 GALI-6 GALI-6 GALI-6
DSI-PMMA-POF
Asahi AC 1000 Si pin PD
BIAS LD 650 nm
core = 980 m, AN = 0.25 S 5052
SLD 1133VL
Above all the capacitance over short distances was investigated on standard
SI-POF. The fiber used was Toray PFU-CD-1001. Over 10 m and 20 m first
1,220 Mbit/s and then 820 Mbit/s at 650 nm was transmitted ([Ziem03g],
[Vin04b], [Vin05c], [Ziem05j]).
For lengths over 25 m it is surely necessary to fall back on other index profiles
for high bit rate applications. The limits of SI-POF with the components described
above on the POF-AC will also be investigated in an internship experiment,
whereby fiber pieces each 5 m in length are connected to FSMA plug connectors
and the maximum bit rate is measured with each length. Measurements up to 85 m
are possible - then with 15 plug connectors. Gigabit transmission is reached up to
about 30 m to 35 m. The results of two internship work groups are shown in
Figure 6.78.
Fiber type: Toray PFU-CD1001, 980 m PMMA SI-POF
Length: 15 m to 85 m (one connector after every 5 m)
Bit rate: up to 1,500 Mbit/s (20 m)
6.3 Overview of POF Systems 453
3000
bit rate [Mbit/s] max. values on
fiber without
connectors
1000 Siemens
POF-AC
PMMA-SI-POF
AN = 0.50
300 Toray
650 nm LD
Even better values can be attained with the same set-up when fewer plug-in
connectors are used. A maximum bit rate of 760 Mbit/s over 100 m of standard
POF could be attained with a two-stage passive equalizer (eye diagram in
Fig. 6.79). This surpasses both the results with optical mode filtering (533 Mbit/s,
[Bat96]) and the first calculations for multicarrier transmission up to 540 Mbit/s
([Ran06a]).
Fig. 6.79: Error free transmission of 760 Mbit/s over 100 m St.-POF
over 100 m of standard POF (one plug-in connection). If an error correction had
been used, even about 1 Gbit/s could have been attained. The eye diagram of the
measurement is shown in Fig. 6.80 (latest POF-AC results of summer 2007, over
1 Gbit/s, see next table).
Fig. 6.80: Error free transmission of 910 Mbit/s over 100 m St.-POF ([Was07])
20
P [dBm] transmitted spectrum
10
-10
-20
-30
-40
0 50 100 150 f [MHz] 200
Fig. 6.81: Equalizing of the frequency response in a multi carrier system
6.3 Overview of POF Systems 455
The measurement technique at hand did not yet permit any real time demo-
dulation. This is why only data packets were transmitted, recorded with a fast
oscilloscope and demodulated in a PC. The constellation diagrams for 2 typical
channels are shown in Fig. 6.82.
3,000
data rate [Mbit/s]
1,000
300
100
30
It is quite obvious that the PMMA SI and DSI-POF cover an essentially greater
area of use than most users perceive. Many of the experiments described above
have indeed been conducted under ideal laboratory conditions, but on the other
hand further improvements in the active components are possible and foreseeable.
The PMMA fiber has the lowest attenuation at wavelengths around 520 nm and
560 nm. Nevertheless most transmission systems and experiments were first
carried out in the third attenuation window around 650 nm. The reason for this
was the lack of suitable LEDs at the time of development of GaN technology in
the second half of the 1990s.
The different stages of transmission systems with blue, green and yellow LEDs
will be presented in the following sections.
573 nm LED
InGaAlP, 9 W
100 m SI-POF
TORX 196
Fig. 6.85: Transmission system according to [Fuk93]
6 power [dBelectr.]
3 HFBR-1527
0
-3
-6
-9 NSPG 500
-12
-15 HLMA-DL00
-18
-21
-24
frequency [MHz]
-27
0.2 0.5 1 2 5 10 20 50 100 200
Fig. 6.86: Bandwidth of different LED ([Ziem98d])
460 6.3 Overview of POF Systems
During this period the setup of the transmission system was not part of the
Berlin research groups work. In a test made in 1999 in cooperation with the
University of Ulm first 125 Mbit/s over 50 m and later 155 Mbit/s over 100 m of
DSI-POF could be transmitted ([Daum01a]).
In [Ino99a] the use of blue LED for a 125 Mbit/s - 100 m of POF-transmission
was under study. The LED was available as a chip and emitted a maximum output
power of 0.92 dBm (1.24 mW). Without peaking, -5.28 dBm (0.3 mW) was
launched into the POF. With optimized lenses and peaking, -3.62 dBm (0.43 mW)
were launched. The SI-POF used has an attenuation of 168 dB/km for the wave-
length used. Thus -22.1 and -20.5 dBm are available after 100 m POF at the fiber
output (without peaking/with peaking and lens optimization). The receiver has
-21.1 or else -22.1 dBm sensitivity (without/with peaking) at BER = 10-12. The
maximum modulating frequency of the blue LED is 120 MHz or 200 MHz respec-
tively without/with peaking.
The transmission can only be achieved with optimized launch. Although the
effective attenuation of the fiber medium is lower than at 650 nm, the parameters
are not yet sufficient for a practical application covering a range of 100 m. It
should be observed that the limits for eye safety for blue light are more restrictive
than for red light.
Fiber type: SI-POF, 168 dB/km at blue light
Length: 100 m
Bit rate: 125 Mbit/s
Transmitter: +0.92 dBm, 120 MHz bandwidth (200 MHz with peaking)
Receiver: -22.1 dBm sensitivity
Reference: [Ino99]
Company: Optowave Inc.
The use of green and blue LED for POF transmission was also described in
[Yago99]. The authors used commercial LED at 475 nm (as a chip) and 520 nm
(in a housing). The available power after 50 m POF (Eska Mega from Mitsubishi)
came to -14.6 dBm (blue) and -17.9 dBm (green). An input power of -5.1 dBm
was specified for the blue LED. The bandwidths of various LED ranged between
70 MHz and 120 MHz so that in each case 125 Mbit/s could be transmitted. No
special peaking was used; the NEC NL2100 (155 Mbit/s transceiver) was used as
the receiver. The following transmission experiments were conducted:
6.3 Overview of POF Systems 461
100 m POF
GaN-LED 126 dB/km Si-PD
495 nm, 100 Mbit/s at 495 nm
510 nm, 200 Mbit/s
Fig. 6.88: Data transmission using green LED at Firecomms
Fig. 6.89: Eye diagram with green LED, 200 Mbit/s over 100 m
Up to 310 Mbit/s could be transmitted over short distances (10 m) with single
LEDs (Agetha final report). In the meantime, Firecomms has been producing
green transceivers as commercial products, for the time being only up to a data
rate of 50 Mbit/s. The guaranteed data are mentioned below.
Fiber type: DSI-POF Asahi AC-1000
Length: 100 m
Bit rate: 60 Mbit/s
Transmitter: 520 nm HSG-LED, -9.7 dBm in the fiber
Receiver: Si-pin-PD
Reference: [Lam03a]
Company: Firecomms
Another company which has been investigating the transmission of data using
green LEDs is Toyota. Using their own LED, they transmitted 125 Mbit/s over
60 m of DSI-POF. This LED is also characterized by very small temperature
coefficients.
6.3 Overview of POF Systems 463
Fig. 6.91: Eye diagram for 250 Mbit/s transmission over 20 m POF
464 6.3 Overview of POF Systems
For some years now there has been new work going on in Italy (Luceat, ISMB:
Instituto Mario Boella and Politechnico di Turino) with the goal of increasing the
range of PMMA POF systems. In addition to using green LEDs optimized modu-
lation procedures, error correction codes and multi level coding have been
employed.
One example is a four level transmission, whereby an additional 5S/6S code
(out of 64 possible codes 54 are selected) in order to keep signals as free of direct
current as possible. For 100 Mbit/s the result is a symbol rate of 60 Mbaud/s and
100 m of SI-POF can be bridged. In Figs. 6.92 and 6.93 the effect of the improve-
ment in signal quality by means of the 5S6S code can be seen after 50 m and
200 m of POF respectively.
Fig. 6.92: Eyes: uncoded and coded for 50 Mbit/s over 100 m ([Gau05a])
Fig. 6.93: Eyes: 50 Mbit/s over 100 m (5S/6S) and 50 Mbit/s over 200 m [Gau05b]
Fig. 6.94: Eye for 150 Mbit/s over 50 m and 100 Mbit/s over 100 m, [Gau04a]
At the 2006 POF Conference in Seoul a data transmission of 100 Mbit/s over
200 m of SI-POF was presented as the latest result ([Nes06a] and [Nes06b]). A
green LED was used as a transmitter (DieMount, +3 dBm fiber-coupled power) as
described in the previous systems. An adaptive equalizer was now used to com-
pensate for the transmission behavior of the POF. Using 8-level coding the symbol
rate was 33 Mbaud/s. The receiving signal was sampled at 66 MSample/s and
further processed.
A noise gap of 19 dB is needed with FEC. It was possible to attain 26 dB
(margin of 7 dB) in experiments. The next step is to integrate the system into a
FPGA.
Fiber type: Luceat SI-POF, AN = 0.50, 105 dB/km (green)
Length: 200 m
Bit rate: 100 Mbit/s (8 level coded, 33 MBaud/s)
Transmitter: green DieMount LED; +3 dBm (pigtail)
Receiver: Hamamatsu-PD + TIA S6468-02
adaptive equalizer
References: [Nes06a], [Nes06b]
Company: Politechnico di Turino
In another experiment a date rate of 10 Mbit/s was transmitted. One particularly
powerful LED (DieMount), a large-surface Si pin-PD with a low-noise trans-
impedance receiver and a special code, allowed a range of 350 m. At the 2006
POF conference it was even possible to demonstrate a transmission over 425 m
(with 8B10B coding, Reed-Solomon code for FEC, [Car06a]).
466 6.3 Overview of POF Systems
Luceat (Italy) also offers a comparable transceiver with green LED as a com-
mercial product.
Fiber type: PMMA SI-POF
Length: 25 m to 200 m
Bit rate: 10 Mbit/s
Transmitter: 520 nm LED
Receiver: Si-pin-PD
Reference: [Luc04a]
Company: Luceat
Astri Technology Centre in HongKong has developed various products for POF
and above all for PCS. Presently in development are components for SI-POF on
the basis of green LEDs. At the 2005 POF Conference a complete module was
presented.
Fiber type: PMMA SI-POF, 70 dB/km
Length: 40 m
Bit rate: 20 Mbit/s
Transmitter: 520 nm LED, transmitter module: 9.7 u 6.2 u 3.6 mm
Receiver: Si-pin-PD
Reference: [Wip05]
Company: Astri Hong Kong
Some time ago Infineon Technologies also presented a development close to
production for data transmission with green LEDs. The system is supposed to
work at 125 Mbit/s over 100 m of DSI-POF and is characterized by low tempe-
rature coefficients.
Fiber type: DSI-POF, Eska Mega
Length: 100 m
Bit rate: 125 Mbit/s
Transmitter: 510 nm LED, 200 W, 0.23 %/K temperature dependency
Receiver: Si-pin-PD, -23 dBm sensitivity
Reference: [Witt03]
Company: Infineon Technologies
468 6.3 Overview of POF Systems
Fig. 6.97: Eye diagram, 125 Mbit/s over 150 m SI-POF, one connection
6.3 Overview of POF Systems 469
In 2006, using new LEDs - set up by DieMount with optimized coupling - even
greater data rates could be attained. 210 Mbit/s were able to be transmitted at
470 nm error free over 50 m of SI-POF. For a back-to-back measurement a modu-
lation of over 1 Gbit/s (eye diagram in Fig. 6.99) could be achieved for the first
time.
Fiber type: PMMA SI-POF, AN = 0.51
Length: 50 m
Bit rate: 210 Mbit/s
max. of 1,090 Mbit/s back-to-back
Transmitter: 470 nm LED, DieMount
Receiver: 800 m Si-pin-PD
Reference: [Ziem06h]
Company: POF-AC
470 6.3 Overview of POF Systems
All experiments in this section are summarized in Table 6.10 and Fig. 6.100.
These systems are of particular interest for distances up to 100 m.
1000
bit rate [Mbit/s]
300
100
blue LED
green LED
30
yellow LED
analog video
10
1 3 10 30 100 300 1000
length [m]
In the previous sections we have shown that you can also use SI-POF for data
rates of over 1,000 Mbit/s, however, for short distances only. Consequently, trans-
mitters in the near infrared range could obviously be used. The attenuation of the
PMMA fiber in this range is indeed considerably greater, however, distances of up
to 10 m as generally used in vehicle networks, can be bridged. The PMMA
attenuation curve indicating losses in 10 m lengths is shown in Fig. 6.101.
10,000
attenuation [dB/km]
|20 dB
3,000 application windows /10 m
for the PMMA-POF
1,000
|1.5 dB
/10 m
|7 dB
/10 m
300 |0.8 dB |0.7 dB
/10 m /10 m
|0.6 dB
/10 m
100
50
350 400 450 500 550 600 650 700 750 800 850 900
wavelength [nm]
Of particular interest is the window around 770 nm for short distance use. As
opposed to a wavelength of 650 nm the powerful VCSELs available are reasonab-
ly priced and can also be employed in a wide range of temperatures. Furthermore,
lasers at 780 nm are normally faster and the Si-PDs have a better sensitivity.
Fig. 6.102: POF for short haul data transmission according to [Schn98]
Fig. 6.104: Eyes for 900 Mbit/s over 15 m SI-POF and 1,000 Mbit/s over 15 m DSI-POF
780 nm LD receiver
125 Mbit/s -32.35 dBm
156 Mbit/s 85 m/ 80m /58 m SI-PC(AF)-POF -31.50 dBm
250 Mbit/s 300 dB/km at 780 nm -26.60 dBm
The maximum data rate over 2 m PMMA POF amounted to 2,000 Mbit/s.
Fiber type: PMMA SI-POF, PFU-CD 1000, 1,670 dB/km
Length: 10 m
Bit rate: 1,700 Mbit/s
Transmitter: 780 nm LD, POF coupled power: +4.7 dBm
received power: -12.0 dBm (10 m)
Receiver: 800 m Si-pin-PD Hamamatsu S5052
References: [Vin02b], [Ziem03f]
Company: POF-AC Nrnberg
10 m St.-PMMA-POF
Si pin PD
BIAS LD 780 nm Toray PFU-CD 1000
S 5052
Laser Comp. 1,670 dB/km
In the following year data transmission over a polycarbonate fiber was then
carried out (Mitsubishi PC-POF, core diameter of 1 mm). The attenuation of this
fiber of 900 dB/km at a wavelength of 780 nm clearly lies below the value of the
previously used PMMA fiber. Thanks to the greater mixing of modes data rates of
1,800 Mbit/s or 1,000 Mbit/s respectively over 10 m or 20 m of PC-POF can be
transmitted error free. The experimental setup is shown in Figure 6.109. The re-
ceived power (fiber) was -4.3 dBm and -14.8 dBm.
6.3 Overview of POF Systems 477
Si pin PD
BIAS LD 780 nm PC-SI-POF
Laser Comp. S 5052
Mitsubishi, 900 dB/km
Fig. 6.109: Transmission experiment with a 780 nm Laser (PC-SI-POF)
Fig. 6.111: Eye diagram for 1,000 Mbit/s over 23 m (SI-mod. PMMA-POF)
For a long time a SI-POF based on modified PMMA from Toray was commer-
cially available (PHKS CD1001 22P, usable up to +115C). As was shown in
[Ziem03e], 1,600 Mbit/s at 780 nm over 10 m could be transmitted. This fibers
attenuation was 1,950 dB/km, the received power was -14.8 dB at the PD after
10 m of fiber.
Another fiber which was tested at the POF-AC was the HPOF-S from Hitachi.
This fiber consists of a silicone material and has a cladding diameter of 1.5 mm.
2,200 Mbit/s can be transmitted over 10 m of the fiber; at 13.5 m it was still
1,700 Mbit/s.
Fiber type: HPOF-S (Hitachi), 1.5 mm
Length: 10 m, 13.5 m
Bit rate: 2,200 Mbit/s, 1,700 Mbit/s
Transmitter: 780 nm LD, +3.4 dBm in the POF
received power: -10.6 dBm (10 m); -15.4 dBm (13.5 m)
Receiver: 800 m Si-pin-PD Hamamatsu S5052
Reference: [Vin04b]
Company: POF-AC Nrnberg
Si pin PD
BIAS LD 780 nm 10 m SI-POF 1.5 mm S 5052
Laser Comp. Hitachi HPOF-S
5000
bit rate [Mbit/s]
2000
1000
500
200
length [m]
100
1 2 5 10 20 50 100
It is not without reason that the multimode, graded index and multi-core POFs are
treated together in one section. These three types of fibers are linked by the idea of
a definitely higher bandwidth than with SI-POFs. The difficulties in the produc-
tion of different index profiles have already been discussed in detail in Chapter 2.
As was to be expected, the system experiments concentrated on particularly high
data rates. Short-wave transmitters are not used because they are too slow and the
attenuation of the three POF types in this area is also generally too great. Wave-
lengths over 650 nm do not play any role either. In this case lengths are limited to
less than 10 m, for which the bandwidth of SI-POF is still enough for Gbit/s data
rates. Thus, all systems subsequently described work with 650 nm lasers.
480 6.3 Overview of POF Systems
90 m PMMA-GI-POF
654 nm LD : 550 m; NA = 0.24/0.30 GRIN Si-pin-PD
+4 dBm Keio University : 400 m
950 Mbit/s S4753
30 m
670 nm VCSEL PMMA-GI-POF GRIN Si-pin-PD
-10 dBm : 400 m
1,062 Mbit/s S4753
Fig. 6.114: PMMA GI-POF system with LD and VCSEL according to [Kuch94]
In 1994, Prof. Koike also presented the first POF transmission system with a
data rate of 2,500 Mbit/s at a range of 100 m ([Koi94], [Yam94], [Koi96c],
[Yam96b], [Ish95b]). A PMMA GI-POF with 200 dB/km attenuation at 647 nm
was used. The bandwidth is specified as 0.5 - 2 GHz 100 m. The POF-NA was
0.21; the core diameter of the fiber was 420 m.
A NEC laser diode with a 647 nm wavelength served as the source. When
coupled to the fiber with a GRIN lens, it was possible to achieve a launched power
of +6.1 dBm in the POF. A Si-pin PD with a diameter of 400 m coupled to a
GRIN lens and FET amplifier served as the receiver with which a sensitivity of
-16.9 dBm at BER = 10-9 was attained. At a fiber length of 100 m, the result was a
deterioration in the sensitivity (penalty) of 0.6 dB through mode dispersion.
Figure 6.115 illustrates the system principle.
Fiber type: PMMA GI-POF; 200 dB/km at 647 nm; AN = 0.21
Length: 100 m
Bit rate: 2,500 Mbit/s
Transmitter: NEC LD 647 nm; +6.1 dBm
Receiver: 400 m Si-pin-PD; -16.9 dBm at 2.5 Gbit/s
References: [Koi94], [Yam94], [Koi96c], [Yam96], [Ish95]
Company: Keio University
PMMA GI-POF was investigated by Boeing as part of the HSPN project for an
optical network in planes ([Krug95]). It was intended for use in the Boeing 777.
An optical network based on 100 m/140 m glass fibers will be used for the first
482 6.3 Overview of POF Systems
time commercially in this plane. In the tests, GI-POF with 750 m diameter
(600 m core) at wavelengths of 650 m were used. Data rates of 10 Mbit/s and
100 Mbit/s were transmitted over a maximum of 30 m. Figure 6.116 shows the
architecture of such an onboard network using optical connections based on POF.
The system was planned with two different transmitters. The available LED are
able to launch -8.5 dBm max. power into the POF; with VCSEL it should be able
to reach 0 dBm. The receiver developed by Honeywell had a minimum sensitivity
of -31 dBm. The PMMA GI-POF used had a typical fiber attenuation of
145 5 dB/km at 650 nm. The typical connector attenuation was 1.5 0.5 dB. At
a permissible temperature range of -40C to +85C, the max. power must not
exceed 1 mW in order to guarantee eye safety. A service life of 20 years should be
reached.
Fiber type: 600 m PMMA GI-POF, 145 5 dB/km at 650 nm
Length: 30 m
Bit rate: 10 and 100 Mbit/s
Transmitter: 650 nm LED, -8.5 dBm (VCSEL planned)
Reference: [Krug95]
Company: Boeing
CMF or MAT
PC with FDDI-interface card
100 Mb/s over 600/750 m
GI-POF at 0.65 m
Fig. 6.116: Boeing POF test for avionic networks according to [Krug95]
-12.3 dBm
650 nm
RC-LED
250 Mbit/s 50 m GI-POF, Mitsubishi
Si-PD
180 dB/km
-23.7 dBm
-4.2 dBm
50 m GI-POF, Mitsubishi
650 nm Si-PD
180 dB/km
RC-LED -18.7 dBm
500 Mbit/s
50 m MC-POF
650 nm LD
1 mm Asahi (37 cores) Si-pin PD
500 Mbit/s
Fig. 6.120: 500 Mbit/s system experiment with multi core POF
The transmission of 2.5 Gbit/s over 200 m was also successfully demonstrated
at the University of Eindhoven in 1998 ([Khoe99]). The Mitsubishi PMMA
GI-POF had an attenuation of 164 dB/km at 650 nm. A 645 nm NEC laser diode
served as the source and had a spectral width of 0.4 nm and a maximum optical
power of +6.8 dBm (4.8 mW). A Si-APD was used for the receiver which made it
possible to attain -29 dBm of sensitivity at a BER of 10-9 (shown schematically in
Fig. 6.121).
Fiber type: 500 m PMMA GI-POF; Mitsubishi; 164 dB/km at 650 nm
Length: 200 m
Bit rate: 2,500 Mbit/s
Transmitter: 645 nm NEC LD; +6.8 dBm
Receiver: Si-APD; -29 dBm
Reference: [Khoe99]
Company: University of Eindhoven
486 6.3 Overview of POF Systems
With the same fiber the Fraunhofer Institute in Erlangen realized a transmission
of a Gigabit Ethernet over 50 m. The transceiver used is so designed that it can be
integrated into a SC-RJ connector (Fig. 6.123). Transmission is also possible with
15 m standard SI-POF.
Fiber type: 900 m, PMMA GI-POF, Optimedia
Length: 50 m
Bit rate: 1.250 Mbit/s
Transmitter: 652 nm LD, 5 mW
Receiver: 800 m Si-pin-PD with commercial TIA
-12.5 dBm sensitivity (BER = 10-12)
Reference: [Off05]
Company: Fraunhofer IIS Erlangen
Fig. 6.123: Transceiver for 1.25 Gbit/s over 50 m GI-POF and eye after 50 m OM-Giga
DSI-MC-PMMA-POF
50 m with 800 Mbit/s Si pin PD
BIAS peaking LD 650 nm
S 5052
filter SLD 1133VL
Fig. 6.124: Data transmission on MC-POF
488 6.3 Overview of POF Systems
From 2003 on, these tests at the POF-AC Nrnberg were continued with im-
proved components. At first a data rate of 630 Mbit/s was transmitted over a
100 m long MSI fiber from Mitsubishi.
Fiber type: 700 m MSI-POF, Mitsubishi ESKA-MIU
Length: 100 m
Bit rate: 630 Mbit/s
Transmitter: 657 nm LD, Sony SLD 1133VL
Receiver: 800 m Si-pin-PD Hamamatsu S5052
Reference: [Vin04b]
Company: POF-AC
The data transmission with multi-core fibers was tested at the POF-AC with
two different versions. First, the 217 core fiber with a simple step index profile
was used. Three fiber samples, 21 m, 44 m, and 90 m in length, were available.
The measurements were carried out with higher power Sanyo lasers and a trans-
impedance receiver.
Fiber type: 1 mm Asahi MC-POF, 217 cores
Length: 21 m, 44 m and 90 m
Bit rate: 900 Mbit/s, 750 Mbit/s, 590 Mbit/s
Transmitter: DL-4147-162 Sanyo, +8.6 dBm
Receiver: 800 m Si-pin-PD Hamamatsu S5052, transimpedance
receiver with HEMT
received power: +4.5 dBm, -0.75 dBm and -10.2 dBm
Reference: [Vin04b]
Company: POF-AC Nrnberg
The second type of multi-core fiber tested was a POF with 37 cores and double
step index profile. Over distances between 30 m to 100 m data rates of
1,400 Mbit/s to 800 Mbit/s could be transmitted, whereby the capacity lay in the
same range as with the MSI-POF. However, the MC POF offers the additional
advantage of having very small bending radii.
Fiber type: 1 mm Asahi MC-POF, MSC-1000
Length: 30 m, 50 m, 64 m and 100 m
Bit rate: 1,400 Mbit/s, 1,300 Mbit/s, 1,200 Mbit/s, 800 Mbit/s
Transmitter: DL-4147-162 Sanyo, POF coupled power: +8.6 dBm
Receiver: 800 m Si-pin-PD Hamamatsu S5052
transimpedance receiver with HEMT
6.3 Overview of POF Systems 489
Later the experiments were repeated on both large fiber lengths using an addi-
tional passive compensation filter (RC high pass), whereby over 1 Gbit/s over
100 m of MC POF could be attained for the first time.
Fiber type: Asahi MC-POF, 1 mm
Length: 64 m, 100 m
Bit rate: 1,270 Mbit/s, 1,150 Mbit/s
1,170 Mbit/s over 100 m with optimized equalizer
Transmitter: 650 nm LD, 5 mW
Receiver: 800 m Si-pin-PD S5052 with transimpedance receiver
References: [Vin05a], [Vin05c]
Company: POF-AC Nrnberg
The latest measurements for two different fibers from [Was07] resulted in
maximum bit rates of 725 Mbit/s over 90 m for the 217-core fiber and
1,170 Mbit/s over 100 m for the 37-core fiber (eye diagram in Fig. 6.127).
Fig. 6.127: Error free transmission of 1,170 Mbit/s over 100 m MC-POF at 650 nm
Since 2005, the PMMA GI-POFs of the Korean manufacturer Optimedia have
been tested. By using red laser diodes, the maximum data rate has been limited to
approximately 1,600 Mbit/s. Even after 100 m with an error-free transmission of
1,550 Mbit/s no serious influence on mode dispersion could be made out. With an
optimized equalizing filter and a new laser 2 Gbit/s were able to be transmitted
over 50 m fiber (Fig. 6.128).
490 6.3 Overview of POF Systems
The bit rate has also been measured with these new components during the de-
partments internship work. An example for the measurement results at different
lengths is shown in Fig. 6.129. Since there are already 9 plug-in connectors in the
link over 100 m, the bit rate is limited by the receiving level necessitating use of a
narrower low pass filters for noise suppression.
Fiber type: OM-GIGA, 900 m PMMA GI-POF
Length: 20 m to 100 m (one connection every 10 m)
Bit rate: up to 2,250 Mbit/s
Transmitter: 650 nm LD
Receiver: 800 m Si-pin-PD
Reference: [Gort06]
Company: POF-AC Nrnberg
3,000
bit rate [Mbit/s]
POF ribbon cables with OM Giga were used also for transmitting the data, as is
described later on. Each of four parallel channels can transmit 1.6 Gbit/s over
50 m whereby the specification for the mask is met quite well in the eye diagram.
We can summarize the potential of the different index profiles as follows:
Multi-core fibers permit transmission of 500 Mbit/s to 1,000 Mbit/s over up to
100 m of fiber, especially when using double step index profile. Furthermore,
they have the advantage that materials and production are similar to standard
POF. MC-POF allows extremely small bending radii which is important for
installation.
Multi-step index fibers are easier to produce than GI-POF (multiple extruder).
At present, only one type from Mitsubishi is available which allows up to
about 500 Mbit/s over 100 m (comparable to DSI-POF).
Graded index PMMA POFs allow 2,500 Mbit/s over 100 m and more. The
fibers most readily available at present are those from Optimedia. The greatest
remaining problem is that of limited bending radius which can, however, be
reduced though an improved primary coating.
have been realized with fibers from Asahi Glass. It was no until the past few years
that fibers from Nexans (Lyon) have also been used. The third manufacturer is
Chromis Fiberoptics whose fibers, however, have not yet been used in published
system experiments (except Prof. Ralph).
Today almost all PF-GI-POF systems are manufactured with a core diameter of
120 m and NA of 0.22 to 0.25. At the beginning of development greater values
for both parameters were also partially used. The essential developments are in
regard to the optimization of the index profile and the continued lowering of the
attenuation.
100 m PF-GI-POF
850 nm Si-PD
VCSEL 120/200 m CYTOP
GaAs
PD
GaAs
PD
50 m PF-GI-POF
865 nm
120/200 m CYTOP
VCSEL
Fig. 6.132: Parallel data transmission over GI-POF according to [Kan98]
In [Imai97] 200 m of fluorinated GI-POF was used to transmit 2.5 Gbit/s. The
fiber has an attenuation of 120 dB/km at 850 nm and 56 dB/km at 1,300 nm (see
also [Khoe99]). A 1,310 nm laser was used as the source (see Fig. 6.133).
Fiber type: 120 m PF-GI-POF CYTOP, 56 dB/km at 850 nm
Length: 200 m
Bit rate: 2,500 Mbit/s
Transmitter: 1,310 nm LD
Receiver: 30 m APD with lens coupling
-25.7 dBm sensitivity (BER = 10-10)
Reference: [Imai97]
Company: Fujitsu Laboratories Inc.
200 m PF-GI-POF
1,310 nm LD 56 dB/km PD
VCSEL 100 m
850 nm PD
100 m PF-GI-POF
: 83 m, 99 m, 147 m, 221 m
Fig. 6.134: Test of different GI-POF with glass fiber components ([Wat99a])
The use of commercial fiber glass transceivers with a 850 nm VCSEL was
tested in [Lin01], whereby fiber lengths up to 300 m and bit rates up to 3.2 Gbit/s
were used.
Fiber type: PF-GI-POF
Length: 300 m
Bit rate: 3,200 Mbit/s
Transmitter: 850 nm VCSEL
Receiver: 120 m GaAs-pin-PD
Reference: [Lin01a]
Company: True-Light Corporation
also 1,250 Mbit/s over 100 m, 200 m and 300 m
6.3 Overview of POF Systems 495
300 m PF-GI-POF
850 nm VCSEL 40 dB/km 120 m
3.2 Gbit/s GaAs-pin-PD
Three fiber pieces each with a length of 100 m were available and connected
with plugs. The total link attenuation at the laser wavelength came to 32.6 dB. A
1 dB penalty was measured through mode dispersion. In addition, there were the
0.6 dB coupling losses at the transmitter and receiver which required a trans-
mission power budget of 34.2 dB. This was possible thanks to the 35.2 dB diffe-
rence between the transmitter power and the sensitivity.
Fiber type: 170 m PF-GI-POF CYTOP, 32.6 dB/300 m at 645 nm
Length: 300 m (2 connectors)
Bit rate: 2,500 Mbit/s
Transmitter: 645 nm Laser NEC; 'O = 0.4 nm, +6.2 dBm
Receiver: 230 m Si-APD; -29 dBm
Reference: [Li98]
Company: University of Eindhoven
A new distance record of 2.5 Gbit/s over 450 m was presented by the same
authors also in 1998 ([Li98]). A 1,310 nm LD was used. The fiber pieces used
were 4 u 100 m and 1 u 50 m GI-POF. The laser was coupled to a 62.5 m GI
glass fiber. An optical amplifier (SOA, Semiconductor Optical Amplifier) in-
creased the power to the required level, which is shown in Fig. 6.137.
The transmitting power attained were not specified, but must have been appro-
ximately 10 mW. With this setup, 5 Gbit/s were transmitted over 140 m and later
200 m, whereby the bandwidth of the receiver was cited as the limiting factor.
Furthermore, a transmission length of 300 m with 2.5 Gbit/s was achieved without
using the SOA.
Fiber type: 170 m PF-GI-POF CYTOP, 31 dB/km
Length: 450 m (4 connectors)
Bit rate: 2,500 Mbit/s
Transmitter: 1,310 nm LD, with SOA amplified
Receiver: 80 m APD
Reference: [Li98]
Company: University of Eindhoven
also 5,000 Mbit/s over 140 m and 200 m
6.3 Overview of POF Systems 497
GOF, 62.5 m
1,310 nm SOA
4 u 100 m, 1 u 50 m
LD 170/340 m PF-GI-POF
APD
Fig. 6.137: 2.5 Gbit/s system according to [Li98] with 450 m reach
The group finally improved the transmission length to 550 m in 1999 (see
[Khoe99], [Li99]) at 2.5 Gbit/s data rate. This was made possible by providing a
550 m GI-POF fiber piece with a core diameter of 170 m without any connectors
(Fig. 6.138).
Experiments with various sources were carried out. The measured attenuation
for the wavelengths was as follows:
110 dB/km at 650 nm (LD as source)
43.6 dB/km at 840 nm (VCSEL as source)
31 dB/km at 1,310 nm (LD as source)
The VCSEL supplies 1.3 dBm of power at a spectral width of 1 nm. It was
possible to couple it directly to the POF (< 1 dB loss). A passive filter for the
VCSEL frequency response compensation was used.
A Si-APD with 230 m diameter was used for the receiver at 840 nm. It
reached -28.6 dBm sensitivity with a BER = 10-9, whereby a budget of 29.9 dB
was available. The experiments resulted in 4.5 dB penalty through mode noise and
dispersion and 24.0 dB attenuation through the 550 m POF link (24.0 + 1.0 + 0.3
+ 4.5 dB yields 29.8 dB).
Fiber type: 170 m PF-GI-POF CYTOP; 43.6 dB/km
Length: 550 m
Bit rate: 2,500 Mbit/s
Transmitter: 840 nm VCSEL, 'O = 1 nm, 1.3 dBm
Receiver: 230 m Si-APD
Reference: [Li99]
Company: University of Eindhoven
Si-APD 230 m
Fig. 6.138: POF system with record transmission distance according to [Li99]
498 6.3 Overview of POF Systems
The transmission of 2.5 Gbit/s over 550 m at 1.3 m was also described in
[Li99]. The used 1310 nm DFB laser had a modulation bandwidth of 5 GHz, a
spectral width of 0.1 nm and max. 0.4 dBm of optical output power (1.1 mW). The
laser is a standard transmitter element for singlemode fiber systems and is
equipped with a corresponding fiber pigtail for singlemode fiber systems. The
singlemode fiber was also used for direct coupling to the GI-POF (< 0.1 dB loss).
With this method, only a small part of the mode field is excited, which increases
the bandwidth considerably.
The receiver used for this wavelength was a InGaAs-APD with a diameter of
80 m. The POF was imaged with a dual lens while changing the NA from 0.25 to
AN = 0.55 (< 0.3 dB loss). The sensitivity was -28.4 dBm with a BER of 10-9.
Thus, a transmission budget of 28.8 dB was available by a loss of 16.3 dB. A
measured penalty of 4.4 dB through mode noise and dispersion resulted in the
required budget of 16.3 + 0.1 + 0.3 + 4.4 = 21.1 dB. The remaining system margin
of 7.7 dB would make a transmission length of up to 750 m possible. Figure 6.139
illustrates the system scheme.
Fiber type: 170 m PF-GI-POF CYTOP, 31 dB/km
Length: 550 m
Bit rate: 2,500 Mbit/s
Transmitter: 1,310 nm LD, 'O = 0.1 nm, 0.4 dBm
Receiver: 80 m GaAsP-APD, -28.4 dBm
Reference: [Li99]
Company: University of Eindhoven
SM-GOF
1,310 nm
DFB-LD
The power budgets of both the 550 m experiments and 1,310 nm are compared
in Fig. 6.140. The clear advantage lies with the 1,310 nm laser diode, since the
attenuation of the POF is essentially lower here. However, the components used
are not in keeping with the low-cost philosophy of polymer fibers.
6.3 Overview of POF Systems 499
Pout
VCSEL sensitivity Si-APD
Pout
DFB-LD sensitivity InGaAs-APD
In 2001 and 2002, the group at the University of Eindhoven showed that
PF-GI-POFs are also suitable for transmission lengths of up to 1 km. In one of the
first experiments an 840 nm VCSEL was used. The fiber consisted of three cas-
caded, 330 m pieces (system setup in Fig. 6.141).
Fiber type: PF-GI-POF, 27 dB/km at 840 nm
Length: 990 m
Bit rate: 1,250 Mbit/s
Transmitter: 840 nm VCSEL with +1.1 dBm average opt. power (1.3 mW)
Receiver: 230 m Si-APD, sensitivity -31.3 dBm (BER = 10-9)
penalty 1.2 dB by mode dispersion
References: [Boo01a], [Nar01]
Company: University of Eindhoven
In the following year a 1 km-long piece of fiber was available. This time a
1,300 nm edge emitter, coupled to a single mode fiber, was used, the output power
of which was enlarged by means of an optical semiconductor amplifier.
Fiber type: 120 m PF-GI-POF
Length: 1,006 m
Bit rate: 1,250 Mbit/s
Transmitter: 1,300 nm LD with semiconductor optical amplifier
Receiver: 80 m InGaAs-APD
Reference: [Khoe02]
Company: University of Eindhoven
500 6.3 Overview of POF Systems
80 m GI-POF
BIAS VCSEL InGaAs
930 nm 155 m CYTOP
pin-PD
The highest data rate of yet for a POF system with 11 Gbit/s over 100 m PF-GI-
POF was demonstrated by Lucent Technologies in [Gia99a]. The source was a
1,300 nm Fabry-Perot laser diode. Via the singlemode fiber coupled to the laser it
was possible to couple 1 mW of power into the GI-POF through a lens. The
attenuation of the GI-POF with a core/cladding diameter of 130 m/300 m was
44 dB/km at 830 nm and 33 dB/km at 1,300 nm.
The receiver was a pin photodiode with fixed glass fiber pigtail (62.5 m
multimode). A lens provides the coupling between the POF and receiver glass
fiber with a loss of 4.8 dB. The error probability is less than BER = 10-10 with
-8.6 dBm received power. Figure 6.144 illustrates the test set-up.
6.3 Overview of POF Systems 501
The laser used fulfilled class 1 with less than +8 dBm of output power. The
penalty due to dispersion amounted to 2.5 dB. This system should also be regar-
ded as a technology test for performance of the GI-POF, since the test setup by no
means met the demand for inexpensive components.
Fiber type: 130 m PF-GI-POF CYTOP; 33 dB/km
Length: 100 m
Bit rate: 11,000 Mbit/s
Transmitter: 1,300 nm LD, +8 dBm
Receiver: pin-PD (with a 62.5 m GI-GOF pigtail)
References: [Gia99a]
Company: Lucent Technologies
33 dB/km
100 m PF-GI-POF
FP-LD 10 m 62.5 m pin-PD
130/300 m, Asahi
1,300 nm SM-GOF GI-GOF
Fig. 6.144: Up to that time highest bit rate for POF systems at Lucent Technologies (2000)
100 m PF-GI-POF
VCSEL 130/300 m, Asahi 70 m
830 nm 44 dB/km at 830 nm pin-PD
850 nm VCSEL (50 :) with pin monitor diode and SiGe driver
TIA receiver with pin-diode
5 mA BIAS current, 7.5 mA Imod,p-p
f3 dB: 5.5 GHz (f6 dB: 8.0 GHz) with mode filter
850 nm VCSEL
10.7 Gbit/s, PRBS 1023-1
coupling with ball lenses
BER < 10-12, BER < 10-10 with optimized launch
100 m
10.7 VCSEL Lucina pin-PD
Gbit/s 850 nm PF-GI-POF
9 -120 m 120 m
butt coupling MM fiber
1300 nm mode
VOA
DFB laser 9 m 9 m 120 m mixer
SM fiber SM fiber MM fiber
120 -120 m
butt coupling
50 -120 m
butt coupling
BER pin-PD
MLSE 50 m 120 m
tester receiver
MM fiber PF-GI-POF
Fig. 6.149: 10 Gbit/s system according to [Lee07a]
10-1
10-2 BER
10-3
10-4
10-5
10-6
10-7
10-8
10-9 fiber length [m]
10-10
60 100 140 180 220
Fig. 6.150: System improvement by MLSE ([Lee07a])
6.3 Overview of POF Systems 505
In principle, this procedure can be used with most dispersion-limited POF sys-
tems in order to gain some additional dB in the system. With PMMA POF this
only corresponds to an increased length of some 10 m.
The greatest bit rates for POF systems realized so far were achieved at the
Georgia Institute of Technology [Ral06], [Ral07] and [Poll07]. A PF-GI-POF
from Chromis Fiberoptics served as the medium. Since transmitters with 1,300 nm
are not sufficiently fast, a 1,550 nm source and a correspondingly fast receiver
were used. As the attenuation of PF-POF at 1.55 m is already relatively large
( 100 dB/km) the transmission length was limited to 30 m. A BER < 10-12 could
be achieved for up to 30 Gbit/s. At 40 Gbit/s you can transmit with an error proba-
bility of 1.45 10-3 (receiver-limited). The fiber diameter permits an offset of
r10 m when coupling-in. The eye diagrams for 10 Gbit/s and 30 Gbit/s are
shown in Fig. 6.151.
Fiber type: 50 m PF-GI-POF
(Chromis Fiberoptics)
Length: 30 m
Bit rate: 10,000 - 40,000 Mbit/s
Transmitter: 1,550 nm fiber laser
with external modulator
Receiver: 50 m multimode fiber detector Newfocus 1454
POF output power: 3.87 dBm
References: [Ral06], [Ral07], [Poll07]
Company: Georgia Institute of Technology
In addition to the bit rate, the impulse response is also determined. A compa-
rison with a measurement on a 50 m glass fiber (GI) is shown in Fig. 6.152.
Thanks to the strong mode mixing in the POF visibly less modal dispersion
occurs.
A comprehensive explanation of the effects of mode mixing on the bandwidth
of PF-GI fibers can be found in Section 10.3.
506 6.3 Overview of POF Systems
All of the system experiments mentioned above with PF-GI-POF are summa-
rized in Fig. 6.153. The fiber limit lay currently in the range of 1,250 Mbit/s km.
The performance is practically identical at 850 and 1,300 nm - a definite advan-
tage over glass fibers. Both the range (higher fiber attenuation) as well as the data
rate (slower lasers) are clearly limited.
Mode multiplex systems are introduced in Section 6.3.7.4. The transmission of
2 u 10.7 Gbit/s over 10 m of PF-GI-POF at a wavelength of 1,550 nm was
achieved by [Sch06].
10,000
3,000
1,000
1,550 nm
1,300 nm
300 850 nm
650 nm length [m]
100
10 20 50 100 200 500 1000
transmission
POF 2
1 POF
light pathes
3 1 2 3
filter
wavelength
All of these methods will be described in the following sections. The problem
is far less critical for GI-POF with its small core diameters and NA common today
compared with 1 mm SI-POF.
WDM systems on PMMA fibers will first be described. In the following part
solutions with PF-GI-POF will then be compiled and finally the components for
bi-directional transmission on fiber will round off the presentation.
The use of wavelength multiplex on PMMA POF will be limited to a few chan-
nels. Laser diodes are available only for 650 nm; in the other transmission win-
dows we have so far had to use LEDs which show a high spectral width as well as
limited bit rates. On the other hand, the large fiber diameter allows for very simple
multiplex constructions without active adjustment.
Bi-directional transmission with a system with a 830 nm source for a 6 MHz
video channel and a 660 nm LED for a 10 kHz control signal was already demon-
strated in [Tak94]. No further information on the test setup is available.
Fiber type: PMMA POF
Bit rate: analog 6 MHz / 10 kHz
Transmitter: 830 nm (6 MHz video signal)
660 nm LED (10 kHz control signal)
Demux: splitter
Reference: [Tak94]
Company: Hitachi research center
In [Ziem97a] and [Ziem97b] a method was proposed that markedly improves
the performance of the system. The fact that the diameter of the photodiode is
markedly larger than that of a typical LED was exploited. By simply "stacking"
the elements, an WDM system for bi-directional transmission is created. This is
show in Fig. 6.155.
photo diode
1 mm 1 mm POF
1 mm SI-POF
0.75 mm GI-POF
grating
The wavelengths used are 645 nm and 675 nm, dependent on the available laser
diodes. A grating with 1,200 lines/mm and 500 nm Blaze wavelength was selected
for the setup. The collimator lens has a focal length of 25.4 mm which results in a
theoretical linear separation of the wavelengths by 995.8 m. This corresponds
very precisely to the 1 mm spacing between the output fibers. The lens diameter is
25.4 mm which is large enough to capture the complete far field of the GI-POF.
Figure 5.74 illustrates the transmission function for both multiplexer outputs,
measured with a white-light source before the GI-POF and with a resolution of
0.1 nm.
-50
rel. power [dB]
-55
-60
-65
-70
-75
-80
wavelength [nm]
-85
610 620 630 640 650 660 670 680 690 700 710
Fig. 6.157: Demultiplexer spectral transmission according to [Hun96]
510 6.3 Overview of POF Systems
The attenuation at the used wavelengths is less than 5 dB. The suppression of
the alternate channel is better than 55 dB which should make error-free operation
possible using a small band laser.
A WDM system with 84 m PMMA GI-POF was set up with this demultiplexer
in [Khoe97]. Both transmitters were a NEC 645 nm laser with a data rate of
2,500 Mbit/s and a Philips 675 nm CD laser with 620 Mbit/s. The sensitivities of
both receivers were -26 dBm and -31 dBm respectively.
Fiber type: 750 m PMMA GI-POF
Length: 84 m
Bit rate: 2 u 2,500 Mbit/s
Transmitter: 645 nm LD/675 nm LD
Demux: grating demultiplexer
References: [Hun96], [Khoe97]
Company: University of Eindhoven
A Tunisian group proposed a further arrangement of the demultiplexer for
POF-WDM systems in [Att96]. The goal was to achieve as compact a form for the
multiplexer as possible. For this purpose, a POF was coupled to a glass fiber
bundle with a round cross-section. It consisted of 61 single fibers with a core dia-
meter of 100 m with AN = 0.28. Approximately 60 % of the 1 mm core diameter
of the POF is utilized. The individual fibers are arranged linearly at the focusing
lenses side. This results in a slit of 6.1 u 0.1 mm. The advantage of the arrange-
ment is that a considerably smaller linear separation of the channels is required
compared with a round 1 mm input, as demonstrated in Fig. 6.158. Smaller lenses
and gratings thus can be used.
0.1 mm
O1
O1
1 mm
O2
O2
input
O1 + O2
input
O1 + O2
A grating with 1,800 lines /mm and a 8 mm u 8 mm size is used for the realized
demultiplexer. Both focusing lens have a 3.6 mm focal width. The wavelengths:
O1 = 632.8 nm, O2 = 650 nm, and O3 = 670 nm are used. The complete additional
loss of the arrangement with these wavelengths is specified as 7 dB. A specific
transmission system has not yet been realized.
6.3 Overview of POF Systems 511
1.0 P [a.U.]
0.8
0.6
0.4
0.2
O [nm]
0.0
400 450 500 550 600 650 700 750
The de-multiplexer was optically realized by expanding with lenses and using
interference filters. The crosstalk was reduced electrically by means of an analog
compensation circuit. To do this, coupling coefficients between the channels were
first measured and afterwards adjusted by tunable controllers (Fig. 6.160). The
system was tested at a bit rate of 10 Mbit/s. In order to demonstrate the crosstalk
compensation, Fig. 6.160 shows pulse trains with and without compensation
circuits.
+
R1 = a11S1 + a12S2 -
+
R2 = a22S2 + a21S1 -
Fig. 6.160: Near end cross talk compensation in the POF-WDM system
512 6.3 Overview of POF Systems
-3
-6
NSPB 500S
NSPG 500S
SFH 757 f [MHz]
-9
1 3 10 30 100 300
I1 I1
Video LD 650 nm receiver Video
I2 I2
Audio LED 520 nm receiver Audio
fiber
I3 I3
Data receiver LED 465 nm Data
O-MUX/DEMUX O-MUX/DEMUX
0
transmission [dB]
-5
-10
-15
-20
-25
O [nm]
-30
400 450 500 550 600 650 700
Fig. 6.163: Transmission of the 3 channel demultiplexer (FhG IIS)
A WDM system for use in the lessons has been developed by the University of
Applied Sciences Harz (Hochschule Harz) as part of their Optomux project
([Fis06a], Fig. 6.165).
Fiber type: St.-SI-POF
Length: 25 m
Bit rate: 3 u 60 Mbit/s
Transmitter: 470 nm, 530 nm and 660 nm LED
Multiplexer: Prism
Reference: [Fis06]
Company: University of Applied Sciences Harz, Harz-Optics
50 m PF-GI-POF
120/200 m CYTOP GaAs-PD
photo
dielectrical diodes
laser diodes interference filters
250 m GI-POF
125 m 125 m
113 m 125 m
O3
O2
O4
1150 1200 1250 1300 1350 1400 1450 1500 1550 1600
wavelength [nm]
Each of the 8 wavelength channels are 10 nm wide and are oriented according
to the available lasers. The large width of the channels and the minimum spacing
of 20 nm make it possible to use non-stabilized sources and relatively simple
filters in the multiplexers. The transmitters operate with LD that transmit the sig-
nal with pulse frequency modulation (PFM-IM) at a carrier frequency of 80 MHz.
pin-PD were used as receivers for wavelengths of 1,200 to 1,600 nm.
Figure 6.171 shows the multiplexers that have been created with planar wave-
guides. A groove was made in the substrate in which the interference band-pass
filters have been inserted. The component can be used to add or drop a specific
wavelength.
O1
O2
waveguides
Due to the relatively small NA of the fibers used (0.2), the angle differences of
the various light paths are not very large. The distance between the wavelengths
used in the experiment was 40 nm, i.e., approximately 3 %. This makes the use of
interference filters possible.
At first, the system setup was successfully operated with 3 wavelengths over
100 m and with 4 wavelengths over 50 m. The more multiplexers are inserted, the
shorter the distance to be covered since the power budget worsens. The author's
estimates [Ueh98] show the following transmission lengths:
with 2 wavelengths, 250 m is possible
with 3 wavelengths, 150 m is possible
with 4 wavelengths, 100 m is possible
with 5 wavelengths, 50 m is possible
According to [Ueh99] the test wavelengths were O2 = 1,265 nm, O3 = 1,305 nm,
and O4 = 1,345 nm. For Point-to-Point transmission 500 Mbit/s and 1 Gbit/s over
100 m are possible with these sources and the GI-POF. The maximum emitted
518 6.3 Overview of POF Systems
power was +3.8 dBm; the sensitivity was -33.5 dBm. If cross-talk suppression is
greater than 36.9 dB, error-free video transmission is possible.
Fiber type: 120 m PF-GI-POF, < 100 dB/km
Length: 50 m to 250 m
Bit rate: Video signals on 80 MHz-carriers
Transmitter: Laser diodes acc. to the Eight-/-Forum, max. +3.8 dBm
Demux: Interference filter in waveguide structures
References: [Ueh98], [Ueh99], [Miz00]
Company: NTT Multimedia-Laboratory
In 1999, Prof. Khoe from the Eindhoven University also presented a proposal
for a 2.5 Gbit/s WDM system at 645 nm, 840 nm and 1,310 nm in ([Khoe99], see
Fig. 6.172).
filter 1 filter 2 PD
GI-POF 1,310 nm
filter 3 filter 4
PD PD
645 nm 840 nm
The transmitter and receiver for this system correspond to the components for
the 2.5 Gbit/s Point-to-Point transmissions (see above). The measured insertion
loss of the demultiplexer was < 1.6 dB with a cross-talk of < -35 dB. In [Khoe00]
practical experiments were shown with this demultiplexer. The 3 lasers described
above at 645 nm, 840 nm and 1,300 nm served as transmitters. Figure 6.173 shows
the experiments that were performed.
645/840/1,310 nm
645 nm LD
2 u 100 m (: 170/340 m) PF-GI-POF
840/1,310 nm
840 nm VCSEL
328 m (: 110 m) PF-GI-POF
840/1,310 nm
Fig. 6.174: 500 nm/650 nm POF WDM system for bi-directional transmission
Table 6.12: Power budget for the 520 nm/650 nm POF system
transmission [dB]
0 operating wavelength
500/650 nm
-5
-10
-15
multiplex filter with
-20 BJC-600e printer
-25
-30
-35
460 480 500 520 540 560 580 600 620 640 660 680
wavelength [nm]
laser mirror
diode
prism POF
photodiode
The laser beam strikes the angled surface of the prism. Thanks to the alignment
of the laser polarization, reflection is nearly complete. The laser light is then
launched into the POF by means of a lens with deflection mirror. The coupling
efficiency of the LD into the POF is specified as 91.4 % (0.26 dB loss). The small
emitting surface and the small emission angle of the LD are, of course, also
utilized. The incoming light from the remote transmitter is unpolarized. This is
why a part of the light focused by the lens is refracted through the prism onto the
photodiode. The coupling percentage from the POF into the PD is 24.0 %, which
is a corresponding loss of 6.2 dB. The degree of polarization of the LD is > 150 at
over 1 mW optical power (0.7 % in the second polarization state).
The limiting factor in this system is the NEXT (Near End Cross Talk), in other
words, the received power of its own transmitter. NEXT was calculated in the
study. For a transmitting power of 1.6 mW (55 mA), this comes to:
only for the LD-PD unit: 2 W (0.13 %)
for the transceiver without POF: 5 W (0.32 %)
for the transceiver with POF: 8 W (0.49 %)
A 125 Mbit/s test with 1 mW of average input optical power ran successfully.
The computer simulation yielded a signal-to-noise ratio (SNR) of 22 dB at a
NEXT of 0.49 % This should allow duplex operation. One problem are connectors
positioned near behind the transceiver. Through the index difference of air and
PMMA, two reflections are created. For shorter lengths, the POF attenuation does
not make much of an impact, the light is also for the most part polarized. Error-
free duplex operation is hence not possible under worst case circumstances. The
authors specified approximately 5 m as the minimum distance for the first plug.
Also not taken into account is the effect that both transceivers of a single line can
have different transmission levels (for example, brought about by different
temperatures). This causes the SNR to deteriorate even more. An active echo
compensation could correct the situation if the reflections only occur at a few
points that are constant in time. In [Kure00] and [Tak00] the calculations for the
signal-to-noise ratio is presented for this type of bi-directional transmission. The
522 6.3 Overview of POF Systems
fact is taken into consideration that the interferences caused by near end cross-talk
are not to be confused with white noise; they are determined by the transmission
level and strength of the reflections. According to the simulations presented,
cross-talk (relative to the transmission level) for a BER < 10-12 of up to 20 % is
tolerable.
Fiber type: Low-NA-POF, AN = 0.32
Length: up to 10 m
Bit rate: 125 Mbit/s, bi-directional
Transmitter: red laser diode
Transceiver: LD and PD with polarization sensitive mirror
References: [Hor98], [Kure00], [Tak00]
Company: Sony
650 nm
650 nm
LED
POF LED
Reference: [Gar99]
Company: Centro Politcnico Superior Zaragoza
transmission line
PD single fiber
duplex up to 10 m
665 nm laser
optical lens S100 to S400
block
lens mirror
LD
prism
Fig. 6.179: Optical multiplexer for bi-directional transmission
A very interesting concept for the bi-directional transmission of data has been
developed by Toyota. In this case the WDM principle made use of a red and green
LED. An interference filter was used as the multiplexer. The setup of the complete
system is illustrated in Fig. 6.180.
524 6.3 Overview of POF Systems
The maximum data rate amounts to 250 Mbit/s for both channels at a trans-
mission length of 10 m of DSI-POF. The eye diagrams of both channels can be
seen in Fig. 6.181.
Fiber type: DSI-POF, Mitsubishi Eska-Mega
Length: 10 m with connector
Bit rate: 250 Mbit/s bi-directional
Transmitter: 495 nm (own development); -5.7 dBm in the POF
and 650 nm LED (Hamamatsu L7726); -1.5 dBm in the POF
Receiver: Si-PD
sensitivity at 495 nm: -17.4 dBm
sensitivity at 650 nm: -20.6 dBm
Multiplexer: self written waveguides with interference filters
85% transmission at 495 nm and 96% reflection at 650 nm
Module: 6 u 7 u 9 mm
References: [Kag03], [Yon04], [Yon05]
Company: Toyota
Fig. 6.181: Eye diagrams of the two channels for the Toyota WDM systems
6.3 Overview of POF Systems 525
One particular feature is the beam guiding in the multiplexer. Instead of a sys-
tem of lenses a waveguide is used which is written self-employed into a polymer
block by UV light. All adjustment steps are thus superfluous. The already written
splitter with filter and coupled POF are shown in Fig. 6.182.
A very economical proposal for bi-directional transmission over POF was made
by an English group in [Kat98]. The authors studied the possibility of using a LED
as a transmitter and receiver simultaneously. It is well known that semiconductors
can also be used as detectors. The maximum emission is produced at shorter
wavelengths. By means of this shift, the authors were able to determine a system
loss of 5 dB.
In comparison to a typical photodiode (Siemens SFH 250), the sensitivity is
approximately 7 dB lower. This makes the system 12 dB worse than a conven-
tional Point-to-Point system. In addition, it can only be used in semi-duplex mode
since the diode operation must be switched over. For shorter distances up to 20 m,
this could be an interesting solution for cost reasons.
Fiber type: PMMA SI-POF
Length: up to 20 m (numerically)
Bit rate: Half duplex operation
Transmitter: LED in photodiode operation
Reference: [Kat98]
Company: University of North London
A similar approach is also described in [Ing06]. A 1.25 Gbit/s half-duplex
transmission with a VCSEL as transmitter and receiver is realized over 500 m of
GI-GOF (50 m). The sensitivity of the VCSEL when operating as a photodiode
is about 0.1 mA/mW (at 850 nm, 0.9 nA dark current) and a receiving bandwidth
of 933 MHz is attained. The sensitivity is -12.3 dBm, as shown in Fig. 6.183.
526 6.3 Overview of POF Systems
10-2
BER 1.25 Gbit/s
10 -4 NRZ
PRBS 27-1
10-6
10-8
back to back
-10 500 m 50 m
10 GI-GOF
10-12
average received power in the fiber [dBm]
10-14
-16 -15 -14 -13 -12 -11 -10 -9 -8
Infineon developed the transceiver SFH800 for the automotive industry (see
[Sch99b]). By using chip-on-chip technology, the LED transmitter is mounted
directly on the photodiode, as shown in Fig. 6.184. The component is intended for
use in passive star-type networks for data rates up to 10 Mbit/s.
Fiber type: SI-POF
Bit rate: 10 Mbit/s, half duplex operation
Transmitter: Photodiode with LED on chip
>300 W at 30 mA, 650 nm
Receiver: -23 dBm sensitivity
References: [Sch99b], [Sch00b], [Gri00]
Company: Infineon Technologies
LED
Fig. 6.184: SFH 800 from Infineon for bi-directional POF operation
In [Bau02] concepts are presented for also using this principle for replacement
systems with significantly higher data rates. By using a RC-LED with below 1 ns
switching time up to 200 Mbit/s can be transmitted. The sensitivities of the photo-
diode are (BER = 10-9):
6.3 Overview of POF Systems 527
100
transmission [%]
90
red filter
80
blue filter
70
60
50
40
30
20
10
0
350 400 450 500 550 600 650 700 750
wavelength [nm]
In the examples mentioned above we have always been dealing with digital signal
transmission, the domain of optical telecommunications engineering. In some
cases, however, the transmission of analog signal makes sense. In the following
segments we will present some ideas for the analog transmission of video signals
at first. Thereafter, we will present some special experiments such as how analog-
modulated digital signals can be transmitted.
One great advantage of POFs in comparison to glass multimode fibers lies in
the large number of modes. As a rule, mode distribution noise does not play any
significant role, unlike, for example, for 50 m GI-GOF. One disadvantage is that
the lasers used are not usually linear like the 1.3 m DFB laser, for example,
which is employed in glass fiber systems. The parameters of singlemode glass
fiber systems cannot, of course, in any way be reached with POFs. POF systems
therefore make sense in applications over short distances in which the main
aspects are simple installation and robust systems.
59 video
channels
AM
LD 200 m GI-POF
channel 10 659 nm : 500 m Si-
BPSK- MSM-PD
modulated
2 Mbit/s
Fig. 6.189: Receiver and the complete system for baseband video transmission
Fig. 6.190: Picture quality after 300 m, 350 m and 400 m standard POF
Fig. 6.191: Noise current density at the receiver input and emission spectrum of the LED
Fig. 6.192: Video transmitter and receiver for up to 470 MHz (FhG IIS)
The complete spectrum of transmitted signals can be seen in Fig. 6.193. In one
of the first experiments in 2003 only two channels (325 MHz and 380 MHz) were
transmitted over 50 m. In later experiments (2004) the complete band up to
470 MHz was transmitted over 35 m of SI-POF.
Fig. 6.193: Video transmission up to 470 MHz (FhG IIS), complete spectrum
The signal quality was analyzed for the transmission of the band up to
470 MHz over 30 m of SI-POF (channels from 147.25 MHz to 335.25 MHz). A
deterioration in CNR of 46 dB (input) to 43 dB (output) was established. The CSO
remained unchanged at 53 dB. The transmission over 100 m of PF-GI-POF (Luci-
naTM) was successfully demonstrated with this system. The values for CNR and
CSO performed as with SI-POF, only the signal level was a few dB smaller
because of the inferior laser coupling.
The transmission of 37 analog and 16 digital channels over 25 m of PMMA
GI-POF was demonstrated in 2005, also with insignificant changes in CNR, CSO
and CTB.
6.3 Overview of POF Systems 533
At the 2006 POF Conference in Seoul a system for transmitting the BK band on
POF was introduced [Kim06b]. A generator with 60 analog video signals (NTSC
format, 55.25 MHz up to 439.25 MHz) serves as a source. A 1.31 m DFB Laser
(10 mW) was modulated with an index of 3.4% per channel. The transmission link
was a 25 m long PF-GI-POF with a 50 m core diameter of Asahi glass. A pin-
photodiode served as the receiver. Only very slight deterioration was determined
for CSO and CNR (Fig. 6.194).
Fiber type: 50 m PF-GI-POF (Asahi Glass)
Length: 25 m
Bit rate: 60 NTSC channels analog TV
55.25 MHz to 439.25 MHz
Transmitter: 1,310 nm DFB laser, +10 dBm
Receiver: Si-pin-PD
Reference: [Kim06b]
Company: National University Kyungpook, South Korea
VDSL VDSL
modem modem
50 m SI-POF
Si-PD : 980 m SLED
AN = 0.47 650 nm
Ethernet Ethernet
signal generator signal analyzer
Figure 6.197 shows the SNR for downstream and upstream after 300 m of SI-
POF (high-quality fiber from Luceat). The bit rate downstream amounts to
approx. 40 Mbit/s. The transmission of 107.42 Mbit/s is possible for over 200 m.
SNR [dB]
32
28
24 DS
20
16
12
8
US US
4
frequency [MHz]
0
0 2 4 6 8 10 12 14 16 18
Fig. 6.197: Signal to noise ratio per carrier after 300 m POF
9
bit/symbol
8
7
6
5 DS
4
3
2
1 US US
0
0 2 4 6 8 10 12 14 16 18
frequency [MHz]
Fig. 6.198: Bit per symbol per carrier after 300 m POF
536 6.3 Overview of POF Systems
In the final system the frequency range would practically be used without any
gaps since you would not have to worry about interference from other services.
The attainable modulation depth per carrier in bit/symbol is shown in Fig. 6.198
(the 8 bit/symbol means QAM 256, 7 bit/symbol QAM 128, etc.).
The transmission of existing data formats over polymer fibers was also the goal
of a test setup conducted by T-Nova in the year 2000 ([Ziem00c]), as shown in
Figs. 6.199 and 6.200. The modules are for transmitting the 192 kbit/s BRI
(S0-bus), as is used between ISDN terminal equipment and the ISDN-NTBA.
inductive FSK-Mod
coupler
FSK-Demod
duplex
40 V/5 V standard
converter PMMA-POF
FSK-Demod inductive
coupler
FSK-Mod
Duplex 40 V/5 V 220 V/40 V
Standard converter converter
PMMA-POF
level using a limiter amplifier, then filtered and reconverted to the original signal
on a discriminator. The bandwidth of the POF is wide enough to transmit a second
channel, for example, bi-directional communication over a single fiber or for
coupling both interfaces on the ISDN-NTBA. Figure 6.201 shows the possible
choice of carrier frequencies.
-1 0 +1 -1 0 +1
channel 1: channel 2:
3.263 MHz 5.470 MHz
3.647 MHz 5.854 MHz
4.031 MHz 6.238 MHz
S0-Bus 1 conversion to
frequencies
f1, f2, f3 POF
S0-Bus 2
conversion to
+ LED
frequencies conversion
f4, f5, f6 from the S0-Bus 1
frequencies
f1, f2, f3
PD
conversion
POF from the S0-Bus 2
frequencies
f4, f5, f6
S0-bus 1 S0-bus 3
electr.
electr. Mux
Mux LED O1 LED O2 3+4
1+2
POF POF
optical S0-bus 4
S0-bus 2 splitter
S0-bus 3 S0-bus 1
POF
electr. electr.
Demux PD PD Demux
3+4 POF POF 1+2
S0-bus 4 S0-bus 2
Fig. 6.203: Combination of WDM for bi-directional transmission
At first sight, the transmission of an ISDN signals over POF instead of copper
wires has no advantage in quality. The connection costs also cannot be lowered.
However, a major benefit is being able to dispense with an electrically conducting
connection. The NTBA and terminal device are usually already connected over the
power supply. The second electrical connection over the BRI produces a loop that
could destroy the components in the event lightening strikes. The POF would
simply eliminate this problem. The better electromagnetic compatibility of POF
should also not be overlooked. This is an attractive alternative for local exchanges
with high security requirements.
Yet another advantage offered by POF-ISDN cabling is the possibility of
migrating to a faster system in the future that offers high data rates without having
to lay new cables. DSI-POF, MC-POF or, wherever available, GI-POF can be
used immediately for the BRI. You can also change over later to Fast-Ethernet,
IEEE 1394 or even Giga-Ethernet.
Fiber type: St.-NA-POF GH 4000, 80 dB/km at 560 nm
Length: 500 m
Bit rate: 0.192 Mbit/s (ISDN S0 bus), frequency modulated
Transmitter: 560 nm LED (Nichia, sample)
Receiver: -45 dBm
Reference: [Ziem00c]
Company: Deutsche Telekom
Multicarrier procedures can also be used for transmitting higher data rates. One
advantage is that each carrier can be modulated with different quantization. Only a
few bits per symbol are used with frequencies having a poor SNR; with a high
6.3 Overview of POF Systems 539
SNR there are however high modulation stages. This splitting up even occurs
dynamically on copper cables (DSL) or in radio networks. The principle of adap-
ting the modulation to the SNR is shown in Fig. 6.204, assuming that the limiting
effect is the low-pass characteristic of the fiber. Alternatively, you can also vary
the power of the carrier so that a constant CNR arises.
-3
-6
noise level
-9
0 50 100 150 200 250 300 350 400 450 500
frequency [MHz]
Fig. 6.204: Schematically operation of multiple carrier transmission on POF (1,650 Mbit/s)
intensity periodically
tunable
modulator optical filter
laser O0 O0
WDM WDM BPF
O0
O1
PD
data
(downlink) O1
fsw MS
LD
LPF O1 mobile
LPF
station
data
PD
(uplink)
The tunable DFB laser is frequency modulated. The data are then added to the
signal in a Mach-Zehnder intensity modulator. The modulation frequency of the
laser lies within the bandwidth limits of the fiber. The frequency is converted up-
ward by the interaction of the modulated signal with the periodic characteristic
curve of an optical filter, e.g. a Fabry-Perot filter (FP).
The back channel can be realized by means of a WDM procedure on the same
fiber and the receiver itself supplies the frequency for the upward conversion so
that a separate oscillator is not needed.
In the experiment a 1,310 nm laser with 10 mW was used. The triangular fre-
quency modulation - here with fsw = 800 MHz and a maximum of 28.8 GHz devi-
ation - was realized by an external phase modulator. The FP filter had a period of
9.6 GHz resulting in a frequency conversion of 5.4 GHz, 10.8 GHz, etc. A carrier
frequency of 225 MHz was given on the modulator which in turn was either BPSK
or QAM modulated (up to 56 Mbit/s error-free transmission). The uplink supports
up to and over 1 Gbit/s.
This procedure enables the use of both glass multimode fibers, demonstrated
with over 4 km, and POF. Other descriptions can be found in [Gie03], [Koo04b],
[Lar06b], [Lar06c], [Ng04a] and [Ng04b].
Fig. 6.206: Principle of MGDM and example for a mode distribution ([Koo03b])
POF
angled
mirror
lens
lens
detector array
The first experimental results for mode multiplex on PF-GI-POF were presen-
ted by [Sch06] at the POF Conference in Seoul. A 62.5 m PF-GI-POF was
used. The signal in different mode groups was coupled into two fibers over a cor-
respondingly positioned singlemode fiber. These mode groups were then brought
together over a coupler maintaining one mode. At the end of the test fiber the
signal was re-detected with a freely positionable singlemode fiber. Both channels
could - separately - be transmitted errors free.
6.3 Overview of POF Systems 543
Fig. 6.208: Near field after 100 m PMMA GI-POF (every left: central launch, every right:
launch with offset)
The mode behavior of SI-POF was investigated in [Jan04]. The far fields after
10 m of POF (Mitsubishi CK-40) when coupling in collimated light (6, 15 and
24 relative to the fiber axis) are shown in Fig. 6.209.
544 6.3 Overview of POF Systems
That MGDM can also function very well in glass fibers was demonstrated in
[Kra00] and [Klu02] (University of Mannheim). Here a 200 m SI-PCS was used
(AN = 0.39). A 632 nm HeNe low divergence laser served as the source. Up to 13
different mode groups can be differentiated after a short piece of fiber (40 cm).
Figure 6.210 shows a picture with every second mode group.
The capacity of the method is described more detailed in [Kra00]. The crosstalk
between the channels is analyzed. A data transmission is investigated in [Klu02]
than. Using PCS length of up to 20 m has been measured. A 200 m POF was
available too. The principal option for data communication was found based on
the measured crosstalk attenuation of more than 10 dB.
Length: 10 m
Bit rate: 4 u 500 Mbit/s, ribbon
Reference: [Hon05]
Company: Honda-Cable
PMMA SI-POF
core/cladding diameter: 980/1,000 m
Uni-directional transmission
100 - 500 Mbit/s per fiber
LED: 650 nm
pin-photodiode
electrical interface: LVPECL (Low Voltage
Positive Emitter Coupled Logic)
operation temperature range: 0 to +60C
Fig. 6.211: 4-fiber parallel transmission by Honda-Cable
The POF ribbon cables used have already been presented in Chapter 2. The
system had the following parameters:
Fiber type: 500 m PMMA SI-POF and PMMA GI-POF, 8-fiber ribbon
Length: 50 m
Bit rate: 4 u 1,600 Mbit/s
Transmitter: 650 nm LD
Receiver: pin-PD
Multiplexer: 8-fiber ribbon
References: [Jun06], [Ziem06g]
Company: OVAL project (Loewe, SGT, FhG IIS, POF-AC)
546 6.4 Other Optical Transmission Systems
One aim of this book is to supplement the first edition by also including informa-
tion on other fibers so that the development of optical short-range communication,
in which other optical fibers are increasingly used, is taken into account.
Polymer fibers which are suitable for high temperatures will be dealt with first.
(The fiber characteristics have already been described in Chapter 2. Some trans-
mission experiments with 780 nm transmitters have already been presented in
Section 6.2.4.3). A description of multiple parallel POF connections then follows.
Finally, systems with PCS and fiber glass bundles are discussed.
Polycarbonate fibers were viewed for a long time as the most likely candidates for
high-temperature POF in ranges up to +130C. The main user for such systems is
the automobile industry since temperatures in certain areas of cars can go up to
over +100C.
System tests on PMMA SI-POF with high data rates have been presented in
Section 6.3.1.3. A test for transmitting high data rates on PC-POF was carried out
as part of the joint experiments by T-Nova and Nexans Autoelectric. Figure 6.213
shows the measurement results at 800 Mbit/s and a pseudo-random sequence at
500 Mbit/s when transmitting over a 10 m fiber. A 657 nm component from Sony
was used as the laser, the attenuation of which lay at about 12 dB.
For the experiment at 800 Mbit/s, a receiver with GPD amplifiers and a
SFH75P photodiode were used. (The maximum data rate of the receiver was
1,200 Mbits.) Thanks to the alternate symbols, the transmission occurred without
errors despite the limited bandwidth of the POF. The received power at the photo-
diode was -11.2 dBm.
In a second experiment a pseudo-random sequence of the length 27 - 1 was
used. The maximum possible data rate here was only 500 Mbit/s because of the
limited bandwidth of the POF. The laser was operated without predistortion and
the receiver setup was unchanged at a maximum of 1,200 Mbit/s. The received
power was also unchanged at -11.2 dBm.
Consequently, the PC-POF can unhesitatingly be employed for applications
such as IEEE 1394 S400 in the event that increased demands on temperature
levels are made, for example, in the engine compartment of vehicles. Higher band-
widths can be achieved by using modified cladding or with PC-MC-POF.
Fiber type: 1 mm PC-POF, 1,200 dB/km at 650 nm
Length: 10 m
Bit rate: 500 Mbit/s
Transmitter: 657 nm LD, Sony SLD 1133 VL
Receiver: SFH75P with amplifier
Reference: [Stei00a]
Company: Deutsche Telekom
6.4 Other Optical Transmission Systems 547
500
receiver signal [mV]
400
300
200
100
0
-100
-200
-300
-400 23.01.2001 Giehmann time [ns]
-500
0 20 40 60 80 100
New experiments with the PC-POF from Mitsubishi have been undertaken at
the POF-AC using an improved test system. Initially, 950 Mbit/s could be trans-
mitted error-free over a 10 m fiber ([Ziem03g]). Later on, 1,000 Mbit/s over 10 m
at a received power of -13 dBm were attained ([Vin04b]). In comparison, the data
rates for PMMA SI-POF under otherwise equal conditions lay at around
1,500 Mbit/s. the reason for this is the greater NA of the PC POF. Furthermore,
the lower receiving level only permits a small penalty.
Other types of fibers used with high temperatures are fibers made of modified
(cross-linked) PMMA and elastomers, both of which were used in various
transmission experiments conducted at the POF-AC.
H-POF (Tver): transmission of 620 Mbit/s over 15 m at 650 nm (received
power is -6.5 dBm), Sony laser ([Ziem03g]).
H-POF (Tver): 1,200 Mbit/s over 11 m and 990 Mbit/s over 23 m of POF at
650 nm (Sanyo laser diode, [Ziem03g], [Vin04b]), see Fig. 6.215.
PHKS CD1001 (modified PMMA POF from Toray), transmission of
1,050 Mbit/s over 10 m and 830 Mbit/s over 20 m at 650 nm (received power
+0.5 dBm and -2.6 dBm, [Ziem02j], [Ziem02k]).
HPOF (modified PMMA, sample from Hitachi): 850 Mbit/s over 24 m at
650 nm ([Vin04b]).
HPOF-S (elastomer), tested only at 780 nm, see above.
548 6.4 Other Optical Transmission Systems
The results of the tests show that all high-temperature fibers investigated per-
mitted similar data rates as the PMMA SI-POFs, however, with clearly higher
losses. There are still numerous problems in the details which have to be solved
before they are actually used in vehicle networks.
POFs offer many advantages for use in optical bus systems as multiple parallel
connections over very short distances. First of all, the fibers have a very good ratio
of cladding to core diameter which makes adjustment easier. Furthermore, the
flexible material and the great NA permit smaller bending radii than for compa-
rable glass fibers. The most important aspect is probably that they are simple to
work on. A POF bundle can simply be cut off with a hot blade or smoothed very
quickly by polishing it.
Extensive developments have been conducted in this area at the University of
Dortmund. In order to realize many channels in a small space, 1/8 mm POF was
used. Since only a maximum of 50 cm were to be bridged, a 850 nm VCSEL
could be used without any problems. In Figs. 6.216 to 6.218 the principle of the
parallel link is first illustrated and then photos of the plug and a PC board with the
POF link. In the principle figure you can see that the fibers are plugged vertically
to a VCSEL array and then diverted in the plug by 90. Thanks to the bending
radius below 2 mm there are no problems with the POF used. 2,500 Mbit/s can be
transmitted over each of the 64 channels.
Fiber type: 125 m SI-POF, AN = 0.50
Attenuation: 1.7 dB at 660 nm (50 cm):
4.5 dB at 870 nm
8.5 dB at 980 nm
Length: 0.5 m
Bit rate: 2,500 Mbit/s
Transmitter: 850 nm VCSEL array (8 u 8)
Receiver: Si-pin-PD array, -23 dBm at 2.5 Gbit/s and BER = 10-11
Multiplexer: SDM up to 128 fibers; module size 3.5 u 10 u 10 mm
References: [Witt98], [Jh98], [Ney01], [Ney02]
Company: University of Dortmund
6.4 Other Optical Transmission Systems 549
Fig. 6.217: Parallel optical link with POF, connector and the complete link
POF bundle
alignement pins
VCSEL array
processor
Similar experiments with thin POF have also been carried out at the University
of Ulm, whereby red VCSELs were used as transmitters. 2,000 Mbit/s could be
transmitted over 1 m of fiber. The laser itself can be modulated up to 5 Gbit/s.
These data rates can also be transmitted over 100 m of graded index glass fibers.
Fiber type: 125 m PMMA SI-POF, AN = 0.50, 500 dB/km
670 MHz bandwidth over 10 m
Length: 1m
Bit rate: 2,000 Mbit/s, also 5 Gbit/s over 100 m MM-GOF
Transmitter: 650 nm VCSEL, 0.79 mW
Receiver: InGaAs pin-PD with lens coupling
Reference: [Sta03]
Company: University of Ulm
550 6.4 Other Optical Transmission Systems
Glass fibers with polymer cladding (PCS) have been used successfully in auto-
mation engineering for many years. The data rates used up to a few years ago lay
at a maximum of 12 Mbit/s, the bandwidth of the fiber playing no role whatsoever.
Only recently has the potential of PCS for much higher data rates been investi-
gated, e.g., for use in future vehicle networks. The advantages of the PCS lie in
their small bending radius, high resistance to heat and useability in the near infra-
red range where better VCSELs are available.
Transceivers for data rates of up to 155 Mbit/s are available from Hewlett
Packard (Agilent and Avagotech later) which can be used with POF as well as
with 200 m PCS. Approximately 6 dB less power can be coupled into the PCSs.
Thanks to the very low fiber attenuation this disadvantage is compensated for
lengths of 50 m and more so that longer distances can be bridged. The comparison
of bit rates and ranges for POF and PCS is shown in Fig. 6.219.
Fiber type: 200 m SI-PCS
Length: 10 m to 700 m
Bit rate: 20 Mbit/s to 125 Mbit/s
Transmitter: 650 nm SLED, -16.2 dBm in the fiber
Receiver: Si-pin-PD
Reference: [HP07]
Company: Hewlett Packard, Agilent, Avagotech
100
application application
range with range with
50 PCS
POF
20
10
10 20 40 60 80 100 10 30 100 300 1000
fiber length [m] fiber length [m]
Fig. 6.219: Transmission distance and bit rates of HFBR components on POF and PCS
For several years now fast components for MM glass fiber systems have been
produced at the Astri Research Centre in Hong Kong. 850 nm components have
been developed for 200 m PCS which permit a data transmission of up to
1,250 Mbit/s over 10 m of PCS. Eye diagrams for three different bit rates are
shown in Fig. 6.220.
6.4 Other Optical Transmission Systems 551
Fig. 6.220: Eye diagrams after 10 m for 300 Mbit/s, 600 Mbit/s and 1,250 Mbit/s
O [nm]
550 600 650 700 750 800 850 900
The POF-AC and the FhG IIS cooperated in setting up an experimental system
for transmitting analog VGA signals on PCS. Conventional red laser diodes served
as transmitters. Up to 100 m of fiber could be bridged with a medium screen reso-
lution. Fig. 6.222 shows the experimental setup.
Fiber type: SI-PCS, AN = 0.37
Length: 100 m
Bit rate: VGA signal, 1,280 u 1,024 pixel
Transmitter: 650 nm LD, 2.5 mW
Receiver: 400 m PD S5973, Hamamatsu
Reference: [Fac04]
Company: POF-AC and Fraunhofer IIS
In later experiments greater transmission lengths with different PCS types were
then investigated. Compensation filters for equalizing mode dispersion were used,
whereby the following results were attained at a wavelength of 780 nm:
2,230 Mbit/s over 10 m
1,040 Mbit/s over 50 m
500 Mbit/s over 100 m
260 Mbit/s over 200 m
Fig. 6.223 shows as an example the eye diagram for the transmission of
350 Mbit/s over 100 m.
The bit rates and ranges of the SI-PCS systems are summarized once again in
Fig. 6.224. Distances of many hundred meters can, of course, be bridged because
the attenuation is much smaller than with PMMA POF. However, the bandwidth
of the SI-PCS only lies approximately in the area of DSI-POF.
1,000
300
100
30
length [m]
10
1 10 100 1,000
In order to be able to utilize the PCSs lower attenuation even at higher bit
rates, PCS with a semi-graded index profile is manufactured by Sumitomo and
OFS ([Sum03]). A work from 1995 ([Kos95]) presents a system of data transmis-
sion with the then new fibers. A 850 nm VCSEL was used as transmitter and a
small-surface InGaAs-APD as receiver. 3,000 Mbit/s could be transmitted over
100 m; over 1,000 m it was still 1,500 Mbit/s. The semi-GI-PCS thus approxi-
mately attained the performance of PF-GI-POF - is, however, considerably more
expensive.
Fiber type: Sumitomo semi-GI-PCS
Length: 100 m, 500 m, 1,000 m
Bit rate: 3,000 Mbit/s; 2,000 Mbit/s; 1,500 Mbit/s
Transmitter: 850 nm VCSEL
Receiver: InGaAs-APD
Reference: [Kos95]
Company: Sumitomo
At the POF-AC the data transmission of Sumitomo semi-GI PCS with a 650 nm
and 780 nm laser respectively was tested over 500 m. The maximum possible data
rates were 600 Mbit/s and 1,000 Mbit/s (eye diagram in Fig. 6.225). An 800 m
Si-pin detector was used that was actually much too big. Consequently, improve-
ments can still be made in the system.
Fiber type: Sumitomo semi-GI-PCS
Length: 500 m
Bit rate: 1,000 Mbit/s, 600 Mbit/s
Transmitter: 780 nm LD, 650 nm LD
Receiver: 800 m Si-pin-PD Hamamatsu S5052
Reference: [Ziem06i]
Company: POF-AC Nrnberg
Fig. 6.225: Eye diagram for 1,000 Mbit/s over 500 m Semi-GI-PCS at 780 nm
of 1,650 Mbit/s, 2,200 Mbit/s and 1,900 Mbit/s respectively could be transmitted.
Fig. 6.226 shows the eye diagrams each for 1,000 Mbit/s. They are opened and
show large system margins.
BIAS Si pin PD
S 5052
LD 780 nm
LD 650 nm MC-GOF (375 cores)
Schott, approx. 200 dB/km
Fig. 6.228: Eye diagrams for 5 m (2,610 Mbit/s) and 30 m (1,840 Mbit/s) from [Was07]
3,000
bit rate [Mbit/s]
1,000 1 mm MC-GOF
AN = 0.50
Schott Glass
780 nm / 850 nm
300
10,000
bit rate [Mbit/s]
3,000
1,000
650 nm
300 780 nm
850 nm
fiber length [m]
100
1 2 5 10 20 50 100
Fig. 6.230: Data transmission on MC-GOF at 3 wavelengths ([Was07])
The first edition of this book contained a chart showing the development of the
capacity of POF systems (Fig. 6.232). The best values at the time was the trans-
mission of 2 u 2.5 Gbit/s over 458 m (University of Eindhoven). This value has
endured to today as a system capacity. Nevertheless, this does not mean that the
development of POF systems would not have made further progress. A large part
of the current developments does not refer to the improvement in the parameters
of PF GI POF, but lies more in the area of reasonably priced PMMA fibers.
Consequently, it was possible in the laboratory to increase the data rates for 1 mm
POF to 2.3 Gbit/s. With a 1 mm PMMA GI-POF 2 Gbit/s over 100 m are possible
and with green LEDs distances of several 100 m of PMMA POF can be possible.
558 6.5 Overview of Multiplex Techniques
100
capacity in Gbit/s100 m
10
0.1
year
0.01
1988 1990 1992 1994 1996 1998 2000 2002 2004 2006
The following list shows the present highest capacity of different polymer and
glass fibers (estimations).
up to 2,500 Mbit/s over short lengths PMMA-POF
up to 40,000 Mbit/s over PF-GI-POF (30 m)
500 m transmission distance for low data rates with PMMA SI-POF
550 m transmission distance for 2.5 Gbit/s with PF-GI-POF
1000 m transmission distance for 1.25 Gbit/s with PF-GI-POF
100 Mbit/s km bit capacity for SI- and DSI-POF
50 Mbit/s bit capacity on 1 mm MC-GOF
50 Mbit/s bit capacity on 200 m SI-PCS
1.500 Mbit/s bit capacity on 200 m Semi-GI-PCS
100 Mbit/s km bit capacity for MSI- und MC-POF
500 Mbit/s km bit capacity for PMMA GI-POF
2.280 Mbit/s km bit capacity for PF-GI-POF
It has always been difficult to assess the future development of PF-GI-POF, but
the successes above all at Chromis Fiberoptics are a cause for optimism. Their use
in home networks could be of increasing interest especially for PMMA GI-POF.
Further improvements in system capacities can be expected with the development
of new multiplex technologies.
Different multiplex systems relevant for POF are compared in Table 6.13 with
typical values according to current publications.
6.5 Overview of Multiplex Techniques 559
The table does not show any clear favorite. In the medium-term the wavelength
multiplex may emerge as the clear winner for PF-GI fibers. Various exceptional
features speak for it:
particularly wide band with low attenuation and dispersion
(600 nm to 1300 nm)
relatively small multiplexer/demultiplexer
large number of available laser diodes at different wavelengths
For PMMA fibers the number of possible channels are relatively limited with
MGDM and WDM. Furthermore, the large core diameter and the large NA lead to
relatively voluminous optical components. Multi carrier procedures are already
widely developed in radio and DSL technologies and can be adapted quite easily
for POF in order to better utilize the limited capacity. However, the development
of better GI POF, faster transmitters and adaptive equalizers offer for the present
even much more potential for very simple solutions.
Above all the development of SDM systems is of interest for many applications
with POF. The extremely simple and reasonably priced cables, uncomplicated end
face treatment and adjustment as well as the availability of very cheap active
components will make multiple parallel systems extremely attractive especially
over short distances. Here PMMA SI-POF can transport straight away many
Gbit/s at distances up to a few meters and thus surpass copper conductors. In
addition, the power requirements go down and available VCSEL in the near
infrared range can be used.
7. Standards
consumer automotive
building elektronics
networks applications
computer
JIS SERCOS
IEC Interbus
VDE/VDI Profibus
Ethernet
measurement automation
methods machine control
The first standards for polymer fibers were established in Japan. The Japanese
Industrial standard JIS-C-6837 set the parameters which were later taken over in
IEC 60793-2. Table 7.1 according to >Wei98@ shows the values.
Parameter Unit mm mm 1 mm
SI-POF SI-POF SI-POF
core [m] 485 735 980
cladding [m] 500r30 750r45 1,000r60
jacket [mm] 1.5 r 0.1 2.2 r 0.1 2.2 r 0.1
core non circularity [%] d6 d6 d6
loss at 650 nm [dB/km] d 400 d 400 d 400
with EMD launch [dB/km] d 300 d 300 d 300
bandwidth* [MHz100m] - - -
bending loss [dB/10 bends] d 0.5 d 0.5 d 0.5
numerical aperture - 0.50 r 0.15 0.50 r 0.15 0.50 r 0.15
* > 10 MHz 100 m defined by the IEC
These values have essentially been defined according to the needs of industrial
applications. For example, all fibers with a NA between 0.35 and 0.65 are in this
class even if it does not make much sense to couple them to each other. The
theoretical coupling loss can amount to as much as 5.4 dB because of the
difference in NA.
7.1 Standards for Polymer and Glass Fibers 563
In the meantime Standard 60793-2-40 has been completely revised and now
contains 8 different classes of polymer fibers (A4a to A4h, see >IEC04@). Tables
7.2 and 7.3 provide the specific parameters. The first three classes are standard
step index profile fibers made of PMMA with unchanged values Class A4d des-
cribes PMMA DSI-POF.
MSI- and GI-POFs made of PMMA fall into Class A4e whereas the last three
classes describe GI-POF made of perfluorinated fibers.
The parameters of this standard are not always comprehensible as to why upper
limits for the bandwidth are given for fibers f to g. This is unnecessary, of course.
Moreover all manufacturers continue to agree that a PF-POF with a 200 m core
diameter will not be produced because it is much too expensive. Core diameter to-
lerances of 60 m are much too big for data-grade quality fibers. More realistic
are maximum deviations of 10 m. Furthermore, specifying a bandwidth from
10 MHz 100 m for A4a is much too pessimistic. Even under negative conditions
564 7.1 Standards for Polymer and Glass Fibers
artificially created there are no values attainable below 30 MHz 100 m. We must
assume here that the fiber manufacturers managed to push through parameters
they established and not those of the users and thus selected a kind of lowest com-
mon denominator.
In order to avoid such problems, the parameters of the fibers to be used are
specified more narrowly, e.g. with MOST, in most standards for applications. This
is actually quite the opposite of the definition of standardization.
The definition of the parameters for glass fibers with polymer cladding is stated in
the standard IEC 60793-2-30 (>IEC06@). The current values are shown in Table 7.4.
The overview of Standard IEC 60793-2 describes the following types of fibers:
Part 2-10: Category A1 multimode fibers (GI-GOF)
Part 2-20: Category A2 multimode fibers (SI-GOF)
Part 2-30: Category A3 multimode fibers (PCS)
Part 2-40: Category A4 multimode fibers (POF)
Part 2-50: Category B singlemode fibers (SMF)
Part 2-60: Category C singlemode fibers for intra connection
The IEC differentiates between step index and graded index on the basis of the
index coefficients, whereby the following is true:
A1: graded index fibers: 1dg<3
A2: step or quasi step index fibers: 3 d g < 10
A3: step index fibers: 10 d g < f
A4: step, multi step, graded index fibers: 1dg<f
The values in Table 7.6 have been specified for the less common SI-GOF in
Category 2 and the core and cladding are composed of silica glass.
You can see from this table that in the meantime a large number of sub-catego-
ries has developed out of the classic singlemode fiber. Consequently, the present
diversity of POF types must not necessarily be viewed as a specific disadvantage.
The optimization goals for singlemode fibers are primarily determined by four
requirements:
As low an attenuation as possible in the entire optical range from 1,260 nm to
1,625 nm - keeping the OH share to a minimum.
As low a chromatic dispersion as possible.
Not too low chromatic attenuation for reducing the effects of four-wave
mixing.
As large an effective area as possible for reducing non-linear effects.
Both the attenuation and the chromatic dispersion can be compensated for
today almost at will through EDFA or a Raman amplifiers and dispersion-com-
pensating fibers. Non-linear effects are intentionally used to compensate for the
dispersion or through special modulation formats. In this way standard SMF in
DWDM systems attain practically the same performance as the many different
classes of special fibers.
What has so far not been specified at all are fiber bundles or microstructured
glass and polymer fibers. One of the most important aims of standardization will
surely be establishing parameters for bend-insensitive fibers which play a particu-
larly important role in home networking.
The following sections describe a number of applications which have more or less
set comprehensive specifications for fibers and active components. First of all, we
would like to compile the most important technical parameters. In the next chapter
we will present the concrete applications with practical examples.
Table 7.7: Worst-case attenuation increase for 50 m POF and 100 m HPCF cable, taking
spectral characteristics of the source and bends into consideration
Even the plug connections for 155 Mbit/s systems have been specified by the
ATM Forum. It is a matter of duplex connector F07/PN and the simplex connector
F05 (Fig. 7.3). The PN connector differs only minimally from F07 due to two
additional raised areas. The couplings have been designed in such a way that both
plugs are compatible with each other. The insertion attenuation should amount to a
maximum of 2 dB. Table 7.9 lists the individual losses.
Table 7.9: Worst case values for the insertion loss of connectors
Processes Loss
extrinsic losses
radial displacement max. 0.1 mm 0.4 dB
endface roughness 5 Pm 0.1 dB
angle offset 1 0.1 dB
Fresnel losses 0.3 dB
environmental conditions(1) 0.3 dB
intrinsic losses
fiber dimension and NA error(2) 0.8 dB
sum 2.0 dB
(1): also high temperature of 70C, vibrations, temperature cycles of -25C to
70C, resulting in a max. increase of the insertion loss of approx. 0.3 dB
(2): from POF specification: AN = 0.30 r 0.05; core diameter = 980 m r 20 m
ces the attenuation less than at 650 nm. Moreover, the temperature dependence of
the spectrum of the 520 nm LED (InGaN) is lower than that of the 650 nm LED
(AlGaInP), whereby changes in temperature also have less effect on the system
power budget.
Fig. 7.3: Examples of a F07 plug (top) and a F05 plug with coupling (below)
The ATM Forum specifications are discussed in detail in Chapter 6.2.1. In the
meantime even red RC-LEDs or blue LEDs would allow a 155 Mbit/s data trans-
mission over 100 m. The ATM standard, however, has not achieved any kind of
importance within building networks so that an expansion of the POF specifi-
cations by the ATM Forum will probably not be discussed.
Table 7.10: Worst case attenuation increase for 50 m POF and 100 m HPCF cable, taking
the spectral characteristics of the source and bends into account (identical with
the ATMF specification)
receptacle
plug
One fundamental difference between the concept of 1394 and other systems is
that from the very beginning this standard was oriented toward use by different
media depending on application and required lengths.
572 7.2 Application Standards
For lengths up to 4.5 m shielded twisted pair copper cables can be used (two
insulated twisted wires for the data and 2 insulated conductors for the power
supply).
Unshielded twisted data cables of Category 5 can be used for S100 up to
100 m - the goal is S800.
Polymer fibers are used for S100 and S200 for up to 50 m (S400 with MSI-
/GI-POF in the new generation).
PCS are intended for lengths up to 100 m and S200 - up to S800 with the new
semi-GI-PCS
Glass multi mode fibers (50 m GI) are used up to 100 m for data rates up to
S3200.
Fig. 7.5: Copper connector for IEEE 1394b ([Har04]) and shielded cable (: approxi-
mately 4 mm)
Master Master
slave
drive
slave drive
slave drive
optical
connections slave drive
slave drive
Fig. 7.6: SERCOS Interface with a ring topology
7.2 Application Standards 573
A master regulates the flow of data in the ring. Slaves are used to connect the
drives to the ring. Data are exchanged only with the master. Up to 254 users can
be controlled with this system. Polymer optical fibers up to a length of maximum
60 m with LED transmitters in the wavelength range of 640 nm to 670 nm are
used with a data rate of 2 Mbit/s. Detailed information can be found at the website
www.sercos.org.
7.2.4 Profibus
Fiber Characteristics
fiber core and cladding diameter [m] 980 / 1,000
numerical aperture 0.50
Transmitter Characteristics
center wavelength min./max. [nm] 640 / 675
spectral width, FWHM [nm] <35
standard increased
maximum transmitting power binary 1 [dBm] -31 -29.5
transmitting power binary 0 max./min. [dBm] -5.5 / -11 -3.5/-8
maximum overshoot binary 0 [dBm] -4.3 -2.3
Receiver Characteristics
center wavelength [nm] 640 / 675
maximum received power binary 1 [dBm] -31
maximum received power binary 0 [dBm] -5
minimum received power binary 0 [dBm] -20
pulse width distortion min./ max. [ns] -20 / 80
jitter min/ max [ns] 0/ 15
BER 10-9
574 7.2 Application Standards
7.2.5 INTERBUS
* Using an LED with a maximum peak wavelength of 676 nm, a minimum transmitting
power binary "0" -8.6 dBm is permissible.
7.2 Application Standards 575
POF HPCF
Type 1 Type 2 Type 1 Type 2
peak wavelength [nm] 635 - 667 635 - 692 635 - 667 635 - 692
spectral width [nm] < 30 < 30
core / cladding diameter of 980/1,000 200/230
the test fiber [m]
NA of the test fiber 0.47 r 0.03 > 0.36
max. received power binary -40 -40
"1" Psmax1 [dBm]
max. received power binary -2.75 -8
"0" Psmax0 [dBm]
min. received power binary -26.4 -28.4
"0" Psmin0 [dBm]
max. rise time [ns] 30 30
max. fall time [ns] 30 30
max. pulse duty factor 12.5 12.5
deviation [%]
The most important standard worldwide for local area networks is the Ethernet
standard. Components have in the meantime become extremely moderately priced
because of the enormously large number of pieces produced. On the other hand
the demands on flexibility and capacity in industrial automation has steadily in-
creasing. It was therefore apparent that the cost and performance advantages of
Ethernet technology should be used. The requirements for reliability and robust-
ness are much higher in the field of automation so that a specific standard had to
be developed.
As part of the work of the ITG sub committee Polymer Optical Fibers a
group of about 20 companies from the German-speaking countries got together
and in 2003/2004 worked out a recommendation for the use of POF in industrial
Ethernet. One primary goal was above all to continue to be able to use the 1 mm
standard step index POF which has since become well established. For the first
time the use of different wavelengths (red and green) with PMMA POF were
specified.
For a transmission length of 100 m the bandwidth of a standard POF at
125 Mbit/s (Fast Ethernet with 4B5B coding) does not suffice without additional
measures such as pre- and post-distortion. That is why the work group recom-
mended use of standard NA POF only up to 50 m. The DSI-POF which is also
commercially available can then be used for up to 100 m. In order to get a suffi-
cient receiving level at 100 m, green LEDs should then be used (Fig. 7.7).
576 7.2 Application Standards
LED PD
650 nm -25 dBm
50 m SI-POF
3 connector
LED PD
510 nm -24 dBm
50 m SI-POF
LED PD
510 nm -25 dBm
100 m DSI-POF
Fig. 7.7: Proposal of the working group for the Fast-Ethernet link specificationen
The lower attenuation of the POF in the green spectral range can also be used
as an alternative in order to be able to use additional plug-in connectors. If you
combine a green transmitter with a standard or DSI-POF up to 50 m, then the
power rating permits up to 3 additional couplings - assuming a maximum loss of
2 dB per plug-in connection (see Table 7.18).
Table 7.18: Power budgets for the different link options (proposal)
The latest outline of Standard IEC 24702 Industry Cabling contains two
different fiber versions and four different link classes with POF. Both fiber
versions are compared in Table 7.20.
Class A4a, i.e. the standard POF, also does not come into question for 50 m
since its bandwidth in fiber Standard IEC 60793 is only specified with
10 MHz 100 m. In reality, of course, this fiber has in any case enough bandwidth
for Fast Ethernet over 50 m and diverse products are sold for it.
The standard defines another four different connection classes with maximum
distances between 25 m and 200 m. Even greater lengths can then be attained with
glass fibers - singlemode or multimode not listed here. The permissible attenuation
values for the links have been compiled in Table 7.21.
578 7.2 Application Standards
The PF-GI-POF can, of course, be used for short connections at 650 nm. One
disadvantage compared with the 1 mm POF is that you have to use a laser because
of the smaller core diameter.
System D2B is actually not a standard but rather represents a specific business
solution.
The D2B Standard specifies a ring system which connects different devices in
vehicles such as navigation computers, car radios, CD changers, telephones, etc.
using POF (Fig. 7.8).
central
interrface unit
active CD changer
speakers
navigation
system
Table 7.22 summarizes the most important data for the D2B standard. A de-
tailed description can be found in [Pet98]. A special plug system is used that per-
mits plugging at 90 or 180. If necessary, the ring can be separated and equipped
with a coupler for hooking up another device.
Parameter Value
minimum LED power -15 dBm
maximum POF attenuation 400 dB/km
range 8m
system margin 5 dB
coupling loss 1.3 dB
decoupling loss 0.3 dB
add-on coupler 1.2 dB
receiver sensitivity -26 dBm
user data rate 5.6 Mbit/s
temperature range -40C to +85C
The D2B has a maximum transmission length of 8 m. If one or two couplers are
used in the line, then the permissible line length is reduced to 7.0 m or 3.6 m res-
pectively (D2B02). The components are booted up and the diagnosis is carried out
with D2B over separate copper conductors.
The polymer fiber used with D2B has a 980 m thick core and a 2.2 nm thick
protective sheath. The minimum bending radius is specified at 25 mm >Her02@.
The protective sheath is in two parts. The inner sheath is 1.5 mm and black to pre-
vent the possible coupling-in of light.
D2B: MOST
2.2 mm
2.3 mm
1.5 mm
1.51 mm
1.0 mm
1.0 mm
The sheaths cannot be separated. The POF for MOST (see following section)
also has a two-part sheath. However, the inner sheath is solidly attached to the
fiber and the outer sheath can easily be cut back. The purpose of this arrangement
is that the connector with MOST is not crimped on the fiber, but on the inner
sheath or is welded with a laser (fiber set-up in Fig. 7.9).
Fig. 7.10: View on the POFs for D2B and MOST ([Her02])
SI-POF
SP 1 SP 4
B = 22.5 Mbit/s
D = 9 dB
MOST SP 2 SP 3 MOST
device device
The sampling frequency of a CD player at 44.1 kHz forms the basis of the bus
clock pulse. The formation of blocks each with 512 bits results in a gross baud rate
of 22.6 Mbit/s. Principally, all components are designed for a frame clock-pulse
rate of between 30 and 50 kHz. Up to 64 nodes can be connected in a MOST link.
A differentiation is made between synchronous data (with permanently
assigned channels), asynchronous data (using available channels) and control data
with permanent assignment within a time frame. Because of the frame clock-pulse
rates the delay times are at a maximum of 25 s.
The time frame for the transmission of the different data is shown in Fig. 7.12.
As can be seen, a division between synchronous and asynchronous data is possible
and can be varied.
Although it was first conceived as a pure ring architecture, the MOST system
can assume other topologies, e.g. combined rings, star, etc., by adding system
master units.
582 7.2 Application Standards
64 bytes
30 to 50 frames per second / 20 s to 33 s
The data structure of the MOST system is oriented extensively toward the
demands of the connectable multimedia terminal devices. Table 7.25 indicates the
various possible data formats with their respective bit rates.
The Standard IEEE 1394 was primarily developed for the networking of devices
in entertainment electronics in apartments (see above). Over the last few years the
applicability of this standard in vehicle networks has been discussed within this
group. In Sections 6.2.2 (the power budget) and 8.1.1.4 (applications) further
information is provided. Except for information on the optical parameters of the
active components hardly any details have been published. We can expect a con-
tinuing development of the MOST standard toward higher bit rates and that trans-
mitters and receivers will be used for both systems. Up till now IDB 1394 has pro-
vided for the use of the SMI connector which, however, does not completely meet
the requirements of automobile technology in its present design, e.g. in regard to
the connection cycles.
Since then there has also been talk in the meantime of the specification of a
POF with a considerably enlarged NA in order to reduce the bending sensitivity.
7.2 Application Standards 583
7.2.10 EN 50173
Standard EN 50173 Generic Cabling Systems is one of the most important stan-
dards at all. It describes the set-up of structured data and telecommunication net-
works. Up till now this standard was of primary interest for public buildings and
companies. Because of the rapid development of broadband connections there will
be structured cabling of residential buildings in the future. Todays telephone and
coaxial cable networks are only oriented toward one specific service and it is not
generic.
The structured cabling in office buildings should make it possible to combine
existing and future applications with one infrastructure. The basis of this approach
has been the use of balanced copper cables. This cabling has, of course, never
really been generic. In commercial networks there are actually only two relevant
services: great amounts of data are transmitted over Ethernet connections and then
there are telephone connections on a 64 kbit/s basis. This low data rate does not
place any demands whatsoever on the cable quality; only two twisted wires are
needed. Consequently, the original EN 50173 is essentially a standard shaped by
Ethernet applications. Cables in Category 5 have been developed for the trans-
mission of Fast Ethernet. Thanks to the 4B5B coding a bandwidth of 62.5 MHz is
584 7.2 Application Standards
used. In order also to be able to use the same infrastructure for 1,000 Mbit/s, the
transmission procedure has been completely revised. A multi level code is now
used instead of the binary coding. The separate transmission in both directions
with a pair of conductors in each direction has now become a bi-directional trans-
mission on all four twisted pairs (Fig. 7.13).
Fast-Ethernet:
100 Mbit/s 4B5B 125 Mbit/s binary: 62.5 MHz
Fig. 7.13: Idea for using the same channel for 100 and 1,000 Mbit/s
Not every Category 5 cable, however, can be used for Gbit. Ethernet, especially
because of the lack of return loss. That is why the intermediate Class 5e has been
introduced. Nowadays cables from Categories 6 and 7 are primarily used which
are capable of 1 Gbit/s without any problems. On the other hand, the transition to
10 Gbit/s will not take place generically since completely new connectors are
necessary and the range of 100 m will presumably not be possible.
The IEC is presently endeavoring to expand the standard to other applications.
The different areas are:
50173-1: general requirements
50173-2: office
50173-3: industry
50173-4: home
50173-5: data centers
The DKE (Deutsche Kommission fr Elektrotechnik/German Committee for
Electrical Engineering) was requested to work out a recommendation for the use
of polymer fibers in the home. A joint work sub-circle (GUK) 715.3 was created
for this purpose.
The situation in the home environment is that much more complicated than in
the commercial area. In addition to telephones and the Ethernet you have to take
analog television as a service into consideration. If you include the satellite
intermediate frequency band, then you have to have a cable with a bandwidth
above 2 GHz with a considerably higher crosstalk attenuation. Only balanced
7.2 Application Standards 585
copper cables in Category 8, which is planned, or good coaxial cables could meet
these requirements. Silica glass fibers - only singlemode fibers offer sufficient
capacity, but are unsuitable for private customers.
The goal of a generic infrastructure which is supposed to meet home demands
in the next couple of decades is extremely ambitious and actually cannot be
fulfilled. For polymer fibers the first problem already arises with the inadequate
directives of the fiber standard IEC 60796-2-40. The capacity of standard POF can
be viewed from three different angles:
According to IEC 60796-2-40 Class 4a has a bandwidth of 10 MHz 100 m.
Therefore, Fast Ethernet can just be transmitted over 15 m.
In reality all commercial 1 mm standard POFs have a bandwidth of about
40 MHz 100 m. Fast Ethernet over 50 m is always possible, with some effort
also 100 m and more, is really not a problem. Copper cables are also strongly
equalized. A Category 5 cable only has a bandwidth of 3 MHz 100 m.
Up to 1000 Mbit/s over 100 m are even possible on standard POF with
methods which correspond to the gigabit transmission on copper cables
(DMT, Siemens 2006, see Chap. 6).
Taking the actual capacity of POF into consideration, the group has recommen-
ded three different link classes. A range of 25 m should be guaranteed within
apartments. For medium-sized and large residential buildings connections up to
50 m or 100 m respectively are planned. Connection lengths over 100 m are prac-
tically never the case in residential buildings. In Germany there are only very few
buildings with more than 15 stories. Typically, 4 to 8 apartments are grouped
around an elevator shaft. Very large residential buildings consist of segments
which are separated by firewalls through which cables may practically never be
pulled.
Table 7.27 illustrates the proposed link classes with the possible corresponding
applications and the fiber classes declared according to IEC 60796-2-40.
Table 7.27: Proposal of the GUK 715.3 for link classes in building networks
Under certain circumstances fibers with a lower category, e.g. with equaliza-
tion, can be used. Since the standard is still being debated, we will not go into a
description of the power budget calculations. It is becoming apparent that only
fibers A4d (DSI-POF) and A4g (PF-GI-POF) will be considered in the final ver-
sion, that link class 25 m will be eliminated and that the use of blue and green
sources is not planned. In effect, this standard is practically useless. Even today di-
verse products for Fast Ethernet over 50 m to 100 m on standard POF are avai-
lable in the market. Fiber A4d does indeed offer nominally sufficient bandwidth
for Fast Ethernet over 100 m. However, the argument against it is that moderately
priced fibers in Category A4e (MC and GI) are already available today which can
also be used for Gbit/s. The recommendation to use Class A4g (120 m
PF-GI-POF) surely makes sense. Now we will have to wait and see how success-
ful this type of fiber is when introduced into the market.
Even today it is apparent that the actual development of building installations is
passing by the concept of this standard:
Standard POF (A4a) is already established in many applications and is used
predominately in home networks in the beginning. Fast Ethernet over 100 m
and Gigabit Ethernet up to 25 m are possible without any problems.
From the editors point of view primarily MC-POF and PMMA GI-POF -
when the bending problem is solved - are becoming apparent candidates for
higher data rates. Both allow the use of existing connectors and connector less
installation respectively.
The PF-GI-POF mainly makes sense for the transmission of CATV signals.
This fiber version is to be recommended in any event even with distances
over 100 m. Bit rates up to 10 Gbit/s have already been realized. If PMMA
POF is considered the first generation, then PF-GI-POF can dominate the
succeeding generations.
As far as the transmission of CATV signals is concerned, we will have to wait
and see as to whether any respective systems will be developed at all. The techno-
logical basis has been available for years. Today, however, a rapid transition
toward a general transmission of all television signals over the internet protocol
(IP TV) is evident. Together with VoIP (Voice over IP) there is again a return to
generic networks because only one network is needed, as a rule Ethernet, for so-
called triple play applications.
In summing up, we can say that different standards apply when dealing with
copper cables and POF as regards standardization. When copper cables move up
to higher bit rates they are granted comprehensive changes in the transmission
procedures which are not granted to POF. There are a number of different versions
(Categories 3, 5, 5e, 6, 7, shielded and unshielded, with 100 : in Europe and
150 : in the USA) whereas only a few versions are allowed for POF. In addition,
the inclusion of analog TV signals - with a questionable remaining propagation
time - tightens up the requirements dramatically. The goal of a long-term
generic approach as opposed to a market-oriented one may possibly be too
greatly overemphasized.
7.3 Standards for Measurement Techniques 587
In comparison with multimode or singlemode glass fibers there are only a few
standardized methods of measurement for POF. They have been set down in the
Japanese Industrial Standard (JIS) and in various IEC standards. Table 7.28 sum-
marizes the standards for measurements of POF.
After the series production of the MOST standard had begun, it was soon dis-
covered that the measurement results for attenuation in prefabricated cables could
deviate considerably. The result of this knowledge was the formation of a working
group which was to first work out unified rules for the measurement of attenuation
in fibers and cables.
Further steps then included working out recommendations for measuring the
mechanical reliability and the influence of climate and chemicals. The document
has appeared as VDE/VDI Recommendation 5570: Testing of prepared and
unprepared plastic optical fiber (POF). It consists of four parts and contains:
Part 1: Terms and Definitions
Part 2: Test Procedures for Optical Characteristic Values
Part 3: Test Procedures for Mechanical and Environmental Characteristic
Values
Part 4: Power Budget
The working out of part 5 for the measurement of transmission characteristics,
i.e. bandwidth, pulse broadening, etc., is planned. Here we simply wish to point
out the main aspects of this recommendation since many of the procedures are
described in Chapters 2 and 9.
The recommendation begins with a definition of the fundamental terms used.
Surprisingly, their use is by no means the same everywhere. The designations for
fiber and vables are shown in Fig. 7.14.
optical core
(PMMA)
fiber
optical cladding
(Fluorpolymer)
jacket cable
(PA, PE, PVC...)
optional secondary
jacket
1.0 mm
2.2 mm
Deff
6=P0 6 = P` Deff
DPOF(ORef)
O O O 0m l O
step 1: s te p 2 : s te p 3 : s te p 4 : step 5:
LED spectral LED spectrum Deff calculated KF calculated
spectrum POF is multiplied Deff
Deff = = D (O )
measured attenuation with the loss POF Ref
10lg(P0/P)/l
measured curve
As can be demonstrated, this correction factor can also be used when the actual
spectral attenuation curve deviates from the typical curve accepted in the first step.
However, the difference should for the most part be wavelength independent and
not too large. In the subsequent Table 7.29 the steps for the formation of the
correction factor in formulas are given.
590 7.3 Standards for Measurement Techniques
Table 7.29: Calculation of the spectral correction factor for loss measurements with LED
In any event, you must keep in mind that the correction factor varies in a non-
linear way with the length of the fiber. One example should describe the method
in which the measurement of the POF attenuation was made at 650 nm with a
laser.
According to the table the PMMA POF has an attenuation coefficient of
132 dB/km at 650 nm.
A red LED with a width of 40 nm, a center wavelength of 650 nm and a
Gaussian-shaped spectrum is used. An effective attenuation coefficient of
185 dB/km, i.e. an additional attenuation coefficient of 53 dB/km, is the result
for a length of 10 m, corresponding to a correction factor of KF = 1.13.
With a reference value of 30 W the result is a measurement value of
Pmeas = 18.5 PW with a 10 m long test fiber (you have to keep in mind that the
correction factor is always valid only for a certain length).
Consequently, the attenuation coefficient measured is:
Dmeas = 10 log (30/18.5)/0.01 km = 210 dB/km.
The reference attenuation quantity per unit length with a correction of
53 dB/km is: 157 dB/km. With a reference value of 132 dB/km the difference
is only 25 dB/km for 0.01 km = 0.25 dB so that the use of the correction
factor is permissible.
7.3 Standards for Measurement Techniques 591
The editor recommends a test set up as described in Section 9.4.5.4 for the
exact measurements of the spectral attenuation under laboratory conditions.
Other parts of the recommendation describe the measurement of the vicinity of
the emission characteristics of a source for distributing the equilibrium mode
distribution of a fiber, the so-called EMDicity.
The different methods of measuring the numerical aperture are also introduced
which include:
Far field method
Reflection method
Inverse far field method
The production and application of reference fibers are described very exten-
sively, followed by definitions of sources of error which are taken into account.
Finally, measurement procedures are described on part 3 for mechanical, clima-
tic and chemical environmental influences. The recommendations here essentially
come from BAM and are described in part in great detail in Sections 9.6 and 9.7
which has not changed since the first edition.
In hardly any other area has the number of applications so rapidly developed as in
the area of optical short-range communication. Applications of polymer fibers,
glass fiber bundles in the field of lighting technology as well as in automation
have been established for many years. Since the end of the 1990s POFs have also
been used in various mobile networks. At the time when our first edition was pub-
lished there had hardly been any information published on MOST applications.
This second edition now contains a detailed description. The use of POF and other
thick optical fibers in the fields of sensor technology and home networks are just
about to be used on a large scale.
Even greater perspectives for optical technologies are becoming apparent in the
fields of interconnection. This area will also be covers in the following chapter
more detailed.
Those areas of use which lie outside the field of data communications and
beyond the scope of this book will only be treated briefly. For these areas we refer
to existing publications ([Wei98], [FOP97]).
The most important media for transmitting high data rates today are electrical
lines, mostly copper, optical fibers and radio. Each one of these channels has its
own special characteristics:
Electrical lines connect the transmitter and receiver directly. Contacts bet-
ween lines can be made easily. The range and data rates are primarily limited
by the skin effect, i.e. the attenuation increases with f1/2.
Optical systems work with light as carrier frequency. The bandwidth is most-
ly limited by the effects of dispersion, whereby the transmission behavior is
2
almost Gaussian-shaped (e-(f/f0) , Fig. 8.1). The distance is limited by the
attenuation of the optical path. In most cases connections inside the link
require great precision.
The special feature of radio is that all users within a cell (the range of the
transmitter) have to share the capacity. Because of the multi-path propagation
and the resulting interferences and external sources of disturbance extremely
complicated channel behavior results which has to be compensated for
through adaptive procedures.
594 8.1 Data Transmission with POF
attenuation [dB/km]
1000
SI GI SM
100
copper cable
10
optical fiber
0.1
105 106 107 108 109 1010 1011 1012
modulation frequency [Hz]
Fig. 8.1: Comparison of the frequency responses for copper cable and different optical
fibers (typical values)
Compared with all other media, optical systems offer by far the greatest
capacity - which can be multiplied at will through parallel lines - the least distur-
bances and the greatest reach. The low need for space and the lower power input
ensue with raising bit rates. In such a case optical lines have to be installed and
opto-electrical converters have to be used at the ends of the transmission path.
Optical data transmission will only then be used when conventional procedures
reach their limits. Up till now this has been the case for glass fiber systems especi-
ally in telecommunications networks and in large corporation networks. The low
cost of polymer fiber technology now opens up entirely new fields for optics,
especially in short-range transmission up to some 100 m. Sometimes conventional
copper wire solutions are substituted, but for the most part applications are being
employed, the realization of which so far had not made much sense. Data commu-
nication with POF can be divided into the essential areas indicated in Table 8.1.
The use of POF in buildings and apartments can be quite problematical. In con-
trast to the automobile industry complete systems are not installed just once; the
networks are constantly being expanded and improved by installing faster com-
ponents. The fiber infrastructure thus has to be dimensioned not only for current
use, but also have the potential for use in new kinds of systems in the future
demands.
8.1 Data Transmission with POF 595
Table 8.1: Applications and requirements for data transmission with POF
Application Typical Parameters Specific Requirements
mobile networks lengths between 10 m complete systems
cars (car) and 200 m (ships, critical environmental
trains/ships airplanes) conditions
airplanes data rates up to extreme high reliability and
1 Gbit/s long live time
LAN lengths 25 - 100 m simple installation
office data rates 100 Mbit/s to different data formats
home 1 Gbit/s mix of different components
condominium
Interconnection many parallel channels very small
on board data rates to 10 Gbit/s low power operation
intra board centimeters to meters automatic equipment
In addition, the building networks can combine the components of many diffe-
rent manufacturers. In an automobile, for example, one manufacturer supplies the
entire cable harness. Connecting extraneous components is avoided when pos-
sible.
The following sections will provide the reader with an overview of data com-
munication with POF in the various areas of application. Please refer to the
chapter on standards for the definition of technical details.
In Europe, the use of polymer optical fibers for the entertainment networks in
DaimlerChrysler vehicles since 1998 represents the first comprehensive appli-
cation of POF in data communications. The following arguments speak for the use
of POF in vehicles (e.g. [Zam00a]):
low cable weight
small cross-section
insensitivity to electromagnetic interferences
The different standards for car networks with POF were described in greater
detail in Chapter 7. The most important representative areas are:
CAN (Controller Area Network)
D2B (Digital Domestic Bus)
MOST (Media Oriented System Transport)
IEEE 1394 (presently not yet specified for the automotive field, but concei-
vable as a future system)
Byteflight (passive star system for vehicular control, [Pan00])
In vehicles, airplanes and rail transportation more and more digital communica-
tions connections are being utilized. As a result, increased demands on the archi-
tecture of the data connections as well as the transmission media are being made.
596 8.1 Data Transmission with POF
In the area of driver information and entertainment systems, less relevant in regard
to safety requirements, serial bus systems are being increasingly used. The indi-
vidual devices are connected in series by means of high-rate connections. The
advantage here is the saving of cables. The disadvantage is the breakdown of an
entire series of devices when a transceiver subassembly is defective.
Figure 8.2 shows the number of cables in a mid-sized car according to
[Zam00a]. Some years ago, the cables for power supply had the bigger portion,
whereas today the strongly increasing number of data connections dominates.
number of cables
1,000
total
power supply
data connections
500
year
0
1980 1985 1990 1995 2000 2005
Fig. 8.2: Number of cables in cars
Since 1997, optical components for use in automobiles are available from
Harman/Becker Automotive Systems ([Sch01]). Since 1998, such components
are standard features e.g. in the DaimlerChrysler Vaneo as of 2001, Fig. 8.3.
Figures 8.4 and 8.5 from [Sch01] show the development of different multi-
media terminal devices in the automotive field. At the beginning of the 1990s CD
changers (CDC) were first installed as a complementary unit to the car radio. Later
on, digital amplifiers (Amp) were added. Combinations of car radios and mobile
telephones (Tel.), in part with separate voice data systems (Voice IO), completed
the features offered.
In the meantime, vehicles are equipped with additional devices such as naviga-
tion systems (Navi), traffic guidance systems (telematics), mobile internet access
and DVD players.
2001
MMI amp
CDC TV
phone DVD
navi voice IO
1998 internet telematics
radio
???
amp
1994 CDC
radio phone
1990 voice IO
amp
CDC CDC
radio
Figure 8.5 (also from [Sch01]) illustrates the development of optical bus sys-
tems. The devices listed above are optically connected to diverse input systems
such as monitors located in different parts of the vehicle. Up until recently the
electronic media in the vehicle have been primarily allotted to the driver and thus
essentially served the purpose of vehicle control support, whereas now the focus is
more and more on the entertainment of the passengers. The first commercially
available products are back seat monitors on which television programs can be
received or DVD can be played.
598 8.1 Data Transmission with POF
8.1.1.1 D2B
The D2B system was developed by DaimlerChrysler in 1998. The primary goal
was the transmission of audio signals between the different components in enter-
tainment electronics. The audio systems were considered to be the most important
application. Figure 8.6 from >D2B02@ shows the configuration in the M Class.
Radio or MCS
D2B master
coupling
POF POF
index gel
The spacing between the fiber end faces is less than 1 mm so that the losses per
connection remain below 2 dB.
8.1.1.2 MOST
The use of D2B is limited to only one manufacturer which is why a drop in price
could not be achieved despite large production quantities. This may have been one
of the main reasons for the development of MOST. German automobile manufac-
turers in particular took the lead in this consortium. The MOST Corporation was
founded in 1998 by BMW, DaimlerChrysler, Becker Radio and OASIS Silicon
Systems. In 2001, the 7-series BMW was the first series production model world-
wide equipped with this data bus.
The first version of the MOST-bus can transport up to 25 Mbit/s. According to
>Thi03a@ the prices for a MOST-link dropped from about 10 to 5 between
2002 and 2003 (the optical components were two-thirds of the costs).
An overview of the development of MOST technology can be found in
>Muy05a@, >Muy05b@ and >Thi03b@. The figures mentioned in the different refe-
rences on the use of MOST technology are:
600 8.1 Data Transmission with POF
Fig. 8.8: Car series with MOST bus (Sept. 2005, [Muy05b])
One great advantage of the MOST technology is the use of standardized trans-
ceivers, fibers and connectors. The transmitters and receivers for MOST are
shown in Fig. 8.9.
The necessary transmission data rates increase parallel to the number of devices
connected. Even now the introduction of 50 Mbit/s and later 150 Mbit/s is being
prepared in the MOST Consortium. At least in the long-term data rates of
400 Mbit/s and more can be expected. Polymer fibers themselves offer sufficient
bandwidth even for these speeds. The development of transmitting diodes makes
such systems appear realizable. The design of correspondingly more sensitive and
faster detectors is still somewhat problematical. Even just a few years ago it was
thought that higher data rates could only be realized with smaller receivers since
the photodiode capacity was otherwise the limiting factor. In the meantime, this
hypothesis has since been refuted. Even with 1 mm thick fibers data rates far
above 1 Gbit/s can be realized.
The following current versions have already been realized and included in the
MOST standard:
LED/POF solution for 50 Mbit/s
RC-LED/POF products for 150 Mbit/s have been certified
Electrical transmission on twisted pair copper cables with 50 Mbit/s (Fig. 8.12)
Transmission of 150 Mbit/s with PCS and 850 nm VCSEL
The use of Gigastar links was proposed in >Kra02b@ as a solution for high data
rates. Up to 1,300 Mbit/s can be transmitted over unshielded twin conductor
cables as a differential CML signal or one-pair STP cables (3 mm). The system
allows a maximum link length of 30 m.
8.1.1.3 Byteflight
The Byteflight system has only been used so far by BMW. It connects airbag sys-
tems with other control components in the Intelligent Safety Integration System
(>Gri00@). The connectors and fibers used correspond to the MOST standard.
However, the data is transmitted bidirectionally on one fiber. The topology is an
active star. The concept of a passive star did not work out because of the large
insertion loss of the central coupler. A BMW of the 6-Series displayed at the
POF2004 in Nuremberg equipped with MOST and Byteflight can be seen in
Fig. 8.13. According to >Fre04c@ a BMW contains 12 sensors for speed, accelera-
tion, and pressure for the airbag system.
Fig. 8.13: 6-series BMW with MOST and Byteflight (POF2004 Nuremberg)
1 2 3 ... 10 35 1 2 3 ... 10 38 75
high priority
messages time for low priority messages
e.g. comfort functions, diagnostics
Fig. 8.14: Data structure for Byteflight
The specified data of the transceivers according to >Sch00b@ and >BFT03@ are:
optical output power at 30 mA: -5.2 dB
rise time and fall time: < 35 ns
peak wavelength at 25 C: 650 10 nm
peak wavelength from -40 C to +85 C: 650 20 nm
receiver sensitivity: < -23 dBm
maximum receiving power: -1.0 dBm (800 W)
power consumption in standby mode: < 10 A
operating temperature: -40C to +85C
At present there are still several directions the on-going development of the
Byteflight system could take. It seems to be certain that the concept of the active
star is being primarily expanded for critical safety applications. The most probable
successor to Byteflight is Flexray with a system bit rate of at least 100 Mbit/s. So
far, Flexray has only been planned as an electrical version although technically
both POF as well as PCS are usable.
motive is primarily the desire to transmit uncompressed video data and the pos-
sibility to connect external devices to the vehicle system via a Customer Conve-
nience Port (CCP). In order to meet the greater requirements in vehicles, the stan-
dard version IDB 1394 has been developed.
The carmakers Renault (Espace model) and Nissan (Fig. 8.16) have demon-
strated prototypes with IDB 1394 in the past few years. Renault uses LED-based
200 Mbit/s connections which can transmit 3 simultaneous video signals for a
DVD player, digital television and a rearview camera.
In the Nissan vehicle seven different cameras can be retrieved over the
400 Mbit/s network. The driver can simultaneously view up to four pictures on
one screen.
A description of the IDB 1394 draft is given in >Tee01@ and >Lit03@. Symme-
trical twisted copper cables as well as POF are planned as media for longer links.
The maximum transmission lengths should be 18 m - 10 m with two plug-in con-
nections. One great advantage of the 1394 specification lies in the free choice of
topology. Tree, mesh, star or ring network topologies are possible. A connector
based on the well-known SMI connector is planned for the CCP. The data struc-
ture of 1394 has been optimized for multimedia applications. Real time applica-
tions such as video transmission as well as data connections with variable bit rates
have been realized.
Components for POF-based IDB 1394 are already available. Firecomms intro-
duced new RC-LEDs in >Lam05@ which can be used for 200 Mbit/s under auto-
mobile conditions. The operating temperature ranges from -40C to +95C with a
POF-coupled power of more than -5 dBm.
A comparison between PCS and POF in regard to the possible power budget is
given in Table 8.2 in >Zeeb02@.
Table 8.2: Comparison of the power budgets for POF and PCS
At first, the PCS seems to come off much better. The advantages of the system
are primarily based on three characteristics:
The fiber attenuation of the PCS with 10 dB/km is practically negligible com-
pared with that of POF (300 dB/km). With a maximum connection length of
10 m, however, this only amounts to a few dB.
The guaranteed output power of the VCSEL is much higher than that of
MOST LED. The reasons for this are a more efficient coupling-in and
especially the output power control. These measures were dropped with the
MOST LED for reasons of cost. An adjusted LED with optimized coupling,
e.g. with micro-mirrors, could also guarantee a very much higher fiber-
coupled power.
Very much better values can be achieved for the attenuation of PCS connec-
tors than are specified for the MOST POF connector. However, much lower
tolerances (about 10 m) have to be met and the surface preparation of PCS
requires much more effort, e.g. cutting with a CO2 laser. The attenuation for
POF connectors could also be reduced considerably if the tolerances of the
connectors and fibers were reduced to the level of PCS.
In any case, it is advantageous that the 850 nm VCSEL can be modulated with
high data rates without any problems. It certainly remains questionable whether
the increased demands on adjustment tolerances and transmitter stabilization can
be realized with the comparably lower POF manufacturing costs.
A possibility for setting up the required central coupler in a passive star is
shown in >Bu00@. The different fibers are coupled to a mixing cylinder with a
diameter of 1,000 m (Fig. 8.18). Insertion losses between 17.10 dB and 18.53 dB
were established in three measurements (Fig. 8.19).
608 8.1 Data Transmission with POF
200 m
PCS fibers mixing
cylinder
reflective
surface configuration
with 16 PCS
For bi-directional data transmission >Bu00@ proposes the use of VCSEL mo-
dules in which light is combined between a photodiode and the transmitter via a
micro-mirror. Up to 250 Mbit/s are possible in half-duplex operation (transmission
power -2.5 dB and the receiver sensitivity -20 dB).
The PCS solution is specified in the MOST standard as Advanced Optical Phy-
sical Layer (aoPHY, >Pof06@). A passive star is no longer planned; point-to-point
connections will now be set up. The specified parameters of the links are:
fiber: 200 m/230 m SI-PCS
transmitter: 850 nm VCSEL
power budget: 20 dB
minimum bending radius: 9 mm
data rate: >150 Mbit/s
connector: MOST compatible
8.1 Data Transmission with POF 609
Sample components for fibers, connectors and active elements are available
from OFS, Polymicro, Infineon, Yazaki, Advanced Optical Components, Delphi,
Leoni and Tyco. There is varying information on the set-up of the PCS lines but
there is uniformity as regards use of 200 m/230 m with a NA of 0.37. The outer
jackets consists of PA12 and has a diameter of 1.5 mm or 2.3 mm respectively.
An example of commercially available VCSEL is given in >Pof02@. Here a
850 nm VCSEL Honeywell 4085-321 was tested. Between -55C and +125C the
output power without adjustment would fluctuate between 10 W and 150 W
(Fig. 8.20).
A relatively constant output power can be adjusted (0.90 0.25 mW) with the
aid of a temperature-dependent resistor network in the VCSEL control, i.e. a com-
bination of NTC and PTC. An error-free transmission of data with 500 Mbit/s
could then be realized.
1,000
bit rate [Mbit/s] IEEE 1394 ?
100
MOST2
10 MOST
D2B
1
CAN
year
0.1
1996 1998 2000 2002 2004 2006 2008
Fig. 8.21: Potential bit rate development in automobiles (prognosis from 2001)
Today it seems to be clear that there is no pressing reason for replacing POF. In
the meantime, there are RC-LEDs among the active components which also attain
150 Mbit/s and clearly improve the power budget and temperature range. SI-POFs
have shown that without qualification they allow up to 2.5 Gbit/s over 10 m. A
PMMA POF with a temperature range up to +105C is available on the market
which has only a slightly higher attenuation compared to standard POF.
On the other hand many experiments have demonstrated copper-based solutions
for high data rates. Both shielded twisted cables as well as coaxial cables have
been tested. An example is described in [Beer05]. Here the data is transmitted at a
maximum of 800 Mbit/s with symmetrical shielded cables. The maximum
operating temperature is +125C and the minimum bending radius of the cable is
10 mm. FCI has developed a corresponding plug-in connector.
Fig. 8.22: Proposal for a MOST copper connector (for up to 800 Mbit/s, [Beer05])
At least under laboratory conditions the newer copper solutions are also con-
form to the limiting values for electromagnetic compatibility (one of the main
arguments for using optics was the protection from disturbances). A very decisive
step for the continued expansion of POF in vehicle networks could be the develop-
ment of very fast (1 Gbit/s) modulable LEDs which have already been achieved in
the laboratory. Only when the link costs in optics come close to those of copper
technology, will they be successful.
8.1 Data Transmission with POF 611
digital
device
data connection
with CMT-POF
42 V power supply over
the metal sheath
Fig. 8.23: Connection of devices with power supply via the protective cladding
Space requirements and the weight of the cable harnesses can be limited in
spite of the increasing number of data connections. Even in the MOST specifica-
tions different hybrid plug-and-socket connectors are planned so that the power
supply and the data transmission can be installed at the same time.
Solutions for the required robust connectors have already been developed for
trains. The 200 m PCS as well as 1 mm POF (SI-POF with MOST specifications
and multicore fibers) are currently being tested. The camera equipment - digital or
analog - with optical fibers is shown in the following sketch.
connector camera 4
camera 2
coupling
trailer
tractor trailer
A flexible connection between the tractor and the trailer naturally constitutes a
great challenge. So far relatively thick coiled cables have been used which are
about 7 m long (cable length), contain around a dozen electrical conductors and
have an outer diameter of 15.5 mm. The spiral diameter amounts to 80 mm
(Fig. 8.26).
8.1 Data Transmission with POF 613
2m 5m 7m 20 m 1m
fiber coupling
Fig. 8.27: Maximum connection length
Corrugated sheath cables are used for the fixed cable connections in the tractor
and the trailer (Fig. 8.28) in order to assure optimum protection of the light wave-
guide. Ratioplasts plug-in connectors guarantee a low insertion loss. Light is col-
limated with the aid of a lens so that a secure connection is guaranteed contact-less
and even at relatively great distances. The necessary angular exactness is ensured
by the guide of the hybrid connector.
Fig. 8.28: Lens connector for camera link with corrugated micro tube cable
A complete box for the multi-pin hybrid cable with protective lid and an inte-
grated optical fiber is shown in Fig. 8.29. The optical fiber in the corrugated
sheath tube can be seen in the cable bundle in the side view.
The use of POF in vehicle networks represented the first large-scale application
and has pushed the development of polymer fibers far ahead. An even greater
potential consists of apartments and building networks which requires differenti-
ating between different fields of application. In office buildings the connection of
8.1 Data Transmission with POF 615
Data networks in office buildings as a rule are planned and set up very care-
fully. The use of shielded cables rather than unshielded cables dominates in
Germany - in contrast to the U.S.A. Paying careful attention to a unified ground
potential throughout the entire building makes an optimal use of the advantages of
shielded cables possible. Consequently, electromagnetic disturbances do not play
a major role in data networks, at least when properly installed.
5 6 8 9
4
7
3
The data cables in office buildings are usually laid on grids below the respec-
tive floor ceilings. The high space requirements for the data cables does not play a
significant role.
Connecting electronic devices to the electric circuit and through data networks
always produces loops which can act as antennas or even create undesired current
paths. In commercial use these problems should always be taken into conside-
ration. Above all, the problem of induction, e.g. caused by lightning striking, has
to be solved by means of appropriate protective grounding. In such a case POF
would be an interesting alternative, which could surely be used in special applica-
tions. Practicable and proven solutions do exist for copper cables, too.
phone
TV
electrical
power
As can be seen, only the electrical power supply effectively connects all rooms.
The telephone and broadband networks do in fact provide a connection to the
access networks, but not the possibility of networking different terminal devices
within an apartment as illustrated in Fig. 8.32.
TV outlet
TV with
DVD-recorder childrens
room with TV
intelligent guest room and computer
refrigerator with TV
advanced and thoroughly affordable. However, the possible bit rates and the
attainable quality are subject to definite limitations. Cable-based systems are pre-
ferable when transmitting high-quality moving pictures in real time or with a
broadband connection of computers, for example, when working at home. Diffe-
rent copper cables as well as optic fibers can be considered. Table 8.3 summarizes
some possible technologies for use in private surroundings.
As can be seen in the table, the PMMA POF lies in the mid-range of perfor-
mance characteristics for the various transmission media. As regards the simpli-
city of installation, radio systems and PLC, of course, cannot be surpassed.
Among the cable-based systems, POF is distinguished as having the easiest cable
8.1 Data Transmission with POF 619
setup and the most reasonably priced connection technology. A size comparison of
different cables is illustrated in Fig. 8.33, clearly demonstrating that POF can be
integrated very well into existing cable duct systems.
coaxial cable
: 7 mm
4 pair
data cable
: 7.5 mm
Fig. 8.34: The red one or the black one? Among the media for inhouse cabling, the custo-
mer has the choice between symmetrical copper cables or POF mainly (Photo:
I. Mnnl, University of Applied Sciences Nuremberg)
620 8.1 Data Transmission with POF
Besides the question of transmission media, the point of greatest interest is the
interface to the consumer. A system can only gain general acceptance when ter-
minal devices are equipped with appropriate connectors, the services desired can
be supported with sufficient quality and the components for setting up the network
are available at reasonable prices. Table 8.4 lists some of the interesting interfaces.
POF systems have already been created for all 4 interfaces mentioned. The
ATM forum has already specified the use of PMMA POF for 155 Mbit/s. Of parti-
cular interest is the inclusion of POF in the IEEE 1394 specification (up until now
100 Mbit/s and 200 Mbit/s over 50 m; 400 Mbit/s over 100 m is in preparation). In
contrast to Ethernet, this interface could gain acceptance not only with computers,
but also in diverse multimedia devices such as game consoles, cameras and video
cameras, televisions and DVD players and with computer peripherals.
The IEEE 1394 standard is intentionally not fixed to a medium, but provides
the user with the option of selecting his own cable. Therein lies great application
potential especially for POF as illustrated in the overview above.
In addition to the question of possible interfaces, the general building network
market in Germany should be considered. In contrast to other countries such as
Japan or the USA most people in Germany live in houses for several families.
Figure 8.35 shows the distribution in building size based on the last apartment and
house count in Berlin. On the left is shown the number of buildings in Berlin
(approx. 200,000) in different sizes (A/B: apartments per building). The right side
of the diagram compares the number of apartments (1,200,000) within the various
building size classes.
8.1 Data Transmission with POF 621
Although almost half of the buildings are single-family houses, they only repre-
sent about 10% of all Berlin apartments. Approximately of the apartments are in
buildings with 7-12, 13-20 or >20 apartments respectively. In Germany around
70% of all apartments are nevertheless in houses for several families.
In conclusion, Fig. 8.36 shows the accumulated frequency of cable lengths in
buildings - calculated for the building size distribution in Berlin. Practically all
cable lengths - measured between the access point in the house and the terminal
device - are under 100 m, typical lengths being 30 m to 40 m. It can also be seen
here that POF fits in well with the requirements not only for networks in apart-
ments, but also in the buildings (see[Kra98]).
cumulated frequency
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1 cable length
0.0
0m 20 m 40 m 60 m 80 m 100 m
Fig. 8.36: Cable length distribution in Berlin
60-70%: Spain
Germany
Italy
50-60%: Schweden
Austria
40-50%: Denmark
France
Finnland
30-40%: Netherland
Portugal
20-30%: Belgium
10-20%: Great Britain
<10%: Ireland
Surely one of the most critical questions is that of the future potential of poly-
mer fibers in the networking of buildings. As explained in detail in Chapter 2,
there has been a number of different fiber versions. Standard SI-POF is the only
version available in large quantities and which is stable and moderately priced.
However, the data rate with the systems available today is limited to 100 Mbit/s.
In order to maintain network expandability up to 1,000 Mbit/s, fibers with a higher
bandwidth should be installed which may then not function optimally with the
existing Fast Ethernet components.
The latest technical developments, however, show a sensible way out of this
predicament. One Gbit/s can also be transmitted over SI-POF with efficient band-
width procedures as they have long been used for copper or wireless. There should
be market availability of such systems in the near future. The need for such high
data rates will presumably only arise when glass fiber connections have proven
successful on a wide front. In Germany at least that will still take a few years.
Another aspect is that POF can be replaced relatively easily. Should the instal-
lation of SI-POF really no longer meet the requirements, then it can be replaced
with GI-POF with relatively little effort. These fibers should be considerably
cheaper in a few years.
The connection lengths will always be below 25 m within the apartments, typi-
cal lengths being 10 to 20 m between rooms. At such short distances 1 Gbit/s can
8.1 Data Transmission with POF 623
be transmitted trouble-free with NRZ coding over a SI-POF. Here, too, the user
can hardly do anything wrong.
Fig. 8.38: Development of broadband access connections world wide (estimation from
Teleconnect Dresden)
VDSL supplies bit rates up to 100 Mbit/s. In many instances the modem is lo-
cated in the cellar when the in-house networks is not of high enough quality.
Wireless LAN and PLC do not possess either the necessary range or the
capacity, especially when several users in the building are online. A cable-
based network infrastructure will be necessary in the building.
FTTH is extremely expensive. With Fiber to the Building (FTTB) each indi-
vidual building has only to be connected when the data is effectively distri-
buted in the building. The capacity of a glass fiber also suffices quite easily
for buildings with many apartments.
Broadband directional wireless connections require an outdoor antenna - on
the roof or an outside wall - in the direction of the base station. Here, too,
basic networks will needed to connect the terminal devices.
Figures 8.39 and 8.40 show that the estimates of bit rates and user numbers,
Germany included, will be fulfilled.
VDSL
ADSL2+
10
ADSL
year
0.1
1998 2000 2002 2004 2006
Fig. 8.39: Available bit rate for private customers in Germany
The first ADSL connection offered a total of about 1 Mbit/s. The possible capa-
city of 6 Mbit/s was soon quickly reached before ADSL and ADSL2+ were ex-
panded to 18 Mbit/s, however, only for subscribers close to the telephone exchan-
ges. In August 2006, after the end of the soccer World Cup, VDSL has also been
available in Germany for private customers. Step-by-step the bit rate is being in-
creased with a theoretical possibility of up to 300 Mbit/s. The first small providers
already offer glass fiber connections. When widespread coverage will be achieved
is more of a political than a technical or economic question. Nevertheless, there is
no doubt that in Germany an expansion of glass fiber must also reach the final
customer. The figures on the development of broadband users in Germany are
shown in Fig. 8.40.
8.1 Data Transmission with POF 625
Even today about 40% of households have such a connection (90% with
ADSL). At the time when the first edition of this book was published it was still
below 1% and many so-called experts doubted the necessity of such a broadband
expansion. What is impressive about this chart is not only the rapid development,
but also the fact that Germany during the given period fell from a leading position
internationally to one of the last places in Europe in regard to connection density
(example in Fig. 8.41).
25%
broadband penetration DSL others
20%
15%
10%
5%
0%
Kor HK NL Dn Can Swz Tai Bel Isl Swe Nor Isr Jap Fin Sin USA Fra UK Aus Por Ger
Without question Germany will recover this lost ground. The necessity for an
efficient networking in buildings should not be viewed as an obstacle, but rather as
a possibility to simultaneously provide many customers with FTTB. However, a
technology which has moderately priced components and installation must form
the basis. The polymer fiber is an outstanding candidate for this task.
In less than a decade the capacity of wireless technologies for the networks of
buildings has increased almost a thousand-fold. Two facts are not taken into
account with this simple comparison. The increase in capacity has for the most
part resulted from a better utilization of the available frequency ranges. For
example, the license-free ISM band from 2,400 MHz to 2,483 MHz is divided into
13 overlapping channels. Only three of these channels can be used at the same
time. With a bi-directional bit rate of 54 Mbit/s two of these channels are needed.
The 108 Mbit/s modems available in the meantime cannot work at all bi-directio-
nally with a full bit rate. In the second license-free band from 5,150 MHz to
5,350 MHz there is somewhat more capacity available, but the attenuation from
walls and other disturbances increases greatly. If several devices compete simul-
8.1 Data Transmission with POF 627
taneously for the available frequencies, possibly also including those of the neigh-
bors, then the attainable bit rates quickly drop.
The second part of the capacity increase results from high quality modulation
procedures (up to QAM 256). The latter require better noise ratios and can there-
fore still only be realized for the most part over short distances.
Figure 8.42 from >Sha04@ shows the relationship between reach (without walls)
and the attainable bit rate for the wireless systems in different generations. Alt-
hough the maximum capacities differentiate greatly, there is still a clear relation-
ship between capacity and reach. One has to keep in mind that the capacity will
drop strongly once again when several walls have to be penetrated, especially with
higher frequencies. Reinforced concrete walls and ceilings are almost im-
penetrable.
100
802.11a, wireless LAN
1
802.15.4, Bluetooth
0.1
0 10 20 30 40 50 60 70 80 90 100
distance [m]
Of course, there have also been real improvements in capacity with wireless
systems, e.g. by means of more efficient error correction algorithms and better
multiple-access procedures. The multiple input - multiple output (MIMO) tech-
nique is particularly effective, although each device has to have several antennas.
Dramatic increases in capacity by orders of magnitude are only possible in
wireless technology when the frequency band is expanded to several GHz - with
other services being switched off, which is illusory - or when the transmission
power is increased immensely which under certain circumstances can be un-
healthy.
Bit rates of several 100 Mbit/s, as can be transmitted by SI-POF without any
problems, are actually only possible for wireless within rooms. All cables form a
point-to-point connection. They thus guarantee the capacity independently of what
628 8.1 Data Transmission with POF
the other devices are doing at the moment. Nevertheless, in order to be able to
profit from the mobility of wireless networks with broadband applications, e.g. a
HDTV device, POF and wireless can be combined as in Fig. 8.43.
The building shown has a broadband connection, initially maybe some VDSL
lines, later a 2.5 Gbit/s glass fiber connection. A POF-based star network distri-
butes the data into the apartments per duplex fiber. There is another switch in
every apartment and from here on the data are transported on simplex fibers in
order to additionally facilitate the installation. The entire system can be set up with
Fast Ethernet with components available on the market today. A later expansion to
1,000 Mbit/s per line is conceivable. In addition, broadband wireless base stations
have been installed in a number of rooms. Since they only have to cover one room
each, they can operate with low transmission power and at high frequencies which
would reduce the disturbances in the neighboring rooms. A handover via the
central building node is possible so that full mobility is given.
broadband optical
radio cell switch apartment network
e.g.: 25 m at 650 nm
simplex fiber
building network
e.g.: 60 m at 470 nm
duplex fiber
DVB-X
receiver
broadband access:
ADSL2+, VDSL,
HFC, WiMax, FTTB..
active
node
The advantages of broadband wireless solutions and fixed POF installations can
be ideally combined in this proposal. The result for the end customer is a desirable
minimization of pollution through radio waves.
8.1 Data Transmission with POF 629
power power
supply supply to
PC
POF
patch
patch
DSL cable
cable
modem
POF-media POF-media
converter converter
Fig. 8.44: Connection of a PC to the DSL modem using POF media converters
PC-POF
plug in
POF POF card
Fig. 8.45: Connection of multiple PCs to one DSL modem with POF interface (fictively)
630 8.1 Data Transmission with POF
Further steps will then be to have special terminal devices equipped with POF
interfaces. A necessary prerequisite for this is of course a stable standard. As des-
cribed above, especially wireless base stations with POF connections, would make
sense.
In addition to the general networking of apartments and buildings there could
be a number of applications in which POF is installed in specific point-to-point
connections. One example is the transmission of uncompromised video data bet-
ween receiver and screen. Other possible POF connections could connect sensors
which could then be optically powered from outside.
As part of the Bavarian Research Foundations sponsored project Loewe Opto,
the Fraunhofer Institute for Integrated Circuits Erlangen, the SGT Weidenberg
Company and the POF-AC Nrnberg as partners have developed a system for
transmitting HDMI video data (3 u 1.6 Gbit/s) over a POF ribbon cable. About
15 m can be reached with SI-POF and about 50 m with GI-POF. Figures 8.46 and
8.47 show the experimental set-up and the ribbon cable used with a prototype
connector.
Fig. 8.46: Demonstrator for HDMI over POF (project OVAL, see [Jun06])
Chapter 5 described what diverse applications optical bus systems could have in
the near future in the field of parallel data transmission. Fiber solutions could be
an alternative to the waveguides integrated into the PC boards. The advantage of
fiber-based versions is that the materials are not subjected to high temperatures
when the PC boards are produced and assembled. Connections between different
plug-in boards present no problems since the fibers or fiber bundles can be bent
almost at will. An example of a POF-based system was shown in Section 6.4.2.
In principle, all conventional fibers can be utilized for these applications. Since
many channels parallel to high data rates are generally necessary, VCSEL arrays
are the ideal sources. Fibers with a relatively small core diameter (125 m or
250 m) are used for an optimum coupling and at the same time with small ben-
ding radii. Compared to standard multimode glass fibers the latter are still relati-
vely thick, but do allow greater tolerances. The most important difference, how-
ever, is surely the extremely easy processing.
Thanks to the short transmission lengths PMMA POF can also be used at wave-
lengths of 780 nm or 850 nm. The use of step index fibers is also possible up to a
few meters with data rates up to 10 Gbit/s per channel.
A comprehensive overview of the details of a POF interconnection solution is
given in >Witt04@ (see also >Jhn98@, >Witt98@ and >Ney02@).
980 nm VCSEL arrays each with a spacing of 250 m were used as emitters
(1 mW output power per laser with a maximum data rate of 1.5 Gbit/s), produced
at the University of Ulm. Arrays of 4 u 8 diodes were produced. The active sur-
face of the VSEL has a diameter of only 13 m.
The receivers were produced by the ETH Zurich. The InGaAs/InP photodiodes
with a diameter of 150 m of the active surface attained a switching time of
300 ps (at 50 :, diode capacity: 1.4 - 1.5 pF. The fibers used were 120 m
/125 m SI-POF from Toray (NA: 0.48). The fiber bundles were formed by
sticking the fibers into pre-drilled, thin plastic discs. The attenuation spectrum
measured is shown in Fig. 8.49.
105
attenuation [dB/km]
104
1 mm POF
2
wavelength [nm]
10
500 600 700 800 900 1000
Fig. 8.49: Attenuation spectrum of a 125 m diameter POF
Very good VCSELs are available with wavelengths of 780 nm and 850 nm.
The attenuation values here lie clearly above the minimum losses around 650 nm,
but do permit however a few meters of a transmission link. Even the use of
980 nm components is possible in the range of several decimeters which corres-
ponds to the dimensions of computer main boards. The increased attenuation in
the short-wave range of the 125 m POF does not play any role in this application
in contrast to standard 1 mm POF.
The fiber is guided and positioned in regard to the emitting and receiving com-
ponents with the pre-drilled acrylic glass plate shown in Fig. 8.50. Production
takes place with tolerances clearly below 10 m so that efficient passive coupling
is easily possible.
Important for maintaining the slight overall height are the tight bending radii of
the POF. The losses for a 360 bend at a wavelength of 650 nm are given as an
example in Fig. 8.51. If an additional loss of 1 dB is permitted in the power bud-
get, then radii less than 1 mm could be used. This would be unthinkable with glass
fibers with the same diameter and also difficult with high frequency copper lines.
Fig. 8.51: Bending loss of a 125 m POF (one 360 bend according to [Witt04])
The latest experiments show that data rates with over 10 Gbit/s even over thick
fibers can be transmitted with suitable multi-carrier procedures and bandwidth-
efficient modulation (QAM). With QAM64, for example, 12.5 Gbit/s can be trans-
mitted at a bandwidth of about 2 GHz. Using 500 m large photodiodes and red
lasers, a system was set up which only drops about 8 dB up to 2.5 GHz; something
that can easily be compensated for in a SCM system. When the data processing is
sufficiently fast, there is nothing that speaks against transporting up to 100 Gbit/s
for some 10 m over an 8-wire ribbon cable with mm POF.
634 8.2 POF in Lighting Technology
Polymer optical fibers are employed in great amounts in many areas of lighting
technology. Two variants in particular are widely used. In the first case, polymer
optical fibers are employed as pure light guides when the light source and the ob-
ject to be illuminated are spatially separated. Secondly, the POF itself is employed
as a means of lighting, which is very decorative, especially for illuminating out-
lines.
A few examples have been added here to those of the first edition. This does
not mean that lighting technology has lost any importance in the past 6 years. In
fact, there are presently a multitude of applications. The second edition of this
POF book, however, concentrates more on data transmission so that there would
not be sufficient space for a complete representation of lighting technology. On
the other hand the demands on the optical parameters of POF in lighting techno-
logy applications are by far not so immense that no great changes in fiber develop-
ment are necessary.
The use of glass fibers for guiding light has been known and has been well estab-
lished for a long time. In communications technologies glass fibers are employed
with attenuation under 1 dB/km. These consist, however, of highly pure silica
glass the use of which for lighting technology would be prohibitively expensive.
In such a case, glass fiber bundles made of reasonably priced material are em-
ployed for greater flexibility. Figure 8.53 shows the comparison of the spectral
attenuation between glass fiber bundles and PMMA fibers in the visible spectral
range.
1.0
attenuation [dB/m]
0.8
GOF
0.6
POF
0.4
0.2
wavelength [nm]
0.0
400 500 600 700 800
Fig. 8.53: Spectral attenuation of POF and GOF for lighting applications
8.2 POF in Lighting Technology 635
The glass fiber shows a clearly better performance from 600 nm and up. There
are advantages for the POF, especially in the blue and green spectral range, which
is important for color reproduction. The possible length of fiber bundles is effec-
tively increased even if less light can be launched into a POF because of the lower
temperature load.
A comprehensive overview of the use of different types of fibers is given for
example in [Mann00a] and [Mann00b]. Table 8.6 summarizes the advantages and
disadvantages of both material options.
In addition to fiber bundles, thick polymer optical fibers for guiding light can
also be used, if no tight bending is necessary. For example, fibers with a core dia-
meter of up to 12 mm are available from Asahi Chemical (Fig. 8.54). Some
examples for POF in lighting technology are shown in Figs. 8.55 and 8.56.
Fig. 8.55: Lamp housing with POF-bundle and color filter disc ([Nich00])
Fig. 8.56: POF bundles with individually capped fiber ends, e.g. for use as a starry ceiling
lights ([Nich00])
By suitably shaping the tapers a particularly large amount of light is directed
toward the lower part of the pillar through which a much more equal distri-
bution of light is achieved. With conventional halogen lamps there are lumi-
nance differences of up to 1 : 10,000.
Fig. 8.57: Timo Reulein with advertising pillar, taper for the illumination
In addition to the purely decorative elements the system can naturally also be
used to display information, e.g. map routes could be marked with points of light.
The entrance door of the POF-AC Nrnberg shows the institutes logo as a sample
of POF ends illuminated with different LEDs (Fig. 8.59). The logo of the
University of Applied Sciences has also been accentuated in this way.
Fig. 8.59: Logos of the POF-AC and the Georg Simon Ohm University of Applied
Sciences with POF illumination
The combined use of LEDs and POF in particular opens up almost unlimited
possibilities. One conceivable application is two-way traffic signs which, how-
ever, have the problem of a limited temperature range with POF.
8.2 POF in Lighting Technology 639
Light can be decoupled laterally by using different procedures. The optical clad-
ding of the POF is transparent and thin. One possibility to achieve lateral emission
is to intentionally disturb the core-cladding interface through mechanical damage
or damage by laser radiation. At the Nuremberg University of Applied Sciences
experiments on light radiation have been carried out by cutting grooves into the
sides (see Fig. 8.60 and 8.62).
An equally practical method for decoupling light is the periodic bending of the
fiber with small radii. Part of the light is then decoupled in the bends.
If many of these fibers are brought together in a plastic tube and illuminated
from one side, even better from both sides, then you get a flexible light element si-
milar to a thin fluorescent strip lamp of more than 10 m in length. Since this bund-
le consists of plastic and does not conduct current it is much safer and has a grea-
ter load-bearing capacity than a fluorescent strip lamp, for example (Fig. 8.61).
Fig. 8.62: Illumination of an acrylic glass plate with the logo of the faculty NF at the Univ.
of Appl. Sciences Nuremberg with POF having cut grooves ([Poi99a])
Figure 8.63 shows components for illuminating the gear shift of an automatic
car transmission ([Nich00]). Here POF is also used to guide light as well as for
direct illumination. LED are being increasingly used as a light source in auto-
mobiles as they are smaller, more efficient and have a longer life expectancy than
light bulbs. However, one problem when using several LEDs in a device is main-
taining both the exact same color during their service life and the temperature
range.
One well-known user of optical fibers is the Hellux Company which prefers
using plastic fibers (>Hell04@) since the color of light already changes with glass
fiber after 3-4 m, with POF, however, only after 8 m.
The average service life of POF for illumination purposes is 20 years as stated
on the product page. The sheaths are halogen-free and flame-retardant according
to Fire Protection Class II of VDE 0207 T24. An example of an application is
shown in Fig. 8.64.
Another vendor of POF illumination systems is the Stiers Company which also
sells fibers with a diameter of 0.75 mm to 3 mm and fiber bundles as sidelight
fibers. Diverse projectors and lenses for the exit of light complement their pro-
gram (examples in Fig. 8.65).
Fig. 8.66: Luminous clothing (Luminex) and side lighting POF according to [Spi05]
8.3 POF in Sensor Technology 643
HPLD PVC
PS DC-DC
810 nm
TCP / IP
Camera
C
MMF
1310 nm
RX TX
FPGA CPLD
Fig. 8.67: Principle of the remote powered camera with data transmission ([Bt06])
Such high power is as a rule not transmitted over polymer fibers. Up to about
10 mW can be coupled into the fiber when using normal LEDs. On the other hand,
modern microprocessors can work at clock frequency of some MHz with currents
below 1 mA. It would be very useful if one could manage without the DC-DC
conversion. Segmented photo receivers connected in series can be added. Even
better would be the efficient use of semiconductors with a high band gap, e.g.
GaN, as a photo element.
The following examples of POF sensors all work according to the same funda-
mental principle. A transmitter, generally a reasonably priced LED, couples light
into a fiber. The light is first coupled out and is then coupled back in again either
into the same fiber or into another POF. The quantity to be measured now alters
the amount of the returning light so that the corresponding process can be
detected.
This principle can be applied with all fibers. However, POF has the advantage
of having a large cross-sectional area so that the sensors can be produced relati-
vely easily. Normally the measurement steps proceed very slowly (in seconds).
The possible measurement speed is only limited by the bandwidths of the trans-
mitter and receiver as well as mode dispersion in the optical path. If necessary,
bandwidths into the GHz range would be possible.
transmitting fiber
receiving fiber
object to be measured
The use of such a sensor for measuring the rotational speed of a wind power
generation rotor is described, for example, in [Zub99].
If typical fibers with a 1 mm core diameter are used then the working distances
can amount to a few centimeters. Significantly greater distances into the meter
range are possible if collimators are attached to the fiber or when the reflecting
object is equipped with retro reflector foil.
A reflection sensor on the basis of a fiber bundle (19 POF: PG-U-FB750), laid
out hexagonally) is described in >Ber05@. A fiber emits light in the center and two
rings from fibers around it detect the signal. The principle of detecting solid
particles in flowing fluids is applied in this work. The advantage of this layout is
that very much of the reflecting light is detected. By comparing the signals in both
rings the effects of different backscattering behavior can be compensated for.
Reflection is also used with the sensor described by >Zub00@. In this case not
the distance of an object is to be measured, but the speed of rotation.
For measuring the wind speed in wind power plants anemometers are normally
used in which the revolutions per unit time are determined by an optical coupler.
The electrical supply cables are problematical because electric currents can be
induced by flashes of lighting. The concept proposed favors supply cables and
outgoing cables of an optical signal via POF (490 m POF). A cylinder with
reflecting segments is mounted on the wind gauge (12 mm in diameter with
12 segments). With a 0.4 mm gap between the two fibers a coupling efficiency of
about 5% is achieved. The evaluation electronics counts the pulses and can
measure wind speeds from 10 to 100 km/h.
A distance sensor using this principle is also described in >Per04@. The aim here
was to determine cracks in concrete structures and POF was integrated into the
corresponding constructional elements. An opening crack separates the fiber and
generates an additional loss (Fig. 8.71).
8.3 POF in Sensor Technology 647
Fig. 8.71: Crack detection due to the distance inside the fiber ([Per04])
pressed foam
foam in the normal state more light will be reflected
This sensor can be used, for example, for detecting accidents. Another pro-
posed application is establishing seat occupancy. Here it is not a question of
whether someone is occupying a seat, but how heavy the passenger is or whether
it may possibly be a piece of luggage. A matrix of corresponding sensors would be
put on the seat. A computer would continuously measure the contact force distri-
bution and determine the seating arrangement (Fig. 8.75).
Optic fibers react sensitively to the most diverse external influences. Surely, the
most well-known effect is the increase in attenuation when bending the fiber. An
entire series of glass fiber sensors has utilized this principle and this procedure has
been realized in practice with POF. The fundamental principle is shown in
Fig. 8.76.
F F F F
A fiber runs straight without external forces and the transmission is then at a
maximum. When external forces arise, bends are produced in the fiber which re-
duce the transmission of the fiber. This effect comes about immediately and the
speed of the sensor is only limited by mechanical inertia and the bandwidth of the
transmitter and receiver.
unstressed: deformed 40
photo current [a.u.]
35
30
25
local bend radius 20
15
r
10
unstressed 5
under pressure force [N]
position 0
0 0 5 10 15 20 25 30 35
Sizes of just a few mm3 can be achieved be using correspondingly thin fibers.
The entire sensor can be cast in order to be able to protect the fiber from environ-
mental influences. A weight scale with four such sensors was built as a demon-
stration model at the POF-AC (Fig. 8.78).
The higher the modulation frequency is, the better is the resolution of the proce-
dure. A schematic representation of the measurement principle is shown in
Fig. 8.79. Figure 8.80 shows a typical measurement signal with which length
changes of at least 10 m can be recognized.
'L
PC f S E
sin(2S f t) POF
A/D U ')
sin(2S f t + M)
920
voltage [mV]
915
207 m
910
905
231 m
900
time [s]
895
0 10 20 30 40 50 60
Fig. 8.80: Measurement example of the strain sensor
If several of these sensors are mounted on the rotor wings, the sagging can then
be measured very exactly. The advantage of POF lies primarily in the very easy
processing and the low component costs. Furthermore, a POF can be expanded
about ten-times more than a glass fiber.
Fig. 8.81: Model of the bending measurement system for wind power station wings
652 8.3 POF in Sensor Technology
An optical sensor which can determine local deflections at several points within
milliseconds is located in the bumper of a vehicle. The time characteristics of
these deflections are typical for a particular kind of accident. If the car computer
registers a collision with a person, the hood opens up a few centimeters in order to
soften the direct blow to the engine block.
8.3 POF in Sensor Technology 653
Polymers fibers in which certain zones of the cladding have intentionally been
damaged - so-called treatments - are used for measuring the deflection. These
treatments have been applied to the side of the fiber. When the fiber is bent in the
direction of the notches the coupling-out of light is diminished, when bent in the
opposite direction the light emission is increased (Fig. 8.83). Not only the degree
of the bend can be determined, but also the direction.
The general usability of single-sided treated fibers for sensor technology is des-
cribed in >Djo03@. According to the author the advantages of the principle are that
the fibers can be embedded easily in the laminate. The measurement of the ben-
ding radii is thereby independent of local stress. Figure 8.84 shows the principle
according to >Djo03@.
In order to achieve the spatial resolution, ribbon cables are used in which diffe-
rent sensitive zones are applied (Fig. 8.85). According to >Tem05@ the treatments,
approximately 100 m wide and 20 m to 30 m deep, are burned in with the aid
of a 266 nm UV laser. The Intelligent Pedestrian Protection System (IPPS) was
introduced in 2007.
654 8.3 POF in Sensor Technology
30 mm
interface
electronics 500 1440 176
(FOSE) leads sensor area turn around
(insensitive) (insensitive)
e.g. 18 cm cell
(bumper dependent)
The sensor from ACTS in >Alb05@ is intended for the same application. In the
latter, however, the increasing bending attenuation is to be measured and there is
no spatial resolution.
impact absorber
electronics
bumper
POF sensor structure foam material
Fig. 8.87: Entering of the optical wave into the surrounding medium ([Poi05a])
8.3 POF in Sensor Technology 655
The fiber core in the sensor proposed and tested by Leoni is surrounded by a
roughly structured material which is attached to only a few points on the fiber
core. Thus, there is almost always total reflection against air and the transmission
is high. Should the surrounding material be compressed by force, the adjacent sur-
face increases at the core and an increasing share of light is either absorbed or
coupled out. In this way the sensor can be used, for example, to recognize objects
caught in a car window when the window is being closed (Fig. 8.88).
Sensor
window
plane
Fig. 8.88: Evanescence field sensor as a jam protection for car windows ([Kodl03])
signal [dB]
0
8
force [N]
10
0 1 2 3 4 5 6 7 8
Fig. 8.89: Typical sensor characteristic and robot hand ([Poi05a])
656 8.3 POF in Sensor Technology
Fig. 8.93: Tuning of a fiber ring laser using a POF grating ([Liu05b])
8.3.5.1 Humidity
A real chemical fiber sensor has been presented by >Mor04@. The goal was to
measure humidity by taking advantage of the fact that certain molecules swell up
when absorbing water and thus change their refractive indices. If such materials
are used as the cladding of optical fibers then humidity can be detected through
the change in light guiding (Fig. 8.95).
n2 n2
5.0
transmission [a.U.]
4.5
4.0 air
70%
3.5 RH
3.0 on
2.5
2.0
1.5
air 15% RH time [s]
1.0
-1 0 1 2 3 4 5 6 7 8 9
Fig. 8.96: Sensor response for the measurement of humidity ([Mor04])
660 8.3 POF in Sensor Technology
In the example shown the transmission increases right after a few tenths of a
second after connecting to the humidity. Saturation was reached after a few
seconds. The typical values of the sensor are a fiber diameter of 0.5 mm and a sen-
sitive length of 5 cm. This principle can be used, for example, to monitor a per-
sons breathing. The transmission of a corresponding sensor for breathing with
20 breaths/minute and 74 breaths/minute is shown in Fig. 8.97. The signal can
surely continue to be improved by correcting the systems time response.
intensity [a.u.]
2.5
2.0
1.5
1.0
0 2 4 6 8 10 0 2 4 6 8 10
time [s] time [s]
8.3.5.2 Biosensors
The detection of biological substances is gaining in importance. Here, too, poly-
mer fibers can be widely used, especially in the field of one-time applications.
The use of microstructured POF for detecting antibodies has been demonstrated
in >Emi05@. The biosensitive layer is applied to the side walls of microstructured
POF. The test liquid fills these holes. Two examples of the MPOF used are shown
in Fig. 8.98.
The authors of the articles describe the production of MPOF from a fiber
preform with a diameter of 20 mm. The drawn POF then has an outer diameter of
300 m with 60 m thick holes. A liquid volume of only 3.4 l fills the 20 cm
long measuring pipe. The evaluation takes place by means of a fluorescence
spectrum, i.e. a 50 m glass fiber is coupled to the spectrometer.
8.3 POF in Sensor Technology 661
Fig. 8.98: MPOF ([Emi05] and [Jen06] with 300 m and 320 m core diameter, respec-
tively (holes: 60 m and 55 m diameter)
8.3.5.3 Liquids
Microstructured POF can generally be used effectively for measuring liquids. A
sensor based on so-called hollow-core MPOF (HC-MPOF) is described in
>Cox06@ (see Fig. 8.99).
The idea of the measurement is that the refractive index of the core filled with
the liquid is higher than the structure of the cladding. As a result, the transmission
ranges of the fiber change when the core hole is filled with liquid.
662 8.3 POF in Sensor Technology
8.3.5.4 Corrosion
An optical sensor for recognizing corrosion on aluminum structures, in this case
military airplanes, is shown in >McA04@. Cations are formed when aluminum
corrodes. They diffuse in the porous cladding of an optical fiber (the core diameter
is 200 m and no data is given on the core material). The cladding with PMMA as
a carrier material is doped with 8-hydroxyquinoline (8-HQ). These molecules
together with the aluminum cations form complexes which generate fluorescence
at 516 nm with UV-irradiation (360 nm to 390 nm). The principle is shown in Fig.
8.100.
Singlemode glass fibers still enable more sensor principles compared to multi-
mode polymer fibers, whereby the filter characteristics of interferometer arrange-
ments are usually utilized. Different examples of such arrangements are described
in >Coo03@. Reasonably priced and reliable optical sensors for measuring tempera-
ture, pressure, rate of flow and sound waves in subterranean systems for oil pro-
duction are being developed in a sponsored project.
8.3 POF in Sensor Technology 663
The F-P sensors produce a periodic transmission function for the reflected light.
An edge of this transmission function is used for the measurement. Multimode
fibers can be used for short distances and singlemode fibers for greater distances.
LEDs and SLEDs serve as broadband light sources.
The deformation of the small tube is used for detection in the pressure sensor
shown. The thermal expansion of the air gap is utilized when measuring the tem-
perature.
Other sensors with glass fibers use Bragg gratings (see above), make use of
bending attenuation or also work with special coatings. Measuring the temperature
can be realized through the temperature-independent frequency shifts with
Brillouin scattering.
The great advantage of glass fibers is the possibility to measure at distances of
many 10 km from the active technology as well as in the fibers excellent stability
resulting in great exactness and resolution.
The advantages of polymer fibers are primarily the extremely easy handling
and in many cases the large cross-section. We wish to point out once again an
article with a comprehensive overview including a look back at the first develop-
ments of POF sensors (>Bar00@). Here a POF sensor developed in 1997 by
Niewisch (>Nie97@) is mentioned in which liquid nitrogen works at 77 K. Different
sensors with fluorescing fibers are also mentioned.
9. Optical Measuring Methods
The measuring techniques for polymer fibers and other thick-core fibers differ in
some essential parameters from those of conventional glass fibers. The main diffe-
rence lies in the dominant mode-dependent effects. The first edition of this book
provided a general overview of POF measurement techniques with just a few
specific results.
The POF-AC Nrnberg has now existed for five years and has adapted and
reworked many methods of measurement. A number of measurement arrays have
been set up and tried out. The following chapter has been supplemented by these
new methods and results while also including the general presentation. The mea-
surement results of fiber bandwidths will not be further dealt with here as compre-
hensive examples were discussed in Chapter 2. This is also valid for the measure-
ment of bending losses.
9.1 Overview
There are three main areas in which optical measuring methods are applied:
for manufacturing, control and product specification,
during and after installation,
for maintenance and for searching faults.
The investigations cover individual components as well as overall transmission
systems. This chapter covers optical measuring methods for polymer fibers rele-
vant to the user; the main aspects are the measurement of fiber attenuation and dis-
persion as well as emission characteristics and connector loss, as illustrated sche-
matically in Fig. 9.1.
The different measured quantities can be determined by different procedures,
for example, for determining the attenuation in transmitted light and back
reflections. All optical characteristic quantities can be measured in dependence of
the launch conditions and the wavelength (spectrally resolved). Furthermore, the
measurements can be combined with variations of the external conditions such as
mechanical or climatic loads.
The basic parameters of the components of a transmission system will be
measured by the manufacturer and should be included in the data sheets and in
application notes. The system designer can use these available data for his infor-
mation and the correct choice of materials and components. The end user, last but
666 9.2 Measuring Power
not least, will perform field tests in order to evaluate the system properties and to
localize occurring faults. The requirement is, that the components have been
produced and characterized in accordance to accepted standards.
near field
far field
numerical detector
light source aperture connector
POF
connector attenuation
loss dispersion
There are only a few specific standard for polymer fibers available. Measure-
ments methods developed and specified for glass fibers will be used in most cases.
Furthermore, there are a limited number of commercial available measurement
devices, optimized for POF. Thats why, measurement devices and methods must
be adapted to the POFs requirements.
Information in data sheets are often minimal. The rating and the reproducibility
of the listed values are difficult, because the measurement conditions are not or
not sufficient described. Under some circumstances however, these values strong-
ly depends on the measurement conditions. For that reason, the following section
shall give the user a tool to interpret and the values of a datasheet and to be able to
repeat the measurements if required.
Where the optical power is calculated on the basis of W, the unit description
for the logarithmic value is dB. Figure 9.2 shows a diagrammatic illustration of
the relationship between mW and dBm.
Positive values for dBm refer to power which is greater than the reference value
of 1 mW whereas negative values refer to power smaller than 1 mW. The diffe-
rence of two power levels is stated in dB, for example, if the level is reduced by
3 dB, the linear power is reduced by 50 %. The advantage of the logarithmic
method of notation can be seen in the fact that the difference in power level for 2
points in a link system represents the attenuation of the respective link in dB.
The typical power levels in POF transmission systems lie between -2 dBm and
-26 dBm, or respectively, 0.63 mW and 0.0025 mW. The power meters used in
this range consist of semiconductor detectors. Silicon photodiodes are the most
sensitive for the range of wavelengths having the smallest attenuation of the poly-
mer optical fiber (approx. 500 nm - 700 nm). Since the responsivity of the detector
depends on wavelength, the power measurement of these detectors is only valid
for the wavelength indicated on the equipment. When measuring different wave-
lengths, a conversion factor must be applied.
In practical applications the optical power measurement is the most important
measurement task. It is used to check whether transmitters maintain the minimum
specifications and whether the necessary minimum receiving power is achieved on
installed transmission links.
Various models of power-measuring devices are available from different manu-
facturers. Normally, they measure the light exiting from fibers with a large-surface
Si photodiode. The display is possible in absolute values in mW or dB. A calibra-
tion function can be used to show the power difference to a reference value in dB.
The various levels of sensitivity for standard wavelengths, mostly, 650 nm,
780 nm and 850 nm, can be programmed. Better devices can also measure the
power of slowly modulated light in order to eliminate the effect of ambient illumi-
nation. In order to be able to use different connectors a number of adapters are as a
rule available which can either be plugged in or screwed on.
All customary devices have the disadvantage of being mode dependent to a
certain degree since flat photodiodes are used. The devices can usually measure to
exactly 0.1 dB for identical modal distributions and 0.5 dB for varying field
668 9.2 Measuring Power
distributions. The customary measuring range starts at -50 dB to -60 dB and ends
at +3 dBm.
An overview of different hand-held devices is shown in Fig. 9.3. The list is by
no means complete and is quickly growing, more proof of the increasing impor-
tance of POF technology.
Fig. 9.3: Hand held power meters for POF (left to right)
above: FO-Systems/Leoni, Tempo, Photom, Senko, OWL
bottom: Scientech, Rifocs, Advanced Fiber Solutions, Fotec, Ratioplast
In addition, there is an entire series of special devices used for continuity tests
and with numerous channels (examples in Fig. 9.4). The manufacturers also sell
stabilized LED and/or laser diode transmitters for almost all power-measuring
devices.
A large number of special devices have been developed for the measurement of
attenuation in cables for vehicle networks. Besides accuracy what matters here are
fully automatic operation and short measuring times. Thanks to the short cables
sensitivity plays a subordinate role. The OptiTest 10 measuring device from
Schleuniger is shown in Fig. 9.6 on the left. The attenuation measuring station of
the IDC 9600 MS processing system from Komax can be seen on the right
(www.komax.ch).
670 9.3 Dependence on the Launch Conditions
A golden fiber, i.e. a reference fiber produced and measured with great pre-
cision, is used in these systems for calibration. Together with active connector
positioning of the fiber to be measured an accuracy of 0.02 dB can be attained
(data from the website www.schleuniger.de).
The measuring devices in the section above do not as a rule use specifically opti-
mized launch conditions. The transmitter is simply positioned directly at the fiber
input. We will subsequently show that this procedure does not suffice for exact
measurements.
The transmission characteristics of an optical fiber are determined by attenu-
ation and dispersion. The measured value for attenuation and/or dispersion de-
pends on the light introduced into the polymer optical fiber; for this reason it is
necessary to create reproducible launching conditions. It means that the distribu-
tion of the optical power onto the modes excited in the optical fiber must be
known.
If the complete core area and numerical aperture are illuminated uniformly (full
filled launching), all modes will carry initially the same power (UMD = Uniform
Mode Distribution, blue curves in Figs. 9.7 and 9.8). During the further passing of
the light through the fiber, the rays propagating at a larger angle to the axis of the
fiber experience greater attenuation than the rays with the lower angle since they
have to travel a longer path and are reflected more often on the interface between
the core and the cladding. For example: in a fiber with AN = 0.5, ncore = 1.497 and a
core radius of 0.5 mm, the core ray, which is still just about reflected completely,
runs at an angle of 19.5; over a length of 1 m the ray will be reflected approxi-
mately 350 times at the interface between the core and the cladding. Due to
inhomogeneities at the core/cladding interface as well as in the core material, it is
possible that power may propagate in two different directions (mode coupling). In
addition mode conversion will cause power to be exchanged between the different
propagation directions at bends within the optical fiber. These effects lead to a
change in the mode distribution excited at the beginning of the fiber. After a cer-
9.3 Dependence on the Launch Conditions 671
tain distance, a steady state distribution is achieved and from thereon mode distri-
bution remains constant (EMD = Equilibrium Mode Distribution, red curves in
Figs. 9.7 and 9.8), provided no such faults occur that would again lead to mode
coupling effects. If the excitation is with a small numerical aperture (green curves
in Figs. 9.7 and 9.8) there will also be EMD after a certain length due to the fact
that higher order modes are created.
optical
power UMD:
all mode guide the
same optical power
EMD:
equilibrium mode
distribution
exitation with small
numerical aperture
Fig. 9.7: Mode distribution for different types of excitation (schematic representation)
600
attenuation [dB/km]
500
200
100
under filled launch
length [m]
0
0 50 100 150 200
Fig. 9.8: Length-dependent attenuation relative to the length of the fiber, for different mode
distributions (schematic representation, no experimental results)
672 9.3 Dependence on the Launch Conditions
Figure 9.9 shows the difference in attenuation between the rays launched paral-
lel to the fiber axis and those at the maximum possible angle of launching (core
radius a = 0.5 mm, ncore = 1.497). Assuming equal distribution of the mode to the
propagation angles, and considering modes at launching angles from 0 to 20
(half of the possible modes) one obtains an average attenuation increase of up to
2.5 dB/km, whilst with the inclusion of the higher order modes a change of up to
6 dB/km occurs - in both cases relative to the basic attenuation of 100 dB/km.
Only that part of the attenuation attributable to the longer path (volume attenu-
ation) has been taken into account here. Further effects which contribute to mode
dependent attenuation (mode conversion and mode coupling) are described later.
This simple example demonstrates the difficult situation that exists when trying to
carry out measurements that are independent of launching conditions.
12
excess loss, related to a basic
attenuation of 100 dB/km [dB/km]
10
2
launch angle []
0
0 5 10 15 20 25 30 35 40
Fig. 9.9: Relative change of attenuation in relation to the launching angle (only caused by
the difference in the propagation path)
For real polymer fibers as well as for PCS, however, there is also the very much
stronger effect of high losses for rays with a large propagation angle caused by the
attenuation of the cladding material. This cladding attenuation with PMMA POF
lies around some 10,000 dB/km. Measurement results for PCS were shown in
Chapter 2.
Figure 9.10 shows the number of modes in an SI-POF with the above
mentioned parameters (650 nm wavelength) in relation to the maximum launching
angle considered.
In order to achieve reproducible results with measurements for attenuation and
dispersion, the modes should be in equilibrium mode distribution. This can be
achieved by using a launching fiber, although this is not very practical for polymer
fibers since the required length of 30 m to 60 m would lead to high additional
attenuation (6 dB to 12 dB for 200 dB/km attenuation). The dynamics of the
measurement setup is reduced by this value. For comparison: for glass fibers, the
length of the launching fiber is in the range of 1 km to 2 km producing an
additional loss of 2 dB to 4 dB (for 2 dB/km of attenuation).
9.3 Dependence on the Launch Conditions 673
4.5
number of modes [u 106]
4.0
3.5
3.0
2.5
2.0 AN = 0.50
1.5
1.0
0.5 launch angle []
0.0
0 5 10 15 20 25 30 35 40
0 5 10 15 20 25
propagation angle in the fiber []
Fig. 9.11: Optical launching arrangement with independently adjustable numerical aperture
and spot diameter
Fig. 9.12: Mode mixer in accordance with the Japanese Industrial Standard JIS 6863
Apart from the mode mixer described above, other set-ups are also in use, for
example the roll mode mixer ([Fus96]). However, these have not been standar-
dized. For DSI fibers such a mode mixer does not achieve the desired result, since
these would require a much smaller bending radius. Analysis shows ([Pfl99]) that
bending radii of less than 15 mm would have to be used, which would lead to a
high attenuation and is therefore not practicable. Consequently, this type of mode
mixer should not be used for DSI fibers, but instead one should only measure
attenuation up to a certain minimum length. This length must be established
individually for each type of fiber using, for example, the cut-back method (see
below).
The methods for measuring the different optic characteristic quantities will be des-
cribed in the following sections together with some examples of measurements.
Here the effects of various launch conditions will be particularly emphasized just
as with the measurement of the bandwidth (Chap. 2). The different experimental
results have diverse practical significance, e.g. for coupling attenuations, for deter-
mining system ranges and capacities, and for the qualitative characterization of
fibers and active components. The following will be dealt with in detail:
Near field distribution
Far field distribution
Inverse far field
Index profile
Optical attenuation
Optical time domain reflectometry
Dispersion
9.4 Measurement of the Optical Parameters 675
The near field describes the power distribution of the light in the output face of the
optical fiber. It can either be measured via an enlarged image or scanned with a
suitable optical fiber. Figure 9.13 illustrates a suitable installation, in which it was
possible to achieve a 55 dB dynamic ([Gie00]). In this installation, a singlemode
fiber (core diameter 9 m), which is driven by a step motor, is guided radially
along the fiber surface. The signal is received by a highly sensitive detector. The
test installation has a dynamic range of approximately 60 dB. Figures 9.14 and
9.15 show the near field patterns of a standard NA-POF and a multicore POF
recorded with this arrangement.
manual
sampling fiber positioning
SM-GOF; 9 m
step drive
POF
-10
-20
-30
-40
-50
Giehmann T-Nova 2001 fiber sensor position [m]
-60
0 250 500 750 1000 1250 1500
Fig. 9.14: Near field pattern of a standard NA-POF, illuminated with an LED (O = 560 nm)
676 9.4 Measurement of the Optical Parameters
0.08
rel. optical power (linear)
0.07
0.06
0.05
0.04
0.03
0.02
0.01
Giehmann fiber sensor position [mm]
0.00
0.00 0.25 0.50 0.75 1.00 1.25 1.50
The scanning path in Fig. 9.15 runs across the center point of the MC-POF and
captures seven individual fibers (see Fig. 2.47). However, it is not aligned to the
center of the individual fibers, which is why some fibers have been only partially
covered.
Commercial devices for measuring the near field distribution have become
available in the meantime, for POF and PCS as well. The LEPAS measuring sys-
tem from the Japanese manufacture Hamamatsu is used at the POF-AC Nrnberg
(see >Bach02@).
Not only is the detection of the entire cross-section of the fiber is important for
correct near field measurements, but also that all propagation angles occurring are
taken into consideration. Image-forming systems with a small NA have been used
in many published measurements of fibers. The consequences of such an error are
shown in Fig. 9.16 for the measurement of a GI-POF.
T
-Tmax T +Tmax
Fig. 9.16: Measurement of the near field of a GI-POF using a to small NA system
9.4 Measurement of the Optical Parameters 677
Modes with different angles occur in the center of GI fibers, almost only par-
axial rays are at the edge. If the measuring system only picks up the small angles,
then a much too small intensity is established in the center. The near field distri-
bution apparently corresponds to that of a SI fiber. The opposite effect can come
about if the optics cannot detect the necessary angular range over the entire cross-
section. What can happen here is that all modes are measured in the center of the
fiber and only part of them on the fiber edges. The near field would then look like
that of a GI fiber.
In order to minimize such effects, a special combination of lenses has been
produced by the Sill-Optik Company for the LEPAS system which permits the
correct measurement of 1 mm POF. The optics and the image-forming principle
are shown in Fig. 9.17.
FOP +
objective lens relay lens CCD-Chip
Fig. 9.17: Near field optics and operation principle
The optics display an image of the output end faces on a CCD chip, with an
enlargement of about 5. The fiber optic plate (FOP) prevents interference patterns
through the chip cover. The technical data of the system are:
Acceptance angle 30
Resolution > 2 m (numerically)
Magnification approx. 5.5
Wavelength range 400 nm - 1100 nm
Working distance 13.8 mm
In order to test whether or not the optics corresponded to the requirements, a
thin (0.2 mm) laser beam (divergence < 0.1) was positioned in the measuring
level of the system, on the sides as well as at different angles. Figure 9.18 shows
the results for 7 different positions.
678 9.4 Measurement of the Optical Parameters
0 0 0 0
-30 +30 -30 +30 -30 +30 -30 +30
1.0
0.8
0.6
0.4
x= x= x= x=
0.2
-0.9 mm -0.3 mm +0.3 mm +0.9 mm
0,0
0 0 0
-30 +30 -30 +30 -30 +30
x= x= x=
-0.6 mm 0.0 mm +0.6 mm
At a distance of 0.9 mm to the system axis only a very slight part of the angle
was detected. Within a range of 0.3 mm practically the complete angular range is
measured. The system can produce satisfactory results for fibers with a diameter
of up to 1 mm and a NA up to 0.50.
Fig. 9.19: Examples for near field measurements (approx. 1 m and 30 m MC-POF)
Two examples of near field measurements are shown in Fig. 9.19. Here a 37-
core MC -POF was measured after a short stretch (on the left) and after a long
link. The attenuation can be determined for each individual fiber. In particular, the
higher attenuation at the fiber edges could clearly be seen in this measurement. If
this measuring system is combined with a corresponding pattern recognition for
identifying the cores, then the individual fiber attenuation can be determined
automatically.
9.4 Measurement of the Optical Parameters 679
The far field is defined through the field distribution at the distance D 2a from
the output face of the fiber (Fig. 9.20).
intensity
B
4max
a
Fig. 9.20: Idealized far field of the step index profile fiber
The angle 4max is calculated from the relationship tan 4max = B/D. The numeri-
cal aperture NA is calculated from the far field angle 4max using the following for-
mula:
All modes carried in the core are excited in this case (full mode launch), i.e.
meridional rays as well as skew rays.
The far field represents the intensity distribution on a spherical surface, in the
center of which the light output face is located. The measurement is carried out
with a photo-detector, selecting the respective angle. The angle resolution depends
on the angle range covered by the detector area. Figure 9.21 shows a possible test
installation for capturing the far field in one plane.
light source
with 4
launching
optics detector
fiber
computer
CCD camera
fiber bundle
fiber array
with 321
fibers rotatable
turntable
(by 180)
polymer fiber
Fig. 9.22: Fast far field measuring setup with the Emitor
Four different 1 mm POFs were measured under the same conditions for a test
(the parameters are in Table 9.1).
9.4 Measurement of the Optical Parameters 681
norm. intensity
1.0
PF-U-CD1000
0.9
AC-1000
0.8
NC-1000
0.7
MH-4000
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-40 -30 -20 -10 0 10 20 30 40
far field angle []
Fig. 9.23: Far field patterns of different fibers, light source LED 650 nm, 50 m fiber,
according to [Hen99]
The 10% far field width of the SI-POF PFU-CD1000, with a numerical aperture
of 0.42, is conspicuously wide. This fiber features an NA of 0.46. [Bun99b] and
[Pei00a] state a coupling length of 36 m for the PFU-CD1000. When using the
length of 50 m with this fiber, the state of equilibrium mode distribution has
already been reached. Also, compared to the other fibers, the intensity distribution
682 9.4 Measurement of the Optical Parameters
rises steeply from zero, whilst the other measured curves show a bell shaped
increase. This indicates that the proportion of power of the leaky waves is small
and hence makes only a negligible contribution to the widening of the far field.
Figure 9.24 shows the 10% far field width in relation to the bending radius for
various numbers of windings. Launching was carried out with AN = 0.5. This
investigation is the basis for setting up a mode mixer, as described, for example, in
the Japanese Industrial Standard JIS 6863 and in the IEC 60794-1-1 (Annex A).
As expected, the numerical aperture reduces most with the bending radius at 10
windings. At a bending radius of 21 mm the numerical aperture is 0.42.
Figure 2.147 shows the excess loss in relation to the inverse bending radius for
different numbers of windings. For the conditions of the above mentioned stan-
dards (bending radius = 21 mm, 10 turns) the excess loss is approximately 2 dB.
Figure 9.25 shows the change in the far field of the DSI fiber MH 4000 for
different numbers of windings, starting at 0.
With two windings, the far field is significantly narrower than without any win-
ding. At this bending radius, the higher order modes are emitted to a larger degree
and hence do not contribute to the far field. This effect occurs with a bending
radius smaller than 15 mm [Hen99]. Due to the length of the fiber, any mode
coupling must be largely precluded. For two windings, the numerical aperture is
approximately 0.30 and for no winding it is 0.44. This is due to the double-step
index profile of the fiber: without the winding, the light is also conducted within
the inner cladding; the bends cause the angle of total reflection between the inner
and outer cladding to be exceeded for a large proportion of the rays so that the
light reaches the outer cladding and is lost through radiation. Now the propagation
characteristics are determined by the refractive index difference between core and
the inner cladding, corresponding to a numerical aperture of approximately 0.30.
The equilibrium mode distribution must still be formed in the core.
9.4 Measurement of the Optical Parameters 683
1.0
norm. intensity 0 turns
0.9
2 turns
0.8
4 turns
0.7
6 turns
0.6
8 turns
0.5
10 turns
0.4
0.3
0.2
0.1
0.0
-40 -30 -20 -10 0 10 20 30 40
far field angle []
Fig. 9.25: Far field patterns of the DSI fiber MH 4000 for different numbers of windings
(bending radius r = 12 mm, light source LED at 650 nm, length of fiber 4 m)
[Hen99] investigates the use of a mode mixer for DSI fibers. As a condition for
equilibrium mode distribution for the fiber MH4000, a bending radius of 9 mm
results, leading to an additional loss of 18 dB for 10 windings. This high value is
not acceptable for attenuation and dispersion measurements so that in DSI fibers it
is only possible to achieve equilibrium mode distribution with the help of mode
mixers at the expense of a high attenuation.
There are also in the meantime a number of commercial devices for the two-
dimensional measurement of far field distributions. The POF-AC uses the LEPAS
system here as well as for near field measurements. The corresponding optics and
the ray trajectory are shown in Fig. 9.26.
FOP + CCD-Chip
f-T lens field lens relay lens
Fig. 9.26: Far field optics of the LEPAS system with ray trajectories
684 9.4 Measurement of the Optical Parameters
The f- lens converts the different angles at the location of the intermediate
image which is then projected onto the CCD chip with the following microscope
array, attaining the following parameters:
Acceptance angle 45
Angle resolution 0.18 (numerically)
Wavelength range 400 nm - 1100 nm
Working distance 2.8 mm
An example in Fig. 9.27 shows the far fields of a 1 mm standard POF with
coupled collimated light each with an angle of 10 after 10 m and 100 m respec-
tively. The ring structure can clearly be seen which after 100 m has for the most
part been resolved through mode mixing.
Fig. 9.27: Far field of a SI-POF for excitation with 10 after 10 m and 100 m
With the inverse far field method described in [Gies98] one can obtain even more
detailed information about the light propagation than with the method described
above. In this method one not only measures the radiated light selectively but also
launches it at selected angles into the fiber (Fig. 9.28).
fiber Tout
Tin
The advantage of this process lies in the fact that the launching spot with a
small diameter (some tens of m) and a small numerical aperture (AN | 0.02) can
be placed at the desired spot of the fiber's end face, thereby making it possible to
excite certain specific mode groups. With this type of excitation only meridional
rays are excited, no skew rays.
Figure 9.29 shows the far field distribution of a standard NA-POF after 10 m
and 50 m length of fiber at launch angles of 15, 20 and 25. The steep rise in the
far field curve indicates that no leaky modes were excited. Whilst at an launch
angle of 15, the power is completely distributed to the smaller angles after 50 m;
at 20 and 25 launch angle, the domination of these mode groups can still be
clearly recognized.
100 20 100
15
25
80 80 20
25
60 60
40 40
20 20 15
0 0
-40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40
far field angle [] far field angle []
Fig. 9.29: Far field patterns of a standard fiber with different launch angles ([Kle98])
For the DSI fiber (Fig. 9.30) the situation is different. In this fiber, all of the
power has gone over into the core from the inner cladding after 50 m so that at a
15 launch angle the far field width has nearly reached the numerical aperture
corresponding to the refractive index difference between core and inner cladding.
In Fig. 9.31, the launching angle is changed by 1 between -30 and +30 each
time and the overall intensity of the emerging light is measured with a large area
detector. In the case of standard NA fibers (diagram on the left), an almost rectan-
gular far field profile for fiber lengths of 1 m and 10 m changes to a parabolic
form for greater lengths; correspondingly, the far field angle reduces (for the
shorter lengths this is approximately 26) (AN = 0.44). This is caused by the stron-
ger attenuation of the higher order modes. After 50 m, equilibrium mode distri-
bution has been reached.
686 9.4 Measurement of the Optical Parameters
40 40
15
20 20
0 0
-40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40
far field angle [] far field angle []
Fig. 9.30: Far field patterns of a DSI fiber under different launch angles [Kle98]
The diagram on the right in Fig. 9.31 shows the result for a DSI fiber. With
increasing length, the rays propagating under larger angles are heavily attenuated
so that at 50 m and at 90 m the far field angle is 18 (AN = 0.32).
The rays with a large angle are reflected at the interface between the inner and
outer claddings. Since the inner cladding has a higher attenuation compared to the
core, the rays propagating in this have been attenuated so strongly after 10 m that
they can no longer be traced in the far field. The manufacturers call the DSI fibers
low-NA fibers; the numerical aperture is stated as 0.30, without making refe-
rence to respective lengths.
9.4 Measurement of the Optical Parameters 687
A very important measurement for polymer fibers is that of the index profile. The
index curve has to be checked regularly in the core, especially with GI-POF which
is produced by diffusion. For the measurement the end face is either directly
scanned or the fiber is radiated through laterally. An overview of the different
methods for measuring the refractive index profile of optical fibers with lateral
radiation is given in >Bun04a@.
One method particularly well-suited for POF is based on the scanning of the
fiber end face with a beam as closely collimated as possible (determine the resolu-
tion). The NA of the light has to be adapted to the NA of the fiber investigated.
Two detectors measure the light passing through the fiber as well as a reference
beam. The refractive index dependant on location can then be calculated from the
difference. This method only produces good results when the fiber end face has
been prepared exactly smooth and flat. The next two figures, 9.32 and 9.33, show
the measurement results on a multistep index POF and a Semi-GI-PCS. This
method of measurement results in only relative differences in the refractive index
unless there has been a calibration beforehand on exact samples which are known.
1.0
nrel.
0.8
0.6 ESKA-MIU
sample 2003
0.4 Mitsubishi
0.2
x [m]
0.0
-400 -300 -200 -100 0 100 200 300 400
Fig. 9.32: Refractive index profile of a MSI-POF
1.0
nrel.
0.8
0.6
0.4
0.2
x [m]
0.0
-120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120
9.4.5 Attenuation
Optical attenuation describes the loss of light between the input and output of an
optical component, i.e. also of a fiber. In principle, only two power measurements
are necessary so that the explanations in Section 9.2 should suffice. The proble-
matical nature of this aspect is actually very much more complicated.
The measuring processes for attenuation provide as a result both the overall
attenuation which is of considerable interest to the system designer as well as the
individual contributions due to absorption and Rayleigh scattering, which provide
the manufacturer with important information for the purpose of process optimi-
zation. Characterization can be carried out for one or more discrete wavelengths
(for example, for the link wavelength) or continually for a larger spectral range.
10 P
Insertion loss = log L1 D fiber D connector
L PL 2
With this method it is possible to measure the attenuation of the inserted cable
including connectors. In order to be able to determine the fiber attenuation, the
excess loss of the connector must be known.
power [dB]
connector loss
insertion loss
connector loss
L
length [km]
A similar process is followed in the substitution method: at first the power PL2
is determined. Subsequently, the fiber is replaced by a short reference length and
the power PL1 is determined. For this purpose the arrangement and characteristics
of test and reference fiber must be identical. In contrast to the insertion method
however, the number of connectors remains constant, which is why it is possible
to determine the fiber losses without connectors. However, it is a requirement that
the attenuation between the used connectors is constant. This is only possible
within in a certain tolerance, since, depending on the type of connector, the values
can diverge by up to several dB. This can therefore lead to significant errors in
kilometric fiber attenuation measurement when measuring short fiber lengths with
this measuring method.
reference fiber
pigtail L1 coupling
PL1
R
T
L2 receiver
transmitter
10 P
Attenuation is calculated as follows: D log L1
L 2 L1 PL 2
690 9.4 Measurement of the Optical Parameters
Level [dB]
L1 PL1
loss of the fiber
reference under test
fiber
PL2
L2 - L1
L2 length [km]
T R
receiver
transmitter L1
PL1
600 P/P0
attenuation [dB/km] LED spectrum at LED spectrum at
the fiber input the fiber output 1.0
500
400 0.8
300 0.6
200 0.4
100 0.2
0 0.0
550 600 650 700
wavelength [nm]
Fig. 9.39: Filter effect of the attenuation spectrum of a POF (ESKA EH 4001, 50 m)
The area under the bell shaped curve represents the total optical transmission
power of the LED. The LED spectrum P(O) is required. This needs to be norma-
lized so that the following applies:
f
PO dO 1
0
692 9.4 Measurement of the Optical Parameters
In addition, one needs to know the attenuation spectrum D(O) of the POF. This
is described, for example, in [Wei98]. However, it is important that the spectrum
used is identical to that of the POF or diverges from this only by a constant attenu-
ation coefficient. In order to determine the correction factor for the attenuation
Dmeasured for a required wavelength O0 (for e.g. 650 nm) and a certain length of
fiber L, one proceeds by forming the following integral:
f D O L
Peff PO 10 10 dO
0
The result obtained is the attenuation of the total spectrum of the LED in dB
(DLED = 10 log (Peff)). The correction factor is then calculated as follows: (see
Section 7.3.1 also):
DLED
correction faktor K F
DO 0
entrance
slit
concave
mirror
grating
exit concave
slit mirror
focal length f
The reflection grating consists of a glass substrate in which parallel grooves are
ruled, either mechanically with a ruling engine or holographically by superimpo-
sing two coherent laser beams. Subsequently, the grating is coated with a highly
reflective medium. There are always several diffraction orders so that, with the
same setting of the grating, wavelengths of, for example, 400 nm and 800 nm
appear at the output slit. In this case, the interfering light must be eliminated with
an edge filter. By giving the grooves a specific shape and depth, one achieves that
a maximum of light is reflected into the first diffraction order. The grating is opti-
mized for a certain wavelength range and is highly efficient in that range. The
wavelength at which the grating is at its most efficient is called blaze wavelength
(Fig. 9.41).
90
grating efficieny [%]
80
blazed grating
70
60
50
40 holographic grating
30
20
10
0
400 600 800 1000 1200 1400 1600
wavelength [nm]
cross area
mode converter conversion
monochromator
Fig. 9.42: Experimental setup for measuring attenuation
diameter: 1 mm
num. aperture: 0.50
60
18
slit
By guiding the light through a lens, it is possible to increase the light spot;
however, the numerical aperture is decreased. This serves to adjust the numerical
aperture. If, for example, the cross sectional area of the fiber is reproduced in
3-fold magnification, the NA is reduced from 0.60 to one third, i.e. 0.20.
Figure 9.45 shows the estimated losses for different mismatches between area
shapes and numerical apertures. The losses are calculated as follows:
area fiber
D Fl 10 log
overlappin g area monochroma tor
2
AN
D NA 10 log fiber
AN
monochroma tor
12 loss [dB]
total
10 mismatch
8
area
mismatch
6
4
mismatch of the
numerical aperture
2
magnification
0
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Fig. 9.45: Mismatch dependent on magnification (slit width 0.5 mm, fiber
diameter 1 mm, NA of fiber 0.50, NA of monochromator 0.16)
696 9.4 Measurement of the Optical Parameters
When coupling the fiber directly to the monochromator slit, a minimum loss of
7 dB is achieved at the stated values of the set-up. Smaller losses result when
using a cross section transformer, which transforms a circular area into a rec-
tangle. It consists of a silica glass fiber bundle with thin glass fibers of approxima-
tely 1 m length, the ends of which are arranged in circular shape at one end and in
rectangular shape at the opposite end (Fig. 9.46).
Due to the numerical aperture of the bundle of 0.22 and a diameter of 3 mm, it
is possible to achieve an optimum adaptation of the POF by using a lens. The end
with the rectangular arrangement of fibers coincides with the width (0.5 mm) of
the monochromator slit. The remaining loss due to the mismatch between the
numerical aperture between the quartz glass fiber bundle and the monochromator
is 2.8 dB.
monochromator
fiber under test detector
Fig. 9.47: Schematic setup for measuring attenuation in polymer optical fibers
ray trajectories
Fig. 9.48: Adaptation of the numerical aperture and the cross section of the ray
The integrating sphere (Fig. 9.49) is a hollow sphere that is often coated with
barium sulphate (BaSO4) on the inside. The coating causes multiple and diffuse
reflections of the incoming light until it is evenly distributed over the sphere
surface. After this integration, effects such as the angle of incidence, polarization,
modes or shadow formation are eliminated.
fiber connector
PMMA - fiber
shutters
detector connection
Hence all of the light entering the sphere from the fiber is captured by the
detector. The two openings for the polymer optical fiber and the detector are
orthogonal to each other. On the inside of the sphere there are shutters to prevent
light from striking the photomultiplier directly from the fiber.
The configuration described last offers the advantages that both insertion and
substitution methods as well as the cut-back method can be used and when using
the integrating sphere as a detector system, all of the radiated light is detected.
This measuring system offers a dynamic range of 30 dB to 35 dB for a wave-
length range of 480 nm to 700 nm with a slit width of 0.25 mm and using a photo-
multiplier as the detector.
substitution method
10
0
400 450 500 550 600 650 700
wavelength [nm]
Fig. 9.50: Standard deviations obtained with the cut-back and substitution methods
([Pei00a])
When using the substitution method, particular care must be taken when pre-
paring the fiber's end faces to ensure good surface quality. The selection of the
connector also has an effect on the reproducibility of the measurement. We have
used FSMA connectors for the tests carried out here.
Figure 9.51 shows the result of an inter-laboratory test for measuring the
attenuation in a standard NA-POF ([Kell98], [Krau98]). The measurements were
carried out using the substitution method over a wide range of wavelengths, with
LED or laser sources and different set up configurations. The attenuation values
compared are in the range of 650 nm wavelength and for fibers of 20 m, 50 m and
9.4 Measurement of the Optical Parameters 699
100 m length. For the assumed attenuation of a normal fiber of 156 dB/km
(ATM Forum specification, [ATM96a], [ATM96b], [ATM99]), 0.5 dB were
added for launching changes for each length. The effects on the kilometric attenu-
ation are very severe for the shorter lengths. One can easily recognize the large
spread of measured values at 20 m, but even for 50 m or 100 m the spread of
measured values is too large so that a reliable statement about the actual
attenuation cannot be made. However, it is worth pointing out that the spread for a
length of, for example, 20 m is approximately 47 dB/km, which - when related to
the 20 m length - means less than 1 dB.
It is an absolute necessity in all attenuation measurement tests to give detailed
information about the test setup and procedure, in particular also about the laun-
ching conditions of the fiber.
attenuation [dB/km]
190
180
156 dB/km
+0.5 dB for launch NA
170
160
120
20 50 100
length [m]
Fig. 9.51: Attenuation in relation to the length of fiber; the measurements were carried out
as part of an interlaboratory test with different measuring set ups ([Kell98])
The attenuation spectra presented below were recorded with the test set up des-
cribed in Fig. 9.47. Figure 9.52 shows the effect of length on attenuation in short
lengths of fiber, measured with the cut-back method. Light launch was carried out
near to equilibrium mode distribution. After a length of approximately 10 m
attenuation remains constant, which means that after approximately 10 m equili-
brium mode distribution has been achieved.
700 9.4 Measurement of the Optical Parameters
180
attenuation [dB/km]
170
160
150 ESKA 4001
140 AN = 0.47
130
length:
120 3m
110 5m
7m
100 10 m wavelength [nm]
90 20 m
500 520 540 560 580 600
Fig. 9.52: Attenuation of a standard NA-POF (EH-4001) in relation to length for short
lengths of fibers, measured with the cut-back method ([Pei00a])
This is confirmed when testing fibers with a greater length (Fig. 9.53) where
attenuation no longer depends on the length.
800
attenuation [dB/km]
700 sample length:
20 m 30 m 40 m
600
50 m 60 m
500
400
300
200
100
wavelength [nm]
0
400 450 500 550 600 650 700
Figure 9.54 shows the spectral attenuation curves for a double step index pro-
file fiber (ESKA MH 4001). The fiber was launched with a numerical aperture of
approximately 0.50 (Fig. 9.48) and the reference length was 1 m in each case
[Pei00b]. The attenuation of this fiber was measured with the cut-back method. It
is clearly discernible that the equilibrium mode distribution is not reached until
after a larger distance of fiber (>40 m). Until then, the measured kilometric attenu-
ation depends on length. For shorter lengths to achieve equilibrium mode distri-
bution, the fiber would have to be light launched with a numerical aperture of
9.4 Measurement of the Optical Parameters 701
approximately 0.30. This would mean that the launching optics in Fig. 9.48 would
need to be modified.
Figures 9.55 and 9.56 show the attenuation spectra of a multicore POF
(37 cores) at different launching conditions.
600
attenuation [dB/km] length:
500 30 m
400 50 m
64 m
300
100 m
200
100
wavelength [nm]
0
400 450 500 550 600 650 700
Fig. 9.55: Attenuation of a multicore POF (Asahi PMC 1000, 37 cores, AN = 0.19) for
different measured lengths; light launched with AN | 0.50
In both measurements, the reference length was 0.68 m; in Fig. 9.55 launching
took place with a numerical aperture of 0.50, in Fig. 9.56 with 0.17 using a glass
fiber bundle. Both measurements were carried out with the substitution method.
The attenuation coefficient depends heavily on length; when light is launched with
AN = 0.17 it was no longer possible to measure the 100 m length, since the
dynamic range was approximately 10 dB lower due to the mismatch of areas bet-
ween the fiber bundle ( 3 mm) and the MC-POF ( 1 mm). In addition, the
702 9.4 Measurement of the Optical Parameters
600
attenuation [dB/km] length:
500 30 m
50 m
400
64 m
300
200
100
wavelength [nm]
0
400 450 500 550 600 650 700
Fig. 9.56: Attenuation of a multi core POF (Asahi PMC 1000, 37 cores, AN = 0.19) for
different measured lengths; light launched with AN | 0.17 (fiber bundle)
attenuation
300 [dB/k ]
250
200
150
100
EH 4001 AN = 0.47 AN launch = 0.50 (St.-NA)
MH 4001 AN = 0.30 AN launch = 0.50 (DSI)
50 PMC 1000 AN = 0.19 AN launch = 0.50 (MC)
PMC 1000 AN = 0.19 AN launch = 0.17 (MC)
length [m]
0
0 20 40 60 80 100
Fig. 9.57: Attenuation for different types of POF in relation to length
9.4 Measurement of the Optical Parameters 703
450
attenuation [dB/km] ESKA GH 4001
400 AN = 0.51
length: 30 m HLMP
350
K155
300
250
HR 430 SHR SHR HLMA
200 525C5 525C3 DL00
150
100
50 LED
Nichia 560nm
0
400 450 500 550 600 650 700
wavelength [nm]
POF sample length: 10 m 20 m 30 m
Fig. 9.58: Measuring attenuation with LED and monochromator (without correction)
450
attenuation [dB/km]
400
HLMP
350 K155
300
250
HR 430 HLMA
SHR DL00
200 SHR 525C3
150 525C5
100
50 Nichia 560nm
wavelength [nm]
0
400 450 500 550 600 650 700
POFsample length: 10 m 20 m 30 m
Fig. 9.59: Measuring attenuation with LED and monochromator (with correction)
704 9.4 Measurement of the Optical Parameters
pulsed
laser
fiber under test
splitter
detector
optical absorber
control unit
signal display
processing
input
signal scattering center
back scattered
signal
A short light pulse is launched into the fiber at t1 and passes the length L2 - L1 at
speed v (approximately 2 108 m/s); the light is reflected at the fiber end and
returns to the beginning of the fiber at time t3; the optical time domain reflectomter
measures the run time of the pulse 't = (t2 - t1) + (t3 - t2) = 2(t2 - t1) (Fig. 9.62) and
converts it into the length. The following formula is used:
L
t t
v 2 1
c t 2 t1
2 nk 2
Because of the pulse propagation time is determined by the speed of light in the
fiber core, for a precize location of an event, e.g. refections at the fiber input or
output, the correct knowledge of the refractive index is required.
t1 t2
t3 length L
L1 L2
Fig. 9.62: Principle of the back scattering method, times and powers
In Fig. 9.63, the backscattered signal is shown in a logarithmic scale over the
length. At the start of the fiber as well as at the end of it, there are reflections
leading to strong backscattering signals. The backscattered power Pr(z) is cal-
culated as follows:
1
Pr ( z) P0 S Ds ti v e 2Dz
2
with P0 being the introduced power, S the backscattering factor, Ds the attenuation
coefficient due to Rayleigh scattering [km-1], ti the time width of the launched
pulse, v the group speed, D the overall attenuation coefficient [km-1], and z the
length of fiber ([Gri89]).
The backscattering factor S indicates the degree of re-coupling, i.e. the propor-
tion of light backscattered into the numerical aperture; only this will get to the
beginning of the fiber and be available for measuring. For step-index fibers the
factor is calculated as follows:
2
3 AN
S
8 ncore
706 9.4 Measurement of the Optical Parameters
A factor of 2 takes account of the fact that the light pulses have to pass the
length of fiber twice. The attenuation coefficient through Rayleigh scattering is
determined as follows ([Ebe00]):
D s >dB@
10 log 0.5 Ds S t i
Whereas in glass fibers the main proportion of attenuation is caused by
Rayleigh scattering, in polymer optical fibers absorption through molecular vibra-
tions and impurities is dominant.
For a POF with AN = 0.5, ncore = 1.497, ti = 1 ns z = 0.0001 km and
Ds = 2.8 km-1 ([Kai81]) one obtains Ds = 52 dB, which means that the power of
the backscattered signal is 52 dB less than the launch signal. To this we must add
the insertion losses of the splitter or coupler of approximately 7 dB. In addition,
there is twice the attenuation of 100 m POF with 30 dB at O = 650 nm. That means
that the OTDR has to cover a dynamic range of approximately 90 dB in order to
be able to measure the attenuation in 100 m POF. For O = 520 nm, the situation is
somewhat more favorable. Since here the Rayleigh scattering attenuation is larger
and the POF attenuation lower than with 650 nm, it is possible to analyze a longer
fiber.
reflection at
the fiber input reflection at the
reflecting connector fiber output
P1 fiber 1
P2
connector fiber 2
loss
splice or non
length fiber 1 reflecting
connector
L1 t1 L 2 t2 length
A plug connection having an air gap between the two fiber ends shows a clear
peak with a step in attenuation in the backscattering signal, similar to the fiber
beginning and end; however, a plug connection without reflection only shows a
step in attenuation. The strong backscattering signal at the beginning and end of
the fiber leads to an overloading of the detector, which means that during a certain
time interval, which is determined by the amount of pulse and reflection as well as
the recovery time of the receiver, no signal can be analyzed. The minimal distance
between a reflecting and a non-reflecting event that can still be resolved by the
OTDR is called the attenuation dead zone, while the minimal distance between
two reflecting events is called the event dead zone. The local resolution is given
by the pulse width ti. An pulse width of 10 ns corresponds to a length of approxi-
mately 2 m, 1 ns corresponds to 20 cm; since double the path (there and back)
must be taken into consideration, the spatial resolution at 10 ns is 1 m and at 1 ns
is 10 cm. However, the reduction of pulse width leads to a reduction in backscatte-
red power and therefore to a reduced dynamic range of the measuring system. In
order to be able to analyze greater lengths one selects a longer pulse duration ti at
the expense of spatial resolution. Figure 9.64 shows the backscattering signal of a
standard NA polymer optical fiber. During the first 10 m one can clearly recognize
a non-linear curve due to the launching process, since excitation did not take place
under conditions of equilibrium mode distribution. These occur after approxima-
tely 40 m; after that distance the curve remains linear and the attenuation coeffi-
cient independent of the length.
-25
back scattered signal [dB]
-30
-35
-40
attenuation 0.16 dB/m
13 dB
-45
-50 80 m
length [m]
-55
0 20 40 60 80 100 120 140 160
Table 9.2: Scope of application for the optical backscattering measuring method
1.2
amplitude
1.0
0.8
0.6
0.4
0.2
time [ns] time [ns]
0.0
0 10 20 30 40 50 0 10 20 30 40 50
Fig. 9.65: Reflected double pulse without (left) and with deconvolution (right), 149 m
The difficulty in this procedure lies in the fact that you have to know the
length-dependent pulse response of the respective fiber investigated and be able to
present it as an analytical printout. The pulse as a superimposition of 4 Gaussian
functions, the intensities and widths of which change depending on the length, was
described in >Otto02@. An example of the approximation of an output pulse after
200 m of fiber (with AN = 0.19) is shown on the left in Fig. 9.66. On the right side
you can see the simulated pulse broadening after 200 m compared with the pulse
response of the system.
amplitude
1.0 1.0 amplitude
output pulse
0.8 after 200 m POF 0.8 system with-
out POF
0.6 0.6
system with
200 m POF
0.4 0.4
pulse parts
0.2 0.2
time [ns] time [ns]
0.0 0.0
0 10 20 30 40 50 0 10 20 30 40 50
Luciol is the sole vendor who can equip the device with blue or green LEDs in
order to be able to measure POF at these wavelengths. A laser is used for 650 nm
for which the manufacturer gives the following parameters:
standard wavelength (POF): 500 nm, 650 nm
sensitivity: -110 dBm
spatial accuracy: 5 mm
spatial resolution: 10 cm
dynamic range (loss): 35 dB
detector: APD, single photon counting
time constant: <500 ps
Another vendor of POF OTDR is the Canadian company Tempo. This device
also works with a 650 nm laser (Fig. 9.68). An center wavelength of 658 nm at
room temperature at a spectral bandwidth of 2.4 nm has been established for this
device.
Lengths up to 150 m can be measured with standard POF (see Section 9.4.6.6).
The device has been optimized for a particularly high spatial resolution, but also
has very good sensitivity. Operating it has been complicated so far, but this will be
improved. Figure 9.69 shows a measurement example with 120 m St.-NA POF.
10.0000
power [a.u.]
1.0000
coupling
0.1000
end
0.0100
0.0010
t [ns]
0.0001
0 200 400 600 800 1000 1200 1400
Fig. 9.69: Measurement example with the Tempo-POF-OTDR (20 m + 100 m POF)
-50 -80
0 50 100 150 200 250 0 50 100 150 200 250
position [m] position [m]
In both cases the end reflex can still be seen clearly after 200 m. The dynamic
range of the device according to >Yago01@ is considerably greater which is why up
to 200 m the Rayleigh signal also remains above the noise. The fiber attenuation is
indeed lower at 532 nm, but the frequency-doubled laser presumably has less
coupled power and/or a too low repetition frequency. It is not known whether the
works cited will lead to a commercial device.
Another commercial OTDR from the manufacturer Scientex (OTDR-2000POF)
is reported on in >Nak04b@. It is not clear whether this device is actually sold
outside Japan. The parameters are similar to those of the Tempo device. The
device is shown in Fig. 9.71 and a measurement example for the attenuation can
be seen in Fig. 9.72.
wavelength: 650 nm
dynamic range: 18 dB
measurement range: 200 m
resolution: 1 cm
Fig. 9.73: Measurement of connector loss with OTDR ([Hut00] and [Bre00])
714 9.4 Measurement of the Optical Parameters
power [a.U.]
1.2
1.0
0.8
0.6 'W = 13 ns
4.74 MHzkm
0.4
0.2
0.0
0 1 2 3 4 5
rel. fiber position [m]
Fig. 9.74: Pulse broadening after a 50 m long test fiber (+20 m pre test fiber)
normalized amplitude
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 20 40 60 80 100 120 140 160
position [m]
The pulse spreading occurring can clearly be seen. At 150 m the pulse is still
sufficiently over the noise level of the device.
The results for a PMMA GI-POF are shown in Fig. 9.76. Because of the
smaller NA much light is lost when coupling in so that the dynamic range is
somewhat smaller. Furthermore, the fiber has a somewhat higher attenuation.
normalized amplitude
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.5 20.0 20.5 40.0 40.5 60.0 60.5 80.0 80.5 100.0 100.5
position [m]
Changes in the pulse width are not optically recognizable in the picture. A more
exact analysis, however, shows that the pulse widths increase significantly by
some tenths of a nanosecond. The results for the bandwidths of both fibers calcu-
lated from this pulse broadening is shown in Fig. 9.77.
bandwidth [MHz]
5000
2000
PMMA-GI-POF
1000
500
200
100 PMMA-SI-POF
50
POF length [m]
20
10 20 50 100 200
Fig. 9.77: Bandwidths of the SI-POF and GI-POF, measured with OTDR
The results tally quite well with the measurements in the frequency range up to
a fiber length of 100 m. You also get realistic values for GI POF - albeit very
faulty. The values over 100 m are too inexact. The reason for this may be less the
noise than the faulty linearity of the method of measurement. All in all, however,
the bandwidth measurements with OTDR are a very viable alternative to trans-
mission procedures.
9.4.7 Dispersion
When measuring mode dispersion, principally the same considerations for laun-
ching light apply as for measuring attenuation (please refer to Sec.9.4.5). It must
be pointed out however, that in spite of uniform mode distribution the bandwidth
is subject not to linear change but to sub-linear change with different lengths of
fiber due to mode coupling and mode conversion. This subject is discussed in
more detail in Chapter 2. Two methods are available for measuring dispersion:
1. Measurements in the time domain
2. Measurements in the frequency domain
pulse is detected at the end of the fiber using a fast receiver with a bandwidth
larger than that of the fiber to be measured. This is then made visible with the help
of an oscilloscope (Fig. 9.78). The pulse changes width and height over the trans-
mission link. The pulse response is expressed as follows:
g( t ) Pout ( t ) Pin ( t ) .
launching
optics
oszilloscope
From the input and output width of the pulse it is possible to calculate the broa-
dening over time. When assuming Gaussian pulse curves, the following simple
formula results:
2 2
't t out t in ,
whereby tout and tin represents the full width at half maximum, at which the pulse
height has dropped to 50 % (see section 2.5.2, Fig. 2.87). 't over the length of
fiber L gives us the dispersion parameter D: D = 't/L [ns/km]. The proportionality
of 't and L applies up to coupling length; for larger lengths, the following applies:
't v LN proportional, whereby N < 1 must be determined for each fiber configu-
ration individually (see Chapter 1). As described in the section on measuring
attenuation, dispersion is also measured with the cut-back and substitution
methods in order to keep launching conditions constant. The transmission capacity
of a fiber is derived from the bandwidth-length-product:
0.44 0.44
B L [MHz km] | L
't D
An example of measurements in the time range are shown in Fig. 9.79. A pulse
width of 5.1 ns is measured on a 50 m long POF which corresponds to a band-
width of 43 MHz 100 m.
718 9.4 Measurement of the Optical Parameters
0.7
power [a.U.]
0.6
0.5
0.4 'W = 5.1 ns
0.3 4.3 MHzkm
0.2
0.1
0.0
t [ns]
0 5 10 15 20 25 30 35 40
fiber spectrum
light under analyzer
sweep source
test
generator
|
launching
optics receiver
The measured curves in Figure 9.81 only serve as examples to show that this
procedure leads to very analyzable and reproducible results over a wide range of
fiber types, lengths and measured wavelengths. An entire series of different fibers
was described in Chapter 2 in regard to bandwidth. These measurements have
almost exclusively been conducted with the procedure described here.
9.5 Connector Measurement 719
0
electr. power [dB]
-3
-6 PFU-U-CD 1000; 10 m
-9 PFU-U-CD 1000; 20 m
-12
-15 PFU-U-CD 1000; 50 m
Fig. 9.81: Frequency response for selected polymer optical fibers at O = 520 nm ([Rit98])
a)
T R P01
L1 = 150 cm
b)
connector set
T A B
R P11
L1 = 60 cm
c)
A B C D
T R P21
L1 = 60 cm L2 = 30 cm
Fig. 9.82 a-c: Measuring method for determining the insertion loss of connectors
P
DK 10 log 0 x ,
P1x
whereby x refers to the test setup.
In the next step, the POF is cut at a distance of 30 cm from the first connector
set, the second connector set is fitted and the power P21 measured (Fig. 9.82c). The
insertion loss of the second connector set in dB is calculated as follows:
P
DK 10 log 1X
P2 X
The above steps are carried out a total of four times in parallel. The four patch
cables made in this way are swapped cyclically and reversed. The test setup
remains the same. That means, for example, that patch cable BC is tested in test
set-up 2, the power P22 is measured and calculated with P12 DK , patch cable FG is
tested in test setup 3, the power P23 is measured and calculated with P13 DK, etc.
From the insertion attenuation values obtained, the mean, the standard deviation
and the maximum and minimum values are determined.
9.5 Connector Measurement 721
3.0
2.5
2.0
1.5
1.0
0.5
0.0
The connector sets for which the fiber end face is polished (HFBR 4501,
HFBR 4531, F-SMA push fit connector, polishing type, and TOCP 155) on
average have a lower insertion attenuation compared with the connectors produced
in the hot-plate process (TCP connectors, F-SMA connectors type hot-plate, F07,
AMP DNP). This results from the fact that in the hot-plate process the fiber core is
expanded and therefore the wave guiding is disturbed for a length of approxi-
mately 0.5 mm (see Fig. 3.68). For a distance of this size the loss is calculated to
be approximately 0.6 dB.
722 9.6 The Reliability of POF
Polymer optical fibers, like other technological products, are subjected throughout
the whole of their service life to a great many kinds of stress from the environment -
mechanical, climatic, chemical, biological, and radiometric. As a result of these
stress factors, physical and chemical alterations may arise in the materials used.
These can in various ways have an effect on the functional behavior, suitability for
use, and serviceable life expectancy (that is to say, the durability) of the POF.
Environmental stress factors, then, have an influence that should not be under-
estimated on the quality and reliability of the fiber optic transmission system. When
polymer optical fibers are used, it is thus imperative to understand and take into
consideration the effects of environmental influences - particularly of industrial
environmental influences - on those properties of POF which enable it to transmit
optical signals.
The most important property in this context is the optical transmission, that is to
say, the optical attenuation. In judging the suitability of a polymer optical fiber for a
particular area of application, it is not the absolute attenuation that is of primary
interest: rather it is the relative changes it undergoes, as subject to the action of the
various environmental influences. The systematic investigation and assessment of
the optical transmission (or attenuation) of a polymer optical fiber when subjected to
extreme environmental conditions can be broken down, as a rule, into these three
steps:
1. analysis of the environmental stress factors,
2. simulation of environmental conditions in carrying out the appropriate test
procedures,
3. quantitative determination of the effect of environmental influences on the
optical transmission.
In the light of the test results, we can then conclude what are the conditions
suitable to the use of this material and what are the limits (e.g. the maximum tem-
perature at which regular operation is possible).
The basis of our environmental simulation must be the analysis of the environ-
mental stress factors to which a polymer optical fiber is liable in any given condi-
tions of use. For this purpose, the environmental influences that are to be expected
have to be registered and defined. We must make distinctions here based on the type
of influence, on its frequency, intensity, and the opportunity it has to cause damage.
In every particular situation of use, there is a plethora of different types of influence
and their possible combinations to be taken into account.
We can distinguish in principle between mechanical, climatic, chemical, bio-
logical, and radiometric influences. Each one of these, as Table 9.4 shows, can occur
in the most diverse forms.
9.6 The Reliability of POF 723
Types of Influence
chemical and
mechanical climatic radiometric
biological
static bend high humidity lubricants UV, X-ray,
repeated bending extreme tempera- fuels and
flexing ture brake fluid nuclear
crush change in climatic hydraulic oil radiation
impact conditions acids and alkalis
torsion thawing solvents
vibration freezing oxygen
tensile tension ozone
reactive gases
micro-organisms
also in various combinations
acceleration is based in this case on the intensification of the stress conditions during
the simulation, as compared with the situation in the real world. This intensification
can be achieved by various methods:
the superimposition of extreme stress conditions, as an ongoing stress situation,
cyclically alternating stress between two opposite extremes,
raising the stress to a value beyond the extreme possible in reality, or
an increase in the rate of change in a situation of alternating stress.
The final choice of a strategy for simulation or testing, and of the parameters that
go with it, is based as a rule on the results of environmental analysis, as also on prior
knowledge and experience in dealing with similar problems, along with foregoing
investigation into the polymer optical fibers that are to be tested.
All the demonstrations that follow in this chapter are based, when not otherwise
stated, on many years of experience in the investigation of polymer optical fibers
and the assessment of their reliability under the auspices of the Federal Institute of
Material Research and Testing (BAM), Berlin. The major part in these tests is
played by the 1 mm SI-POF on a base of polymethylmethacrylate (PMMA), which
is almost the only one to be used in practice. Polymer optical fibers based on other
materials, e.g. polycarbonate (PC), or deuterated or fluorinated polymer (e.g.
CYTOP), if relevant in practice, will also be taken into account when it is appropri-
ate to do so. The simulation and testing procedures that are described are basically
suitable for all POF types, though of course the parameters must be accommodated
in each case to the nature of the material and its distinctive structural characteristics.
RS 232
step drive axis 1 position control
PC control unit axis 2 RS 422 encoder signal
aperture control
detector unit light source unit
PIN
RS 232 photodiode LED power connector
low-noise supply light
A/D converter
amplifier source
O 1 O2 O 3
spectrum aperture
analyzer
connector
incremental
position
measuring linear
system positioning
system
POF connections
input for n ... 3 2 1 n ... 3 2 1 reference
information flow
POF samples fiber
electrical connections
The layout of a typical multiplexer for this purpose will be introduced in the
following description ([Gn00]). The basic layout (Fig. 9.84) can be subdivided
into three functional areas: the light source unit, the detector unit, and the posi-
tioning system. The light source unit consists of three LED that attain their maxi-
mum optical power at wavelengths of 525 nm, 590 nm, and 660 nm respectively.
This gives us the possibility of carrying out transmission measurements at the
wavelengths that are relevant to the test. The individual LED are launched into the
polymer optical fiber to be investigated or the reference fiber, by means of a con-
trollable aperture and of a 4 u 1 coupler. An additional optical input makes it
possible to carry out transmission measurements with an external light source, e.g.
laser diode or white light source.
The detector unit consists of a pin-photodiode followed by a low-noise ampli-
fier. To enable the use of additional detectors, e.g. an optical spectrum analyzer,
an asymmetrical 1 u 2 coupler is incorporated into the optical path so that an addi-
tional external optical output is available. Both the detector and the light source
units are set on the adjustable platform of the positioning system.
The actual switching between the POF to be tested is executed by the linear
positioning system. As can be seen in Fig. 9.84, the coupling of the POF sample is
executed as a front surface coupling with an in-between gap in the order of
100 m. The POF samples are inserted into the multiplexer from without and
fixed in special receptacles, uniformly and on a single plane. Measurement of the
optical transmission by the insertion method is then carried out in the following
way. First the optical power of all the LED is measured by means of a short
reference fiber. Thereafter the platform is positioned for the other POF samples
and their transmission values are likewise measured.
9.6 The Reliability of POF 727
Fig. 9.85: New generation of multiplexers for PCS and POF (POF-AC)
t micro pore
fiber break
micro crack
I I detector system
t
t=f(L,n) imperfection display
Fig. 9.86: Detection of imperfections by measuring backscattering
The basic principle is shown in Fig. 9.86. An optical pulse from a laser diode,
e.g. with OP = 670 nm; FWHM = 4 nm; 't < 100 ps, is launched into the polymer
optical fiber to be tested. The pulse passes through the fiber. If there are imper-
fections present that cause a Fresnel reflection, a part of the pulse is scattered and
reflected in a backward direction. By recording the optical power of the reflected
light for the duration of the test, we can arrive at the information we need on the
location and extension of the imperfections, provided that the instrument has been
properly calibrated at the outset (Fig. 9.87 and Fig. 9.88). Information gathered in
this way plays a part not only in reliability investigations, but also in locating and
analyzing damages within installed POF cables ([Zed98]).
0.6
0.4
0.2
0.0
4 8 12 16 20 24 28 32 36
cable length [m]
Fig. 9.87: Diagram showing the backscattering of a POF cable without imperfections
9.7 Reliability under Various Environmental Influences 729
0.6
0.4
0.2
0.0
4 6 8 10 12 14 16
cable length [m]
Fig. 9.88: Diagram showing the backscattering of a POF cable with several imperfections
climatic chamber
drive
weight drive shaft
unit
200 g
data
light recording
source receiver system
110
transmission [%]
100
90
80
R = 5 mm
70 R = 10 mm
R = 20 mm
60
R = 40 mm
50
10 100 1,000 10,000 100,000
number of repeated bending cycles
110
transmission [%]
100
90
80
70
60
50
10 100 1,000 10,000 100,000
T = -40C T = +23C number of repeated bending cycles
T = +85C/85% r.H.
Fig. 9.91: Optical transmission of a 1 mm SI-POF with PE jacket subjected to repeated
bending stress under varying temperature/climatic conditions and with
R = 10 mm ([Daum93])
732 9.7 Reliability under Various Environmental Influences
A peculiar reaction of the fiber can be seen at high temperatures. In this case it
is true that a longer durability can be observed, but as regards the functionality of
the cable, it can be seen that after only 100 repeated bending cycles a rapidly
increasing deterioration in the transmission already appears. This effect on the
optical transmission can be explained by an increasing and irreversible geometric
change in the vicinity of the bending point, the optical fiber being still relatively
flexible at these temperatures (Fig. 9.92). If the constriction becomes too great for
the stress at any time, the fiber will break. In practical applications this means that
at high operating temperatures the polymer optical fiber should not be exposed to
repeated bending stress or only to a very limited degree. At lower temperatures
component failure is generally characterized by a clean break in the fiber
(Fig. 9.93, [Daum93]).
The estimate of the smallest allowable bending radius with a predetermined num-
ber of repeated bending cycles is based on the following considerations. As
mentioned already, it can be taken as a first approximation that in bending stress the
mechanical stress at the periphery varies in inverse proportion to the bending radius.
With a decrease in the stress, that is, with a greater bending radius, the number of
cycles needed to cause component failure is higher. By extrapolating the results of
the tests - the number of repeated bending cycles up to the point of component fai-
lure with various bending radii - to any specified number of repeated bending cycles,
one can estimate the corresponding smallest allowable bending radius (Table 9.6).
9.7 Reliability under Various Environmental Influences 733
Table 9.6: Estimate of the smallest allowable bending radius in repeated bending
with a given number of cycles and T = +23C
9.7.1.2 Flexing
It is characteristic of the use of POF in mechanical engineering that in many cases a
transmission of signals takes place between the control unit and the moving com-
ponents of the system. A typical case of this is the transmission of data between a
stationary machine control unit and the various drive units in large automated mani-
pulator systems. In systems like this both cables that supply energy and those that
transmit signals are carried by means of trailing chains.
The mechanical stress arising from this is characterized by a cyclically repeated
unrolling movement where a certain length of the cable executes a U-shaped turn
through 90. When this occurs, the polymer optical fiber is subjected to an alter-
nating bending stress, which extends over the whole length of the trailing chain. As
with repeated bending, this means that the POF undergoes alternating states of
734 9.7 Reliability under Various Environmental Influences
mechanical stress at its periphery over the entire length. In this case as well, the
degree of the stress varies in inverse proportion to the bending radius.
In order to simulate this characteristic stress condition and to test the ease of
handling, we can make use of an established flexing test as a basis for our testing
procedure. This test method has been effectively used for quite sometime as a
means of testing cables and insulated lines. The basic principle of an apparatus for
carrying out this simulation is shown in Fig. 9.94.
weight 200 g
data
receiver recording
light system
source
The POF is installed around the two pulleys in the form of an S and by a pre-
determined movement of the pulley arrangement is subjected to a repeating cycle
of bending. This subjects the polymer optical fiber to alternating states of stress at
its periphery (tensile stress caused by stretching, absence of stress, and compres-
sion stress caused by compression). In contrast to repeated bending where stress is
focused just at a single point in the fiber, a whole section of the polymer optical
fiber is subjected to extreme mechanical stress in this test. In order to ensure that
the polymer optical fiber adheres securely to the bending radii, the POF is loaded
with a weight (typically 200 g) at both ends and subjected to tensile load.
During the test the optical power is measured at intervals after the execution of
a certain number of cycles of flexing. For this purpose, the movable carriage is
first brought to a resting position. After a relaxation time of about 60 seconds has
passed, the optical power can be measured and the relative transmission value
determined.
Figure 9.95 records typical test results for the measurement of transmission in
fibers subjected to stress due to flexing with varying bending radii.
9.7 Reliability under Various Environmental Influences 735
110
transmission [%]
100
90
80
70
R = 20 mm (PE)
60 R = 40 mm (PE)
R = 40 mm (PA)
50
1 10 100 1,000 10,000 100,000 1,000,000
number of flexing cycles
Fig. 9.95: Optical transmission (1 mm SI-POF with PE or PA jacket) under flexing with
varying bending radii and T = +23C
9.7.1.3 Torsion
A twisting of the polymer optical fibers can occur, for example, in the manufacture
of cables or directly during the installation of a POF. In this case, too, we must make
a distinction between static and dynamic stress. Static torsion is found, for instance,
when polymer optical fibers are fixed in position. This has no determining signi-
ficance for the reliability of the fiber. Dynamic twisting on the other hand puts the
polymer optical fiber under considerably greater stress. This kind of stress arises
mostly from a cables being subjected to movement, which happens with industrial
robots or automatic manipulation systems, for example.
As in the example of repeated bending, dynamic torsion leads to a cyclical
repetition of the build-up of mechanical stress and its release in the polymer
optical fiber. As possible effects of this stress on the optical transmission we may
mention here the damage of the interface between the optical cladding and the
fiber core; development of micro-cracks in the fiber, possibly leading to a break; a
crack in the jacket (Fig. 9.98), leading to the direct influence of humidity or other
aggressive substances on the fiber. At high temperatures an irreversible geometric
change is also possible in view of the softening of the fiber.
For the investigation of optical transmission under torsion stress, possibly in
combination with climatic stress, a test apparatus like that shown in Fig. 9.96 is
736 9.7 Reliability under Various Environmental Influences
the most useful. This consists of an installation for the testing of torsion, which is
incorporated in a climatic chamber, while the drive unit attached to it is situated
outside the chamber. The actual torsion testing device consists of two mounts for
holding the POF, one fixed and the other capable of rotation, both of which in the
form of a guide pipe with a clamp fitting at the pipes end. The left guiding pipe is
mounted on a moveable carriage, which is drawn in the direction of the POF axis
with a weight of 200 g so as to subject the sample length under test to a certain
amount of tensile stress.
600 mm
data
weight 200 g recording
system
climatic chamber
detector
The following testing cycle has proved to be suitable for investigating the
optical transmission under torsion stress. The POF sample is first turned in a
clockwise direction for a specific number of revolutions. Then the sample is
brought back to its starting state and turned through the same number of revolu-
tions in a counterclockwise direction before finally being brought back again to
the starting position. The measurement of the transmission always takes place in
the unstressed state after a relaxation time of about 60 seconds. As a result of the
test the relative change in the transmission is determined, that is to say, the change
that has set in after a given number of torsion cycles have been carried out as
compared with the starting value of the sample that has not yet been exposed to
torsion stress.
Typical effects on the reliability of polymer optical fibers under torsion stress are
shown in Fig. 9.97. Starting from the unstressed condition, the transmission at first
continues almost unchanged in spite of the increasing number of repetitions of the
torsion cycle until a particular threshold is reached, after which the transmission falls
sharply. Component failure (transmission < 50%) occurs at room temperature at
some point between 2,000 and 3,000 cycles. The transmission curve at low tempe-
rature (-40C) is in principle similar to that at room temperature, but here component
failure occurs as early as somewhere between 400 and 500 cycles. The notably
9.7 Reliability under Various Environmental Influences 737
shorter durability can be ascribed to the brittle condition of the polymer optical fiber
at this temperature as compared with that at room temperature - temperature being a
factor determining the modulus of elasticity and the mechanical stress properties.
110
transmission [%]
100
90
80
70
T = -40C
60 T = +23C
number of
T = +85C/85% r.H.
torsion cycles
50
10 100 1,000 10,000 100,000
Fig. 9.97: Optical transmission of a 1 mm SI-POF with PE jacket subjected to torsion stress
(1 cycle: r 10 u 360) under varying climatic conditions
climatic chamber
top view
force sensor
R R drive
F unit
100 mm
light data
detector recording
source
system
top view
F
R = 40 mm R = 40 mm
Fig. 9.99: Test equipment for testing the tensile strength
The left chuck drum of the tensile testing device has a radius of R = 40 mm. It is
fixed in place and connected with a firm framework structure situated outside the
climatic chamber. During the tensile test the right chuck drum (R = 40 mm) is
moved by a drive unit in the direction of the POF samples length, which results in
9.7 Reliability under Various Environmental Influences 739
the sample being stretched between the two drums and is subject to tensile stress.
While the tensile stress increases, the optical power and the traction force that is
being applied are measured continuously. The result of the test consists in a
calculation of the relative transmission during the increasingly intense tensile stress,
correlated with the stress-strain curve.
Typical results after the execution of tensile tests are seen in Fig. 9.100 to
Fig. 9.101. Particularly noticeable, especially at room temperature is the behavior,
specific to polymers, of the stress-strain curve.
300
force [N]
T = -40C
250 T = +23C
T = +85C/85% RH
200
150
100
50
strain [mm]
0
0 50 100 150 200 250
Fig. 9.100: Stress-strain curve of a 1 mm SI-POF with PE jacket subjected to tensile stress
under varying climatic conditions
110
transmission [%]
100
90
80
70
T = -40C
60 T = +23C
T = +85C/85% RH strain [mm]
50
0 50 100 150 200 250
In the first phase there is at once a sharp increase in the traction with only a slight
increase in the strain. In this zone there is initially an almost linear ascent of force
over strain. This zone comes to an end when the yield point is reached - visible as an
overshooting (decrease in stress with increase in strain). Here the zone of plastic
deformation begins. In this zone after a brief drop in force, a gradual increase in ten-
sile force accompanied by an increase in the strain can be observed. The sample is
pulled out of shape as a result of cold flow. When the deformation of the POF
sample can go no further, the sample breaks. This is the point at which the maxi-
mum traction force for the respective POF sample is reached.
During the tensile test a steady diminution in the optical transmission can
generally be observed, up to the point where the sample breaks. Before the zone of
plastic deformation is reached, the transmission is only insignificantly affected to
the order of 2% to 3%. With regard to ensuring reliability in practice, it is abso-
lutely essential that the limit of extension is not be reached if the POF is subjected
to an intense short-term tensile stress. In regard to the various climatic conditions
this means that the various limiting maximum values for short term traction force
must not be exceeded, either in the installation or during the period of its use, to
prevent irreversible deformation or breaks in the fiber. In addition, one must take
care that the development of micro-cracks does not lead to a deterioration in the
transmission quality.
transmission [%]
110
100
90
80
70
T = -40C
60 T = +23C
T = +85C/85% r.H.
50
0 50 100 150 200 250
strain [mm]
force [N]
400
T= -40C
350
T= +23C
300 T= +85C/85% r.H.
250
200
150
100
50
0
0 50 100 150 200 250
strain [mm]
Fig. 9.103: Stress-strain curve of a 1 mm SI-POF with PA jacket subjected to tensile stress
under varying climatic conditions
Since risks such as tensile stress of varying magnitude and the development of
cracks cannot fundamentally be excluded, we recommend in practical situations that
the allowable short-term maximum tensile force Fmax be limited using a safety
coefficient S, to Fmax/S, where S = 1.5. Because the mechanical stress properties
are affected with an increase in temperature and with the risk of a loss of dimen-
sional stability, Fmax should absolutely be avoided at higher temperatures, or at
least considerably reduced.
With continuous tensile stress it must be noted that on account of the properties
of the material only substantially lower tensile forces are permissible. It is recom-
mended in the literature ([Schmi92]) that the zone of linear visco-elastic deformation
not be exceeded with continuous tensile stress. This zone corresponds to an exten-
sion of 0.1% to 0.5% with thermoplastics. There are ways of improving the resis-
tance of the cable to tension load either by changing the material of the jacket or the
composition of the cable, e.g. with an additional strength member.
In addition, it is possible that the optical fiber may suffer irreversible geometric
change in shape. The burst of the jacket as a result of impact stress can enable humi-
dity or other aggressive substances to act upon the optical fiber without impediment.
An apparatus for the testing of impact effects to polymer optical fibers is shown
in Fig. 9.104. The installation used basically consists of a free falling mass with a
weight of 1 kg to create the energy of the impact. The drop weight is guided by a
guiding rod, practically without any friction, for the duration of the free fall before
making contact with an impact piece, which lies, on the POF sample. This impact
piece then transmits the entire energy of the impact to the polymer optical fiber.
To intensify the stress, the underside of the impact piece has a radius of 10 mm
which is positioned straight across to the POF sample. After the impact has been
effected, an electromagnet picks up the drop weight which is then brought back by
means of a drive unit to the prescribed drop height and allowed to fall again.
climatic chamber
drop weight
1 kg
data
impact piece detector recording
20 mm system
light
source
If impact tests are carried out on POF, we find, as was to be expected, that when
the drop height is greater, fewer impacts are needed to induce component failure,
that is, a reduction of the transmission to less than 50% of the starting value. This is
the case in all climatic conditions. At temperatures of +23C and -40C the optical
transmission of the POF samples basically shows a similar pattern as the number of
impacts rises. Starting from the unstressed state, the transmission remains almost
constant or is reduced only by a minute gradient until the critical number of impacts
is reached. From this point on the transmission is rapidly reduced, leading to
component failure. A visual examination of the POF samples after the end of the test
often shows that the jacket of all the samples has split open. In addition, it can be
observed that the fiber shows signs of brittle splintering with zones of pronounced
fiber-like cracks or it may even break entirely.
9.7 Reliability under Various Environmental Influences 743
A summary of typical test results for impact is shown in Fig. 9.106 and
Fig. 9.107. The illustrations represent the number of impacts needed at room
temperature for a typical POF sample to reach the 50 % transmission critical point,
the drops taking place from varying heights.
90
drop height [mm]
transmission < 50%
80
transmission > 95%
70 drop weight 1 kg
60
50
40
30
20
10
number of impacts
0
1 10 100 1,000
Fig. 9.106: Component failure of a 1 mm SI-POF with PE jacket subjected to impact stress
by an object falling from various heights
90
drop height [mm]
80 transmission < 50%
transmission > 95%
70 drop weight 1 kg
60
50
40
30
20
10
number of impacts
0
1 10 100 1,000
Fig. 9.107: Component failure of a 1 mm SI-POF with PA jacket subjected to impact stress
by a drop weight from various heights
9.7 Reliability under Various Environmental Influences 745
data
recording
system
detector
drive
100 mm unit
F drive
shaft
light
source
climatic chamber
The transmission values are repeatedly taken after the pressure has been released
from the sample and after a suitable time of relaxation has been allowed to elapse.
The result of the test consists of the relative transmission values in a rising number
of lateral pressure cycles, taking the unstressed state at the start of the test as a refe-
746 9.7 Reliability under Various Environmental Influences
rence point. As a rule the test is carried out until material damage to the sample sets
in, e.g. burst of the jacket, or until transmission declines to a level of 50%.
transmission [%]
110
100
90
80 damage of
the jacket
70
F = 4500 N (PE)
60 F = 1950 N (PE)
F = 4500 N (PA)
50
0 200 400 600 800 1,000 1,200 1,400
number of crushing cycles
9.7.1.7 Vibration
Stress caused by vibration has a particular relevance to automobiles but can also be a
factor in industrial applications. Experience with glass fibers shows that this kind of
vibratory stress in extreme cases can lead to a failure of the optical transmission
following breaks in the fiber. In view of their greater flexibility, however, such
behavior is not to be expected from polymer optical fibers. Tests have shown that in
the example of automobiles [SAE78] the typical vibratory stress lies in a frequency
band of 10 to 2,000 Hz.
In order to simulate these stress conditions we can - following [IEC95] - fasten
the POF samples in the shape of a ring on a vibrating table (shaker), taking into
9.7 Reliability under Various Environmental Influences 747
account the permissible radius of bending in accordance with the type of installa-
tion intended, e.g. fastening in place by means of cable ties. While the trans-
mission is concurrently measured, the samples are then subjected to the vibratory
stress described above. Tests like these carried out in the frequency range of 10 to
2,000 Hz over 100 hours which correspond to 1636 frequency cycles at 100 sec
per decade resulted in no alteration in the optical transmission, thereby confirming
the optimal qualities of polymer optical fibers in this respect.
Typical test results for various POF samples are shown in Fig. 9.112. The alterations
in temperature and humidity shown in Fig. 9.111 form the basis of these results. A
complete cycle of temperature change lasts 8 hours, of which 2 hours are to be
envisaged at each of Tmin and Tmax. The relative humidity (95% RH), being deter-
mined by the properties of the climatic chamber, can only be sustained with reaso-
nable accuracy when the temperature is between +23C and +90C.
Fig. 9.111: Temperature and humidity curves (2 cycles) in the simulation of extreme stress
caused by climatic change
9.7 Reliability under Various Environmental Influences 749
Within a time span of approximately ten hours after the start of the experiment,
both POF samples show a decline in transmission in the order of 10%. Here we are
faced with a characteristic effect which can be observed in all investigations where
high temperature and high relative humidity occur in conjunction (see also the
following section). However, the transmission thereafter remains practically con-
stant over the remaining time span of the test (125 cycles of change or 1,000 hours).
For practical use this has the implication that in a strictly limited time-frame of
exposure to stress caused by extreme changes of temperature coupled with high rela-
tive humidity, apart from a fixed falling-off in the transmission, no further change in
the optical transmission is to be expected. When the stress of extreme climatic
change continues over a longer period of time, however, we must take into account
the accelerated aging of the polymer optical fiber as is explained in more detail in
the following section.
100
90
80
70
PE
60
PA
50
0 200 400 600 800 1,000 1,200
time [h]
according to the material and the wavelength, can occur when water is absorbed.
This increase in attenuation is caused, essentially, by the O-H absorption of the light
at 750 nm (third harmonic of the O-H stretching vibration mode) and at 850 nm
(combination of the second harmonic of the O-H stretching vibration mode and the
O-H deformation vibration mode).
In connection with the absorption of humidity, the thermal aging that occurs at
high temperatures is the essential determinant of the usability and durability of
polymer optical fibers. Decomposition effects can be shown to occur with thermal
aging which, as a consequence of transference of energy in the form of heat, lead to
a degradation or splitting of the polymer chain. Every polymer has its own weak
points where decomposition effects are most likely to occur. In this class belong
lateral chains, for example, and substitutes that are connected to the main chains
with a low level of binding energy. With PMMA there is particular risk of depoly-
merization. This term stands for the splitting off of end groups and the loosening of
monomer components from the end of the chain. As the parts that break off have
unattached valences - being known as free radicals - they try to form new combina-
tions, for example with oxygen. This leads to oxidization resulting in the accumu-
lation of submolecular debris. The consequence is brittleness and disintegration,
with a direct effect on the mechanical and optical properties of the polymer optical
fiber (see for instance [Stru66], [Bros89]). With perfluorinated, graded index profile
polymer optical fibers, a thermally determined alteration in the dopant material can
come about, leading to changes in the refractive index. New materials that have just
recently become available, however, behave with admirable stability in this regard.
As [Kog00] and [Oni99] show in relation to CYTOP, thermal aging at 70C over
10,000 hours leads to no significant alteration of the refractive index or to any
deterioration in terms of attenuation and bandwidth.
It is characteristic of the thermal aging process that the relevant property, in this
case the optical transmission, does not deteriorate continually, but stays approxi-
mately constant at first over a greater or lesser period of time. Only after this initial
phase has come to an end, a steadily accelerating deterioration of the transmission
does set in. In order to predict the durability of the fiber, we can make use of the
principle of correspondence between time and temperature as it has been carried
over into polymer testing, seeing that the aging processes, so far as is known, obey
the laws of reaction kinetics. On this basis a mathematical extrapolation of the
decline in transmission due to aging can be made ([McK94]).
The procedure we are about to describe has proved effective for investigating the
effects of absorbed humidity and thermal aging on the optical transmission. The
conditions of the simulation or of the aging are determined in such a way that the
aging process is accelerated, bringing about an artificial (accelerated) aging. This
strategy of speeding-up is based on an intensification of the simulated stress in the
relatively short time span of the test. Here we must be careful that no other aging
factors that have anything to do with the aging effect we wish to produce should
enter the picture. For this reason, in investigating the life expectancy of polymer
optical fibers there is an experimental strategy, which has proved effective that takes
extreme, but still allowable environmental conditions as its starting point. This
means essentially that the temperatures chosen to produce the accelerated aging
9.7 Reliability under Various Environmental Influences 751
must lie clearly below the glass transition temperature Tg (in the order of 115C for
PMMA SI-POF).
As explained in the preceding section, the POF samples are positioned on a grid
in a climatic chamber and loosely rolled up for a prescribed length of time to induce
accelerated aging. The inputs and outputs of the fibers are taken out of the climatic
chamber and connected to a suitable measuring device capable of high resolution
measurement of the transmission (see also section 9.6.2.2). There then follows a
relaxation phase, lasting perhaps several hours, in which samples and the measuring
equipment can settle before going into operation. After this rest period the fibers are
subjected to temperature and humidity stresses while the transmission is measured at
the same time.
To begin with, the room temperature is kept constant while the humidity is raised
to the desired maximum value and then also kept constant at this level for the entire
duration of the test. After the maximum value has been reached, the temperature is
gradually raised over a period of 4 hours until it reaches the selected aging
temperature. During the entire test the optical transmission of all POF samples is
measured and the relative transmission in relation to the unstressed starting state is
determined.
Figure 9.113 shows the typical transmission pattern for three different wave-
lengths as shown by a 1 mm SI-POF under stress from high temperature and humi-
dity. Developments in the optical transmission can be broken down into four phases.
So far as current findings indicate, these phases can be correlated with certain
definite time-dependent aging effects ([Ziem00b]):
1) Within the first 24 to 48 hours a definite deterioration in the transmission
occurs following on the first assimilation of humidity.
2) In the following preliminary phase there is only a very slow deterioration in
the transmission to be observed. The length of this period, and the gradient of
the transmissions falling-off, depend on the manufacturer, on the material of
the jacket, and on the type of POF, as shown for example in Fig. 9.114. The
aging effects that take place during this period have at this stage only a slight
effect on the transmission.
3) This phase is characterized by a rapid deterioration in the transmission. The
reason for this is thought to be the increase in the free volume and a
heightened absorption of humidity following from that.
4) If the humidity is changed while the temperature is kept constant, the trans-
mission responds to these modifications. When the humidity is reduced to
normal values, the transmission almost returns to the values current at the end
of the second phase. If the humidity is raised again, the transmission immedi-
ately deteriorates, with the humidity now being very quickly absorbed into the
polymer optical fibers.
752 9.7 Reliability under Various Environmental Influences
1.0
rel. transmission
0.9 POF sample No. 4
0.8
2
0.7
0.6
1
0.5 4
0.4 3
0.3
0.2
525 nm
0.1 590 nm
0.0 650 nm
1.2
rel. transmission POF-No. 1
at 650 nm POF-No. 4
1.0 POF-No. 7
T= 92C/95% r.H.
POF-No. 10
0.8 POF-No. 13
0.6
0.4
0.2
0.0
aging time [h]
Figure 9.115 shows the way in which physical aging effects as a result of the
action of temperature and humidity can impact on the entire attenuation/transmission
spectrum of a polymer optical fiber. It can be clearly seen that the decline in trans-
mission occasioned by aging is not consistent across the entire spectrum
([Daum97]). A sharper decline in transmission can be seen in the zone of the lower
wavelengths in particular. This effect must be given particular consideration when
polymer optical fibers are used in displays and variable traffic signs where strong re-
quirements in terms of the long-term reliability of color transmission need to be met.
9.7 Reliability under Various Environmental Influences 753
0.6
rel. transmission 15 m POF before aging
0.5 after aging T= +85C/85% r.H.
0.4
0.3
0.2
0.1
0.0
300 400 500 600 700 800
wavelength [nm]
Fig. 9.115: Typical spectrum-related changes in the optical transmission of a 1 mm SI-POF
before and after aging caused by high temperature and humidity ([Daum97])
where
aT = acceleration factor (time shift factor)
T = selected aging temperature (K)
Ts = reference temperature (K)
The critical value in this connection is the reference temperature Ts. There is no
single definition for this given in the literature. Generally the value is characterized
in such a way that the reference temperature gives the value at which deliquescence
of the polymer can first be observed. On the strength of the knowledge we have to
date, this phenomenon can well be related to the glass transition temperature TG of
the polymer fibers. Consequently TG is generally substituted for the reference
temperature Ts.
The maximum possible operating temperature Tmax for a given durability tL can
be obtained by conversion from the same equation as:
754 9.7 Reliability under Various Environmental Influences
For the practical assessment of the thermal aging process, this equation is set out
and applied in the following form (DIN ISO 2578:1994) [DIN94]:
tA A e B T
Thus in practice there is a linear connection between the logarithm of the aging
time required to effect a given decline in optical transmission, and the conversion
value of the absolute aging temperature associated with it. On the basis of this
9.7 Reliability under Various Environmental Influences 755
connection, expected results for higher temperatures can be extrapolated on the time
taken to cause breakdown at lower temperatures.
The Arrhenius hypothesis has been used with success in predicting the durability
of electronic components, for example, and has been called in to determine the time
and temperature limits in cases of long-term effects of heat on polymeric materials
(DIN ISO 2578:1994).
10,000
time [h] maximum operation temperature of a given
durability of 15 years: T = +80C +/- 2 K
1,000
88C
89C 90C
100
91C
92C
10
aging time
linear interpolation aging temperature 1/T [1/K]
1
2.7810-3 2.7710-3 2.7610-3 2.7510-3 2.7410-3 2.7310-3
Fig. 9.116: Estimation according to Arrhenius, of the maximum permissible operating tem-
perature for a special 1 mm SI-POF (L = 10 m) without jacket for a given
durability
As with the WLF hypothesis, so here the Arrhenius equation can be used to
estimate either the durability for a given operating temperature, or the maximum
permissible operating temperature for a given durability. Figure 9.116 shows a typi-
cal result of this, determined by following the procedure described in DIN ISO
2578:1994 for a special 1 mm SI-POF without jacket. What is depicted is the esti-
mate of a maximum operating temperature for a given durability of 15 years for a
special polymer optical fiber, which is to be used for lighting purposes.
With reference to the envisaged method for a prediction of durability it must be
stated as a matter of principle, that the results - as is the case with all extrapolatory
procedures - are accompanied by a degree of uncertainty which we should not be
oblivious to. The more unfavorable the relation between the time taken for the test
and the envisaged durability, the more unreliable the prognosis. In dealing with
cable technology one is usually limited with a prediction to a duration of
20,000 hours - 25,000 hours (about 3 years), on the assumption that the constant
operating temperature so determined will either not be reached at all or else only
fleetingly and for short periods. We must likewise take into consideration that the
tests needed here take a great deal of time, calling for some thousand hours of
laboratory use. More rapid testing procedures for POF such as the investigations
based on chemoluminescence for example [Scha99] are only just in the process of
being developed.
756 9.7 Reliability under Various Environmental Influences
optical
1
multiplexer
with light
source and
detector
n
data recording
system
glass vessel
with chemical
heating chamber
Table 9.7 shows an overview of typical test results in testing the resistance to
chemicals of polymer optical fibers. Depending on the chemical and the jacket, we
can specify three classes of substances:
1. harmless,
2. harmful,
3. less harmful.
If a chemical is ranked under the heading harmless, a safe and reliable trans-
mission of signals in normal operating conditions can be guaranteed.
Stress caused by contact with a chemical classed as harmful must be avoided
at all costs. Here further protective measures are absolutely necessary: for instance
a protective tube made of metal, or an additional jacket or protective covering, or
one with higher resistance to chemicals.
With less harmful chemicals, contact over an extended period of time should
be avoided. A short term contact (e.g. dripping, followed by immediate removal)
would appear to cause no lasting damage.
Jacket
Chemical
PE PA
In addition to the material of the jacket, the material of the cable sheath also
plays an important role in regard to the resistance to chemicals of optical fibers on
a POF base. The following Table 9.8 gives an overview of the most commonly
applied materials and their resistance to aggressive substances, cf. also Tables 2.26
to 2.30.
758 9.7 Reliability under Various Environmental Influences
Abbre-
Chemical Properties
viation
resistant to oils, fats, diluted acids and alkalis up to
Polyvinylchloride PVC 50C; special mixtures are resistant to solvents
and fuels
In addition the cables can be better protected, not just against aggressive sub-
stances but against thermal and mechanical stress as well, through the use of
special laser-welded corrugated micro tubes (CMT) as cable sheaths ([Schei00]).
CMTs (Fig. 9.118) are immediately effective as primary protection, and can,
depending on requirements, be made of the following materials: copper, alumi-
num, brass, bronze, steel, or special steel alloys.
A significant amount of stress due to radiation can occur when polymer optical
fibers are installed in nuclear power plant, high-energy physics laboratories, linear
accelerator or synchrotron facilities, or even in medical or industrial radiation
equipment. Admittedly this form of stress has not been very comprehensively
investigated hitherto; at any rate, only a few generally known findings and publi-
cations on this issue are to be found. Special aspects of the resistance to radiation
of POF materials (PMMA, PFMA, P4FFA) have been written up in [Lev94].
Experiments by [Hen93] with polymer optical fibers made of PMMA, some un-
protected, others with jacket, indicate a high resistance to radiation at a radiation
of 60Co with energy doses < 100 krad. Measurements of the optical attenuation at
670 nm and 780 nm show only very slight changes in the attenuation when the
polymer optical fibers are exposed to this type of radiation. Slightly higher
changes in the attenuation result in combination with both high (+80C) and low
temperatures (-40C). With similar tests using POF on a PC base a significantly
higher sensitivity to radiation can be demonstrated.
If polymer optical fibers with jacket are exposed over a long period of time to
sunlight or artificial light with a high ultraviolet element, discoloration (a
bleaching effect, or a change of color) is the result (Fig. 9.119), and in extreme
cases it may even lead to micro-cracks on the surface. Discoloration is not on the
whole connected with any essential changes in the properties of the material. With
the formation of micro-cracks, however, it can lead to an impairment of the
mechanical properties. As the depth to which the UV radiation can penetrate the
jacket is strictly limited, the optical fiber will not be adversely affected.
Unprotected optical fibers (e.g. without protective covering, or the unprotected
fiber end surfaces) that are exposed to high UV radiation tend to go yellow. Their
optical transmission deteriorates, the longer they are exposed to the radiation.
760 9.7 Reliability under Various Environmental Influences
Optical waveguides and fibers have been used for many years to transmit signals.
A number of models have been developed to describe their characteristics. In the
field of optical fibers one differentiates between singlemode fibers (SMF), in
which one eigenwave (mode) per polarization is capable of propagating, and
multimode fibers in which the light can propagate in the shape of different eigen-
waves at varying speeds. How many modes can propagate in a fiber depends on its
diameter compared with the wavelength of the light and the numerical aperture.
The greater both are, the more modes can propagate. The fiber parameter V des-
cribes the fundamental properties of a fiber:
2S a 2 2
V a AN 2S ncore - ncladding
O O
Here stands for the wavelength of the light, a is the core radius of the fiber AN
the numerical aperture, ncore and ncladding describe the refractive indices at the fiber
axis and in the fiber cladding. The greater V is, the more modes are guided. The
number of modes is approximately proportional to V2.
Polymer optical fibers have a large diameter and a high numerical aperture so
that an extremely large number of modes can propagate in them. In the case of a
standard step index polymer fiber with a numerical aperture of AN = 0.5 and a dia-
meter of about 1 mm several million modes can propagate.
The optical signal is distorted and attenuated when it propagates over the fiber.
These effects have to be modeled when describing the signal transmission. They
behave quite differently in different types of fibers. Whereas signal distortions in
singlemode fibers are primarily caused by chromatic dispersion, i.e. the different
speeds of individual spectral parts, the description of dispersion in multimode
fibers is considerably more complex. Not only does chromatic dispersion occur in
them, but also has the generally much greater multimode dispersion. This effect
comes about because the individual eigenwaves here propagate at different speeds.
The signal is split up into the different modes which then arrive at the receiver at
different points in time because of the varying speeds, thus transmitting a distorted
signal. In contrast to chromatic dispersion multimode dispersion is influenced by
different effects which do not exclusively depend on the fiber. The delay between
the modes are determined by the refractive index profile of the fiber and depend
exclusively on the fiber. How the signal power is split up into the individual
764 10.1 Modeling of Polymer Optical Fibers
modes, however, is determined by the coupling of the light into the fiber. For
example, when the light is coupled very narrowly around the fiber axis and almost
parallel into the fiber, then only one individual mode would be excited, i.e. the
entire signal power would be coupled into this mode and would not even notice
the time delay between the individual eigenwaves at all since only one mode is
providing power. The other extreme would be a very wide coupling of the trans-
mission angle and the surface. In this case the power would be divided evenly to
all modes whereby the skews of all modes would play a role. This example clearly
shows that when modeling multimode fibers you not only have to take the fiber
itself into consideration but also the coupling in and out of the light. Furthermore,
modes in polymer fibers do not spread independently of each other but are
coupled to each other. A part of the power can be coupled over to other preferred,
neighboring modes because of impurities and non-ideal interfaces between the
core and the cladding. This mode coupling arises particularly strongly in polymer
fibers which is why current models for the propagation of light in optical fibers
have to be expanded >Whi99@ and >Shi97@. The most important types of fibers will
subsequently be presented and fundamental boundary conditions for their
modeling are given.
Finally, the different approaches of modeling for polymer fibers will be presen-
ted which will include a short description of the procedure, performance and com-
plexity. Since mode coupling or mode mixing is a very specific effect for polymer
fibers and influences the performance of the modeling, the most important
approaches for describing them and their possibilities for integrating them into
existing models for light propagation will be presented.
Polymer fibers are generally multimode fibers with large core diameters and a
high numerical apertures for the easy coupling in and out of the light. They have
good mechanical characteristics as well. There are different types of multimode
fibers which differ in their refractive indices and the rotation-symmetrical profile
of the refractive index in a radial direction. The refractive index profile of a fiber
determines the speed and the time delay of all guided modes. In general, one
differentiates between step index profile fibers (SI fibers) which have a constant
refractive index in the core and a somewhat lower, but also constant refractive
index in the cladding and graded index profile fibers (GI fibers) in which the
refractive index continuously decreases from the core to the cladding and thus
reduces the mode delay.
Graded index profiles can be produced in which the propagation times of the
modes are practically the same. However, producing such fibers takes a lot of
effort, especially with very thick fibers (>Yab00a@ and >Ish96@). Consequently,
there are still in-between forms in which the optimal refractive index profile can
be approximated through many small steps, so-called multistep index fibers (MSI
fibers). The more steps used to approximate the optimal profile, the better the pro-
pagation times of the modes tally with each other. The complexity of the manufac-
turing process, however, is further increased. It is always a compromise between
cost/effort and performance (>Lev99@ and >Irie01@).
a) b) c)
n (r) n (r) n (r)
r r r
-a 0 +a -a 0 +a -a 0 +a
Fig. 10.2: Refractive index profiles of different fibers in a comparison:
a) step index, b) graded index, c) multi step index with 5 layers
766 10.1 Modeling of Polymer Optical Fibers
Light guiding in polymer fibers is based in all conventional optical fibers on the
principle of total internal reflection. The fiber core has a greater refractive index
than that of the surrounding cladding. Basically, the same approaches and methods
can be used for describing wave guidance as with glass fibers. However, polymer
fibers have characteristics which make some approaches difficult or even impos-
sible.
For example, many approaches proceed from very weakly attenuating fibers
which can no longer quite be guaranteed with a basic attenuation of about
120 dB/km at a wavelength of 650 nm.
The greatest limitation when describing the propagation characteristics is surely
the extremely large number of modes which are capable of propagating. In prin-
ciple, all modes of a polymer fiber can be calculated. This, however, requires a lot
of memory, computation time and very good resolution, too. For these reasons
many simplified descriptions have been established in which either the mode
groups are calculated or the light propagation is dealt with the ray theory and addi-
tional wave phenomena outside the ray theory are described. Such hybrid
approaches in particular are often used. The most current procedures for des-
cribing light propagation in polymer fibers will subsequently be described.
tially valid. Nevertheless, calculations based on the scalar wave equation only
show very small inaccuracies in regard to group delay.
Fig. 10.3: Examples for field distributions of the LP0,2- und LP2,2 modes in a graded index
profile fiber
The models based on the solution of the wave equation in the form of a model
solver differ fundamentally only in regard to the solution method and whether or
not you are proceeding from a more computer-intensive vector wave equation or
the more usual scalar wave equation.
In the technical literature solutions for the vector wave equation with the aid of
finite element method (FEM, >Bha00@ and >Liu95@), with finite differences (Finite
Difference Time Domain Method - FDTD, >Xiao06@) and the beam propagation
method (BPM, >Hua93@) are well-known. These are generally used for very small,
mostly singlemode waveguides in which polarization characteristics play a role.
Polymer fibers are quite large and receive the polarization of light for only a few
centimeters. That is why analytical estimates of the scalar wave equation, the so-
called WKB Method and Ray Tracing, are primarily used for the modeling of
POF.
Analytical solutions for the wave equation exist for different refractive index pro-
files such as step index and parable profile. Then you only have to determine the
propagation constants and propagation times of the modes in order to describe the
fiber. In most cases the so-called WKB Method is used with which the propaga-
tion times of very many modes can be calculated efficiently (>Ish05c@ and
>Ohd05a@).
r2 1 l2
r k r dr p - S with : k r k 02 n2 (r ) - E2 -
1 2 r2
Here l stands for the circumferential order of the mode or the skewness of the
ray. With an ever increasing circumferential order the beam path continues to go
along the edge of the fiber while the lowest circumferential order 1 = 0 describes
the so-called meridional rays which intersect the fiber axis. Both integration
boundaries r1 and r2 are the inner and outer reversal radii (caustic). They describe
the region in which the light is concentrated within the fiber. A ray can therefore
never come closer to the fiber axis than the inner reversal radius r1 and also not
further away than the outer reversal radius r2.
Fig. 10.4: The inner and outer reversal radius (caustic). Left: side view on the fiber, right:
view on the end face.
10.1 Modeling of Polymer Optical Fibers 769
N(r) stands for the profile of the group index over radius r.
u a k 02 ncore
2
E2 ; u a E 2 k 02 ncladding
2
Here the first term describes the effective refractive index neff which expe-
riences the guided wave. The second term describes the group delay gr which
requires the light in a material with group index Ncore. This description is equi-
valent to beam propagation in free space with the refractive index neff and the
group refractive index Ncore. We shall see in the section on ray tracing that you can
describe quite well the light propagation in a fiber in which very many modes can
propagate by using the beam model.
770 10.1 Modeling of Polymer Optical Fibers
g
r
2
ncore 1 - 2' for r d a
n2 (r ) a
2
ncladding ncore 1 - 2' ,
2
else
Here the profile exponent is g which determines the steepness of the profile. If
g becomes very large, then the profile approximates more and more that of a SI
fiber. Parameter describes the profile height which stands for the difference in
refractive index between the core and the cladding and also influences the nume-
rical aperture of the fiber. There are exponential profiles as direct solutions for the
group propagation times and indirectly for the bandwidths of such fibers. Fibers
with power-law profiles possess the characteristic that the modes can be put in
mode groups which have the same propagation constant and also similar mode
delay (at least for exponents close to g = 2). The following relationship exists for
propagation constants of the mode groups and their orders:
m
E ncore k 0 1 - 2 '
M
whereby M represents the highest mode group order. The propagation times of the
modes are only then dependent on the propagation constant and no longer on the
circumferential order. Assuming an power-law profile as indicated above, then
group delay can be determined with the aid of the WKB method by differentiating
the propagation constant from the angular frequency (>Mar77@).
2g g 2
' >4 - 2P@ m
1-
Ncore g 2 M
W
c0 1 - 2' m M2g g 2
Here m stands for the order of the mode group and M for the highest mode
group. Parameter P describes the so-called profile dispersion (see e.g. >Pre76@).
ncore O d'
P
Ncore ' dO
The normalized propagation times ' = ( - 0)/0 for different exponents g and
without profile dispersion (P = 0) is illustrated in Fig. 10.5. You can see that the
skews become very small for parabolic profiles with g = 2. For greater exponents
(g > 2) the higher modes are slower and with smaller exponents they are faster.
10.1 Modeling of Polymer Optical Fibers 771
+0.0004
g = 2.1
+0.0002
g = 2.0
0.0000
g = 1.9
-0.0002
g = 1.8
-0.0004
norm. mode group m/M
0.0 0.2 0.4 0.6 0.8 1.0
Fig. 10.5: Normalized propagation time = ( - 0)/0 over the mode group (m/M) for dif-
ferent profile exponents without profile dispersion (P = 0)
1.58
transit time [a.u.] transit time [a.u.]
1.56
1.54
1.52
1.50
1.48
1.46
1.44 E invariant E invariant
1.42
1.40 1.42 1.44 1.46 1.48 1.50 1.40 1.42 1.44 1.46 1.48 1.50
Fig. 10.6: Transit times of meridional rays in MSI fibers with two (left) and 6 layers (right,
[Zub04])
772 10.1 Modeling of Polymer Optical Fibers
whereby Ee represents the electric field at the transmitter and E,l the electric field
of the respective mode. ,l is the coupling efficiency in the respective mode with
propagation constant and circumferential order l.
In addition, the field distributions E,l of all modes are needed, something
which is generally not possible. Rather you try to determine the launching con-
dition and the resulting mode distribution in the fiber from the near and far field of
the transmitter.
SA (t) Km SE t - Wm
m
Here SA(t) and SE(t) are the output or input signal respectively. Counting index
m goes over all mode groups whereby m and m are the share of the total power
and the propagation time of the mode group.
The impulse response h(t) results analogously by accepting a Dirac (unit) im-
pulse as the input signal.
h( t ) Km G t - Wm
m
spectrum of which lies within this bandwidth. You calculate the bandwidth by ob-
taining the frequency response through Fourier transformation of the pulse res-
ponse.
H ( f ) F h ( t )
The bandwidth is then the frequency f3db with which the amplitude of the fre-
quency response has dropped to 50% (3 dB):
H f3 dB 1
2
10.1.4 Ray-Tracing
When calculating the propagation characteristics, a very large number of rays are
generated at the transmitter and each ray follows its own path along the fiber. In
doing so, its attenuation as well as the elapsed propagation time is logged. At the
end of the fiber a histogram about the rays received is drawn up which includes
propagation time, attenuation, location or propagation direction (>Zub02a@). This
procedure is based on tracing very many rays to obtain a statistic and it is here that
you can see the main disadvantage of this procedure: it requires great computing
power and the result approximates the exact solution in step with every further ray
calculated according to the Law of Large Numbers.
Fig. 10.7: Emission characteristics (left) of a light source in the Ray-Tracing model and
simulated output signal in comparison with a measurement result ([Zub04])
In this case L corresponds to the fiber length and is the angle of the ray in
reference to the fiber axis. The greater the propagation angle in regard to the fiber
axis, the longer the path becomes. The ray tracing procedure is then used to calcu-
late the different propagation times of the eigenwaves dependent on the propa-
gation angle in regard to the fiber axis.
L(T) ncore L ncore
W(T)
c0 cos (T) c 0
r
cladding
n(r)
core
&
d dR &
n(r ) n
ds ds
Here s is the length of the ray path and lR describes the radius vector so that its
derivative to s corresponds to the tangent at the trajectory of the ray. In SI fibers
the right hand side of the equation would be absolutely zero so that the propa-
gation direction of the ray does not change.
r cladding
n(r)
core
r cladding
n(r)
core
Fig. 10.10: Splitting of the rays in MSI fibers at the interfaces between the different layers
776 10.1 Modeling of Polymer Optical Fibers
10.1.4.4 Bends
Bent fibers can be dealt with quite easily with the Ray Tracing method. If the fiber
to be calculated is not straight, then the ray will nevertheless still always go
straight in SI fibers. However, the fiber axis and the core-cladding interface will
no longer run straight. The same principle is also true for MSI fibers which do
have more than one interface. It is still valid for these reflections that the angle of
incidence is the same as the angle of reflection, but the direction of the fiber axis
changes in the bends so that the propagation direction of the ray changes in regard
to the fiber axis. At the end of the bend the rays direction of propagation is gene-
rally another one, i.e. mode conversion has taken place. Guided modes can also be
converted into radiation modes in this process. The latter modes can then leave the
core and become noticeable as additional bending attenuation (>Arr01b@ and
>Durr03b@).
propagation direction
before the bend
propagation
direction after
the bend
Fig. 10.11: Description of the mode conversion in the beam propagation model; this effect
will be included in the Ray-Tracing
tion higher modes have more intensity in the vicinity of the core-cladding inter-
face and even in the cladding itself which is why higher losses could be expected
in non-ideal interfaces. In addition, some effects can also be explained as part of
the Ray Tracing with some very intuitive graphic models.
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
normalized mode group x
Fig. 10.12: Measurement of the mode dependent attenuation dependent of the normalized
mode group x = m/M ([Yab00b])
Fig. 10.13: Influence of the excess loss on the far field distribution due to the longer path
length (lengths 1 m, 10 m, 20 m, 50 m and 100 m, UMD launch, AN = 0.50,
core = 120 dB/km).
Fig. 10.14: Influence of the excess attenuation on the far field due to the longer path length
and the reflection losses (UMD launch, AN = 0.5, core = 120 dB/km,
R = 0.9999).
O
dGH (T)
2
2 S ncore cos 2 (T) - ncladding
2
Fig. 10.15: Influence of the excess loss on the far field due to the longer path length, the
reflection losses and the Goos-Hnchen effect (UMD launch, AN= 0.5,
core = 120 dB/km, cladding = 5000 dB/km, R = 0.9999)
Mode mixing is a process which is rather strong in polymer fibers and decisively
influences the propagation characteristics of these fibers (>Ish96@, >Bun99a@,
>Rud95@ and >Sav06@). While an entire mode is transformed in bends into a new
eigenwave or a ray deterministically changes its direction, mode mixing is then
rather a statistical process in which modes exchange power with each other. This
effect generally occurs through irregularities in the fiber, whether they are rough-
ness of the core-cladding interface or impurities in the core material. These irregu-
larities are microscopic and lead to light scattering. This effect can therefore only
be described with statistical means.
The main effects are Rayleigh and Mie scattering which differ in the size of the
scattering centers (>Cam03@). Rayleigh scattering arises through the molecular
structure of matter which is why no material can have perfectly homogenous
properties. Its optical density fluctuates around a mean value which represents the
10.1 Modeling of Polymer Optical Fibers 781
refractive index of the material. These fluctuations are very small and have typical
sizes in the range of molecules (< m). Rayleigh scattering depends on the wave-
length and decreases with greater wavelengths as of the fourth power (~-4). Mie
scattering comes from the fluctuations of the refractive index which has greater
typical lengths that mostly come about because of impurities in the material such
as air bubbles or specks of dust which are large compared with the wavelength of
light. The ensuing scattering has more of an effect on the direction of propagation
of the light and is independent of the wavelength. A typical example for Mie
scattering is the white color of emulsions such as milk which comes about because
of the wavelength-independent scattering of the light.
Fig. 10.16: Mode mixing by scattering on the core cladding interface (left) and inside the
core material (right, [Bun06])
Here Pm, m and cm,n stand for the power in mode m, its mode-dependent
attenuation per unit length and the coupling coefficient between modes m and n.
The first term describes the power drain through losses arising and the second
term the power flows between the modes which on the one hand are proportional
to the coupling coefficient and on the other proportional to the difference between
the guided power of both modes. The system thus described makes sure that there
is a power flow of modes which guide much power up to modes which are only
stimulated a little bit (>Kahn92@, >Sav02a@ and >Sav02b@). The final state aimed for
would be a uniform distribution of the power over the modes. However, since the
modes incur different attenuations, the stationary state which the system aims for
is more similar to a Gaussian distribution with which the higher order modes can
guide less power.
The coupling coefficients which describe the coupling between modes can
either be described by analytical attempts which are based on observations of
mode overlapping (>Djo00@, >Djo04@, >Kov05@ and >Oha81@) or are defined in a
more phenomenological manner. Examples of such phenomenological descrip-
tions could be models for Mie and Rayleigh scattering or approximations based on
measurements.
Fig. 10.17: Application of the mode coupling matrix between short peaces of ideal fiber
The power flow equation shown above for a mode represents one part of an
equational system which can be described concisely with the aid of a matrix multi-
plication (>Kru06a@).
P ( z 'z) M ( 'z) P ( z)
Here the components of the power vector p(z) correspond to the power distri-
bution over the modes at location z while M(z) is the mode coupling matrix.
modes effectively interact with each other so that a simpler description of the
mode coupling suffices (>Kit80@). Investigations have shown that neighboring
modes, i.e. those with similar propagation constants primarily show strong mode
coupling. Analytical observations have led to the conclusion that the strength of
the mode coupling decreases with the fourth power of the difference. From these
observations Gloge developed the diffusion model which solely differentiates bet-
ween mode groups and only describes the mode coupling of neighboring mode
groups (>Glo72@). If you also consider the fact that higher mode groups also con-
tain more modes and that these are stimulated almost evenly within the mode
group then you obtain the following description from the equation above:
d Pm
m - m Dm Pm m c m Pm 1 - Pm (m - 1) c m -1 Pm 1 - Pm
dz
As already described above, the first term describes the power dissipation of the
mode group m by attenuation. The second term stands for the coupling with the
neighboring, higher mode group m + 1 and the last term for the coupling with the
neighboring, lower mode group m - 1. If you now assign a propagation angle to
each mode group, as is possible in a step index fiber, and assume that very many
modes are capable of propagating so that the individual propagation angles only
differ very slightly, then you can sum the last two terms to a derivative and get:
dP 1 w wP
- D (T) P (T) 'T2 T d ( T)
dz T wT wT
Here d() describes the mode coupling in the form of a diffusion constant
which can indeed be dependent on the angle. You can take the mode-dependent
attenuation (), as described above, into account through the additional path. This
results in
1 D T2
D(T) D D 1 | D 1
cos T cos T 2
With this approximation and approach, i.e. that the coupling constant between
the modes is constantly D, then the diffusion equation can be written in the
following, frequently used form.
dP D w wP
- A T2 P T
dz T wT wT
This equation can be solved numerically for different launch conditions in order
to determine the transition to the equilibrium mode distribution. The latter behaves
as a stationary solution for dP/dz = 0 and in the case of equilibrium you get a
Gaussian distribution the width of which gets wider with the increasing diffusion
constant D and lower attenuation A (>Jia97@ and >Zub03@).
784 10.1 Modeling of Polymer Optical Fibers
z=1m z = 20 m z = 50 m
rel. power rel. power rel. power
1.0
0.8
0.6
0.4
0,2
0.0
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
D [] D [] D []
analytical
EFDM
Fig. 10.18: Examples for far field distributions, calculated using the diffusion model for
four different launching conditions after lengths of 1 m, 20 m and 50 m fiber.
The comparison shows a good agreement with the theoretical results (see
[Djo00])
0.4 1m 0.4
10 m
0.2 30 m 0.2
50 m
0.0 0.0
0 5 10 15 20 0 2 4 6 8
D [] t [ns]
Fig. 10.19: Calculated far field distributions (left) and impulse responses (right),
using the split-step algorithm for POF with AN = 0.5, core = 50 dB/km,
cladding = 50,000 dB/km, D = 7 10-4 rad2/m, ([Bre06])
Fig. 10.20: Modeling of the mode mixing using the Ray-Tracing: beam propagation und
random change of the propagation direction during entering (1) and exiting (2)
the fiber, in the core after a characteristic length (4) and at the reflection on the
core cladding interface (3), ([Zub04])
The different kinds of dispersion that limit the bandwidth of optical fibers have
been summarized in Chapter 1. The polymer fibers used today are generally multi-
mode fibers so that wave-guide dispersion and polarization-mode dispersion can
be neglected.
This leaves mode dispersion and chromatic dispersion as the relevant processes
to be considered. In [Bun99a] a comprehensive investigation of the bandwidth in
SI-POF is undertaken. The work looks at different mechanisms, making the
following basic assumptions:
Due to the large number of modes (approx. 2.4 million for 1 mm SI-POF at
650 nm) it is assumed that the angles are of continuous distribution.
All calculations were carried out under launching conditions that only depend
on the angle.
The calculations were based on uniform mode distribution (UMD) launch -
constant far field over the range of guided rays.
Fixed values were entered for the attenuation in the core and cladding.
No inhomogeneities in the core diameter, NA or the geometry of the core-
cladding interface were taken into account.
The following processes were investigated:
Geometric beam propagation in a cylindrical wave guide
Losses due to homogenous attenuation in the core
Additional losses due to differences in propagation paths
Additional losses due to core-cladding interface
Goos-Hnchen effect
Mode coupling
Effect of leaky waves
10.2 Examples for Simulation Results 787
Figure 10.21 shows the result of a calculation for POF with an NA of 0.40 and
0.50. The values for basic attenuation were entered at 120 dB/km and 220 dB/km.
A significant effect on the result came from the attenuation of the cladding
material that was entered here with 50,000 dB/km and 65,000 dB/km.
Above 30 m, the effect of mode related processes significantly cuts in and leads
to an increase in the bandwidth of the POF.
bandwidth [MHz]
500
basic attenuation
200 120 dB/km, AN=0.4
cladding attenuation
100
50,000 dB/km
50 basic attenuation
220 dB/km, AN=0.5
20 cladding attenuation
65,000 dB/km
10
5 10 20 50 100 200
length [m]
Figure 10.22 shows a simulation that takes account of the effect of mode coup-
ling. The mode coupling constants correspond in principle to the reciprocal of the
coupling length, here that would be 30 m, 300 m and f).
bandwidth [MHz]
1000
Jf = 0.033/m
Jf = 0.0033/m
100
10
no coupling
length [m]
1
1 10 100 1000
Fig. 10.22: Influence of mode coupling to the bandwidth according to [Bun99a]
788 10.2 Examples for Simulation Results
The calculations in [Bun99a] for fibers with a small diameter provided particu-
larly interesting results, as demonstrated in Fig. 10.23. Although current data
transmission relies almost exclusively on fibers that have a 1 mm core diameter,
significantly smaller diameters are also of interest where parallel data links or
multi-core fibers are being used (see Chapter 3). Theoretically, in a step index
fiber the diameter has no effect on the bandwidth. However, in practice the diame-
ter does play a role for the effect of the mode related process. Both mode-related
attenuation as well as mode coupling are primarily determined by the core-clad-
ding interface. Reducing the diameter increases the number of reflections and
thereby the effect of these mechanisms. Also, the wave guiding of leaky beams
changes.
bandwidth [MHz]
1,000
POF-NA = 0.30
500 POF: 50 m
POF: 100 m
20
4 6 8 10 20 50 100
length [m]
Fig. 10.23: Bandwidth for POF with different core diameters ([Bun99a])
Figure 10.25 shows the bandwidth and attenuation for fibers with an AN = 0.50
and AN = 0.30 for core diameters between 100 m and 1,000 m, both again cal-
culated on the basis of UMD launch. In the multi-core fibers available today the
single core diameters are approximately 140 m. Based on theoretical calculations
one can expect a doubling of the bandwidth while, however, the attenuation is also
significantly increased.
250 25
200 20
POF AN = 0.30
POF AN = 0.30
150 15
100 10
POF AN = 0.50 POF AN = 0.50
50 5
0 0
100 200 500 1000 100 200 500 1000
core diameter [m] core diameter [m]
Fig. 10.25: Bandwidth and attenuation for different fiber NA and diameter
The effect of mode related processes becomes even clearer in Fig. 10.26, which
shows the simulated far field distributions for a 50 m thick SI-POF for lengths of
up to 50 m.
790 10.2 Examples for Simulation Results
rel. intensity
1.0
launch point
0.8 10 m
0.6 20 m
50 m
0.4
0.2
0.0
-25 -20 -15 -10 -5 0 5 10 15 20 25
propagation angle []
Fig. 10.26: Far field distribution of a 50 m thick POF ([Bun99a])
After 20 m the far field width has already dropped to , resulting in a band-
width increased by a factor of 4. When light is launched from the start at a small
angle, the bandwidth advantage is maintained without too great an increase in
losses compared to the 1 mm POF.
The qualitative descriptions of theoretical models available to date coincide
very well in describing the behavior of SI-POF. There is not yet a universally
applicable model available for making a quantitative estimate. Finding such a
model is the aim of a work group of the European FoTON project which was
formed at the end of the year 2000 (for further information see www.pofac.de).
In a second step a model had to be created which calculates the bending losses
from these two input parameters. The assumption that the losses could simply be
added on, i.e. the bending losses could be modeled as linear processes, gave the
model its name. On the other hand this assumption ignores the influence of pre-
vious bends on the power distribution in the fiber and therefore on the following
bending losses. You get more or less an upper estimate of the losses occurring.
Measurements of the bending losses on MOST fibers were conducted in depen-
dence on the bending radius and bending angle. All measurements took place
under equilibrium mode launch. The results are fitted to a simple model and
applied to a simulation model.
Some of the results were presented at the 2003 POF Conference in Seattle
(>Bun03b@) and were subsequently compared with other simulation results on a
ray tracing basis.
First the attenuation at a bend with different bend angles was determined
(Fig. 10.28). The bending radius was uniformly 20 mm and the bend was made
once at the beginning and once at the end of the fiber. In the latter case somewhat
higher losses resulted since several higher modes are come about in the fiber
through mode coupling.
The measurement curves show that the bending attenuation increases almost
linearly with the bending angle. However, the curves do not intersect the y-axis at
exactly zero. The assumption in this model is that in a bend losses occur at the
transition between the straight fiber in front of and behind the bend and the bent
piece of fiber.
792 10.2 Examples for Simulation Results
Figure 10.29 shows the dependence of the bending losses on the bending
radius, for measurements at different positions in the fiber. The coupling-in NA
was adjusted here at 0.34, a value close to the equilibrium mode distribution.
2.5
bending loss [dB]
after 5 m
after 10 m
2.0
after 15 m
after 20 m
1.5
AN = 0.34
1.0
0.5
0.0
0 5 10 15 20 25
bending radius [mm]
The losses for smaller bending radii increase approximately inversely propor-
tional with r (See Chap. 2 for other examples.). How big the influence of the nu-
merical aperture at the coupled-in location actually is on the bending losses is
illustrated in Figs. 10.30 and 10.31. For radii of 5 mm to 25 mm the bending
losses were determined after different fiber lengths for the launchin NA up to 0.65
(overfilled).
5.5
bending loss [dB]
5.0
4.5
4.0
after 5 m
3.5 after 10 m
3.0 after 15 m
2.5 after 20 m R = 5 mm
2.0
1.5
1.0
0.5
launch-NA
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Fig. 10.30: NA dependent bending loss at 5 mm bending radius (360 bend)
10.3 Measurement and Simulation of Bandwidth of PF-GI-POF 793
0.6
R = 25 mm
0.5
0.4 after 5 m
after 10 m
0.3 after 15 m
after 20 m
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
launch NA
With smaller bending radii NAs up to 0.30, i.e. approximately up to the value
of the equilibrium mode distribution, are relatively problem-free. In combination
with tight bends coupling in with large angles leads to high losses.
The measurements show that the bending losses primarily depend on the mode
distribution. Very exact calculations of the bending losses on a given link without
exactly knowing the mode distribution of the transmitter are in principle not
possible. To compound matters the fibers from different manufacturers have
mode-dependent losses which also find expression in the bending behavior.
The greatest error, however, lies in the fact that every tight bend also changes
the mode distribution itself. Subsequent bends thus actually acquire another
launch NA. A possible expansion of this procedure is therefore the so-called non-
linear model in which the bends in their combinations are taken into account. The
model parameters here too were calculated from experimentally established data.
The exact knowledge of the bending losses allows a guarantee of the power rating
in construction, even with tight bends, i.e. optimal utilization of the components at
hand.
Graded index profile glass fibers can completely eliminate multimode disper-
sion only when the index coefficient amounts to exactly 2 and the chromatic dis-
persion can be ignored. In reality both requirements cannot be met so that different
propagation times still always appear for the modes. These differences in propa-
gation time are measured as differential mode delay (DMD). With GI fibers a
small spot of light is coupled in at different locations over the fiber cross-section
and the propagation time of a short impulse is measured.
In SiO2 GI fibers practically no mode coupling arises (the mode coupling co-
efficient is 1.5 10-4 m-1). The consequence of this is that the individual modes
clearly run apart after long fiber lengths and are visible as peaks in the impulse
response. An example of a measured impulse response (1.1 km of a 50 m
GI-GOF) is shown in Fig. 10.32.
signal [a.u.]
1.0
0.8
50 m MM-GOF
length 1,100 m
0.6
0.4
0.2
0.0
0 1 2 3 4 5
time [ns]
After a link of a good kilometer the different mode groups have recognizably
run apart and form individual maxima in the pulse response. The launch of only
one mode group can, however, significantly increase the possible capacity.
A very much larger mode coupling arises in PF-POF which is stated in >Ral06@
as being > 1.5 m-1, i.e. at least four times the order of magnitude above the value
for silica glass fibers. A value of 10 m-1 has been established in >Ral07@.
Figure 10.33 shows the influence of such a large mode coupling on the pulse res-
ponse of a GI fiber in a simulation.
10.3 Measurement and Simulation of Bandwidth of PF-GI-POF 795
Fig. 10.33: Simulated pulse responses for different mode coupling values ([Ral07])
If the mode coupling lies in those ranges typical for glass fibers, then the peaks
belonging to the different mode groups will be somewhat rounded, but the width
of the pulse response will remain more or les constant. With very strong coupling
a significantly narrower impulse, almost in Gaussian form, will arise. The reason
for the lower pulse broadening is that individual modes can no longer propagate
constantly over the entire fiber length at maximum or minimum speed, but are
interchanged again and again on their way. The influence of growing mode coup-
ling on the maximum impulse broadening (DMD) with different profile coeffi-
cients is shown in Fig. 10.34.
DMD [ps]
400
mode coupling length
for PF-GI-POF:
10 m to 100 m
320
240
D = 1.9/2.1
160 D = 1.9
D = 2.0
D = 2.1
80
In addition to fibers with profile coefficients of 1.9, 2.0 and 2.1 a fiber with a
refractive index shape with different profile coefficients inside and outside was
viewed. With a small mode coupling you see a strong influence of the deviation of
the profile coefficients from an optimal 2.0. With greater mode coupling the influ-
ence is diminished considerably and the skews are generally reduced.
This result does not only explain the comparably large bandwidth of the
GI-POF compared with glass fibers with identical index profiles, but also explains
the independence of the bandwidth and the DMD of the launching conditions.
Measurements of the pulse response for different launch positions on a 200 m long
PF-GI-POF (core: 50 m; Chromis Fiberoptics) is shown in Fig. 10.35.
offset [m]
35
30
25
20
15
10
5
0
-5
-10
-15
-20
-25
-30
We have already pointed out in Chapter 4.4 (receivers) and Chapter 6 (systems)
that the capacity of large area photodiodes are generally viewed as the limiting
factor for the use of large core fibers with high data rates. Many experts were
therefore surprised when several years ago the transmission of several gigabits
with thick fibers and the corresponding large area diodes was stated in a series of
published reports (see Chapter 6.3). Had all the experts been wrong?
The reason for this misunderstanding lies in some fundamental differences bet-
ween photoreceivers in optical telecommunication engineering so far and in POF-
specific solutions. The influence of substrate resistance on the function of the
receivers has for the most part been overlooked so far. In order to achieve high
bandwidths despite the high capacity of photodiodes, receiver designs with low
input resistance have to be chosen which means increased importance for the sub-
strate resistance of the photodiodes. In a joint project among Schott Mainz, the
Fraunhofer Institute for Integrated Circuits Erlangen, DieMount Wernigerode and
the POF-AC different large photodiodes (diameter between 250 m and
1,600 m, produced by CIS Erfurt) are currently being investigated in combi-
nation with different receiver circuits. By modeling the receivers, the parasitic
parameters can then be determined as exactly as possible (>Skl07@).
798 10.4 Simulation of Optical Receivers and Large Area Photodiodes
The different diodes have substrate layers about 300 m thick. With a diameter
of 1,300 m, for example, and a specific resistance of the silicon of 50 :cm a
bulk resistance of the substrates of about 113 : results.
This is greater than the typical input resistance of a HF amplifier. If you reduce
the size of the photodiode significantly, the capacitance will indeed drop, but at
the same time the substrate resistance will increase so that little changes at the RC
time constant.
A simple electrical equivalent circuit diagram was used for the simulation of
the measured impulse responses (shown in Fig. 10.36). The photodiode has been
reproduced as the ideal current source with parallel capacitance, a relatively high
insulation resistance and direct-axis inductance. In addition, there is the substrate
resistance as well as the serial resistance and the input capacitance of the follow-
ing amplifier stage.
RB LS
Iph CPD CP
Rp Rload
25 :
50 M:
amplitude [V]
0.25
simulation
0.20
0.15
0.10
0.05
0.00
measurement t [ns]
-0.05
0 1 2 3 4 5
Fig. 10.37: Simulated and measured impulse response for a 1,300 m large photodiode
Table 10.1 illustrates the parameters with which three of the different photo-
diodes used so far were simulated.
In this model the photodiode capacitance changes proportionally to the area, the
parasitic influences, however, are in a similar order of magnitude. Advantageous
with larger diodes is the substrate resistance which clearly drops. These results
also explain quite well the relatively slight dependence of the measured bandwidth
of the receiver from the reverse voltage and the diameter of the photodiode used as
shown in Fig. 10.38.
800 10.4 Simulation of Optical Receivers and Large Area Photodiodes
1000
BW3 dB [MHz]
900
PD:
800
700 850 m
1,300 m
600
1,600 m
500
400
UPD [V]
300
0 4 8 12 16 20
Fig. 10.38: Dependence of the measured receiver bandwidth from the photo diode size and
reverse voltage
The following important rules ensue when using thick fibers at high data rates:
In addition to the low capacitance you have to above all pay attention to the
low substrate resistance in the production of the photodiodes.
High bandwidths can be attained mainly with receiver concepts which make
small input resistance possible.
The size of the photodiode only plays a minor role. As a rule, a larger photo-
diode allows a more efficient coupling of the fiber so that there is all told an
improvement over a small photodiode.
The dependence of the receiver bandwidth on the bias voltage is relatively
small. Should the available bandwidth be completely used, then you can also
work with low voltages.
Fig. 10.39: Data transmission with a 1.6 mm photo diode, 12 V reverse voltage, 1 Gbit/s
over 50 m OM-Giga
10.4 Simulation of Optical Receivers and Large Area Photodiodes 801
Proceeding from this knowledge, the transmission of a 1 Gbit/s data rate was
realized using a receiver with a 1,600 m large photodiode. The eye diagram after
a 50 m fiber link (OM-Giga) is shown in Fig. 10.39.
The eye is completely open although no equalizer was used. The results are
hardly worse than the results achieved so far with 800 m photodiodes. As Figure
10.40 shows, error free transmission is even possible with a photodiode voltage of
only 3.3 V.
Fig. 10.40: Data transmission with a 1.6 mm photo diode, 3.3 V reverse voltage, 1 Gbit/s
over 30 m OM-Giga, world wide first demonstration!
A reduction in the eye opening can clearly be seen. However, using a passive
equalizer could compensate for this without any problem. This equalizer should be
considerably less technically complex than generating a high photodiode voltage
with a charge pump.
If a receiver with a higher input resistance is used then the diode capacitance
will have a stronger effect which will be investigated in the next phase of this
project.
11. POF Clubs
This chapter is intended to provide a look at the international scientific and techni-
cal activities in the field of polymer fibers. Many established groups have in the
meantime become aware of this technology and have been dealing with it. A
detailed treatment of all these aspects is beyond the scope of this book. We shall
therefore only present the most important groups and events which have to do
with POF.
Of the interest groups existing today in the field of polymer optical fibers the
Japanese POF Consortium can look back on the greatest amount of activity. It was
founded in 1994 and has been led since by Professor Yasuhiro Koike who has
achieved international recognition for his numerous publications, especially on
graded-index profile polymer optical fibers. A total of approximately 70 institutes
and manufacturers are represented in the Japanese POF Consortium. Table 11.1
shows its status in 1999 ([Pol99], [Koi96d]).
In the U.S.A., the IGI Company in Boston can be viewed as the most important
representative of POF-interested parties. IGI regularly publishes POF News and
sells different studies on developments within the telecommunications field, inclu-
ding polymer optical fibers. IGI organizes the annual POF World events which are
intended primarily for commercial users. The current managing director is Paul
Polishuk who is also the head of the POF Interest Group with members world-
wide.
Of international importance were two consortiums in the U.S.A. which have
been working in succession for several years on the development of polymer opti-
cal fiber systems, primarily for use in avionics.
The HSPN Consortium (High Speed Plastic Network) was founded in 1994, the
aim of which, among others, was the development of 650 nm VCSEL by Honey-
well. PMMA-based GI-POF were to be developed concurrently. Both products
were able to be demonstrated under laboratory conditions, but have not yet been
developed for series production. Figure 11.1 shows the structure of the HSPN
project. At the end of the project in 1997 the successor organization PAVNET
(Plastic Fiber and VCSEL Network) was founded. The latest member was Lucent
Technologies (see Fig. 11.2). The goals are:
PF-GI POF with < 60 dB/km at 500 to 2,000 nm
Expansion of the temperature range to +125 C
use of existing VCSEL technology at 850 and 1,300 nm
622 Mbit/s over 30 m, later 2,500 Mbit/s over 100 m
11 POF Clubs 805
Packard-Hughes
I n t e rc o n n e c t
Program Management
Packard-Hughes Interconnect
Program Management &
Administration
Project Integration
This group has not been internationally active over the past years. The reasons
for this may be problems with the technologies for red VCSEL and the GI-POF. In
contrast to the Japanese approach, Boston Optical Fiber used a Teflon-based mate-
rial, but so far, the losses have still been in the vicinity of some 1,000 dB/km (cf.
[Ily00]). During the some years in the late 90ies Mitel in particular has published
in the field of VCSEL.
The greatest amount of work on the North American continent was published
by Lucent Technologies at the beginning of this decade. In the summer of 2000
they announced their own GI-POF production on a CYTOP-basis [Luc00]. A few
years ago the production of PF-GI-POF was detached to a subsidiary and has since
been carried on by OFS. A continuous production procedure already described in
Chapter 2 has in the meantime been developed by Chromis Fiberoptics.
In the meantime the POF Trade Organization (POFTO) has developed out of
the former POF Interest Group into an international trade and information plat-
form. Current activities were presented in [Pol06a]. Members of POFTO are indi-
cated in Fig. 11.3.
806 11 POF Clubs
The French POF Club was founded as early as 1987. The director of the group for
many years was Michel Bourdinaud. The first international POF conference took
place in Paris in 1992 and was organized by IGI Europe. By 1994, approximately
200 members were registered in the French Plastic Optical Fibre Club (FOP Club,
[Bou94]). It is part of the French Optical Society (SFO) and is supported by the
French Atomic Energy Commission. The background to this involvement is the
idea of using scintillating polymer optical fibers for detection of elementary
particles (e.g. [Far94], [Des94], [Bar96]).
Membership in the FPO is free of charge. Financing is taken care of by the SFO
(French Optical Society), the CEA (Commissariat lEnergie Atomique) and
small contributions at conferences. Participants include representatives from
universities, research institutions, industry and government or military institutions
respectively. There are 50 to 80 participants at the bi-annual meetings. In 1994,
the FOP published the first comprehensive book on polymer optical fibers
[FOP94] which has been available in an English translation [FOP97] since 1997.
In Germany, there has been considerable interest in POF for some time now, in
particular through the activities of the chemical industry (Hoechst, Bayer). Until
1996 there was no national interest group in the field. The creation of just such a
group goes back to a meeting of various German participants at the POF Confe-
808 11 POF Clubs
rence in Paris (October 1996). After some preliminary preparations it was decided
by the sub department 5.4 Communication Cable Networks of the Information
Technology Society (ITG) within the Association for Electrical, Electronic &
Information Technologies (VDE) to found the sub committee (FG) 5.4.1 Polymer
Optical Fibers on December 3rd, 1996. The head of the sub committee and chair-
man of the ITG-FA 5.4 since 1999 has been Olaf Ziemann of the POF-AC Nrn-
berg (all information on the sub committee and the sub department can be found
under www.pofac.de). The 24 meetings held so far are as follows (see Fig. 11.4):
16.01.1997 Berlin (Federal Institute for Materials Research and Testing, BAM)
12.05.1997 Nuremberg (Nuremberg University of Applied Sciences)
12.04.1997 Cologne (in cooperation with the cable conference)
28.04.1998 Darmstadt (Technology Center of the Deutsche Telekom)
05.-08.10.1998 International POF Conference in Berlin (BAM)
10.12.1998 Ulm (University of Ulm in cooperation with Daimler/Chrysler)
20.04.1999 Jena (Fraunhofer Institute for Optics and Fine Mechanics)
16.09.1999 Stuttgart (Lapp Kabel GmbH)
09.03.2000 Mnchengladbach (Alcatel Cable)
19.10.2000 Potsdam (University of Potsdam)
27.03.2001 Gelsenkirchen (University of Applied Sciences)
24.10.2001 Giessen-Friedberg (University of Applied Sciences)
24.04.2002 Leipzig (Telekom Univ. of Applied Sciences)
10.07.2002 Munich (BMW)
10.12.2002 Colonge (in cooperation with the cable conference)
26.03.2003 Offenburg (University of Applied Sciences)
25.06.2003 Munich (during the Laser 2003 exhibition)
05.11.2003 Mainz (IMM)
09.03.2005 Erfurt (DieMount, CIS and IMMS)
27.-30.09.2004 Nuremberg (International POF conference)
08.03.2005 Wetzikon, Switzerland (Reichle & De Massari)
21.11.2005 POF-AC Nrnberg
12.05.2006 Oldenburg (BFE)
25.10.2006 Munich (on the Systems 2006 exhibition)
17.07.2007 Erlangen (Fraunhofer Institute for Integrated Circuits)
17.09.2007 Berlin (POF-Day on the ECOC 2007 exhibition)
Between 30 and 130 visitors attend the respective meetings. The large number
of people attending past meetings attests to the increased interest in polymer opti-
cal fibers in Germany. In Europe, Germany is at present the country with the most
POF research and application work which was also mirrored in the number of
German papers given at international conferences Plastic Optical Fibers & Appli-
cations since 1992 (Fig. 11.5).
The highlight in the work so far of the ITG Sub Committee and the POF-AC
Nrnberg was the organization of the 13th International POF Conference in 2004
in the Nuremberg Conference Center. In addition to the scientific session program,
a trade exhibition with over 30 exhibitors was held for the first time.
11 POF Clubs 809
12. SCM
11. SCM
6. SCM
17. SCM
23. SCM
4. SCM
POF2004: CCN
7. SCM
2. SCM
15. SCM
20. SCM
5. SCM
13. SCM
19. SCM 22. SCM 16. SCM
Fig. 11.4: Recent meetings of the ITG sub committee 5.4.1 (as 2007)
110
100
Papers from Germany
90
Papers other countries
80
70
60
50
40
30
20
10
0 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Fr. NL Jap. USA Fr. USA Ger. Jap. USA NL Jap. USA Ger. HK Kor.
In October 2000, the POF-AC (The Polymer Optical Fiber Application Center)
was founded in Nuremberg as an institute of the Nuremberg University of Applied
Sciences. The project receives financial support amounting to 2.3 Mio from the
High-tech Offensive of the State of Bavaria. The institutes goals are:
providing support when introducing the new technology
offering measuring equipment for characterizing POF
carrying out contractual investigations and developments
setting up of demonstration and pilot systems
database for all POF relevant information
simulating components and systems
maintaining close contacts to universities and other research institutes
11 POF Clubs 811
trainings
characterising connector
techniques
pilot projects
POF illumination
optoelectronics systems
applications
coupler
tools splitter
measurement interface
devices cards others
Since 2006, the institute has been financed exclusively through industrial pro-
jects and sponsored research projects. Between 2001 and 2005 approximately 200
individual projects were carried out. The projects have been divided among diffe-
rent working areas and are shown in Fig. 11.7.
passive components
training/consulting
active components
sensors
simulation
data communication
devices
general optics
fiber measurements
Fig. 11.7: Splitting of the working areas at the POF-AC
All work dealing with lighting technology and non-fiber measuring techniques
fall into the category of general optics. With fiber measurements, investigations of
the optical characteristics as well as other factors, e.g. long-time and climate
measurements, are taken into account.
812 11 POF Clubs
When the first large-scale use of POF cables for data communications in auto-
mobiles on the basis of the MOST specifications was introduced as of the autumn
of 2001 was undertaken, there also resulted a great need for exact test specifi-
cations for measuring optical and transmission parameters of fibers and prepared
cables respectively as well as the testing of mechanical characteristic values and
environmental resistance.
Between 2002 and 2006 the comprehensive recommendation (VDE/VDI 5570:
testing connectorized and non connectorized plastic light waveguides) was worked
out with the cooperation of about 20 companies and institutes. The contents of this
recommendation are included in Section 7.3.1. In the meantime representatives of
this group work mostly as part of the DKE and the working out of standards for
home networks.
Compared to other optical technologies POF still represents only a relatively small
segment which is why POF products are very difficult to find in the best-known
trade directories. Since September 2005, the POF-AC has worked out an Internet-
based overview of manufacturers. The project is being conducted by the Bavarian
competency network Bayern Photonics and is supported by the BMBF.
Figure 11.9 shows the surface of the search mask.
The European Union has promoted a series of projects over the past few years
which have also focused on POF technologies. These included:
IO: Interconnect by Optics (IST-2000-28358) with the participants Alcatel,
Optospeed, Avalon, Helix, FCI, Nexans, RCI, PPC, LETI. The goal was the
development of parallel optical connections for the direct connection of
CMOS switching circuits and on PC board level including optical backbones.
Agetha: Amber/Green Emitters Targeting High Temperature Applications
(IST-1999-10292) with the partners CRHEA, CNRS, Thales, University of
Madrid, Trinity College Dublin, University of Surrey, Infineon, BAE Systems
and the Institute of Electron Technology Warsaw. The goal was the develop-
ment of 510 nm and 570 nm RC-LEDs with data rates up to 500 Mbit/s and
operating temperatures up to +120C.
Home Planet: Home Plastic Fiber Networks based on HAVI (IST Optimist)
with the participants NMRC, Nexans, Firecomms and Grundig. The goal was
the development of IEEE 1394 S200 and S400 systems for home networks
according to the HAVI Standard.
The POF-ALL project has been promoted since the beginning of 2006 and is
geared directly toward the development of POF systems. The official press state-
ment made at the start by project management is presented here again for eluci-
dating the project:
Ask anybody in Europe what broadband access means, the answer will most
likely be ADSL or cable (Fig. 11.10). xDSL technologies actually dominate
the broadband offer because of legacy telcos existing copper-based infrastruc-
tures, and cable modem technologies likewise in countries with a high CATV
cabling density. The telecom bubble at the end of last century made clear how
hazardous it can be to invest in a new and innovative (e.g. optical) infrastructure.
Even the few local operators that survived those days huge CAPEX are now
recovering under the umbrella of copper legacy infrastructure.
30
subscibers/100 inhabitants DSL cable others
25
20
15 OECD average
10
0
Kor. Dn. Swi. Fin. Nor. Jap. USA st. Aust. Ital. Sp. Ung. Pol. Slov. Mex.
NL Isl Can. Bel. Swe. UK Fra. Lux. Deu. Por. NZ Irl. Tsch. Trk. Gri.
Still, the explosion of peer-to-peer (P2P) services is rapidly heading for a band-
width bottleneck. At the end of 2004, 60% of all internet traffic was P2P-based;
due to its symmetrical nature, on average 80% of upstream capacity is consumed
by P2P daily (source: CacheLogic Peer-to-peer in 2005, Fig. 11.11).
Fig. 11.11: P2P flooding: 60% to 80% of existing bandwidth is being used by P2P
The major advantage of POF is that anyone can install it in 30 seconds with
common tools: a pair of scissors to cut it, a stripping tool to remove the jacket and
a crimping tool to connect it (Fig. 11.13).
Some devices even work without connector, by just cutting the cable with a
blade and inserting it - remarkably easier than handling and terminating a glass
fiber cable.
POF cables are extremely thin and flexible and can be laid down in electrical
conduits or alongside walls. Moreover, POF uses visible light instead of infrared,
avoiding eye-safety related issues and pioneering a revolutionarily simple test
procedure: if you see light coming out the fibers tip, then the system works.
For some years now multiple POF activities have also been carried out in South
Korea. A series of private investments have been made in addition to publicly pro-
moted projects. The Optimedia Company is such an example which under the
direction of Prof. Park has developed the OM-Giga fiber already described in
several sections.
The Korean POF Club KPCF was founded in February 2004 by a number of
interested partners. The members are companies, research institutes and private
individuals interested in POF technology. In 2006, the KPCF already had 20 mem-
bers from major industries, medium-sized companies and scientific institutions.
The companies represent both fiber and cable manufacturers as well as the produ-
cers of communication technology equipment.
The activities of the KPCF include:
Cooperation in the standardization for POF use.
Cooperation in governmental regulation in order to represent polymer fibers
in the corresponding committees and to support the development of new poly-
mer fibers.
Conducting seminars and workshops in order to popularize POF technology
and its applications in Korea.
Offering training programs for certifying POF specialists (in cooperation with
the Korean Society of Information and Communications Technology Engi-
neers).
The high light so far has been the hosting of the 15th International POF Confe-
rence in Seoul in September 2006. The POF-AC Nrnberg in the meantime looks
back on cooperative work with the Optimedia Company for several years now.
Korean work was presented at the meeting of the ITG sub commitee 5.4.1
Polymer Optical Fibers in Oldenburg in May 2006 ([Park06a] and [Park06b]).
South Korea today is among those countries with the greatest density of broad-
band connections in the world. Figure 11.14 from [Eng05] shows for example the
share of broadband-equipped households in a worldwide comparison. In 2004,
predominantly ADLS and HFC connections were installed; in 2007, however,
VDSL was dominant. In the meantime there is an ever increasing transition to
glass fiber connections with at least 100 Mbit/s. In this respect the building net-
works today in Korea play a very great role.
The Korean Ministry for Communication and Information started its so-called
IT 8-3-9 Strategy in 2004 which concentrates on the promotion of new services,
infrastructure and new growth markets. A constituent element of the new services
is also the home network. By 2010, 20 million users are to be provided with
connections from 50 Mbit/s to 100 Mbit/s.
An essential part of the strategy is the broadband certification of the apart-
ment buildings. Of the four possible classes the two highest platinum and first
require the connection of all apartments with optic fibers. Both glass MM fibers as
well as POF are planned.
820 11 POF Clubs
As was shown in the last sections POF is no longer only a topic in individual
countries. Over 20 countries are regularly represented at the international POF
conferences. Important national centers for POF activities are shown in Fig. 11.16.
In addition, internationally active groups such as the ITG sub committee 5.4.1
represent the Central European area.
[Schl06] Schleifring und und Apparatebau GmbH: Fiber Optic Rotary Joints,
Data sheet, Frstenfeldbruck, 2006
[Schm92] H. Schmiedel (Hrsg.): Handbuch der Kunststoffprfung, Carl
Hanser, Mnchen, 1992
[Schm00] K. Schmieder, K.-J. Wolter, D. Krabe, W. Scheel, S. Patela: Efficient
Technologies for Board-Level Optical Interconnection, POF2000,
Boston, 05.-08.09.2000, pp. 60-63
[Schm05] M. Schmatz, B. J. Offrein: Optical Interconnects System-interne
Datenbertragung mit Licht, Transparencies for Internet-Download
and media, IBM Zrich Research Laboratory, April 2005
[Schn98] P. Schnitzer, M. Grabherr, R. Jger, R. King, R. Michalzik, D.
Wiedenmann, F. Mederer, K. J. Ebeling: Vertical cavity surface
emitting lasers for plastic optical fibre data links, POF1998, Berlin,
05.-08.10.1998, pp. 157-162
[Schn99] P. Schnitzer, F. Mederer, H. Unhold, R. Jger, M. Kicherer, K. J.
Ebeling, M. Naritomi, R. Yoshida: 7 Gb/s Data Rate Transmission
Using InGaAs VCSEL at O = 950 nm and Perfluorinated GI POF,
POF1999, Chiba, 14.-16.07.1999, pp. 209-212
[Schn03] B. Schneider: VCSEL Technologie - die Chance kostengnstiger
Lichtquellen zur Kommunikation, 16. Meeting of the ITG-SC 5.4.1,
Mnchen, 25.06.2003
[Scho88] F. W. Scholl, et. al.: Application of plastic optical fiber to local area
networks, EFOC1988, Atlanta, pp. 338-343
[Sch99a] O. Schnfeld, K. Panzer, J. Wittl, H. Essl: Transceivers for In-Car
Optical Busses, POF1999, Chiba, 14.-16.07.1999, pp. 201-204
[Sch99b] O. Schnfeld, K. Panzer, J. Wittl, H. Essl, H. Hurt, G. Mller: Trans-
ceivers for In-car Optical Buses- Multimedia and Safety Applica-
tions, Internet 1999
[Sch00a] D. Schnefeldt: Charakterisierung von LED fr POF-Systeme,
optische Eigenschaften, diploma thesis at the Leipzig University of
Applied Sciences and the T-Nova GmbH, May 2000
[Sch00b] O. Schnfeld: Infineon Transceiver Components for byteflight,
Internet 2000
[Sch01] H. Schpp: Principles and Applications of the MOST Network,
Meeting of the ITG Focus project ITF, Frankfurt/M., 11.05.2001
[Sch03] G. Schtz, K.-F. Klein: Quarzglas-Dickkernfasern, 15. Meeting of
the ITG-SC 5.4.1, Offenburg, 25./26.03.2003
[Sch06] S. Schllmann, W. Rosenkranz: Mode Group Diversity Multiplexing
as a Possible Cost Efficient Alternative to Coarse WDM for LAN
over MMF, ITG conference Photonic Networks Leipzig,
27./28.04.2006
[Schr02] H. Schrder, F. Ebling, E. Strake, A. Himmler: Heigeprgte Poly-
merwellenleiter fr elektrisch-optische Schaltungstrger (EOCB) -
Technologie und Charakterisierung, DVS/GMM conference
Elektronische Baugruppen - Aufbau- und Fertigungstechnik,
06.02.2002, Fellbach
862 References
Stephen Economides
Berlin University of Applied Sciences (Fachhochschule fr Technik und
Wirtschaft Berlin (FHTW Berlin)
Prof. Dr. Stephen Economides, born in New York in 1944, received his B.A. and
M.A. degrees in German and history from the University of New Hamsphire in the
U.S.A. He acquired his doctorate from the Free University of Berlin, writing his
dissertation on a topic in Modern German History. He was self-employed for 9
years as a professional, fully accredited translator and interpreter working pri-
marily for business and industry. In 1987, he became Professor of Technical
English at the University of the German Federal Post Office at Berlin before
assuming a professorship at the Berlin University of Applied Sciences (FHTW
Berlin) in 1996. His courses in Technical English involve various aspects of
telecommunications and data communications. He has been a member of the
German Translators Association (BD) since 1982.
Most parts of the first edition have been used for the second edition, published by
Springer in 2001 (German) and 2002 (English).
The first edition was translated by
Stephen Economides
Berlin
Werner Behnke
Berlin
Index of Key Terms
Avago Technologies
Customer Information Service
c/o Promotion Team Wetzlar
(www.avagotech.com/pof-ad) p. 358
Kalsmuntstr. 14b-c, 35578 Wetzlar
++49 6441 92460
info@promotionteam.de
Optimedia, Inc.
(www.optimedia.co.kr)
204 Byuksan Technopia, Sangdaewon-Dong,
Joongwon-Gu, Seongnam-Si, Kyonggi-Do, p. 386
462-716, Korea
++82-31-737-8151
support@optimedia.co.kr
Polymicro Technologies,
a subsidiary of Molex Incorporated
(www.polymicro.com)
p. 560
18019 N 25th Ave., Phoenix, AZ 85023-1200, USA
++1 602 375 4100
polymicrosales@molex.com
Reichle & de-Massari AG
(www.rdm.com)
Binzstrasse 31, CHE-8620 Wetzikon, Switzerland p. 592
+41 44 933 81 11
andreas.bloechlinger@rdm.ch
fiberware GmbH
(www.fiberware.de)
Bornheimer Strae 4, 09648 Mittweida p. 664
++49 30 5670 0730
office@fiberware.de
DieMount GmbH
(www.diemount.de)
Giesserweg 3, 38855 Wernigerode p. 762
++49 3943 625 9760
hans.kragl@diemount.com
POF-AC
outer
(www.pofac.de)
front
Fachhochschule Nrnberg
cover
Wassertorstrasse 10, 90489 Nrnberg
page
0911-5880 1070, Fax: 0911-5880 5070
pofac@pofac.fh-nuernberg.de
Biographies
Olaf Ziemann
Polymer Optical Fiber Application Center of the FH Nrnberg
Prof. Dr.-Ing. Olaf Ziemann (42) studied physics at the
University of Leipzig. Between 1990 and 1995 he did
his doctorate degree at the Technical University of
Ilmenau in the field of optical telecommunications engi-
neering. His fields of work were optical superhetero-
dyne reception and optical code multiplexing. During
this period he was recipient of a scholarship from the
University Foundation of the German People. Between
1995 and March 2001 he worked in the research center
of the Deutsche Telekom (T-Nova) in the specialized
areas of hybrid access networks and building networks.
Since 1996 he has been the chairman of the Information
Technology Society-Sub-Committee Polymer Optical
Fibers (ITG-SC 5.4.1). Since the beginning of 2001 he
has been the scientific director of the POF-AC at the
Nrnberg University of Applied Sciences (FH Nrn-
berg) and Prof. since Nov. 2001.
Jrgen Krauser
University of Applied Sciences Leipzig
Professor Dr.-Ing. Jrgen Krauser (60) studied physics
at the Technical University of Berlin. From 1975 - 1980
he was academic assistant in the Institute for Solid-State
Physics where he received his doctoral degree in 1981,
writing his dissertation on a subject in the field of sold-
state physics. Subsequently, he played a substantial role
in the building up of the Integrated Optics research sec-
tion at the Heinrich Hertz Institute in Berlin and was
head of the research group on Optical Measuring Tech-
niques. In the beginning of 1986 he was offered a pro-
fessorship at the Univ. of the German Federal Post Of-
fice at Berlin in the field of optical telecommunications
engineering and physics. Since 2000 he has assumed
duties in this field at the Deutsche Telekom University
of Applied Sciences at Leipzig. He has published nume-
rous works and papers in the field of optical telecommu-
nications engineering, especially polymer optical fibers.
Peter E. Zamzow
Nexans Deutschland Industries AG & Co KG
Dipl.-Ing. Peter E. Zamzow (67) is an independent tech-
nical consultant for research and development of cable
systems. After completing his studies in telecommuni-
cations engineering in Munich and Graz, he started in
1970 to work for AEG Kabel (Cable). In 1980, he
became director of the product area for light waveguides
and in 1982 chief engineer. In 1985 he was appointed
director. From 1990 on he was director of the new glass
fiber and glass fiber cable plant. At the beginning of
1994 he assumed leadership of CATV-Cable and sys-
tems for Alctel. Since 1998 he has been based in
Hannover and was responsible for marketing and sales
in the area of worldwide licensing and production facili-
ties for Alcatel. Between Jan. 2001 and 2004 he has
been responsible for research and technology for the
company headquarters in Mnchengladbach. In 2005 he
started an own national and international consulting
career with the focus on cable systems.
Werner Daum
Federal Institute for Materials Research and Testing (BAM) Berlin
Director and Professor Dr.-Ing. Werner Daum, born in
1956, studied electrical engineering with a major in
measuring techniques at the Technical University of
Berlin. Upon graduation in 1984 he started work for the
Federal Institute for Materials Research and Testing
(BAM). Until 1989 he was academic assistant in the
specialist group non-destructive testing. Thereafter, he
assumed the directorship of the laboratory for optical
methods of measurement and experimental tension ana-
lysis and commenced with the development of test pro-
cedures on reliability evaluation of polymer optical fi-
bers. Since 1996 he has been head of the specialist
group Measurement and Test Techniques; Sensor Tech-
nology. He belongs to the founding members of the In-
ternational Cooperative of Plastic Optical Fibers
(ICPOF) which has organized the international POF
conferences since 1992. In 1998 and 2001 he was res-
ponsible for organizing the scientific aspects of these
conferences. In Germany he played a decisive role in the
founding of the VDE/ITG Sub-Committee Polymer
Optical Fibers.