Вы находитесь на странице: 1из 6

Comparison of optical back propagation schemes for ber-optic communications

Mahdi Malekiha

, Dong Yang, Shiva Kumar


Electrical and Computer Engineering Department, McMaster University, Canada
a r t i c l e i n f o
Article history:
Received 17 May 2012
Revised 18 September 2012
Available online 25 October 2012
Keywords:
Fiber-optic communications
Highly nonlinear ber-optic
Optical back propagation
Nonlinear phase noise
a b s t r a c t
Optical back propagation techniques that utilize two highly nonlinear bers to compensate for transmis-
sion ber nonlinear effects are analyzed. When the step-size is equal to the amplier spacing, inline com-
pensation of ber nonlinearity without inline dispersion compensation provides the best performance.
2012 Elsevier Inc. All rights reserved.
1. Introduction
The impairments caused by ber dispersion and nonlinear ef-
fects can be mitigated by solving the nonlinear Schrodinger equa-
tion in digital domain with the sign of distance being reversed. This
is known as digital back propagation and it has drawn much atten-
tion [14]. Recently, optical back propagation (OBP) using highly
nonlinear bers is proposed [5]. An advantage of the optical back
propagation is that impairments due to inter-channel nonlinear ef-
fects in wavelength division multiplexing (WDM) systems can be
compensated whereas digital back propagation require large
amount of computation resources for mitigation of the WDM non-
linear impairments. In this paper, we compare various types of
optical back propagation techniques such as inline OBP, receiver-
based OBP and inline nonlinear compensation.
In receiver-based OBP, the OBP module is placed at the end of the
ber-optic link. It consists of dispersion compensation bers (DCFs)
and nonlinear compensators (NLCs). Nonlinear compensator is
realized using two highly nonlinear bers (HNLFs). The rst HNLF
imparts a nonlinear phase shift that is equal to that due to transmis-
sion ber propagation. The second HNLF reverses the sign of the
phase shift. The combined effect is that the sign of the nonlinear
coefcient of the NLC is opposite of that of the transmission ber.
In inline-based OBP, the OBP module is placed at each amplier site.
The DCF and NLC compensate for the dispersion and nonlinear ef-
fects of the preceding transmission ber, respectively. In this paper,
compensation of dispersion and nonlinearity of a single polarization
system is studied. The proposed scheme required modications to
compensate the impairments of a dual polarization system which
will be the subject of future study.
In the case of inline nonlinear compensators, the NLC module
is placed at each amplier site. The ber dispersion is compen-
sated at the receiver either optically or electrically. When the
propagation step-size Dz of the back propagator is equal to the
amplier spacing, we found that inline nonlinear compensation
outperforms receiver-based OBP and inline-based OBP. This is
probably due to the fact that in the case of inline NLC, pulses
signicantly broaden due to the absence of inline DCF and the
system becomes more quasi-linear. However, as the step-size
becomes much smaller than the amplier spacing, the inline
OBP outperforms other schemes. This is because if the inline
OBP is ideal (Dz ?0), it would fully compensate for dispersion
and nonlinear effects of the preceding ber system and the
system becomes linear. In other words, the nonlinear interaction
between signal and amplier spontaneous emission (ASE) leading
to GordonMollenauer phase noise [6] is absent in the system
with inline ideal OBP.
In Section 2, theoretical background on optical/digital back
propagation is reviewed and various types of OBP schemes are
discussed. Section 3 focusses on numerical simulation of digital
and optical back propagation schemes. The sensitivity of the OBP
scheme to the ber dispersion variations is also discussed in
Section 3. Finally, Section 4 provides the conclusion.
2. Theoretical background on backward propagation
The evolution of eld envelope in a ber-optic system is
described by the nonlinear Schrodinger equation (NLSE):
i
@q
@z

b
2
2
@
2
q
@t
2
ca
2
zjqj
2
qi
a
2
qiG1

N
n1
dznL
a
qt; nL
a
; 1
where b
2
, c, a are the ber dispersion, nonlinear and loss coefcients,
respectively. Here G and L
a
represent the amplier gain and spacing,
1068-5200/$ - see front matter 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.yofte.2012.09.007

Corresponding author.
E-mail address: mahdi.malekiha@mail.mcgill.ca (M. Malekiha).
Optical Fiber Technology 19 (2013) 49
Contents lists available at SciVerse ScienceDirect
Optical Fiber Technology
www. el sevi er . com/ l ocat e/ yof t e
respectively. For simplicity, we assumed that the bers in the link
are identical and the ampliers are equally spaced and amplier
gain compensates the loss exactly. Using the transformation
qz; t azuz; t; 2
where
az exp
modz; L
a
a
2
_ _
: 3
Eq. (1) can be written as
@u
@z
i

N

D
_ _
u; 4
where

D denotes the ber dispersion effect

Dt
b
2
2
@
2
@t
2
; 5
and

N denotes the ber nonlinear effect

Nt; s ca
2
sjut; sj
2
; 6
The formal solution of Eq. (4) can be obtained as follows.
du
u
i

N

D
_ _
; 7
or
ut; L
tot
Mut; 0; 8
where
M exp i
_
Ltot
0

Dt

Nt; s
_ _
ds
_ _
; 9
and L
tot
is the total transmission distance. Multiplying Eq. (8) by
M
1
on both sides, we obtain
ut; 0 M
1
ut; L
tot
: 10
u(t,L
tot
) represents the received eld envelope which is distorted
due to ber dispersion and nonlinear effects. So, If we multiply
the received eld by the inverse ber operator, M
1
, distortions
due to ber propagation can be completely undone. Since
exp~x exp~x 1 11
for any operator ~x, we nd
M
1
exp i
_
Ltot
0

Dt

Nt; s
_ _
ds
_ _
: 12
Eq. (10) with M
1
given by Eq. (12) is equivalent to solving the fol-
lowing equations
@u
b
@z
i

N
b


D
b
_ _
u
b
13
with u
b
(t, 0) = u(t, L
tot
) and

D
b
t

Dt
b
2
2
@
2
@t
2
; 14

N
b
t; s ca
2
sju
b
t; sj
2
: 15
Now, Eq. (10), becomes
u
b
t; L
tot
M
1
u
b
t; 0 M
1
ut; L
tot
16
with
M
1
exp i
_
Ltot
0

D
b
t

N
b
t; s
_ _
ds
_ _
: 17
Eq. (13) can be solved in digital domain to reverse the effects of
ber dispersion and nonlinearity. In this paper, we consider the
optical techniques to realize the back propagation. To realize M
1
,
we utilize the symmetric split-step Fourier technique with step-size
of Dz, [5,7]
M
1
% At; 0; Dz=2Bt; DzAt; Dz=2; 3Dz=2Bt; 2Dz . . .
Bt; L
tot
At; L
tot
Dz=2; L
tot
; 18
where
At; x; y expiDty x; 19
Bt; x exp i
_
x
xDz
a
2
zju
b
t; zj
2
dz
_ _
: 20
Here, A(t, x, y) represents the ber dispersive effects over the inter-
val [x y] and B(t, x) represents the nonlinear phase shift accumu-
lated over a ber length Dz at x. Typically, the function a
2
(z)
varies more rapidly than ju
b
(t, z)j
2
and therefore, in the interval
[x Dz x], ju
b
(t, z)j
2
can be approximated to be independent of z,
i.e. ju
b
(t, z)j
2
% ju
b
(t, x)j
2
and now, the integral in Eq. (20) can be
evaluated analytically using Eq. (3) as
Bt; x exp
_
icDz
eff
expmodx; L
a
aju
b
t; xj
2
_
; 21
where
Dz
eff
m1 expaL
a
=a f1 expmodDz; L
a
ag=a; 22
with m = oor(Dz/L
a
). When the step-size is an integral multiple of
the amplier spacing, i.e., Dz = m L
a
, m being an integer. The opera-
tor B takes the form
Bt; x expficm1 expaL
a
jut; xj
2
=ag: 23
2.1. Optical back propagation at the receiver
Fig. 1a shows the schematic of the ber-optic system that em-
ploys the optical back propagation module at the receiver. The
operator A can be realized using a dispersion compensating ber
(DCF) or ber Bragg grating (FBG). The rst DCF of OBP compensates
for the half of the accumulated dispersion of the last span of the -
ber-optic link assuming that the step-size is equal to the amplier
spacing. The launch power to DCF is low so that nonlinear effects
in DCF can be ignored. If there exists a ber with negative nonlinear
coefcient, the operator B can be easily realized through self-phase
modulation (SPM) in such a ber. Unfortunately, such a ber is not
available. Instead, the operator B is realized using a nonlinear com-
pensator (NLC) whose schematic is shown in Fig. 1b. An amplier is
used to compensate for the loss of DCF and NLC. First the output sig-
nal of the ber-optic link, u
s
ut; L
tot
=

2
p
and a CW pump beam
(pump 1) of frequency X (relative to optical carrier frequency) are
launched into a highly nonlinear ber HNLF1. The eld envelope
of the pump beam after propagating a distance L
1
in a dispersion-
free HNLF1 is given by [7]
u
p1
t; L
1
u
p1
t; 0 expi2c
1
L
eff ;1
ju
s
tj
2
ih
1
iXt; 24
and
h
1
c
1
L
eff ;1
ju
p1
t; 0j
2
; 25
where L
eff,1
and c
1
are the effective length and nonlinear coefcient
of HNLF1, respectively. In Eq. (24), the rst and second terms on the
exponent represent the cross-phase modulation (XPM) and SPM,
respectively. Since the pump is CW, the phase shift due to SPM, h
1
is a constant. L
eff,1
and c
1
are so chosen that the nonlinear phase
shift due to XPM is equal in magnitude to the phase shift given
by Eq. (23), but of opposite sign,
c
1
L
eff ;1
cm1 expaL
a
=a: 26
M. Malekiha et al. / Optical Fiber Technology 19 (2013) 49 5
In the other words, HNLF1 introduces a phase shift that is the same
as transmission ber. To reverse the sign of the phase shift, we pro-
ceed as follows. First a band pass lter is introduced which attenu-
ates the signal, but transmits the modulated pump. The signal u
s
and the pump beam at the output of HNLF1 are multiplexed using
a WDM coupler and its output is launched into a highly nonlinear
ber HNLF2. A second CW pump beam, pump 2 that has the same
frequency as pump 1 is launched into HNLF2. The polarization of
pump 2 is chosen to be orthogonal to signal polarization. Typically,
highly nonlinear bers are birefringent bers and the eld envelope
of pump 2 at the output of HNLF2 is given by [7]
u
p2
t; L
2
u
p2
t; 0e
i2c
2
L
eff ;2
justu
p1
t;L
1
j
2
=3ih
2
iXt
; 27
and
h
2
c
2
L
eff ;2
ju
p2
t; 0j
2
; 28
where L
eff,2
and c
2
are the effective length and nonlinear coefcient
of HNLF2, respectively. As before, rst and second terms on the
exponent of Eq. (27) represent the XPM and SPM effects, respec-
tively. We choose the nonlinear coefcient c
2
, effective length L
eff,2
,
powers of signal and pump1 such that the phase shift due to XPM in
HNLF2 is quite small,
2c
2
L
eff ;2
ju
s
t u
p1
t; L
1
j
2
=3 (p: 29
Under these conditions, Eq. (27) can be approximated as
u
p2
t; L
2
%u
p2
t; 0e
ih
2
Xt
12ic
2
L
eff ;2
ju
s
t u
p1
t; L
1
j
2
=3: 30
So, Eq. (30) can be expanded as
u
p2
t; L
2
% u
p2
t; 0e
ih
2
1 2ic
2
L
eff ;2
ju
s
j
2
ju
p1
t; 0j
2
e
iXt
=3
_
2ic
2
L
eff ;2
u
s
u
H
p1
t; 0e
i2c
1
L
eff ;1
jus tj
2
ih
1
=3
2ic
2
L
eff ;2
u
H
s
u
p1
t; 0e
i2c
1
L
eff ;1
jus tj
2
2iXtih
1
=3
_
: 31
The rst, second and third terms on the right hand side of Eq. (31)
correspond to optical signals centered around X, 0 and 2X, respec-
tively. If we introduce an optical band pass lter (BPF2) with its cen-
ter frequency at zero and bandwidth equal to the bandwidth of
u
s
(which is much smaller than X), frequency components centered
around X and 2X will be rejected by BPF2. Therefore, the output of
BPF2 is
u
out
Ku
s
t expi2c
1
L
eff ;1
ju
s
tj
2
; 32
where
K i2c
2
L
eff ;2
u
p2
t; 0u
I
p1
t; 0 expih
2
h
1
=3 33
is a complex constant. Thus, the phase shift provided by the OBP is
of the form given by Eq. (23) which is required for back propagation.
The phase of K is a constant and it has no impact on direct detection
systems. When coherent detection is employed, this phase shift is
removed by digital signal processing (DSP) unit of the coherent re-
ceiver. In this paper, we choose
jKj 1; 34
so that there is no loss due to NLC. The desired component of output
has a polarization that is orthogonal to the original signal u
s
(t) and
hence it can be removed using a polarization lter.
2.2. Inline optical back propagation
In this paper, we investigate the possibility of introducing an
OBP module at each amplier site. The system schematic is shown
in Fig. 2a. An optical signal acquires phase that is proportional to
power because of the Kerr effect. Therefore the power uctuation
due to amplied spontaneous emission (ASE) noise, can be trans-
lated into phase uctuations which is known as GordonMol-
lenauer phase noise [6]. Although the DBP and OBP at the
receiver can compensate for deterministic nonlinear impairments,
it cannot compensate for GordonMollenauer phase noise. How-
ever if OBP is placed before an amplier and if it is ideal, it would
exactly compensate for the ber nonlinearity and the signal output
of the amplier would be linear. As a result, the Gordon-
Mollenauer phase noise does not occur in such an ideal system.
This issue is discussed further in Sections 3 and 3.4.
Fig. 1. (a) Block diagram of a ber-optic link with optical back propagation. (b) Block diagram of NLC. Tx = Transmitter, RX = Receiver, HNLF = Highly nonlinear ber,
BPF1 = Band-pass lter and BPF2 = Band-pass lter and polarization lter.
Fig. 2. Schematic of inline compensation techniques: (a) Inline OBP. (b) Inline NLC.
6 M. Malekiha et al. / Optical Fiber Technology 19 (2013) 49
2.3. Inline nonlinearity compensation scheme (INC)
In Fig. 2a, inline OBP compensate for both dispersion and non-
linearity. In Fig. 2b, the nonlinearity of the preceding ber span
is compensated by the inline NLC and dispersion is compensated
at the receiver either in optical domain or in electrical domain.
NLC is realized using two HNLFs as discussed before.
3. Numerical simulation and results
In this section, we compare the performance of the DBP and var-
ious types of OBP. The numerical simulation of a single channel -
ber-optic link with DBP/OBP is carried out with the following
parameters: symbol rate = 25 GSym/s, modulation = 32 quadrature
amplitude modulation (32-QAM), transmission ber dispersion,
b
2
= 21 ps
2
/km, transmission ber loss coefcient = 0.2 dB/km,
nonlinear coefcient c = 1.1 W
1
km
1
, amplier spacing L
a
= 80 -
km, spontaneous emission noise factor n
sp
= 1.5 for inline ampli-
ers and OBP ampliers, dispersion of DCF = 130 ps
2
/km, loss of
DCF = 0.4 dB/km, nonlinear coefcient of DCF = 4.2 W
1
km
1
, non-
linear coefcient of HNLF1 ber = 2 W
1
m
1
, effective length of
HNLF1 = 232 m, pump1 power = 10 mW, frequency of the pump1
(and also pump 2), X = 400 GHz, the full bandwidth of the optical
band pass lters = 87.5 GHz. A pseudo-random bit sequence (PRBS)
of length 2
18
is simulated. The nonlinear coefcient and effective
length of HNLF2 are chosen as 2 W
1
m
1
and 200 m, respectively,
and power of pump 2 = 5.625 mW. In Sections 3.1,3.2,3.3, we as-
sumed that the step-size is equal to the amplier spacing. To test
the validity of the approximations done in Eq. (31), NLS equations
corresponding to HNLF1 and HNLF2 are solved numerically.
Although the proposed technique could compensate for inter-
channel nonlinear impairments of a WDM system, we have fo-
cused only on the single channel since the simulation of a WDM
system with OBP takes enormous amount of computational time.
The simulation of a WDM system is deferred to a future study.
3.1. NLC parameters optimization
Fig. 3 shows the bit error rate (BER) as a function of nonlinearity
coefcient of HNLF2, c
2
. When
2c
2
L
eff ;2
ju
s
t u
p1
t; L
1
j
2
=3 (p: 35
the receiver-based OBP and DBP should lead to the same solution for
the NLS equation for the given step-size, and therefore, they should
have the same performance. When the product of c
2
L
eff,2
is small en-
ough (61 W
1
m
1
), these two schemes should have the same per-
formance, but when the product is large, the linear approximation
done in Eq. (31) for XPMbreaks down leading to performance degra-
dations. On the other hand, if the product is too low, NLC would not
compensate for transmission ber nonlinear effects. In our simula-
tions for OBP, we have optimized all of the parameters to get the best
performance. FromFig. 3, we see that the BER is minimumwhen c
2
-
= 2 W
1
m
1
. In this case, Eq. (35) is not satised and higher order
terms represented in Eq. (27) are contributing to the nal solution
which surprisingly leads to the better system performance.
3.2. Optical BP vs. digital BP
Fig. 4 shows the BER as a function of average launch power to
the ber optic link for DBP, receiver-based OBP, inline OBP and in-
line nonlinearity compensation, for transmission distance of
1040 km. As can be seen, OBP at the receiver outperforms DBP.
To understand the difference between two, we simulated the sys-
tem with DBP by taking 2, 4 and 8 samples/symbol at the receiver.
We found that the BER is roughly same in all cases. The perfor-
mance advantage of the receiver-based OBP over DBP is actually
due to the optimization of c
2
discussed in Section 3.1. From
Fig. 4, we see that the inline nonlinearity compensation (INC) has
the best performance. First, let us compare INC with inline OBP.
The INC performs better because when there is no inline dispersion
compensation, the pulse broaden signicantly and the system be-
come more quasi-linear. We expected that inline OBP to perform
better than receiver-based OBP since the ideal inline OBP could
fully compensate for Gordon-Mollenauer phase noise. However,
when the step-size is equal to the amplier spacing, the deter-
ministic (or bit-pattern dependent) nonlinear impairments are
larger than the stochastic nonlinear impairments due to
signalnoise interaction. When the step-size becomes very small,
the inline OBP provides a signicant performance advantage which
will be discussed in Section 3.4.
Next, let us compare INC with receiver-based OBP and DBP. In
every span, the ber imparts a nonlinear phase shift which is even-
tually converted to amplitude uctuations due to ber dispersion.
If the nonlinear phase shift imparted by the ber in a span is com-
pensated at the end of each span, the system would operate closer
to the linear regime and hence nonlinear impairments can be min-
imized. The signal available to the receiver-based OBP and DBP is
already corrupted by noise due to a long chain of ampliers.
Although the receiver-based OBP and DBP (with step-size = ampli-
er spacing) provide nonlinearity compensation similar to INC, the
signal used in nonlinearity compensation of the receiver-based
OBP and DBP has more noise.
0 0.5 1 1.5 2 2.5 3 3.5
10
4
10
3
10
2
10
1

HNLF2
(W
1
m
1
)
B
E
R
ReciverBased OBP
FEC limit, 2.1 10
3
Fig. 3. Bit error rate vs. nonlinearity coefcient of HNLF2. Transmission
distance = 800 km and L
eff,2
= 200 m.
8 6 4 2 0 2 4 6
Power (dBm)
DBP
Inline OBP
ReciverBased OBP
Inline Nonlinearity Compensation
FEC limit, 2.1 10
3
10
4
10
3
10
2
10
1
B
E
R
Fig. 4. Bit error rate vs average launch power to the ber-optic link.
M. Malekiha et al. / Optical Fiber Technology 19 (2013) 49 7
In Fig. 4, the performance of inline OBP is worse than that of re-
ceiver-based OBP and DBP with step-size = amplier spacing. This
can be understood from the fact that the DBP and receiver-based
OBP do not have inline dispersion compensation making them
more quasi-linear. However when the step-size is very small, inline
OBP can undo most of the nonlinear impairments of the preceding
span and the system would operate closer to a linear regime.
Fig. 5 shows the BER as a function of transmission reach. Each
point in Fig. 5 is obtained after optimizing the launch power. The
transmission reach without any type of back-propagation (but
with DCF) is limited to 240 km at the forward error correction
(FEC) limit of 2.1 10
3
. This is because the multi-level QAM sig-
nals are highly sensitive to ber nonlinear effects. The maximum
reach can be increased to 720 km, 1040 km and 1440 km using in-
line OBP, receiver-based OBP and inline nonlinearity compensa-
tion, respectively and in the case of DBP the maximum reach can
be increased to 800 km.
For a linear system, there is no limit on the maximum reach.
The signal power can be increased arbitrarily to combat the noise.
The OSNR required for QAM-32 to reach a BER of 2.1 10
3
in a
linear system is 26.8 dB. When we turned off nonlinearity, the
numerically measured OSNR at 1040 km (corresponding to simula-
tion results shown in Fig. 4) with 0 dBm launch power is 29 dB. In
Fig. 4, the BER at 0 dBm launch power for receiver-based OBP is
close to 2.1 10
3
. The additional OSNR of $2 dB (as compared
to the theoretical OSNR of 26.8 dB of a linear system) required in
the nonlinear system is due to the nonlinear interference noise.
3.3. Sensitivity analysis
Eqs. (24) and (27) are only valid if the dispersion is equal to
zero, but in reality it is hard to maintain zero dispersion through-
out the ber length during the manufacturing process and disper-
sion value varies slowly along the ber. Let us introduce the term
called dispersion correlation length (DCL) which is the length of
the ber segment over which dispersion can be assumed to have
a constant value and dispersion can be modeled as a Gaussian ran-
dom variable with zero mean and the standard deviation r. With
the advanced fabrication techniques r can be smaller than
0.5 ps
2
/km [8].
We use the split-step Fourier scheme (SSFM) to investigate the
impact of dispersion of HNLFs. Fig. 6 shows the BER as a function of
dispersion standard deviation of HNLF1 and HNLF2 for receiver-
based OBP. As can be seen, system performance is hardly affected
for the r smaller than 0.5 ps
2
/km, and therefore, these systems
have reasonably good tolerance to random uctuations in ber dis-
persions, when r < 0.5 ps
2
/km.
3.4. Nearly ideal optical back propagation
So far we assumed that the step-size is equal to amplier spac-
ing. In this section, we compare the inline and receiver based OBP
when the step-size is very small. The transmission reach can be
signicantly enhanced as shown in Fig. 7, using the step-size,
Dz = 80 m. Note that the step-size of 80 m does not correspond
to physical length of DCF, HNLF or erbium-doped ber amplier
(EDFA). It refers to the length of the transmission ber whose dis-
persion and nonlinear effects needs to compensated using the OBP
module. For example, to compensate the dispersion of 80 m long
transmission ber, ber Bragg grating (FBG) can be used (which
is only a few cm long) whose length is independent of the amount
of dispersion compensation. To compensate the nonlinear phase
shift due to 80 m long transmission ber, HNLFs are used. Their
lengths need not be of the order of 80 m, but the nonlinear phase
shift provided by the NLC should have the same magnitude as that
imparted by the 80 m long transmission ber. This can be achieved
by lowering the pump powers in NLC without changing the lengths
of HNLFs.
In receiver-based OBP, when the step-size is very small, the
deterministic (bit-pattern dependent) nonlinear effects can be
compensated. However, the signal-amplier noise nonlinear inter-
actions leading to Gordon-Mollenauepr phase noise cannot be
compensated. In contrast, the inline OBP with very small step-size
would compensate for the nonlinear effects before the amplier so
that signal-amplier noise nonlinear interactions are absent in this
case. In other words, the system is linear and transmission capacity
is limited only by signal-to-noise ratio.
400 600 800 1000 1200 1400 1600

FEC limit, 2.1 10


3
Transmission Distance (Km)
DBP
Inline OBP
ReciverBased OBP
Inline Nonlinearity
10
4
10
3
10
2
B
E
R
Compensation
Fig. 5. Bit error rate vs transmission reach.
3600 3800 4000 4200 4400 4600
10
2.8
10
2.7
10
2.6
FEC limit, 2.1 10
3
Transmission Distance (Km)
B
E
R
Inline OBP
RxBased OBP
Fig. 7. Bit error rate vs transmission reach. Dz = 80 m.
0 0.1 0.2 0.3 0.4 0.5 0.6
10
2.7
10
2.6
10
2.5
Dispersion Standard Deviation (ps
2
/Km)
B
E
R
Dispersion correlation length = 1m
Dispersion correlation length = 5m
FEC limit, 2.1 10
3
Fig. 6. Bit error rate vs dispersion standard deviation of HNLF1 and HNLF2 for
receiver-based OBP. Transmission distance = 1040 km.
8 M. Malekiha et al. / Optical Fiber Technology 19 (2013) 49
Fig. 8 shows the BER as a function of average launch power to the
ber optic link at a transmission distance = 4400 km. As can be seen
from Figs. 7 and 8, inline OBP outperforms the receiver based OBP.
In Fig. 7, the maximum reach is limited by the nonlinear signal
noise interactions for the receiver based OBP rather than by
deterministic nonlinear impairments. If the step-size is reduced
further, the maximum reach for the receiver based OBP does not
increases, but in the case of inline OBP, the reach increases inde-
nitely (limited only by the launch power) as the system becomes
nearly linear.
4. Conclusion
We compared various optical back propagation techniques to
mitigate ber transmission impairments. When the step-size is
equal to the amplier spacing, inline nonlinear compensation tech-
nique provides the best performance. When the step-size is much
smaller than the amplier spacing, inline OBP outperforms the re-
ceiver-based OBP as the signalnoise nonlinear interactions can be
suppressed by inline OBP.
The DBP requires intensive computation and therefore, it is cur-
rently limited to off-line signal processing. In contrast, OBP could
be used in real time and it is compatible with WDM.
References
[1] E. Ip, J. Kahn, J. Lightw. Technol. 26 (2008) 3416.
[2] X. Li, X. Chen, G. Gokifarb, E. Mateo, I. Kim, F. Yaman, G. Li, Opt. Express 16
(2008) 880.
[3] S. Oda, T.Tamimura, T. Hoshida, C. Ohshima, H. Nakashima, T. Zhenning, J.C.
Rasmussen, in: Optical Fiber Conference, San Diego, CA, 2009.
[4] D.S. Millar, S. Makovejs, C. Behrens, S. Hellerbrand, R.I. Killey, P. Bayvel, S.J.
Savory, IEEE J. Sel. Top. Quant. Electron. 6 (2010) 1217.
[5] S. Kumar, D. Yang, Opt. Lett. 36 (2011) 1038.
[6] J.P. Gordon, L.F. Mollenauer, Opt. Lett. 15 (1990) 1351.
[7] G.P. Agrawal, Nonlinear Fiber Optics, fourth ed., Academic Press, New York,
2007 (Chapters 2, 6, 7 and 11).
[8] V. Srikant, Corning Inc., Private Communication.
3 2 1 0 1 2 3 4 5 6 7
10
3
10
2
10
1
Power (dBm)
B
E
R
Inline OBP
RxBased
FEC limit, 2.1 10
3
OBP
Fig. 8. Bit error rate vs average launch power to the ber-optic link. Dz = 80 m and
Transmission distance = 4400 km.
M. Malekiha et al. / Optical Fiber Technology 19 (2013) 49 9

Вам также может понравиться