Open Access

Dispersion and Nonlinearity Distortion

Compensation of the QPSK/16QAM Signals
Using Optical Phase Conjugation
in Nonlinear SOAs
Volume 12, Number 1, February 2020

Aneesh Sobhanan
Lakshmi Narayanan Venkatasubramani
R. David Koilpillai, Member, IEEE
Deepa Venkitesh, Member, IEEE

DOI: 10.1109/JPHOT.2020.2964414
IEEE Photonics Journal Dispersion and Nonlinearity Distortion Compensation

Dispersion and Nonlinearity Distortion

Compensation of the QPSK/16QAM
Signals Using Optical Phase
Conjugation in Nonlinear SOAs
Aneesh Sobhanan , Lakshmi Narayanan Venkatasubramani ,
R. David Koilpillai, Member, IEEE,
and Deepa Venkitesh , Member, IEEE

Department of Electrical Engineering, Indian Institute of Technology, Madras 600036, India

DOI:10.1109/JPHOT.2020.2964414
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
https://creativecommons.org/licenses/by/4.0/

Manuscript received November 17, 2019; revised December 18, 2019; accepted January 2, 2020. Date
of publication January 6, 2020; date of current version January 24, 2020. This work was supported in
part by the Office of the Principal Scientific Adviser, Govt. of India, in part by the Department of Science
and Technology, Govt. of India, in part by the Ministry of Electronics and Information Technology, Govt.
of India, and in part by the Visvesvaraya PhD scheme, Govt. of India. Corresponding author: Deepa
Venkitesh (e-mail: [email protected]).

Abstract: We experimentally demonstrate the simultaneous compensation of both disper-

sion and nonlinear effects in a 100 km optical fiber link using optical phase conjugation
of a 21 GB aud QPSK and 16QAM signal with nonlinear SOAs. Error-free performance is
recorded for a launched power of up to 12 dBm, without any digital signal processing to
compensate for distortions due to chromatic dispersion and nonlinear effects in fiber. The
performance is verified for operation across the C-band.

Index Terms: Optical phase conjugation, semiconductor optical amplifier.

1. Introduction
Implementation of advanced modulation formats such as PM-QPSK and PM-16QAM has been
possible in optical communication systems through efficient digital signal processing, typically at the
receivers. The most commonly compensated impairments in such optical coherent communication
links include phase noise of the transmitter and receiver lasers, frequency offset between the
lasers, polarisation de-multiplexing, chromatic dispersion and clock recovery. In case of long fiber
links, the overhead in digital signal processing (DSP) is the highest for chromatic dispersion
compensation [1]. Nonlinear impairments in the fiber are typically avoided by controlling the optical
power launched into the fiber. Even though advanced machine learning algorithms have been
used in the recent past for nonlinearity compensation [2], the corresponding signal processing
overheads are typically large. In this context, all-optical techniques - especially Optical Phase
Conjugation (OPC) - with mid-span spectral inversion has proved to be effective in compensating
for impairments induced by both dispersion and nonlinearity in fibers [3]–[7]. In addition, phase
conjugation finds applications in time reversing the scattering processes for countering atmospheric
turbulence and biological imaging in highly turbid samples [8]. Most of these demonstrations have
used highly nonlinear fibers or periodically poled Lithium Niobate as the nonlinear medium - both
of which require power levels, typically larger than 100 mW to initiate nonlinear effects. Nonlinear

Vol. 12, No. 1, February 2020 7800107

IEEE Photonics Journal Dispersion and Nonlinearity Distortion Compensation

semiconductor optical amplifiers (NL-SOAs)-on the other hand are compact and chip-integrable,
where efficient nonlinear effects are observed even for pump power levels of ≈0 dBm. These have
been used in the past to demonstrate various optical signal processing functionalities such as
wavelength conversion [9]. Demonstration of optical phase conjugation with SOAs are however
limited to intensity modulated signals [10], [11]. The most commonly quoted deterrents for these
applications are the nonlinear chirp and the noise figure introduced by the SOA. However, we
recently presented an experimental demonstration of OSNR retention of conjugate and signal
along with chirp-free amplification of the signal with QPSK modulation in the presence of saturating
pump [12]. These results indicate that the nonlinear chirp introduced by the SOA may not offer a
significant penalty in conjugate generation.
In this work we experimentally demonstrate the mid-span spectral inversion (MSSI)-aided non-
linearity and CD compensation of a 21 Gbaud QPSK and 16QAM signal, using a nonlinear SOA
in a two-span link, with each span of length 50 km. The residual penalty due to the nonlinear chirp
introduced by SOA is evaluated by measuring the BER as a function of optical signal to noise
ratio (OSNR) at the receiver. The nonlinearity compensation is demonstrated by evaluating the
performance of the link for different power levels launched in to the fiber. We believe this is the first
demonstration of simultaneous dispersion and nonlinearity compensation using a nonlinear SOA
for QPSK and 16QAM signals.

2. Theory
The mathematical analysis of OPC based Kerr nonlinearity and dispersion compensation is well
established in literature [13]. The signal transmission through a nonlinear (γ ), dispersive (β2 ) and
lossy (α) medium is described by nonlinear Schrodinger equation (NLSE) given as,
∂A(z, t ) i ∂ 2 A(z, t ) α
+ β2 − iγ |A(z, t )|2 A(z, t ) + A(z, t ) = 0. (1)
∂z 2 ∂t 2 2
Equation (1) shows the evolution of the complex envelope of the signal (A) through the first span
of the fiber. Conjugate of the signal (A∗ ) is generated through OPC. The NLSE for the propagation
of conjugate through the second span of the fiber can be written as,
∂A∗ (z, t ) i ∂ 2 A∗ (z, t ) α
− β2 + iγ |A∗ (z, t )|2 A∗ (z, t ) + A∗ (z, t ) = 0. (2)
∂z 2 ∂t 2 2
Note that the signs of the second and third terms in Eqn (1) are inverted for the propagation of con-
jugate. Thus, the dispersive and nonlinear phase accumulated in the first span can be compensated
when the conjugate is transmitted through an equal length of the second span of fiber with identical
β2 and γ , assuming identical launched powers [13]. The optical phase conjugate is generated
using partially degenerate four wave mixing in SOA. Considering a signal at frequency ωs with line
width νs , and pump at frequency ω p with line width ν p are the input to the SOA, the conjugate
at 2ω p − ωs is generated in the SOA with a line width of 4ν p + νs [14]. In addition, the SOA
is expected to introduce nonlinear phase distortions depending on the bias current and the input
power levels. Thus the optical phase conjugate generated in SOA is expected to have a penalty
with respect to the back-to-back back signal performance due to the inherent phase distortion in
SOA and the phase noise transfer during the OPC generation process. However, considering the
compact size and the relatively smaller power levels required for conjugate generation, we proceed
to demonstrate dispersion and non-linearity compensation with an SOA.

3. Experimental Setup
The schematic of the experimental set up to demonstrate dispersion and nonlinearity compensation
with mid-span spectral inversion using SOA is shown in Fig. 1. A narrow-line laser of wavelength
1551.11 nm is modulated with 21 Gbaud QPSK using an IQ modulator driven by an arbitrary wave
generator (AWG, 64 GSa/s). The modulated signal is fed to the first span of standard single mode
fiber (G.652D) of length 50 km through an erbium doped fiber amplifier (EDFA1) and an in-line

Vol. 12, No. 1, February 2020 7800107

IEEE Photonics Journal Dispersion and Nonlinearity Distortion Compensation

Fig. 1. Schematic of the experimental setup to demonstrate dispersion and nonlinearity compensation
using mid-span spectral inversion with a nonlinear SOA. Spectra at the input and filtered conjugate
output are shown as insets for reference.

power monitor (PM) such that the average power at the input of the first span is maintained at
0 dBm. The signal power launched into the OPC stage is maintained at −8 dBm, by appropriately
amplifying it through EDFA2. The pump laser at wavelength 1550.11 nm (corresponding to a
pump-signal detuning of 125 GHz) with 0 dBm power is used to saturate the NL-SOA and initiate
the four wave mixing processes in the SOA. The nonlinear SOA used in our experiments has an
output saturation power of 11.5 dBm and a carrier recovery time of 25 ps. The phase conjugate
is generated at 1549.11 nm due to the partially degenerate four-wave mixing (FWM) process. The
spectra at the input, output of SOA and after filtering is shown as insets in Fig. 1 for reference.
The phase conjugate is filtered, amplified using EDFA3 and is fed to second span of a similar
G.652D fiber of length 50 km. The OSNR of the signal/conjugate is controlled by introducing ASE
noise from an EDFA at the coherent receiver.
The received signal is further digitized using a real-time scope (80 GSa/s). The digitized output is
post-processed using off-line DSP, and the specific algorithms used are discussed in later sections.
For the C-band experiment, both the signal and the pump wavelength are swept over C-band while
maintaining the detuning of 125 GHz between them.

4. Results and Discussion

Figure 2(a) shows the BER performance as a function of OSNR for a different operating conditions.
The “b2b signal,” (shown as points with open red circles) represents the performance of the modu-
lated signal without including the fiber and OPC stage; these results are further used as reference
to calculate the OSNR penalty. The sequence of DSP algorithms used for post processing are:
frequency offset compensation using the periodogram technique, phase noise correction using
the Decision-Directed Least Mean Squares (DD-LMS) algorithm [15], followed by clock recovery
and demodulation [16]. The signal is now fed to the OPC stage and the BER performance of the
conjugate is studied as a function of OSNR of the conjugate. These results are shown as blue open
circles with the legend “b2b-OPC”. The off-line DSP algorithms used is the same as before, and
the BER is estimated with respect to the conjugate data in this case.
Figure 2(a) shows a penalty of about 1 dB for “b2b-OPC” with respect to the back-to-back
case because of (a) phase noise transfer from the pump laser, inherent to the degenerate mixing

Vol. 12, No. 1, February 2020 7800107

IEEE Photonics Journal Dispersion and Nonlinearity Distortion Compensation

Fig. 2. (a) BER performance as a function of different OSNR values for 40 Gbps QPSK signal- b2b:
back-to-back performance without the 100 km link. (b) 21 GBaud QPSK signal Constellation after
50 km for an OSNR of 27 dB for both without digital CD compensation (BER = 0.5, fails in proper
demodulation), and with digital CD compensation (BER = 0).

process [17] and (b) due to the inherent nonlinear phase distortion in SOA. The performance of
the system with 100 km fiber is now ascertained without OPC and are shown in open diamonds
in red. In addition to phase noise compensation and frequency offset correction the off-line DSP in
this case also includes dispersion compensation using frequency domain equalization [1]. These
results are identical to that of the back-to-back case since the dispersion introduced by the fiber is
fully compensated by DSP.
The total ISI in a 50 km link is about 6 symbols while it is about 12 symbols in a 100 km link.
Note that, even in case of 50 km, the data cannot be demodulated without applying a dispersion
compensation algorithm. Fig. 2(b) shows the constellation of 21 Gbaud QPSK signal at the output
of 50 km fiber for an OSNR of 27 dB for the case (a) without digital CD compensation algorithm
(b) with digital CD compensation. It is evident that ISI corresponding to even 6 symbols leads to a
significant distortion. Thus, SOA-based OPC indeed has a significant role to play even when the
ISI is as small as 12 symbols (6 symbols from first fiber span and 6 symbols from next fiber span).
We now include the OPC stage in the mid-span, where the signal after one span is launched at
the input of the nonlinear SOA, along with a CW pump of power 0 dBm, with a detuning of 1 nm
with respect to the signal. A conjugate conversion efficiency of −3 dB is obtained. The conjugate
is further propagated through the second span, and it is expected that the phase accumulated in
the signal due to chromatic dispersion will be compensated in the conjugate wavelength at the end
of the second span. The off-line DSP in this case does not include dispersion compensation. BER
performance is ascertained as a function of conjugate OSNR, and we find a penalty of less than
1.5 dB (at FEC limit BER) with respect to the “b2b-OPC” case. This additional penalty is potentially
due to the nonlinear chirp accumulated in the SOA. An overall penalty of <2.5 dB is observed with
respect to the back-to-back case. This OSNR penalty can be potentially compensated by launching
larger power into the system. However, this may increase the nonlinear penalties. We now proceed
to evaluate the resilience of nonlinearities to the SOA-based OPC generation system.
The power launched into each span is now increased in steps of 2 dBm, and the performance of
the system is evaluated for the case without OPC and with OPC. Care is taken to ensure that the
launched power in each span is identical, while the signal power at the input of the OPC stage is
maintained at −8 dBm. The pump power is maintained at 0 dBm in all cases. The EVM is found

Vol. 12, No. 1, February 2020 7800107

IEEE Photonics Journal Dispersion and Nonlinearity Distortion Compensation

Fig. 3. EVM as a function of input launch power to each 50 km fiber span, in the presence and absence
of OPC (a) 40 Gbps QPSK and (b) 80 Gbps 16 QAM.

out for both cases and shown as a function of power launched in each span. We carry out this
experiment for both QPSK and 16QAM modulations with corresponding data rates of 40 Gbps and
80 Gbps and the results are shown in Fig. 3(a) (a) and 3(b) (b) respectively.
Drop in EVM with the increase in the power per span is due to a corresponding increase in the
OSNR at the input. The drop is more evident in 16-QAM case (Fig. 3(b)) as the phase margin of
the 16QAM format (16.90 ) is much smaller compared to the QPSK format (450 ) [18] and is more
sensitive to even the smaller variation of OSNR. As the input power increases further, the nonlinear
distortions would result in an increase in EVM, as indicated in Fig. 3(a) and 3(b). In the presence
of OPC, the power at which the EVM starts to degrade is larger, indicating the nonlinear tolerance
of the system.
As expected, for smaller values of input powers, the performance of the OPC system is poorer for
both QPSK and 16QAM, which is primarily due to the phase noise transfer from the pump and the
nonlinear chirp due to the SOA. The largest penalty in the smallest launch power (−4 dBm) can be
also attributed to the lower conversion efficiency and the lower output OSNR of the optical phase
conjugate at lower input signal OSNR even though the input signal to SOA is maintained at −8 dBm.
For QPSK and QAM signals, however, the EVM is maintained constant (14% for QPSK and 9.5% for
16QAM within 0.5% change) up to 8 dBm launch power. The data is retrieved with 0 BER for QPSK
and below FEC limit for 16QAM signals even at a launch power of 12 dBm when OPC is used in the
link. For the system without OPC, the EVM penalty starts increasing significantly for input power
levels of larger than 8 dBm due to nonlinear effects introduced in the fiber. The nonlinear effects
accumulated by the signal in the first span is compensated by the conjugate propagation in the
second span, thereby providing reduced penalties even at larger input power levels. To emphasize
the results, QPSK constellations at 12 dBm launch power for both with and without OPC systems
are shown in Fig. 3(a). The BER is increased to 6.4 × 10−2 , which is above FEC limit when system
has no OPC, whereas the system gives error free performance when OPC is included. We thus
prove that, even though the SOA-based OPC has an OSNR penalty of 2.5 dB, the same can be
easily compensated using larger input power levels. We additionally observe that the link distances
can be extended further since a power level of even up to 12 dBm can be launched into the fiber.
Note that even an input power of 12 dBm per span has not resulted in an EVM corresponding to
the FEC limit. Referring to Fig. 3(a) and 3(b), considering an EVM of 20 % for QPSK (and 10 %
for 16QAM), the power margin with OPC is found to be 4 dB (and 2 dB for 16QAM). This power
margin between the cases with OPC and without OPC could be slightly higher if the total mid span
length is higher (say 500 km each) as the nonlinear distortion increases with length. This can be
further experimentally ratified in long-haul systems.
We now proceed to evaluate the performance of the SOA-based OPC across the C-band. Fig. 4,
shows the EVM evaluated for both QPSK and 16 QAM signals in a 100 km link when the launch

Vol. 12, No. 1, February 2020 7800107

IEEE Photonics Journal Dispersion and Nonlinearity Distortion Compensation

Fig. 4. EVM throughout the C-band when OPC is used in the link with a per span launch power of
10 dBm (a) 40 Gbps QPSK and (b) 80 Gbps 16QAM.

power per span is 10 dBm each throughout the C-band. For QPSK, the EVM is around 16%
throughout the C-band and for 16 QAM it is 10% with a ±1% variance. However, all the points
in the Fig. 4(a) has an error free performance, and the points in the Fig. 4(b) has BER below FEC
limit. Thus the OPC generation and its utility for impairment compensation using nonlinear SOA for
entire C-band is experimentally demonstrated successfully.

5. Conclusion
We have experimentally demonstrated the feasibility of nonlinear SOA as a medium for optical
phase conjugation process. Using OPC in NL-SOA a simultaneous compensation of nonlinearity
and chromatic dispersion is achieved with a performance improvement which gives an error free
performance up to 12 dBm launch power. The SOA-based OPC has a penalty of around 2.5 dB in
a link operating in the linear regime. The phase noise transfer in OPC generation and the nonlinear
phase distortion in SOA are the major performance degrading factors. An independent study has
been carried out to reduce this phase distortion from the nonlinear SOAs during the conjugation
process. However, by launching larger power, overall increase in reach can be obtained, and
that too without any digital signal processing for dispersion and nonlinearity compensation. This
scheme can be further extended to polarization multiplexed configuration based on the polarization
insensitive implementation reported in [19]. We also demonstrate conjugate generation in the
entire C-band with the SOA, and hence the results presented in this paper provides a pathway
for all-optical dispersion compensation and nonlinearity compensation for a practical link, using
compact and energy efficient nonlinear SOA. Even though the demonstration in this paper is limited
to 100 km link we do not expect any change in penalty with the use of NL-SOA in a long-haul link,
so long as SOA-OPC stage is placed in the mid-span. Compared to other optical signal processing
media such as HNLF, PPLN, Silicon waveguide which consume several Watts of optical pump
power for efficient conjugate generation, SOA requires only <1 dBm pump power for a good
conversion efficiency. The consequent reduction in the energy per bit for achieving the nonlinear
operation and its compact implementation are compelling reasons to use SOAs as an attractive
choice for optical signal processing applications.

Acknowledgment
The authors would like to thank Sterlite Technologies for providing the fiber and K. Vijay and
Anirudh.V for fruitful technical discussions.

Vol. 12, No. 1, February 2020 7800107

IEEE Photonics Journal Dispersion and Nonlinearity Distortion Compensation

References
[1] S. J. Savory, G. Gavioli, R. I. Killey, and P. Bayvel, “Electronic compensation of chromatic dispersion using a digital
coherent receiver,” Opt. Express, vol. 15, no. 5, pp. 2120–2126, Mar. 2007.
[2] A. Amari, X. Lin, O. A. Dobre, R. Venkatesan, and A. Alvarado, “A machine learning-based detection technique for
optical fiber nonlinearity mitigation,” IEEE Photon. Technol. Lett., vol. 31, no. 8, pp. 627–630, Apr. 2019.
[3] T. Umeki et al., “Simultaneous nonlinearity mitigation in 92× 180-Gbit/s PDM-16QAM transmission over 3840 km using
PPLN-based guard-band-less optical phase conjugation,” Opt. Express, vol. 24, no. 15, pp. 16945–16951, Jul. 2016.
[4] M. A. Z. Al-Khateeb et al., “Experimental demonstration of 72% reach enhancement of 3.6Tbps optical transmission
system using mid-link optical phase conjugation,” Opt. Express, vol. 26, no. 18, pp. 23960–23968, Sep. 2018.
[5] M. Al-Khateeb, M. Tan, T. Zhang, and A. D. Ellis, “Combating fiber nonlinearity using dual-order raman amplification
and OPC,” IEEE Photon. Technol. Lett., vol. 31, no. 11, pp. 877–880, Jun. 2019.
[6] K. R. H. Bottrill, N. Taengnoi, F. Parmigiani, D. J. Richardson, and P. Petropoulos, “PAM4 transmission over 360 km of
fibre using optical phase conjugation,” OSA Continuum, vol. 2, no. 3, pp. 973–982, 2019.
[7] M. A. Z. Al-Khateeb, M. E. McCarthy, C. Snchez, and A. D. Ellis, “Nonlinearity compensation using optical phase
conjugation deployed in discretely amplified transmission systems,” Opt. Express, vol. 26, no. 18, pp. 23945–23959,
Sep. 2018.
[8] A. Mosk, A. Lagendijk, G.Lerosey, and M. Fink, “Controlling waves in space and time for imaging and focusing in
complex media,” Nature Photon. vol. 6, pp. 283–292, 2012.
[9] A. P. Anthur, R. T. Watts, K. Shi, J. O’ Carroll, D. Venkitesh, and L. P. Barry, “Dual correlated pumping scheme for phase
noise preservation in all-optical wavelength conversion,” Opt. Express, vol. 21, no. 13, pp. 15568–15579, Jul. 2013.
[10] M. C. Tatham, G. Sherlock, and L. D. Westbrook, “Compensation fibre chromatic dispersion by optical phase conjuga-
tion in a semiconductor laser amplifier,” Electron. Lett., vol. 29, no. 21, pp. 1851–1852, Oct. 1993.
[11] W. Pieper et al., “Nonlinearity-insensitive standard-fibre transmission based on optical-phase conjugation in a
semiconductor-laser amplifier,” Electron. Lett., vol. 30, no. 9, pp. 724–726, Apr. 1994.
[12] A. Sobhanan, V. A. M. Karthik, L. V. Narayanan, R. D. Koilpillai, and D. Venkitesh, “Experimental analysis of noise
transfer in optical phase conjugation process in nonlinear SOA,” in Proc. Opt. Fiber Commun. Conf. Exhib., 2019,
pp. 1–3.
[13] C. Lorattanasane and K. Kikuchi, “Design theory of long-distance optical transmission systems using midway optical
phase conjugation,” J. Lightw. Technol., vol. 15, no. 6, pp. 948–955, Jun. 1997.
[14] R. Hui amd A. Mecozzi, “Phase noise of fourwave mixing in semiconductor lasers,” Appl. Phys. Lett., vol. 60, no. 20,
pp. 2454–2456, 1992.
[15] I. Fatadin, D. Ives, and S. J. Savory, “Blind equalization and carrier phase recovery in a 16-QAM optical coherent
system,” J. Lightw. Technol. vol. 27, no. 15, pp. 3042–3049, Aug. 2009.
[16] S. J. Varughese, V. Mathew, S. Swain, D. Venkitesh, and R. D. Koilpillai, “200G system with PDM-16QAM: Performance
evaluation and trade-offs,” in Proc. 21st Nat. Conf. Commun., 2015, pp. 1–6.
[17] J. Zhou, R. Hui, and N. Caponio, “Spectral linewidth and frequency chirp of four-wave mixing components in optical
fibers,” IEEE Photon. Technol. Lett., vol. 6, no. 3, pp. 434-436, Mar. 1994.
[18] Mp.Paskov, “Algorithms and Subsystems for Next Generation Optical Networks,” 2015.
[19] A. Sobhanan, L. N. Venkatasubramani, R. D. Koilpillai, and D. Venkitesh, “Polarization-insensitive phase conjugation
of QPSK signal using Bragg-scattering FWM in SOA,” IEEE Photon. Technol. Lett., vol. 31, no. 12, pp. 919–922,
Jun. 2019.

Vol. 12, No. 1, February 2020 7800107