Академический Документы
Профессиональный Документы
Культура Документы
Faculty of Engineering
Dept. of Electrical Engineering
Thesis
By
2014
Abstract
The importance of semiconductor optical amplifiers (SOAs) as key
components in optical communications and integrated optics, covering a wide
range of applications for the 1550- and 1300-nm optical windows, has grown in
recent years. All-optical signal processing, including wavelength conversion,
optical logic gates and signal regeneration, etc, is one of the most important
enabling technologies to realize optical switching, including optical circuit
switching, optical burst switching and optical packet switching. SOAs are very
promising in all-optical signal processing since they are compact, easy to
manufacture and power efficient. The need for all-optical elements for
increasing the capacity of current and future communication networks and
optimizing the operation of optical switching networks has been one of the
main motivations for considering SOAs as essential elements in all-optical
switching scenarios in recent years.
This present work is concerned with the analysis of the QD-SOA-based Mach-
Zehnder Interferometer (MZI) and its fundamentals and applications. This
analysis is carried out using the rate equations model.
The dissertation starts with the dissection of foundation of the quantum dot
laser and amplifier. This includes a brief introduction to the development of
quantum dots and quantum dot lasers. Advantages and disadvantages of the
latter are discussed as well as the progress in manufacturing. In the discussion
of the QD amplifier we concerned on the gain and gain saturation. Some
important parameters are also considered in this background such as the
Linewidth Enhancement Factor, the Amplified Spontaneous Emission, and the
simulation methods of Quantum-Dot Semiconductor Optical Devices.
Then, this dissertation proceeds to study the QD laser using the Multi
Population Rate Equations MPRE model. In which, we have solved the rate
equations for the quantum-dot laser considering homogeneous and
inhomogeneous broadening of optical gain numerically using fourth-order
Runge-Kutta method. The dynamic (relation with the time) and static (relation
with the current) characteristics of proposed QDL are presented in chapter
three. This chapter studies also the effect of the FWHM of homogeneous
broadening and the injected current on the rise and fall time, hence on the bit
rate. These later results aims to study the possibility of using the QDL as a
pulse source for high bit rate data transmission.
Then the all-optical QD-SOA-based MZI switch is used to design three all-
optical logic gates; XOR, AND, and OR. For each all-optical logic gate the
principle of operation, the proposed design, and the simulation results are
presented.
Finally and for the first time, a new scheme for all-optical full adder using fife
QD-SOA based MachZehnder interferometers is theoretically investigated
and demonstrated. The proposed scheme consists of two XOR, two AND, and
one OR gate. The impact of the peak data power as well as of the QD-SOAs
current density, maximum modal gain, and QD-SOAs length on the Extinction
Ratio ER and Q-factor of the switching outcome are explored and assessed.
The operation of this system is demonstrated with 160 Gbit/s.
Table of Contents
ABSTRACT II
Table of Contents IV
List of Symbols IX
Chapter (1):
Introduction
1.1 Context 1
1.2 Motivation 2
Chapter (2):
Chapter (3):
3.3.3 Semiconductor Quantum Dot Lasers as pulse Sources for High Bit rate 47
Data Transmission
Chapter (4):
Chapter (5):
Chapter (6):
Chapter (7):
Appendix (A):
Appendix (B):
Design Metrics
Arabic Abstract
List of Symbols
1.1 Context
1.2 Motivation
The optical fibre is the principal enabler of the large data transmission
capacity we have today, with reported capacity times distance product of
1121015kmbit/s [4], achieved with 155 m wavelength channels at 100 Gb/s
over 7200 km. This record represents an increase of over seven orders of
magnitude over 35 years, since the beginning of optical telecommunications
[5]. Dense wavelength division multiplexing (DWDM) and time domain
multiplexing (TDM) are key technologies for achieving such high transmission
bit rates, of the order of tens of terabits per second. Typically, at the network
node, the incoming data are converted to the electrical domain, the data headers
are processed and the data are then forwarded to the correct output port. With a
data rate of tens of terabits per second per fibre arriving at a network node, it is
a hard task for the electronics to handle such amount of data due to the limited
electronic bandwidth of the order of tens of gigabit per second per channel.
Therefore, there is a bottleneck effect, where the fibre provides very large
bandwidth capacity, but at the network nodes there are not enough processing
capabilities to take advantage of the fibre capacity.
Despite the very large increase of fibre data transmission capacity and
different technologies involved, the optical network nodes have always been
based in optical-electrical- optical (OEO) conversion, until the adoption of
optical add/drop multiplexers (OADM). This was the first optical solution to be
adopted, where the traffic is routed in the optical domain, on a wavelength
basis. The next solutions, which are currently being adopted, are the
reconfigurable OADM (ROADM) and the optical cross connect (OXC), where
the OADM functionality is retained, with the added capability of remotely
rerouting the traffic without local user intervention. Although, the OADM,
ROADM and OXC provide wavelength routing, neither regeneration nor
routing at a finer granularity are possible.
Introduction
In this chapter, the background of the quantum dot laser and amplifier is
introduced. This includes a brief introduction to the development of the
semiconductor laser down to the quantum dot lasers. Section 2.1 discusses the
concept of the QD starting from semiconductor lasers down to QD laser and its
progress in manufacturing taking into account the Inhomogeneous broadening
property of the QD. A review of advantages and disadvantages of the quantum
dot laser QDL is presented in section 2.2. The quantum dot amplifier and its
gain as well as the gain saturation concept are discussed in section 2.3. Sections
2.4 and 2.5 are devoted to the Linewidth Enhancement Factor and the
Amplified Spontaneous Emission, respectively. Finally, Section 2.6 introduces
the numerical methods used to model the QD laser and amplifier.
Semiconductor devices are among the most promising candidates for all-
optical processing devices due to their high-speed capability up to 160Gb/s,
low switching energy, compactness, and optical integration compatibility.
Their performance may be substantially improved by using quantum dots in the
active region characterized by a low threshold current density, high saturation
power, broad gain bandwidth, and a weak temperature dependence as
compared to bulk and multi-quantum well (MQW) devices. This section
discusses the concept of the QD starting from semiconductor lasers down to
QD laser
Although low threshold current and high temperature stability [17] have
been predicted for QD lasers in the 1980s, the realization of lasing was lagged
due to the lack of an appropriate technology at that time. Lasing properties of
three-dimensionally quantized carriers were first investigated indirectly by
placing QW lasers in strong magnetic fields [17], which demonstrated an
increase in the characteristic temperature. There have been several approaches
to the realization of QD structures. A traditional way was to selectively etch
QW structures. In 1987, Miyamoto et al [18].demonstrated electrically pumped
light emission from a QD structure, which was realized from an MOCVD-
grown GaInAsP/InP QW by applying the holographic process, etching, and
overgrowing [18]. The authors reported a Fabry-Perot-mode-like spectrum,
which suggested a possible stimulated emission. Later efforts yielded a laser
operation, but the devices exhibited high jth (7.6 kA/cm2 at 77 K under pulsed
operation), most probably due to a high density of surface states created during
etching.
from a laser mode by less than their homogeneous linewidth (in the region of 5-
10 meV at room-temperature and increases for higher carrier densities [20]) are
able to contribute to its oscillation having an inhomogeneous line-width that is
greater than the homogeneous broadening will limit the lasers efficiency.
Therefore growth recipes that result in lower inhomogeneous broadening are
desirable. For quantum dots, typical value of homogeneous broadening of few
eVs are normally reported in the literature for low temperature [21]. At room
temperature the rate between homogeneous and inhomogeneous broadening is
around 4, typical value of homogeneous broadening is on the order of 10-20
meV [20] and inhomogeneous broadening around 40-60 meV.
The gain of the SOA results from transitions between the conduction
and valence bands which depend on the carrier density and distribution in both
bands. There dynamical processes that determine the gain variation after
propagating an optical pulse through the SOA can be categorized into interband
and intraband classes and the gain coefficient may be described as a
combination of these processes [8, 50]
( ) ( ) ( ) (2.1)
where the input and output optical powers of the amplifier are denoted by Pin =
P (z = 0) and Pout = P (z = L). The amplifier length in the growth direction (z)
is given by L. The propagation of light power along the z-axis in an amplifying
medium can be described by
P (2.3)
The solution of the above equation for a constant gain g is in the form of P(z) =
Pinexp (gz) which results in the amplifier Gain by extracting output power at z
= L as
exp( ) (2.4)
Since the optical gain depends on the input signal intensity and frequency, it is
common to model the frequency-dependant gain coefficient with a
homogeneously broadened two-level system
( ) (2.5)
( )
where g0 is the maximum value of the gain, is the optical angular frequency
of the incident signal, is the atomic transition angular frequency, T is the
dipole relaxation time, P the optical power of the incident signal and PS is the
saturation power of the gain medium. Thus, the frequency-dependant relation
between the amplifier gain G and the optical gain g can be written as G ( ) =
exp (g ( ) L). For input signal powers much smaller than the saturation power
(P << PS), the gain coefficient reduces to
( ) (2.6)
( )
which describes a Lorentzian-shape gain spectrum with maximum at = .
The optical gain bandwidth is defined as the full-width-at-half-maximum
(FWHM) of the gain spectrum given by . Hence, the
amplifier bandwidth (FWHM of G( )) can be obtained through Eq. (2.4)
(2.7)
( )
It is clear from the above equation that the amplifier bandwidth is smaller than
the optical gain bandwidth due to the exponential dependence of the amplifier
gain to the optical gain. Also, this simplistic model of a homogeneously
broadened two level system cannot be applied to QD-SOAs with
inhomogeneously broadened gain spectrum [48].
Increasing the input signal power to the amplifier will result in depletion
of the carriers in the active region and consequently decreasing the amplifier
gain. This fact is referred as gain saturation which is common between
amplifiers and lasers and leads to signal distortion. A typical SOA gain versus
output signal power characteristic is displayed in Fig. (2.7) where the 3 dB
saturation power is defined as the amplifier output power at which the amplifier
gain is half the small-signal gain. Combining Eqs. (2.3 & 2.5)at the maximum
gain frequency, , yields
In bulk and QW amplifiers the peak gain depends linearly on first order carrier
density given by
( )( ) (2.11)
where is the confinement factor, V is the volume of the active region, N0 is
the transparency carrier density and a is the differential gain coefficient (a =
dg/dN). The carrier density rate equation can then described by
( ) (2.12)
Using a 2-level rate equation model for QDs in CW operation mode, the
saturation power in the limit of high-inversion condition under high current
density where the QD states and the WL band edge are completely filled, can
be expressed by [27]
(2.14)
( )
Introduction
Fig. (3.1) The interaction between the cavity-mode photons and the carriers in
the quantum dots via homogeneous broadening of gain.
( ) (3.9)
( ) (3.10)
( )
(3.12)
( )
(3.13)
with being the refractive index and is the optical confinement factor.
In order to calculate the entire emission spectrum, the model has been
extended considering also the presence of many cavity longitudinal modes, Fig.
(3.1), hence the photon number with resonant energy of the mode is
depicted by
(3.14)
( ( ) )
(3.15)
( ( ) )
Where and are the number of photons emitted from exited and
ground state at the mode number m, respectively, and is the total number of
photons at the mth mode which equals to . The rate of
Where is the band gap, is the electron effective mass, is the spin-orbit
interaction energy of the QD material [43].
Furthermore, let us emphasize that the various QD population are coupled by
the homogeneous broadening of the stimulated emission process assumed to be
Lorentzian such as
( )
(3.20)
( ) ( )
with being the full-width at half-maximum (FWHM) of the homogeneous
broadening and( ( ) ) being the mode energy.
All parameters used in the calculations are summarized in Table I.
Parameter Value
The Multi Populations Rate Equations MPRE model together with the
photons rate equations Eq.(3.11-3.15) have been solved using the fourth order
Runge-Kotta method with the help of the optical gain equations, Eq.(3.16,
3.17). In this section there are three sets of analysis are presented. These sets
can be described as follows,
In the first set, we achieve dynamic response and analyze photon time
evolution of mentioned QD laser at different injected currents and FWHM of
homogeneous broadening. Then, the effects of the initial relaxation oscillation
time are treated on the number of photons characteristics.
While in the second set,we achieve static response and analyze power-
current evolution of mentioned QD laser at different injected currents and
FWHM of homogeneous broadening. Then, the effects of the initial relaxation
oscillation time are treated on the power-current and gain-current
characteristics.
Finally, in the third setwe study the effect of the FWHM of homogeneous
broadening and the injected current on the rise and fall time, hence on the bit
rate. Finally, the effects of the initial relaxation oscillation time are treated.
In this section the dynamic analysis will be studied. Fig.(3.3) shows the
simulation results of photon-characteristics when the inhomogeneous
broadening will be neglected, N=M=0. As can be seen from the transient
response of the GS and ES shown in Fig.(3.3), the turn on delay decreased as
current increases. The ground state lasing starts first and the lasing of the ES
delayed when the current is 160A/cm2 and the steady state of the GS is higher.
As the current increases the ES lasing becomes faster and the steady state of the
two levels increases and comparable to each other. as can be seen from the
figure at current density equals 380A/ cm2 the steady state of the ES is higher
than that of the GS which known in the published papers by the two states
lasing or double wavelength lasing. All the above can be explained by that as
the current increases the carriers supplied to the active region required to
starting the lasing is faster and the turn on delay decreased. The two lasing
phenomena is due to the direct relaxation process of the carriers from WL to
GS. The results shown in Fig.(3.3) is in a good agreement of the that published
in [44].
influence of any group on the others. The quantum-dot laser at this situation
behaves in the same way as if it included independent lasing media in the same
cavity. When homogeneous broadening is comparable to inhomogeneous
broadening, lasing mode photons are emitted not only from energetically
resonant dots, central group, but also from other non resonant dots within the
scope of the homogeneous broadening of the central group. Since carriers of
non resonant dots are brought into the central lasing mode by stimulated
emission, lasing emission with a narrow line takes place and the steady state of
the lasing mode will be slower [50]. So, in quantum-dot lasers, homogeneous
broadening of optical gain connects spatially isolated and energetically
different quantum dots, leading to the collective lasing of dot ensemble.
In this section the static analysis will be studied. Fig.(3.9) shows the
simulation results of light-current characteristics when the inhomogeneous
broadening will be neglected; N=M=0. In this figure, the calculated ES and GS
output power are reported as functions of the injection current density. It shows
two thresholds corresponding to the two laser emissions. When the ES
stimulated emission appears, only a slight decrease of the GS slope efficiency
is predicted. At the same time, the global slope efficiency increases. Here, the
double laser emission seems to result from the efficient carrier relaxation into
the GS due to the increase of the Auger effect for larger injection rates [105].
Although the competition between GS and ES transitions of different QDs is
not taken into account in this figure, these numerical results give a good
qualitative understanding of the experimental results recently reported for an
optical pumped InAs/InP (113)B diode laser [45] and it is in a good agreement
of the that published in [44].
Fig.(3.10) shows optical gain-current characteristics for the same case of
the results shown in Fig.(3.9); N=M=0. The first point, it seems that optical
gain increases to the threshold gain and then becomes fixed with enhancing of
the injection current. Actually, what happens when the current is increased to a
value above threshold is that the carrier density and gain initially (by the order
of a nanosecond) elevate to values above their threshold levels, and the photon
density grows. But then, the stimulated emission rate also heightens that leads
to reducing of the carrier density and gain until a new steady-state balance is
created [35]. The second point, due to the threshold current of the GS is small
than the ES, the GS gain reaches its steady-state value faster than the ES.
(b)
All the above can be explained by that as the current increases the
number of carriers supplied to the active region required to start lasing is
reached faster and the turn-on delay decreases. As a result, the rise time
decreases.
On the other hand, for high current, when the current is switched off, the
number of existing carriers is high enough to increase the rate of stimulated
emission and as a result, the fall time decreases.
Figure (3.18) shows dependence of the rise, fall time and bit rate on the
injected current. As can be seen from this figure, as the injected current
increases the rise time decreases and the fall time alsodecreases. The rate of
decreasing the fall time is small with respect to the rate of the decreasing the
fall time. Therefore, the bit rate increases with the increase of the injected
current.
Fig. (3.18) Depends of the Rise time (a), fall time (b) of the number of photons,
and Bit rate (c) on the injected current at =20 meV,
Figures (3.20), (3.21) and (3.22) show the dependence of the rise time,
fall time, and the bit rate on the FWHM of the homogenous broadening, .
In these results, the current density is used as a parameter and its value are;
J=1000, 500, 250, 100 A/cm2. Again, these figures explain that, as the
increases, the rise and fall times increase. Hence, the bit rate decreases.
Fig. (3.24) Rise(a), fall(b) time and Bit rate(c) versus the coefficient of phonon
relaxation(Aw) at =20meV,
The final result is the study of the initial carrier relaxation which
corresponds to phonon bottleneck. Fig. (3.23) shows the photon-characteristics
at =20meV and and Current Density J= 100 A/cm2 at
different coefficient for phonon relaxation; Aw=0.35*1010, and 10.35*1010s-1.
As shown in Fig.(3.23), as the Aw coefficient decreases which means the initial
carrier relaxation time increases, referred to Equations.(3.3, 3.5), the turn on
delay increases and the steady state value becomes lower. This is because, the
injected carriers are consumed in the WL and thus do not contribute to lasing
oscillation. It can be seen also that, the effect of this parameter on the rise and
fall time is weak. To see this effect, we plot in Fig.(3.24) the dependence of
rise and fall time and the bit rate on the Aw coefficient.
As can be seen, the rise time increases and the fall time decrease as the
Aw coefficient increases. As a result, the bit rate increases. These results can be
explained by noting that, the effect of increasing Aw coefficient is equivalent
to the increase of the injected current, which means increasing the supplied
carriers to the QD region
Chapter (4)
Introduction
Practical implementation of all-optical signal processing unit requires
integrated all-optical devices for ease of manufacturing, installation, and
operation. The semiconductor optical amplifier MachZehnder interferometer
(SOA-MZI) is an integrated all-optical logic gate which can fulfill these
requirements. Conceptually, SOA-MZI-based logic gate operation is
straightforward, relying on optically inducing XGM, XPM or other nonlinear
phenomena between the SOAs located in each of the two interferometer arms.
This chapter proposes a theoretical model of a QD SOA-MZI based ultrafast
all-optical signal processor which under certain conditions can simultaneously
carry out wavelength conversion WC and 3R Regeneration of the distorted
optical signals. The QD SOA-MZI operation has been analyzed theoretically
by solving the rate equations of the QD-SOA dynamics, optical wave
propagation equations in an active medium, and the MZI equations. Section 4.1
introduces a brief review for the QD SOA-MZI and its transfer function that
will be used to analyze this device. The rate equations are introduced in section
4.2 while the QD-SOA characterizations are theoretically investigated and
demonstrated in section 4.3. Sections 4.4 and 4.5 are devoted to the all-optical
W.C. and 3R where the principle, design, and the numerical simulation results
will be introduced. Finally, section 4.6 summarizes the concluded remarks that
may be obtained from the displayed results.
Fig. (4.1) SMZ configuration and nonlinear phase response cancel out
mechanism
Short control pulses cause depletion in the carrier density and the slow
recovery in the carrier density is compensated by exciting both arms with an
interval of . In this timing interval the gating window for the probe beam
opens while outside this interval the destructive interference for the probe beam
at the output of the MachZehnder interferometer is maintained even though
the carrier density in the SOAs on both arms is slowly recovered. Therefore,
the rising and falling of the gate window are defined by the control pulse
duration [50]. The technology of integrating optical circuits is an important key
for developing practical MZI-based all-optical gates. SOAs incorporated into
MachZehnder interferometers can be used as appropriate unit call for optical
logic gates. The integration of active and passive waveguides for SMZ gates
has been developed both in a hybrid manner and in a monolithic manner.
Figure (4.2) depicts a hybrid integrated device where SOAs are mounted on a
silica-based planar lightwave circuit (PLC) [48]. Monolithic integration of
SMZ gates with other active devices, such as input and output optical
amplifiers and fixed or tunable diode lasers and significant reduction in the
footprint of MachZehnder optical circuits using photonic-crystal waveguides
[51] are important technological outcomes in the way of realizing integrated
all-optical signal processing devices.
( )is the active medium length. Considering each pair of input and output
lightwaves one can write [127]
( ) ( )( )
(4.1)
An optical signal travelling through the amplifier will experience an
amplification of and a phase shift of . Therefore,
( ) ( )( )
(4.2)
Then, the transfer function can be described as
( ) ( )( )
(4.3)
Where
( ) ( ) ( ) ( )
( ( ) ( ) ( ) ( ) )
( ( ) ( ) ( ) ( ) )
( ) ( ) ( ) ( )
(4.4)
Denoting the input and output signal powers with PA1, PA2, PD1 and PD2
and assuming an ideal 3 dB coupler ( ( ) ( ) ) the transfer
function reduces to
|[ ] | [ ]
(4.5)
|[ ] | [ ]
(4.6)
(4.7)
As mentioned above, when the control signals A1 and/or A2 are fed
intothe two SOAs they modulate the gain of the SOAs and give rise to the
phase modulation of the co-propagating CW signal due to the linewidth
enhancement factor (LEF)[8, 51]. LEF values may vary in a large interval
from the experimentally measured value of LEF =0.1in InAs QD lasers near
the gain saturation regime [51] up to the giant values of LEF =60measured in
InAs/InGaAs QD lasers [52]. However, such limiting cases can be achieved for
specific electronic band structure [54]. The typical values of LEF in QD lasers
are [53]. Detailed measurements of the LEF dependence on
injection current, photon energy, and temperature in QD SOAs have also been
carried out [31].
For low-injection currents, the LEF of the dot GS transition is between 0.4 and
1 increasing up to about 10 with the increase of the carrier density at room
temperature [31]. The phase shift at the QD SOA-MZI output is given by [10]
(4.8)
It is seen from (4.8) that the phase shift is determined by both LEF and
the gain. For the typical values of LEF , gain , and
the phase shift of about is feasible
The system of the rate equations accounts for the following transitions:
1) the fast electron transitions from WL to ES with the relaxation time
;
2) the fast electron transitions between ES and GS with the relaxation time
from ES to GS and the relaxation time from GS to ES
;
3) the slow electron escape transitions from ES back to WL with the
electron escape time .
The balance between the WL and ES is determined by the shorter time
of QDs filling. Carriers relax quickly from the ES level to the GS level,
while the former serves as a carrier reservoir for the latter [40]. In general case
the radiative relaxation times depend on the bias current. However, it can be
shown that for moderate values of the WL carrier density
this dependence can be neglected [36, 50]. The spontaneous
radiative time in QDs remainslarge enough: ( ) [55, 56].
For the case of the signal detuning smaller than the QD spectrum homogeneous
broadening the electron rate equations have the form [99]
(4.9)
( )
( ) (4.10)
( ) ( ) ( ) (4.11)
( ) ( ) ( ) (4.12)
where variable Z is the longitudinal direction along the QD-SOAs length L, i.e.
Z=0, means for the input and Z=L for the output facet of each QD-SOA,
variable t is the local time measured in a coordinate system moving with the
pulse group velocity . The functions used in the derivatives [Eqs. (4.9-4.12)]
are the photon density of input data signals, which is related to their power
( ), with the equation as ( ) ( )( ),where
is the effective cross section of the QD-SOAs, is
the group velocity of the propagating signal, and is the photon energy
( , where is the speed of the light in vacuum and is the
wavelength of the incident light). The electron density in WL is , and the
electron occupation probability in the ES and GS is h and f , respectively.
maximum modal gain, where l, and is the number, surface density and
effective thickness of the active QD layers, is the homogeneous linewidth
of the resonant QDs, the inhomogeneous linewidth of the QD ensemble
and is the resonant cross section of the carrier-photon interaction [57].
Furthermore, is the material absorption coefficient, J is the injection current
density, e the electron charge, the electron relaxation time from the WL to
the ES, the surface density of QDs, the effective thickness of the active
layer, the electron escape time from the ES to the WL, the spontaneous
radiative lifetime in the WL, the electron relaxation time from the ES to the
GS, the electron escape time from the GS to the ES and the spontaneous
radiative lifetime in the QDs [41 - 43].
Fig. (4.5) (The considered spatio-
temporal grid on which the 4th order
Runge-Kutta method is applied. L: the
QD-SOAs length. N=10 the number of
uniform segments that the QD-SOA is
divided into. z: the length of each
spatial segment. t: the temporal
interval between two consecutive
samples of an input pulse.
Given the central role of the QD-SOA in the operation of the MZI
configured as logic gate it is important to characterize the dynamical behaviour
of this device with respect to several critical operational parameters. This is
necessary in order to be able to properly interpret in next sections and chapters
the simulation results obtained when QD-SOAs are incorporated in the MZI.
In fact the information given by most of previous reported works [58-
62] as part of the study on the performance of a QD-SOA-based MZI intended
for use in the implementation of all- optical logic gates at ultrafast bit rates is
not sufficient or complete for the pursued goal. More specifically, [59] have
dealt with the QD- SOA carrier dynamics and the temporal evolution of the
occupation probabilities in the discrete levels of the QD system. In [62] the
temporal dependence of the gain has been obtained both for a single input pulse
and a non return-to-zero (NRZ) input pulse train. In the first case the result
concerns the saturation and quick recovery of the QD-SOA caused by the pulse
as it enters and leaves the device, respectively. In the other case the focus is on
the gain that is modulated by the NRZ input pulse train and it is shown to
respond directly to the input data pattern, i.e., the gain is saturated as long as
logic ones enter the QD-SOA and it recovers when one or more logic zeros
appear at the input. The same study examines also the change of the QD-SOA
gain dynamics by a NRZ input pulse train for different values of which
shows that the decrease of this parameter speeds up the gain recovery of the
QD-SOA. Moreover in [58], the temporal dependence of the gain is derived for
the case of a single input RZ pulse and for different values of J and The
results for J indicate that the increase of this parameter accelerates the gain
recovery of the QD-SOA, while the results for reveal that this parameter is
a limiting factor for the QD-SOA gain dynamics. Finally, regarding the QD-
SOA saturation properties, in [62] the steady-state static gain has been obtained
as a function of the input power for different values of J, which shows that the
gain begins to saturate as the input power is increased and that the saturation
power is increased with J. This dependence on J can also be noticed in [61],
where the 3 dB input saturation power, Pin,sat, is calculated as a function of J for
different values of QD-SOA length. This unveils that if the current density is
kept constant the QD-SOA saturation power is decreased as its length is
increased. This means that the longer the device is, the more easily it can be
saturated by a smaller input power.
Despite the significant contribution of the aforementioned works, a more
comprehensive characterization of a QD-SOA intended for use as nonlinear
element for interferometric switching should take into account the impact of a
whole set of critical parameters. This task is of a greater importance when the
mode of operation is pulsed and the strain imposed on the operating conditions
is quite demanding [63], as in our case. Therefore we have focused on the QD-
SOA gain dynamics and their dependence on these parameters, since it is
expected that they will affect the performance of the proposed logic function.
Specifically, in Figs. (4.7& 4.8) we examine the instantaneous QD-SOA gain
variation for different values of the critical parameters, Ppeak, GmaxJ, L, and .
This task is accomplished for a few pulses, which have been selected to be the
last four pulses in the longest run of these bits inside the PRBS.
This is done in order to ensure that a saturation equilibrium has been
established for the QD-SOA gain dynamics and thus that the conducted
characterization is as realistic as possible. For this purpose the process that is
followed is to change each time one parameter while keeping the rest constant.
Specifically, the typical values chosen to be fixed for, Gmax, J, L, and are
15/cm, 3 kA/cm2, 3mm, and 0.15 ps, respectively. On the other hand the choice
for the peak input data power has been dictated by the gain saturation of the
QD- SOA that occurs under pulsed mode of operation. In this context the
saturated gain, which is defined as the minimum of the instantaneous gain that
is dropped after the pulse has acted on the QD-SOA has been plotted as a
function of this parameter in Fig. (4.6).
Fig. (4.6) Variation of minimum instantaneous QD-SOA gain versus peak input
data power. The difference P = 4 dBm denotes the deviation of the QD-SOA
bias point chosen as fixed for the simulation analysis from the 3 dB input
saturation power.
From this figure we see that the 3 dB saturation input power [32] is
Pin,sat= 7.5dBm. Thus in order for the QD-SOA to be adequately saturated for
the needs of this characterization while at the same time keep Ppeak reasonable
we set Ppeak,fixed= 11.5 dBm. The incremental deviation of this fixed value from
the 3 dB QD-SOA saturation input power under pulsed mode of operation
defines an interval, which is indicated in Fig. 4.6 by the vertical dotted lines
and the arrows attached on either side of them. The extent of this interval is
determined by the, dimensionless, relative difference P = Ppeak,fixed - Pin,sat= 4
dB, within which the QD-SOA single pass gain has been reduced by 5.96 dB.
Therefore the remarks that will be extracted for the QD-SOA characterization
conducted under the specific bias condition will also hold for a lower saturation
level. Figure (4.7a) shows the instantaneous QD-SOA gain variation for
different values of the peak input data power. First, it is observed that the initial
level of the gain is not its maximum modal gain value, which also holds for
Figs. (4.7. b, c, dand4.8. a, b). This happens because as already mentioned the
examined pulses are the last four pulses. For this reason when this pulses enter
the QD-SOA it encounter a gain that has been partially recovered after it have
been saturated by the preceding mark. Now the amount that the gain is dropped
from its maximum value, G, becomes larger as Ppeak is increased.
Fig. (4.7) Instantaneous QD-SOA gain variation for different (a) Peak input
power (b) maximum modal gain (c) current density (d) Length
This is consistent with Fig. (4.6) and the fact that as Ppeak is changed in
this direction the QD-SOA becomes more saturated because of carrier
depletion. Furthermore, this gain drop is favourable to impart a differential
phase between the decomposed components of probe pulse (Fig.(4.1)) as close
as possible to and hence achieve full switching. Then the dynamic gain
response for different values of the QD-SOA maximum modal gain is
illustrated in Fig. (4.7. b). Notably the drop of the gain from its unsaturated
value is greater for larger maximum modal gain. More specifically, for gmax =
11/cm the drop is G 0.97 dB, while for gmax =15cm-1 G 2.7dB. Thus
similarly to the case of the peak input data power the extent of the QD-SOA
saturation quantified by G becomes larger as gmax is increased, which as
mentioned is helpful for creating the desired phase difference between the MZI
arms. Moreover, Fig. (4.7c) shows the instantaneous QD-SOA gain variation
for different values of current density. The increase of J from 2 to 4 kA/cm2
supplies the GS with more carriers that are available for the gain recovery
process and thus the latter is accelerated. However, if J is further increased to 5
kA/cm2 the effect on the dynamical behavior of the gain is not so pronounced.
This happens because the ES and WL, which act as carrier reservoirs for the
GS, have been adequately filled and the additionally supplied carriers do not
fully participate in the QD-SOA gain recovery process. Finally, in Fig. (4.8a)
we present the instantaneous QD-SOA gain variation for different electron
relaxation times from the ES to the GS and for the chosen fixed value of
current density. With the decrease of from 1.3 to 0.3 ps carriers relax faster
from the ES to the GS and so the process of gain recovery is made faster. At
the same time there are more carriers involved in the amplification process
which leads to less intense saturation and gain modulation. This also holds for
the case in which is decreased down to 0.15 ps. But when the QD-SOA
becomes saturated subject to the consecutive bits of the PRBS, more carriers
relax to the GS. Therefore they are depleted in the upper levels, i.e., the ES and
the WL, eventually becoming insufficient for speeding up the QD-SOA gain
recovery. Nevertheless, if the current density is
Fig. (4.8) Instantaneous QD-SOA gain variation for (a) J = 1 kA/cm2, (b) J = 3
kA/cm2.
concurrently increased, the behavior of the gain is altered in a common manner
for all considered changes of , as shown in Fig. (4.8b). This means that when
investigating the impact of on the switching capability of the logic gate it is
also necessary to take into account the simultaneous effect of J.
(4.15)
[ ]
where is the CW or the clock stream optical signal divided and introduced
via the symmetric coupler into the two QD-SOAs.
are the non inverted converted and the
inverted converted output, respectively. System of equations (4.9-4.12)
constituting a complete set of equations describing XGM and XPM in QD SOA
are essentially nonlinear and extremely complicated. Their analytical solution
in a closed form is hardly possible, and for this reason, this system of equations
has been solved numerically for the same typical parameters values of the QD-
SOA which are stated in previous section.
Fig. (4.10) Schematic of SOA-MZI configuration for wavelength conversion
based on cross phase modulation effect
The situation here is like that, there is only one data signal interacts with
the CW-light signal as shown in Fig. (4.10). In such a case, WC occurs
between the input signal at the wavelength propagating through
the upper arm of QD SOA-MZI and the CW-light with . Figure
(4.11) shows the input waveform (data A) and the simulated output waveforms,
specifically, the non-inverted converted version of input data and the inverted
version. It can be seen from this figure that the quality of the non-inverted
output is better than the quality of the inverted one. Therefore, in the remaining
of this section we will study the effect of the important parameters on the ER
for the non-inverted output.
Figure (4.12) illustrates the effect on the ER with; (a) peak power of the
input data signals, (b) maximum modal gain, (c) the current densities, and (d)
QD-SOA length. In each case the other parameters are kept fixed.
As it can be observed from Fig. (4.12.a), the obtained curve exhibits a
bell-like variation with a maximum point at around 11dBm, on either side of
which the ER is decreased. In order to interpret this behavior we recall from
Figures (4.6, 4.7) that the peak input data power determines the extent of the
QD-SOA gain excursions, G, which in turn makes the phase difference
between the MZI arms lie in different intervals [82]. This affects analogously
the magnitude of switching and accordingly the ER. Thus initially the ER is
Fig. (4.12) Variation of extinction ratio (ER) with (a) peak data power, (b) the
maximum modal gain, (c) the current densities, and (d)QD-SOA length.
increased with the peak input data power, because QD-SOA1 is progressively
brought into deeper saturation and the phase difference approaches closer to its
optimum value of . However as the examined parameter is increased further
beyond 11dBm, the additional differential gain that is induced causes the phase
difference to diverge away from [82]. As a result the ER does not continue to
improve but it is declined with a steeper slope than that of its rising part due to
the stronger carrier depletion. The ER is acceptable within a total input power
dynamic range of roughly 5.5dBm, whose central peak power of 11dBm which
can be provided by commercial erbium doped fiber amplifier.
Fig. (4.12.b) shows the dependence on maximum modal gain. It can be
noticed that there is a similarity between the obtained curve and that of Fig.
(4.12.a). This is attributed to the common impact that both parameters have on
the QD-SOAs dynamical behaviour, as demonstrated in Section 4.3. Therefore,
as this parameter is altered the phase difference created between the replicas of
CW-light undergoes a variation analogous to that described in the context of
Fig. (4.12.a). This means that in order to achieve switching as anticipated
according to the requirements of wavelength conversion an efficient level of
maximum modal gain is necessary. The ER remains above 10 on either side of
a maximum modal gain of approximately 15 cm-1, where it becomes a
maximum. For given QD- SOAs length this maximum modal gain range can be
achieved in a feasible manner by intervening in the number of the QD layers
when designing the QD-SOA structure [71], as discussed in section 4.2.
Figure (4.12.c) illustrates the ER versus the QD-SOAs current
density. For small current densities, the ER is sharply increased, and after
exceeding its required minimum, it becomes almost independent of this
parameter. This happens because a lower current density facilitates the
saturation of a QD-SOA [61]. As a result the gain of QD-SOA1 is dropped to a
greater extent and it becomes more difficult for it to recover closer to its
unsaturated value, which is additionally verified by Fig. (4.7c). Consequently,
the ER is very low and hence totally inadmissible. In contrast, a larger current
density offers a redundancy of supplied carriers and thus permits the dynamical
optical properties of QD-SOA1 to reach an equilibrium state, which has a
positive impact on the considered metric. Fig. (4.12.c) shows that if J is
adjusted to be over 2.17 kA/cm2 then the ER is made acceptable. The
corresponding bias current is 260.4 mA, which lies within reasonable limits
and can be practically supplied by commercial current sources [61]. Therefore
a moderate current density is fine for allowing the proposed wavelength
conversion circuit to be realized at least with an adequate performance.
Figure (4.12.d) shows the ER as a function of the QD-SOAs length. It can be
noticed that there is an apparent similarity between the obtained curve and that
of Figures (4.12.a & b) for the peak data power and the maximum modal gain,
respectively. This is attributed to the common impact that these parameters
have on the QD-SOAs dynamical behavior and subsequently on the switching
performance. In particular the QD-SOAs length determines the extent that the
gain of each QD-SOA is dropped from its unsaturated level [58] and
subsequently the amount of the differential gain created between the MZI arms.
Since this quantity imparts through the line width enhancement factor a
differential phase shift, it can be realized that the changes of also affect
and hence the ER in a way analogous to that described in the context of Fig.
(4.12.a). this reason the three ER diagrams resemble in shape both having a
rising and a falling part, although the effect on the ER magnitude is more
pronounced for the QD-SOAs length because the gains G1 and G2 that enter in
the expressions of the characteristic transfer functions are influenced directly
by its variation ,Eq. (4.5). The ER remains above 10 dB on either side of a
QD-SOAs length of 3mm, where it becomes maximum, and within a range of
approximately 1.1mm, which for given QD-SOAs maximum modal gain can be
achieved by intervening in the number of the QD layers when designing the
QD-SOA structure.
Design Parameters
According thus to Figs. (4.12a-d) it can be inferred that the requirements
for the critical parameters, to have a good performance with reasonable
amplification, are ,
, mm, and . By
following these guidelines and using the combination of values
, J= , L=3mm, , respectively, which
obviously is not unique and it fall within the specified boundaries, a more than
adequate ER of about 15dB can be obtained, which is reflected on the high
quality of pulse stream obtained at the output [8].
On the other hand, when we use another combination of values such
as , J= , L=4mm, , respectively. These
values are not falling within the specified boundaries. ER in this case equals
4.8dB, which is reflected on the low quality of pulse stream obtained at the
output. The input waveforms and the simulated output waveforms are shown
in Figs. (4.13a-c), respectively.
Short optical pulses propagating in optical fibers are distorted due to the
fiber losses caused by material absorption, Rayleigh scattering, fiber bending,
and due to the broadening caused by the material dispersion, waveguide
dispersion, polarization-mode dispersion, intermodal dispersion [8, 64]. 3R
regeneration is essential for the successful logic operations because of the
ultrafast data signal distortions. 3R regeneration requires an optical clock and a
suitable architecture of the regenerator in order to perform a clocked decision
function [48]. In such a case, the shape of the regenerated pulses is defined by
the shape of the clock pulses [65, 66].
Fig. (4.16) Waveforms of 3R-based QD-SOA MZI, where (a) Noisy input data
stream A, (b) Regenerated output, (c) Clock
Figure (4.16) shows the noisy input waveform (data A), the regenerated
output waveforms, and clock waveforms. It can be seen from this figure that
the quality of the regenerated output is excellent and it depends basically on the
clock waveforms. Therefore, in the remaining of this section we will study the
effect of the important parameters on the ER for the regenerated output.
Figure (4.17) illustrates the effect on the ER with; (a) peak power of the input
data signals, (b) maximum modal gain, (c) the current densities, and (d) QD-
SOA length. In each case the other parameters are kept fixed.
As it can be observed from Fig. (4.17.a), the obtained curve exhibits a
bell-like variation with a maximum point at around 12dBm, on either side of
which the ER is decreased. In order to interpret this behaviorwe recall from
Figures (4.6, 4.7) that the peak input data power determines the extent of the
QD-SOA gain excursions, G, which in turn makes the phase difference
between the MZI arms lie in different intervals [82]. This affects analogously
the magnitude of switching and accordingly the ER. Thus initially the ER is
increased with the peak input data power, because QD-SOA1 is progressively
brought into deeper saturation and the phase difference approaches closer to its
optimum value of . However as the examined parameter is increased further
beyond 12dBm, the additional differential gain that is induced causes the phase
difference to diverge away from [82]. As a result the ER does not continue to
improve but it is declined with a steeper slope than that of its rising part due to
the stronger carrier depletion. The ER is acceptable within a total input power
dynamic range of roughly 1.9dBm, whose central peak power of 12dBm which
can be provided by commercial erbium doped fiber amplifier.
Fig. (4.17) Variation of 3R extinction ratio (ER) with (a) peak data power, (b)
the maximum modal gain, (c) the current densities, and (d)QD-SOA length.
Fig. (4.18) Waveforms of 3R-based QD-SOA MZI, where (a) Noisy input
data, (b) Output with ER=10.5dB, (c) Output with ER=2.8dB
On other hand, when we use another combination of values such as
, J= , L=4mm, , respectively. These values
are not falling within the specified boundaries. Therefore, ER in this case
equals 2.8dB, which is reflected on the low quality of pulse stream obtained at
the output. The input waveforms and the simulated output waveforms in the
two cases are shown in Figs. (4.18a-c), respectively.
Introduction
[ ] (5.1)
Fig. (5.2) Variation of Q-Factor with peak data power for different current
densities, keeping other parameters fixed.
As it can be observed, the obtained curve exhibits a bell-like variation
with a maximum point at around 11dBm for J=3kA/cm2, on either side of
which the Q-factor is decreased. In order to interpret this behavior we recall
from Figures (4.6, 4.7) that the peak input data power determines the extent of
the QD-SOA gain excursions, G, which in turn makes the phase difference
between the MZI arms lie in different intervals [82]. This affects analogously
the magnitude of switching and accordingly the Q-factor. Thus initially the Q-
factor is increased with the peak input data power, because QD-SOA1 is
progressively brought into deeper saturation and the phase difference
approaches closer to its optimum value of . However as the examined
parameter is increased further beyond 11dBm, the additional differential gain
that is induced causes the phase difference to diverge away from [82]. As a
result the Q-factor does not continue to improve but it is declined with a steeper
slope than that of its rising part due to the stronger carrier depletion. The Q-
factor is acceptable within a total input power dynamic range of roughly 4dBm,
whose central peak power of 10.8 dBm which can be provided by commercial
erbium doped fiber amplifier; we consider the Q- factor = 10 as a limit.
Another point, the common characteristic of all curves is that the maximum
value changes and shifted to the right as the current density becomes larger.
From a physical perspective this happens because the current density
determines the power required to alter the optical properties of a QD-SOA and
properly saturate its gain, and the higher it is the more power is necessary for
this purpose [61]. This fact also explains that as we move well enough into the
falling slope of the curves a larger current density is necessary to enhance the
Q-Factor and hence improve performance for a given power. This in turn
allows selecting the peak data power from a wider range of permissible values,
which potentially offers greater flexibility in the design of the AND gate.
Figure(5.3) shows the Q-factor against the QD-SOAs maximum modal
gain for three different QD-SOAs length, when the other parameters are kept
fixed. It can be noticed that there is a similarity between the obtained curve and
that of Fig. (5.2).This is attributed to the common impact that both parameters
have on the QD-SOAs dynamical behaviour, as demonstrated in Section 4.3.
Therefore, as this parameter is altered the phase difference created between the
replicas of input data signal B undergoes a variation analogous to that
described in the context of Fig. (5.2).This means that in order to achieve
switching as anticipated according to the requirements of AND operation an
efficient level of maximum modal gain is necessary. The Q-factor remains
above 10 on either side of a maximum modal gain of approximately 16 cm-1 for
L=4mm, where it becomes maximum. For given QD- SOAs length this
maximum modal gain range can be achieved in a feasible manner by
intervening in the number of the QD layers when designing the QD-SOA
structure [71], as discussed in section 4.2. It can also note in this figure; the
common characteristic of all curves is that the maximum value is shifted tothe
right as the QD-SOAs length becomes smaller. The smaller value of the
maximum modal gain requires the longer QD-SOAs length to obtain the
maximum value of Q-factor.
Fig. (5.3) Variation of Q-Factor with the maximum modal gain for three
different QD-SOAs length, keeping other parameters fixed
Fig. (5.4) Variation of Q-Factor with current densities for two different QD-
SOAs length, keeping other parameters fixed
Figure (5.5) illustrates the effect on the Q-factor with QD-SOAs length
for two different peak powers of the input data signals, when the other
parameters are kept fixed. The common characteristic of these curves is that the
Q-factor is increased with QD-SOAs length to a certain value after that it is
dropped. The maximum value is shifted to the left as the peak power becomes
larger. The smaller value of the QD-SOA length requires the larger peak
control power to obtain the maximum value of Q-factor.
Fig. (5.5) Variation of extinction ratio (Q-Factor) with QD-SOA length for two
different peak data power, keeping other parameters fixed.
Fig. (5.6) Variation of Q-Factor with electron relaxation time from the ES to
the GS for two different QD-SOAs length, keeping other parameters fixed
Figure (5.6) depicts the Q-factor as a function of the QD- SOAs electron
relaxation time from the ES to the GS for two QD-SOAs length, when the other
parameters are kept fixed. This figure demonstrates that in order for the
proposed scheme to operate without logical errors, this parameter must not be
chosen arbitrarily fast, as it would be expected from the evidence available on
the implementation of another basic logic function using the QD-SOA-based
MZI [59]. In fact must not be too small neither too large so as to avoid
impairing the quality of switching in terms of the Q-factor. More specifically,
for >0.32ps the Q-factor starts becoming degraded and as is increased
further in this region the performance of the proposed AND gate is strongly
deteriorated due to the slower QD-SA1 gain recovery process described in the
QD-SOA characterization. On the other hand for <0.32ps the very fast
relaxation of the carriers from the ES to the GS causes the carrier density in the
upper energy levels of the discrete QD-SOA energy diagram to be dramatically
decreased. This results in lack of supplied carriers, which limits the ultrafast
operation of the MZI and renders the Q-factor unacceptable. Thus we need to
take jointly into account the contribution of the current density, as inferred
from the discussion of the QD-SOA characterization (section 4.3). To this end
the current density must be properly chosen as specified from Fig. (5.4). The
Q-factor remains over 6 against within approximately 1.17 ps. The values
lying in this range as well as the maximum of 0.95 ps are typical for [83].
Design Parameters
Fig. (5.7) waveforms of AND gate -based QD-SOA MZI, where (a) input data
stream A, (b) input data stream B, (c) Output with Q=15.15, (d) Output with
Q=2.5
Figs. (5.8 a-c) demonstrate Simulated waveforms with pseudo-eye-
diagram (PED) for (a) input data stream A or B, (b) output with Q=15.15, and
(c) output with Q=2.5. As can be seen from these figures, the quality of case (b)
is better than that of case (c). The relative eye opening (O) in case (b) equals
92%, whilst it equals 22% for case (c).These values indicate a quite good
response of the circuit under consideration at its output terminals in case (b).
Concurrently, for the case shown in Fig. (5.7c), the marks are sufficiently
balanced and the difference between their peak amplitudes is only 0.1 dB, as
verified by Fig. (5.9.a). In this figure, the uniform drop of QD-SOA1 gain in
response to data signal A can be easily noted. On the other hand, for the case
shown in Fig. (5.7.d), the marks are unbalanced and the difference between
their peak amplitudes is higher than 0.1 dB, as verified by Fig. (5.9.b). In this
figure, the inconsistent drop of QD-SOA1 gain in response to data signal A can
be easily seen.
Fig. (5.8) Simulated waveforms with pseudo-eye-diagram (PED)where (a)
input data stream A, (b) output with O=92%, (c) output with O=22%
Fig. (5.9) Gain response of QD-SOA1 to data stream A of Fig. (5.7). for (a)
Q=15.15, (b) Q=2.5
XOR logic gate is one of the most applicable gates in optical signal
processing and is a key element to implement primary systems for binary
address and header recognition, binary addition and counting, pattern matching,
decision and comparison, generation of pseudorandom binary sequences,
encryption and coding. This gate has been demonstrated at 40 Gb/s and 80 Gb/s
[80] using SOA-MZI differential schemes that have been deployed to
overcome the speed limitations imposed by the bulk SOAs slow recovery time.
However, it is predicted that QD-SOA based XOR gates have the potential to
operate above 160Gb/s bit rates [53].
A B XOR
0 0 0
0 1 1
1 0 1
1 1 0
(b)
(a)
Fig. (5.10) (a) Simulated setup of QD-SOA-based MZI configured for Boolean XOR
operation between data A and B and (b) its truth table.
[ ] (5.2)
Fig. (5.11) waveforms of XOR gate -based QD-SOA MZI, where (a) input data
stream A, (b) input data stream B, (c) A XOR B
Fig. (5.13) (a) Gain response of QD-SOA1 to data stream A, (b)Gain response
of QD-SOA2 to data stream B
Fig. (5.15) Variation of (a) ER, (b) AM, and (C) Q-factor with the delay
Figure (5.14) shows the meant by delay. The delay is time difference
between the two input data in one bit duration. Fig. (5.15) demonstrates the
influence of this delay on the ER, AM, Q-factor, respectively. As it is expected,
the performance of the XOR gate worse as the delay increases. The ER and Q-
Factor degrade while the AM increases. As can be also seen from this figure,
the delay must be below 0.45 ps because this value is the minimum critical
value for the three metrics. Above this value the ER is under 10dBm which is
the lower limit for the good communications. Although, the delays up to 0.58
ps the Q-factor is still reasonable and up to 0.73 ps the AM is still acceptable,
but we will take the lower limit for the three metrics. We compare in Fig.
(5.16) between the XOR outputs for the two input stream represented in
Fig.(5.11.a, b) in two cases; case (a) delay equals zero, and case (b), delay
equals 1 ps.
Fig. (5.16) waveforms of XOR gate -based QD-SOA MZI, where (a) delay
equals zero, (b) delay equals 1 ps.
Fig.(5.20) shows the effect on the three metrics with the difference in lengths
between the two QD-SOAs. As is shown in this figure, the AM is the first
metric affected by this difference then the Q-Factor. Differences up to 0.3mm
the performance is acceptable with respect to ER and Q-Factor. Above this
value the Q-Factor will be less than 6 and the performance no longer
reasonable. We compare in Fig. (5.21) between the XOR outputs for the two
input stream represented in Fig. (5.11.a, b) in two cases; case (a), difference in
Length equals zero, and case (b), equals 0.5mm. It can be noted from this
figure that high AM is very clear but still ones and zeros can be distinguished
from each other which means the ER is still acceptable.
Fig. (5.18)Variation of (a) ER, (b) AM, and (C) Q-factor with the difference in
the peak input power
Fig. (5.19) waveforms of XOR gate -based QD-SOA MZI, where the
difference in the peak power equals (a) zero, (b) 0.5dBm.
Figure (5.22) shows the effect on the three metrics with the difference in
the maximum modal gain between the two QD-SOAs. Also it is obvious from
this figure; the three metrics are very sensitive to the difference in the
maximum modal gain between the two amplifiers. This difference has to be
less than 0.025cm-1 to avoid higher AM. Whilst, it has to be less than 0.08cm-1
to obtain a good performance with regarding to ER and Q-Factor. We compare
in Fig. (5.23) between the XOR outputs for the two input stream represented in
Fig. (5.11. a, b) in two cases; case (a), difference in maximum modal gain
equals zero, and case (b), equals 0.2 cm-1. It can be noted from this figure that
the performance in case (b) is very bad; zeros and ones cannot be distinguished
from each other and this comes from lower ER and Q-Factor.
Fig. (5.20) Variation of (a) ER, (b) AM, and (C) Q-factor with the difference in
length
Fig. (5.21) waveforms of XOR gate -based QD-SOA MZI, where the
difference in length equals (a) zero, (b) 0.5mm.
Finally, the effect of the difference in the electron relaxation time from
ES to GS between the two amplifiers on the three metrics is explained in Fig.
(5.22). Like all the above parameters except delay in time between the two
data, the AM is the first metric affected by this difference. Differences up to
0.28ps the performance is acceptable with respect to ER and Q-Factor. Above
this value the Q-Factor will be less than 6 and the performance no longer
reasonable.
From the observation and interpretation of Figs.(5.14 to 5.24) it can be deduced
that the difference in each parameter between the two QD-SOAs has to be less
than some values to avoid degradation the XOR performance.
Fig. (5.22) Variation of (a) ER, (b) AM, and (C) Q-factor with the difference in
maximum modal gain
Fig. (5.23) waveforms of XOR gate -based QD-SOA MZI, where difference in
maximum modal gain equals (a) zero, (b) 0.2 cm-1.
Fig. (5.24) Variation of (a) ER, (b) AM, and (C) Q-factor with the difference in
relaxation time from ES to GS
5.4 All-Optical OR Gates
A Boolean logic OR gate is one of the essential logic gate devices and is
capable of forming more complex all-optical functional blocks, modules, or
subsystems. Several SOA-based logic OR gates have been demonstrated. An
OR gate based on cross-gain modulation effect in a single SOA has a simple
configuration but low extinction ratio and relatively large chirp [84]. OR logic
gates based on cross-phase modulation (XPM) [85] own advantages of high
extinction ratio or high contrast ratio at the cost of complex interferometer
configurations. In this section all optical OR gate based on QD-SOA MZI will
be investigated.
5.4. 1 Principle of Operation of Proposed OR Gate
A B OR
0 0 0
0 1 1
1 0 1
1 1 1
(b)
(a)
In this context the most suitable metric for comparing their level and
assessing the degree of their uniformity is by definition the amplitude
modulation (AM) [87]. In general there is no absolute upper limit for the AM
but according to the information that is available on the performance of SOA-
based switches [88] a value that does not exceed 1dB should be acceptable for
the OR gate [89]. The satisfaction of this requirement depends in turn on the
peak power of the data signals A and B, Ppeak as well as on the QD-SOAs
injected current density, J, maximum modal gain gmax, electron relaxation
time from the ES to the GS, and QD-SOAs length. For this purpose in the
following the impact of these critical parameters on the AM is investigated and
assessed through the observation and interpretation of the numerically derived
curves shown in Figs. (5.26-5.30). These figures have been obtained by setting
them to the fixed values Ppeak =12 dBm, J=3kA/cm2, gmax =16 , L=4mm,
and , which are representative of practical QD-SOA devices [58, 59],
and then scanning one after the other.
Fig. (5.26)Variation of AM with the peak input power for two different current
density, keeping other parameters fixed.
Fig. (5.26) depicts the AM against the equal peak power of data signals A and
B for two values of J. As it can be observed, the AM becomes minimum at
around 11dBm for J=2kA/cm2 and 12dBm for J=3kA/cm2 and on either side of
this point there are two distinct areas where the AM is drastically varied in a
parabolic like manner. In order to interpret this behavior we note first that the
change of the data power launched into QD-SOAs 1 and 2 defines their degree
of saturation and subsequently the reduction of their gains whose relative ratio,
G1/G2, determines in turn the amount of the phase difference,
( ) ( ), Eq. (4.8), induced between the copies of the clock
stream. Thus depending on the level of the driving data power this quantity is
made to lie in different intervals [88], which affects analogously the magnitude
of switching and accordingly the AM. In this context we note further with the
help of the truth table of the Boolean OR operation and the transfer function
that can be extracted from Eq. (4.5), (
( )), that there are three values to R depending on the input data binary
pairs. Specifically, ( ) ( ) ( ) ( ) ( ) is not in AM interest
because it corresponds to the zero output. In all times, G2=Gss because QD-
SOA2 receives no data. Therefore, the value of G1 controls the phase difference
and this in turn change the value of cosine in the transfer function. Since
the cosine is a trigonometric function it has minima and maxima. Therefore the
effect of the data power on the AM can be qualitatively estimated by means of
the impact that different at border points of the trigonometric circle have on
the previous transfer functions. More specifically, when the provided peak
Fig. (5.27)Variation of AM with the current density for two different lengths,
keeping other parameters fixed.
Figure (5.27) illustrates the AM versus the QD-SOAs current density.
As it can be seen, for values less than 0.8 kA/cm2 the AM is very high and thus
inadmissible. This is explained by the fact that the current density determines
the saturation power of a QD-SOA [61] in such way that the lower it is the
easier can the QD-SOA be saturated. The result of this correlation is that the
gain deviates much away from the small signal gain and according to
the previous paragraph so do the corresponding transfer functions in which
they are involved, ( ) and ( ) . It can be noted that, as J is increased the
AM is sharply decreased and when it exceeds the threshold of 1 kA/cm2 the
AM becomes almost constant and independent of this critical parameter. This
evolution is a byproduct of the satisfaction of a sufficient and of a necessary
condition, which together make the AM take the specific form. The first
condition is that the amount of injected carriers is such that it can force the
excursions of the QD-SOAs gains to occur closer to the same average level.
The second condition is that the mark density of the OR output is high, i.e.0.75,
and thus it increases the probability of having a pair of marks within the
switched sequence whose amplitude fluctuation is less than 1dB, as required
for the AM. Then provided that J is adjusted to be over its identified minimum
its final selection is dictated by the requirement that the equivalent bias current
can be supplied by the commercially available current sources while it is not
prohibitive for the practical use of QD-SOAs [61]. Therefore a moderate
current density value of the order of 3kA/cm2should be fine for this purpose.
Design Parameters
From the observation and interpretation of Figures (5.26 to 5.29) it can
be deduced that the requirements for the critical parameters are
, , mm,
. Thus using the combination of values
, J= , L=4mm, , and ,
which fall within the specified boundaries, the OR gate can be realized
according to the desired performance standards. This can be seen in Fig. (5.30)
where the OR logic has been executed for convenience of visual representation
between two input data stream shown in Fig. (5.30.a, b). In fact pulses emerge
only at the bit slots of A and B that are not simultaneously occupied by '0' or
equivalently if just one of these inputs is in the 1 logical state, and are
extinguished when both A and B are 1. This proves that the OR operation is
correctly executed for 160 Gb/s RZ data according to its truth table in Fig.
(5.25.b), as it can be alternatively confirmed by comparing the resultant
sequence in Fig. (5.30.c) with the expected one Moreover, despite the non-
uniformity among the output marks, the quality of the logical outcome is more
than adequate since the difference in their amplitudes is strongly suppressed
and its maximum does not exceed 0.5 dB. The existing pattern dependence is
inevitable and cannot be fully eliminated due to the special nature of switching,
which demands combating the amplitude fluctuations of marks that correspond
to a high percentage, 75%, of the output logical possibilities. Despite the
difficulty of this task, the proposed scheme, provided that it is designed by
following the extracted guidelines, can render the AM acceptable to a greater
extent than other SOA-based MZI circuits having the same aim but more
complicated configuration and much lower speed capability [89]. The above
observation is further verified by the pseudo-eye-diagram [10]shown in Fig.
(5.31). An eye-diagram with large eyes indicates a clear transmission with a
low bit rate. Here, we get PED O=89.3%. This value indicates a good response
of the circuit under consideration at its output terminals.
Introduction
Interferometric devices have drawn a great interest in all-optical signal
processing for their high-speed photonic activity. Quantum-dot semiconductor
optical amplifier (QD-SOA) - based gate has added a new momentum in this
field to perform all-optical logic and algebraic operations. In this chapter, for
the first time, a new scheme for all-optical full adder using fife QD-SOA based
MachZehnder interferometers is theoretically investigated and demonstrated.
The proposed scheme is driven by three input data streams; two operands and a
bit carried in from the next less significant stage. The proposed scheme consists
of two XOR, two AND, and one OR gate. The impact of the peak data power
as well as of the QD-SOAs current density, maximum modal gain, and QD-
SOAs length on the ER and Q-factor of the switching outcome are explored
and assessed. The operation of the system is demonstrated with 160 Gbit/s.
section 6.1 introduces a brief review for the all-optical full Adder. Principle and
design of proposed all-optical full adder will be explained in section 6.2.
Section 6.3 is devoted to introduce the numerical simulation results of our
proposed design. Finally, section 6.4 summarizes the concluded remarks that
may be obtained from the displayed results
The demand for faster optical communication networks has been on the
rise in recent years. As mentioned in chapter one to accommodate this demand,
the new generation of optical communication networks is moving towards
terabit per second data rates. Such data rates can be achieved if the data remain
in the optical domain eliminating the need to convert the optical signals to
electronic signals and back to optical signals. Therefore, to be successfully able
to achieve higher data rates, advanced optical networks will require all optical
ultra-fast signal processing such as wavelength conversion, optical logic and
arithmetic processing, adddrop function, etc [90, 91]. All -optical
combinational circuits are required for managing of the contentions and the
switch control in a node of an optical packed switched network. Calculating the
addition of Boolean numbers is an important functionality to perform packet
header processing [92].
In recent years, optical logic gates based on several different schemes
are demonstrated and reported, which are based on dual semiconductor optical
amplifier (SOA) MachZehnder interferometer (MZI), semiconductor laser
amplifier (SLA) loop mirror, ultrafast nonlinear interferometers, and four-wave
mixing process in SOA [72], etc.All-optical binary adders have been reported
by using many optical designs such as terahertz optical asymmetric
demultiplexers (TOADs) [93] and ultrafast nonlinear interferometers [94].An
all-optical half adder using an SOA-assisted Sagnac interferometer has been
suggested and demonstrated by several groups of researches [95-97].
A scheme for an ultrahigh-speed all-optical half adder based on four-wave
mixing in SOAs has been demonstrated by Li [96]. The operation of a half
adder/subtractor arithmetic using the darkbright soliton conversion control has
been reported by Phongsanam et al. [95]. Menezes et al. have suggested all-
optical half adder using the symmetric planar three-core non-linear directional
coupler, operating with a short light pulse [97]. Finally, M. Scaffardi et al. have
introduced all-optical full adder using a single SOA as a basic building block
[92]
Among different topologies, monolithically integrated SOA-based MZI
switches are the most promising candidates for all-optical signal processing
devices due to their compact size, thermal stability, high-speed capability, low
switching energy, relative stability, and optical integration compatibility [98].
The technology of quantum-dot SOAs (QD-SOAs) is very appropriate
owing to its remarkably ultrafast response, which, combined with its attractive
characteristics, distinguishes them from conventional SOAs [99, 100]. They
have larger unsaturated gain than bulk SOAs, resulting in higher power optical
amplifiers, but also have enough gain at low injected currents to enable
operation with low power consumption. Their saturation power can be
controlled by changing the injected current with the unsaturated gain kept
constant, thus enabling easy tuning of the operating point for signal
amplification and processing. QD-SOAs exhibit much faster gain recovery
after gain compression than bulk SOAs, enabling amplification and processing
of short pulses with negligible pulse-shape distortion. The all-optical logic
gates, optical wavelength conversion, and optical regeneration based on a QD-
SOA MZI are promising candidates for faster speed of operation compared
with bulk and MQWSOAs due to the comparatively small electron relaxation
times in QDs [77 - 79].
This chapter introduces for the first time (to the knowledge of the
authors) a theoretical model of an ultrafast all-optical full adder based on fife
QD SOA-MZIs. The all-optical adder is potential to execute the addition in the
optical domain up to 160 Gbit/s.
A B C Sum Carry
0 0 0 0
m 0
y
0 0 1 1 0
0 1 0 1 0
0 1 1 0 1
1 0 0 1 0
1 0 1 0 1
1 1 0 0 1
1 1 1 1 1
(a) (b)
Fig. (6.1) (a) Full adder logic diagram and (b) its Truth table
6.2 Principle and Design of Proposed All-Optical Full Adder
A full adder adds binary numbers and accounts for values carried in as well
as out. A one-bit full adder adds three one-bit numbers, often written as A, B,
and C; A and B are the operands, and C is a bit carried in from the next less
significant stage. The full-adder is usually a component in a cascade of adders,
which add 8, 16, 32, etc. bit wide binary numbers.
As it is known, the full adder is a combinational logic circuit performing
addition of three binary digits. The carry-bit is logic 1 when at least two inputs
have logic 1. Otherwise, it is logic 0. The sum-bit represents the least
significant bit of the three bits binary summation. The truth table shown in Fig.
(6.1.b) explains all the expected cases for the full adder. While, Fig.(6.1.a)
shows the concept upon which the all-optical full adder circuit is designed
[101].
From the standpoint of optics, Fig.(6.2) shows the configuration of the
proposed all-optical full adder. It consists of fife symmetrical QD-SOA-based
MZIs (QD-SOAMZI-1 to QD-SOA MAZI-5) with the same QD-SOA placed
in each of its arms. In the first and second gates (MZI-1, and MZI-2), the first
XOR and the first AND, respectively, the input data are A and B. On the other
hand, the third and fourth gates, (MZI-3, and MZI-4), the second XOR and the
second AND, respectively, the input data are C and the XOR output of the
MZI-1. Finally, the fifth gate, OR (MZI-5), the input data are the AND outputs
of the MZI-2 and MZI-4.The circuit described in Fig. (6.2) can be divided into
three types of gates; XOR, AND, and OR gates which have been explained and
studied in chapter (5). It has to be noting that, the output of any one of these
gates has to be attenuated before using it as input to the next gate to avoid the
cascading amplification.
WSC MZI 3
MZI 1
WSC C1 QD-SOA1 C2
Data A CLK
SUM
C1 QD-SOA1 C2 WSC
CLK Data C
WSC QD-SOA2
Data B
C1 QD-SOA1 C2
Data C
WSC
QD-SOA2
WSC MZI 2
Data A WSC
MZI 5
C1 QD-SOA1 C2
Data B C1 QD-SOA1 C2
CLK
WSC Carry
WSC
QD-SOA2
QD-SOA2
Fig (6.2) Configuration of the proposed all-optical full adder using fife symmetrical
In this QD-SOA model the values of the different parameters are taken from
the literature on other QD-SOA based interferometric gates [77, 79].
All parameters used in the calculations are summarized in Table 6.1.
In order to assess the performance of the full adder circuit at 160 Gbit/s, we
have to choose which output (Carry or SUM)) we will study?And what is the
suitable metric for this purpose?
We will study the Carry bit because it deals with a higher number of gates (4
gates) than the SUM bit (2 gates). With respect to chose the suitable metric,
this can be done with the help of the truth table of the full adder shown in Fig.
(6.1.b).More specifically, we observe that of the eight logical possibilities,
marks (1) and spaces (0) are equal (4 times for each case).Therefore, we must
focus on how to distinguish marks from the spaces. Hence, the most suitable
metric for this purpose is the extinction ratio and the Quality factor[8].
Fig. (6.3) illustrates the effect on the ER with peak power of the input
data signals for three different current density values, when the other
parameters are kept fixed. The common characteristic of all curves is that the
ER is increased with power up to a certain value after that it is decreased. The
maximum value changes and shifted to the right as the current density becomes
larger. From a physical perspective this happens because the current density
determines the power required to alter the optical properties of a QD-SOA and
properly saturate its gain, and the higher it is the more power is necessary for
this purpose [61]. This fact also explains that as we move well enough into the
falling slope of the curves a larger current density is necessary to enhance the
ER and hence improve performance for a given power. This in turn allows
selecting thepeak data power from a wider range of permissible values, which
potentially offers greater flexibility in the design of the full adder circuit.
Fig (6.3)Variation of extinction ratio (ER) with peak data power for different
current densities, keeping other parameters fixed.
Fig (6.4) Variation of extinction ratio (ER) with QD-SOA length for different
peak data power, keeping other parameters fixed.
Figure (6.5) shows the variation of ER with current densities for three different
QD-SOA lengths. From this figure, it is observed that among the three QD-
SOAs lengths the ER becomes acceptable only for 3.5 and 4 mm. This in turn
allows keeping the current density below 4 kA/cm, since after reaching its
defined minimum the ER becomes almost independent of this parameter
because there is an oversupply of carriers and the QD-SOA is sufficiently
biased to the desired point.
Figure(6.6) shows the variation of ER with the maximum modal gain for
three different QD-SOAs lengths. The common characteristic of all curves is
that the ER is increased with the maximum modal gain to a certain value after
that it is dropped. The maximum value is shifted to the right as the QD-SOAs
length becomes smaller. The smaller value of the maximum modal gain
requires the longer QD-SOAs length to obtain the maximum value of ER.
Fig (6.5) Variation of ER with current densities for three different QD-SOAs
length, keeping other parameters fixed
Fig (6.6) Variation of ER with the maximum modal gain for three different
QD-SOAs length, keeping other parameters fixed
Figure (6.7) shows that the ER is very sensitive to the variations of the electron
relaxation time from the ES (excited state) to the GS (ground state) since the
slope of the curve is decreased in an exponential-like manner as this relaxation
time is increased, finally becoming more smoother near the left edge of the
diagram. So the transition time between ES and GS must be kept below some
value and ideally be as fast as possible. In this curve it has to be below 0.25ps
to obtain ER a round 10dB.
Figure (6.8) illustrates the effect on the Q-factor with current densities
for three different maximum modal gains, when the other parameters are kept
constant. The common characteristic of all curves is that the Q-factor is
increased with current density up to a certain value after that it is decreased.
The maximum value of Q-factor is dropped when the value of gmax outside the
certain range. So the value of gmax must be fixed at particular value to achieve
the maximum Q-factor value.
Fig (6.7) Variation of ER with electron relaxation time from the ES to the GS.
Fig (6.8) Variation of Q- Factor with current densities for three different
maximum modal gains, when the other parameters are kept constant.
Fig (6.9) Variation of Q- Factor with current densities for three different QD-
SOA lengths, when the other parameters are kept fixed.
Figure (6.9) shows the variation of Q-factor with current densities for three
different QD-SOA lengths. From this figure, it is observed that among the three
QD-SOAs lengths the Q-factor becomes acceptable only for 4 mm. Thus from
Figs.(6.8 and 6.9), we say that increasing of injection current up to a value
approximately 3.5 kA/cm , more carriers are injected to the wetting layer. Thus
each energy level in the QD can recover to its initial carrier density faster after
carrier depletion by the injected pulse. So, up to a certain value approximately
3.5 kA/cm of the injection current, the Q-factor value increases. After a value
(approximately, above 3.5 kA/cm) of the injection current, more carriers in the
conduction band will be depleted due to stimulated emission. So the QD energy
states will take much longer time to recover their carrier density level, hence
decrease the output quality.
Design Parameters
According thus to Figs. (6.3-6.9) and their interpretation, it can be
inferred that the requirements for the critical parameters are
, , mm,
, and 0.3ps. By following these guidelines
and using the combination of values , J= ,
L=4mm, , and , respectively, which obviously is
not unique and fall within the specified boundaries, a more than adequate Q-
factor of about 12.5dB and ER of about 11.7dB can be obtained, which is
reflected on the high quality of pulse stream obtained at the output [8]. The
input waveforms and the simulated output waveforms are shown in Figs.
(6.10.a-e), respectively. The eye-diagram is the superposition of the outputs for
the repetition period of the inputs. Fig.(6.11) is not a classical eye-diagram
because it is not as informative in the sense that degrading effects, normally
observed in the point-to-point communication links[72], such as noise source,
are added by the detector and optical fibers. This diagram is called a pseudo-
eye-diagram [10]. The relative eye opening (O) is defined as (
) , where and are the minimum and maximum powers at 1-
state and 0-state, respectively. An eye-diagram with large eyes indicates a clear
transmission with a low bit rate. Here, we get PED O=92%. This value
indicates a quite good response of the circuit under consideration at its output
terminals. The cascadeability refers to the ability of a switch to drive directly
from its main output to another input where the signal responsible for switching
is launched and constitutes a key requirement for the perspective of building
combinational circuits. In the cascading stages, the extinction ratio will be
decreased and Q-factor value maybe increased at the output. In the cascading
circuit, high input power at the previous stages is required to provide sufficient
pump power at the input of the subsequent cascading stages. This high pump
power degrades the extinction ratio at the subsequent stages of the subsystem.
Moreover, it leads to improvement of the output Q-factor value and the power
transfer function of the QD-SOA becomes steeper due to the deep gain
saturation caused by the incoming pulses in cross gain modulation operation. It
is also noted that the performance of the subsequent system, which are limited
by the degradation of the extinction ratio is expected to be significantly
enhanced using saturable absorbers[104].
Fig (6.10) waveforms of all-optical full adder, where (a) input data stream A,
(b) input data stream B, (c) input data stream C ( ), (d) output sum-bit S and
(e) output carry-bit ( ).
Fig (6.11)Simulated output waveforms with pseudo-eye-diagram (PED)
Chapter (7)
7.1 Conclusions
The Multi Populations Rate Equations model can be used to study the
semiconductor optical amplifiers and absorbers. In turn, we can use it to
model the Quantum Dot Mode-Locked Lasers using Quantum Dot
Saturable Absorbers
Using the all-optical QD-SOA-based MZI switch introduced in chapter
four to design another optical logic gates such as Not, NOR, NAND, .
Using the all- optical logic gates to design more complex circuits such
as all optical multiplier, Multiplexer, Demultiplexer, All-Optical Flip-
Flop,
4thOrder Runge-KuttaIntegrator
( ) ( )
( )
( )
(A2)
( )
( )
( )
( )
( ) (A5)
[ ( )]
( )
( )
( ) (A6)
[ ( )]
( )
( )
( ) (A7)
[ ( )]
( )
( )
(A8)
( )
[ ]
( )
( )
As it can be seen in Eqs. (A3-A8) our data structure for the output
variable y consists on a matrix having 4 rows (1st forphoton density, 2ndfor the
electron density in WL, 3rdfor the electron occupation probability in the ES and
the fourth for the electron occupation probability in the GS) and as many
columns as the time vector is long (because our implementation is fixed-step-
like). The variables assigned to k0. . . k3are column-vectors because they
contain the solution of every rate equation; these variables are supposed to be
cleared after the end of each step (only after being computed in the output y). In
equations. (A4 A7) the notationkS(Nw, h, f)
stands for the solution of the
corresponding rate equation for photon density (the photon density, the electron
density in WL, electron occupation probability in the ES, electron occupation
probability in the GS).
Multi Population Rate Equation MPRE
For the Multi Population Rate Equation MPRE model, the same
procedure as the simple rate equations model is followed except for that each
equation replaced by a number of equations represents the modes and groups.
Specifically,
Equation of Nwl , Eq. (3.11) remains one equation
Equation of NES, Eq. (3.12) is replaced by n equations, (n number of groups)
Equation of NGS, Eq. (3.13) is replaced by n equations, (n number of groups)
Equations of SES and SGS;Eqs. (3.14, 3.15) each is replaced by m equations, (m
number of modes)
Therefore, if we have 15 groups and 15 modes, the rate equation actually are 61
equations. This means the program will solve 4 x 61 = 244 equations per
iteration. In this situation the function k in Eq. (A1) is one of the rate equations;
Eq. (3.11) to Eq. (3.15) and is represented by
( )
( )
( )
( ) (A9)
[ ( )]
Appendix (B)
Design Metrics
Through this work there are four design metrics have been used to study
the influence of some important parameters on the performance of the device
under test. These metrics are Extinction Ration ER, Amplitude Modulation
AM, Q-Factor, and Pseudo-Eye-Diagram PED. This appendix introduces a
briefreview for these metrics.
(B1)
( ) ( )
which means that this extinction ratio (ER) is performed between the minimum
and maximum peak output powers of the marks and spaces, and ,
respectively, occurring at the output of the device.
This metric must be over 10 dB [8, 106] so that the 1s can be clearly
distinguished from the 0s, which is the basic measure of the capability to
execute Boolean logic functions according to the expected performance
standard. In other words, the higher the ER the more clearly can the 1s be
distinguished from the 0s,
B.3 Q-Factor
Where are the average value of all the out coming "1" and "0" data's
peak power, respectively. are their standard deviations. The Q-factor
is related to the bit error rate in terms of [8]
( ) (B4)
( )
where and are the minimum and maximum powers at 1-state and 0-
state, respectively. An eye-diagram with large eyes indicates a clear
transmission with a low bit rate.
References
( )SOA
0551 0011.
,
.
, .
.
QD-SOA .
.
""
" "
.
.
.
.
.
.
" "
" MPRE
."Multi Populations Rate Equations
.
.
.
" "
.
.
.
.
( ).
" "
. AND XOR .OR
.
"
"
. XOR
AND . OR ,
,
""Extinction Ratio .
.160 Gbit/s.
" "
.
-
( )
( )
. .
-
.. .
2104