Design of Probabilistic Shaping 4D Ultra High Order Modulation Format With 8APSK Pilot Aid
Design of Probabilistic Shaping 4D Ultra High Order Modulation Format With 8APSK Pilot Aid
Abstract—A new four-dimensional (4D) probabilistically shaped communication services, the internet traffic is growing rapidly
(PS) ultra-high order modulation format is proposed with pilot for center interconnection. Consequently, the development of
aided carrier phase recovery (PA-CPR) algorithm to improve the high-capacity optical communication is necessary. The tech-
general mutual information (GMI) performance. While the struc-
ture of the 4D-PS ultra-high order modulation format uses constel- nology of ultra-high order modulation (UHM) is used to in-
lation set partitioning (SP) based on amplitude translation, and the crease the spectral efficiency (SE) toward the Shannon limit
PA-CPR applies 8-level amplitude phase shift keying (8APSK) pi- and meet the growing demand for high transmission capacity.
lot. To demonstrate the feasibility of the scheme, the PS-1024QAM Many experiments for ultra-high order signals on single-carrier
system is designed on the simulation platform. The results show that have been demonstrated [1]–[10]. The first 1024 QAM single-
the proposed 8APSK-PA-CPR algorithm outperforms the QPSK-
PA-CPR algorithm. The gain of the optimal 8APSK-PA-CPR is carrier coherent optical transmission is demonstrated in [10],
∼0.4 dB compared with the quadrature phase-shift keying (QPSK) in which a 60 Gbit/s polarization-multiplexed signal at a SE
PA-CPR algorithm. The steep drop of blind phase search (BPS) of 13.8 bit/s/Hz was transmitted over 150 km [1]. Similarly,
with different PS factors is overcome by 8APSK-PA-CPR when probabilistically shaped 4096-QAM has been transmitted over
the signal-to-noise (SNR) ratio is lower than 24 dB. The maximum 50.9 km at a potential SE of 15.9 bits/s/Hz and a net bit rate of
improvement of GMI is 1.3 bit/symbol when the SNR is higher
than 26 dB and the complexity of the 8APSK-PA-CPR algorithm 484.4 Gb/s [9]. So far, the highest-order QAM constellations
is reduced by 99.9% compared with the BPS algorithm. The gain that have been demonstrated in optical systems are 10-GBd
obtained from the proposed 4D modulation scheme is ∼2.8 dB probabilistically shaped square 16384-QAM [10] at a net bit
under the condition of GMI = 17.7971 bit/symbol. It demonstrated rate of 223.8 Gb/s. However, as the modulation format order
that this scheme has a strong tolerance to noise, low computational increases, the minimum Euclidean distance decreases, and thus
complexity, and high coding gain. It is a feasible and flexible scheme
for PS high-order modulation format transmission systems. the ability to resist noise declines. Therefore, it is critical to
propose a novel modulation format to expand the Euclidean
Index Terms—Carrier phase recovery, coherent optical distance of constellation points and optimize the DSP algorithm
communications, four-dimensional, pilot aided, probabilistic
shaping, set partitioning.
to improve the system performance.
Therefore, the four-dimensional (4D) modulation format is
I. INTRODUCTION an advanced optical modulation format that is generated by
drawing into the check bit through modulation coding in the two
ITH the rapid development of mobile communication
W networks, live video broadcast, telemedicine, and other
quadrature-phase components and two polarizations of the opti-
cal field. It has attracted a lot of attention in recent years because
Manuscript received November 8, 2021; revised January 12, 2022 and Febru- it can increase the minimum Euclidean distance between sym-
ary 10, 2022; accepted February 13, 2022. Date of publication February 23, bols [11]–[14]. The 4D modulation format uses Ungerboeck’s
2022; date of current version June 16, 2022. This work was supported in part by
the National Key R&D program of China under Grant 2020YFB1805805 and in set partitioning strategy to generate the parity by XOR operation
part by the National Natural Science Foundation of China (NSFC) under Grants of information bits [11], such as 128-level SP quadrature ampli-
61875248, 62021005, and 62027819. (Corresponding author: Feng Tian.) tude modulation (128-SP-QAM) and 2048-SP-QAM. Moreover,
Feng Tian, Tianze Wu, Yiqing Ji, Chuxuan Wang, Qi Zhang, and Qinghua
Tian are with the School of Electronic Engineering, Beijing University probabilistic shaping (PS) is a technology that can increase the
of Posts and Telecommunications, Beijing 100876, China (e-mail: tian- mutual information of signals toward the Shannon limit and
feng@[Link]; wutianze@[Link]; jyiqing@[Link]; wangchux- provide a flexible platform for rate adaptation. PS-based multi-
uan@[Link]; zhangqi@[Link]; tianqh@[Link]).
Ran Gao and Zhipei Li are with the School of Information and Elec- dimensional signaling schemes have recently been reported [15],
tronics, Beijing Institute of Technology, Beijing 100081, China (e-mail: gao- [16] to provide higher fiber nonlinear tolerance. The PS scheme,
ran198412@[Link]; lizhipei@[Link]). however, is incompatible with the SP-QAM modulation format
Xiangjun Xin is with the School of Electronic Engineering, Beijing University
of Posts and Telecommunications, Beijing 100876, China, and also with the due to the presence of parity bits.
School of Information and Electronics, Beijing Institute of Technology, Beijing Besides, the scheme of UHM is more sensitive for channel
100081, China (e-mail: xjxin@[Link]). noise, especially for phase noise. Moreover, the reported phase
Color versions of one or more figures in this article are available at
[Link] recovery algorithms are not suitable for the 4D-PS-MQAM
Digital Object Identifier 10.1109/JLT.2022.3153321 signal. It’s necessary to develop the advanced carrier phase
0733-8724 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See [Link] for more information.
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
TIAN et al.: DESIGN OF PROBABILISTIC SHAPING 4D ULTRA HIGH ORDER MODULATION 3689
√
recovery (CPR) algorithm to improve transmission performance. M-QAM√ into two independent M -level pulse amplitude mod-
Many CPR algorithms have been reported, such as blind search ulation ( M -PAM) signals. The PAS structure of each com-
algorithm (BPS), quadrature phase-shift keying (QPSK) parti- ponent includes a constant component distribution matcher
tioning, maximum likelihood estimation (BPS/ML), quadratic (CCDM) and a forward error correction (FEC) encoder [24].
approximation (BPS/QA) algorithm, and Viterbi-Viterbi (V-V) The n binary bits are divided into n1 and n2 parts. Firstly,
algorithm [17]–[21]. The BPS algorithm have the mutual in- n1 bits are entered√into CCDM to generate m non-uniform
formation punishment with the PS signal at low signal-to-noise positive amplitude M /2-PAM signals,√ and the information
ratio (SNR) [22]. The QPSK partitioning algorithm based on rate of the obtained positive amplitude M /2-PAM signal is
constellation-dependent is inappropriate to phase recovery for β=n1 /m bit/symbol. Then, the obtained symbols from CCDM
high-order modulation format signals. The two-stage BPS al- are reflected into binary labels, which are combined with n2
gorithm has large computational complexity after introducing bits together to enter the FEC encoder for generating parity
a two-stage configuration. The pilot-aided carrier phase re- bits. These
√ parity bits and n2 are used as the sign bits to
covery (PA-CPR) algorithm has high noise tolerance and low form a M -PAM signal following Maxwell-Boltzmann (MB)
computational complexity. However, the main drawback is the distribution [25]. The MB distribution can be expressed as:
inherent overhead (OH) which will reduce the SE of the system. 2
2
Therefore, it is necessary to optimize the PA-CPR scheme for PX (x) = e−λx /Z(x ), Z(x ) = e−λx (1)
x ∈X
PS-4D-SP-1024QAM. √
In this paper, the scheme of PS-4D-SP-1024QAM combined where x represents the M -PAM constellation points and λ
with the 8-level amplitude phase shift keying aided pilot carrier denotes the shaping factor.
phase algorithm (8APSK-PA-CPR) is proposed. The amplitude
B. Generation of PS-4D-SP-1024QAM
translation method is used to generate the parity for the SP
modulation format. The redundancy generated from the parity The structure of PS-4D-SP-1024QAM is shown in Fig. 2,
of PS-4D-SP-1024QAM will be carried by the pilot to improve which can maintain probability distribution characteristics after
the SE. At the receiver, the information of the pilot is taken out SP encoding. The n bits enter the processing unit of DP-PAS
for signal demapping after preliminary recovery and decision and pass through the SP to perform the amplitude translation in
of pilot symbols. Different laser linewidths are used to simulate the first stage which generate m symbols and 2m binary label
the 20-GBaud PDM-PS-1024QAM and PS-4D-SP-1024QAM bits respectively. In the second stage, the mentioned 2m binary
transmission systems. The results demonstrate that the proposed label bits are used to generate 2k symbols and 4k label bits in
8APSK-PA-CPR algorithm outperforms the QPSK-PA-CPR al- the same process. Finally, the bits coming from the second stage
gorithm at average power. The gain improved by the optimal are mapped into 2k/3 PDM-8APSK symbols, which is used as
8APSK-PA-CPR is ∼0.4 dB compared with the quadrature the pilots for CPR in optical fiber communication systems.
phase-shift keying (QPSK) PA-CPR algorithm. 8APSK-PA- Fig. 3 shows the generation of SP-QAM modulation which is
CPR overcomes the steep drop in BPS with different PS factors performed by XOR coding on the information bits, cf. Fig. 3(c).
when the SNR is less than 24 dB. The maximum improvement In PS systems, the coding after shaping at the transmitter
of GMI is 1.3 bit/symbol when the SNR is higher than 26 dB causes the shaped symbol distribution to be distorted, shown in
and the complexity of the 8APSK-PA-CPR algorithm is reduced Fig. 3(a). On the other hand, XOR coding before shaping at the
by 99.9% compared with the BPS algorithm. The SNR gain transmitter, has no practical effect on decoding at the receiver,
obtained from PS-4D-SP-1024QAM with 8APSK-PA-CPR is Fig. 3(b). In this section, the amplitude translation method is
∼2.8 dB. proposed to overcome the incompatibility between SP and PS
for ultra-high order constellations.
II. PRINCIPLE OF PS-4D-SP-1024QAM Fig. 4 shows that the 1024QAM constellation is divided into
four subsets (A, B, C, and D) named subset constraint (SSC). The
A. Probabilistic Amplitude Shaping corresponding 4D-1024QAM signals are generated according to
Probabilistic amplitude shaping (PAS) is a mainstream stan- the subfamily constraints (SFC) in Table I.
dard probabilistic shaping scheme [23]. As shown in Fig. 1, it The minimum Euclidean distance of the 4D-1024QAM can be
divides the in-phase and quadrature components of the square expanded by twice. Firstly, the non-uniform amplitude set of one
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
3690 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 40, NO. 12, JUNE 15, 2022
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
TIAN et al.: DESIGN OF PROBABILISTIC SHAPING 4D ULTRA HIGH ORDER MODULATION 3691
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
3692 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 40, NO. 12, JUNE 15, 2022
III. PS-4D-SP-1024QAM PHASE RECOVERY SCHEME Fig. 9. The implementation of pilot aided CPR scheme.
A. Phase Noise
The DSP algorithm in coherent optical communication in-
cludes dispersion compensation, clock recovery, polarization
demultiplexing, frequency offset estimation, and phase recov-
ery. Except for phase noise, all the channel impairments are
compensated in this paper. The received signal at discrete time
k can be written as:
yk = xk ejθk + nk (8)
where xk is the transmitted symbol, θk is laser phase noise, nk
is AWGN additive white Gaussian noise. The phase noise can
be modeled as a Wiener process, and is given by (9):
t
θt = wi (9)
i=−∞
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
TIAN et al.: DESIGN OF PROBABILISTIC SHAPING 4D ULTRA HIGH ORDER MODULATION 3693
Fig. 11. Simulation setup for the 20Gbaud PDM-PS-1024QAM and PS-4D-
SP-1024QAM systems.
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
3694 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 40, NO. 12, JUNE 15, 2022
Fig. 14. GMI results versus λ with different phase recovery algorithms in PS-1024QAM systems when the SNRs are (a) 20 dB. (b) 23 dB. (c) 26 dB. (d) 29 dB.
ceiling because the effect of AWGN for the pilot format has algorithm with PS, even though increasing the window size
been minimized when it increases to a certain extent. In Fig. 12, reduces the sudden drop in GMI. The impact of different PS
a 15 level is adopted in QPSK-PA-CPR to maintain the average factors on GMI in the 8PASK-PA-CPR scheme is evaluated.
transmitted power in the system. Due to the increase in the Fig. 14 shows the GMI performance of 8PASK-PA-CPR and
average energy of the transmitted signal, the system requires BPS (N = 100, B = 128) in different PS factors λ. The SNRs
higher transmitted power to keep the same operating SNR. The are 20 dB, 23 dB, 26 dB and 29 dB respectively. For SNR
selection of CPR pilots at the average energy is more appropriate. = 20 dB (Fig. 14(a)) the sudden drop of GMI under BPS is
In Fig. 13, the AoIC = 13 and AoOC = 17 is chosen as close obvious from 0 to 0.06. 8APSK-PA-CPR does not cause a sudden
as possible to the average power to carry out the amplitude of drop in the whole measurement interval of PS factor λ. In the
the 8APSK-PA-CPR scheme to avoid the change of the overall condition of low SNR, the additive white noise is dominant and
average power. the 8APSK-PA-CPR has accurate recovery on phase noise. The
BPS algorithm is a blind estimation of AWGN and phase noise.
The performance of 8APSK-PA-CPR is inferior to that of blind
B. Compatibility Evaluation of 8APSK-PA-CPR in phase recovery algorithm when the AWGN is too large to be
Probabilistic Shaping System accurately estimated. In this situation of SNR, the performance
In PS systems, the GMI will suddenly decline at the optimal of 8APSK is greatly affected by AWGN compared with BPS,
value of λ [17] when using the BPS algorithm for phase recovery resulting in a gap of 0.5 bit/symbol. Moreover, the FEC threshold
at low SNR. The BPS algorithm has a strong dependence on of GMI corresponding to 3.8×10−3 cannot be reached. For
PS, even for long noise rejection windows. The decisions made SNR = 23 dB (Fig. 14(b)), the sudden drop caused by the
inside the BPS algorithm are affected by shaping. The impact of BPS algorithm is alleviated, and the GMI of 8APSK-PA-CPR
PS on BPS can affect the overall system performance, specially increases by ∼0.1 bit/symbol compared with that of BPS. The
at low SNRs [21]. It reflects the incompatibility of the BPS GMI performance of 8APSK-PA-CPR is better than that of
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
TIAN et al.: DESIGN OF PROBABILISTIC SHAPING 4D ULTRA HIGH ORDER MODULATION 3695
Fig. 15. GMI results versus SNR with different polarizations in PDM-PS-1024QAM and PS-4D- SP-1024QAM systems when the value of shaping parameter
is 0.02, the linewidths are (a) 10 kHz. (b) 50 kHz. (c) 100 kHz. (d) 200 kHz. .
BPS as the value of SNR increases. For example, the maximum there is a larger gap between PS-1024QAM and Shannon limit
GMI increases by 0.7 bit/symbol when SNR = 26 dB, and the [4] The Euclidean distance is expanded by multidimensional
maximum GMI increases by 1.3 bit/symbol when SNR = 29 set-partition, so the noise tolerance of PS-4D-SP-1024QAM is
dB, approaching the limit (see Fig. 14(c) and (d)). The GMI higher than that of PS-1024QAM. As shown in Fig. 15, the
with 8APSK-PA-CPR can reach the FEC threshold at this SNR, SNR gain of PS-4D-SP-1024QAM is ∼2.8 dB compared with
when the shaping parameter is greater than 0.02. It is acceptable PDM-PS-1024QAM. The gain of 2.8 dB is the theoretical gain of
to obtain significant gain at higher SNR (greater than 26 dB). modulation format. For linewidth = 10 kHz (Fig. 15(a)) the gaps
between Shannon limit with the GMI of PS-4D-SP-1024QAM
and that of PDM-PS-1024QAM is 0.5 dB and 3.5 dB respec-
C. Evaluation of PS-4D-SP-1024QAM and 8APSK-PA-CPR tively when the GMI is 8 bit/symbol. For linewidth = 200 kHz
Performance (Fig. 15(d)) the gaps rise to 1.3 dB and 4.6 dB, respectively.
In the 8APSK-PA-CPR scheme, the overhead generated by The sensitivity of PS-4D-SP-1024QAM decreases by only 0.7
the second stage is put into the pilot symbol to obtain additional dB from the linewidth 10 kHz to 200 kHz, while the sensitivity
mutual information and increase the transmission efficiency of of PDM-PS-1024QAM decreases by 1.1 dB, indicating that the
the system. Fig. 15. shows the SNR and GMI curves of PDM-PS- proposed scheme has a high phase noise tolerance and better
1024QAM and PS-4D-SP-1024QAM applied with the 8APSK- pilot phase recovery performance behavior. The penalty between
PA-CPR scheme. The PS factor is 0.02, and the linewidths are X polarization and Y polarization of PS-4D-SP-1024QAM for-
set to 10 kHz, 50 kHz, 100 kHz, and 200 kHz respectively. mat is ∼0.15 dB due to the amplitude translation taken in the
In the simulation, the HD decoder is used at the receiver and 4D format. The maximum GMI of PS-4D-SP-1024QAM can
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
3696 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 40, NO. 12, JUNE 15, 2022
TABLE II REFERENCES
COMPUTATIONAL COMPLEXITY OF DIFFERENT ALGORITHMS
[1] Y. Koizumi, K. Toyoda, M. Yoshida, and M. Nakazawa, “1024 QAM (60
Gbit/s) single-carrier coherent optical transmission over 150 km,” Opt.
Exp., vol. 20, no. 11, pp. 12508–12514, 2012, doi: 10.1364/oe.20.012508.
[2] S. Beppu, K. Kasai, M. Yoshida, and M. Nakazawa, “2048 QAM (66Gbit/s)
single-carrier coherent optical transmission over 150 km with a potential
SE of 15.3 bit/s/Hz,” Opt. Exp., vol. 23, no. 4, pp. 4960–4969, 2015,
doi: 10.1364/oe.23.004960.
[3] R. Maher, K. Croussore, M. Lauermann, R. Going, X. Xu, and J. Rahn,
“Constellation shaped 66 GBd DP-1024QAM transceiver with 400 km
transmission over standard SMF,” in Proc. Eur. Conf. Opt. Commun.,
Gothenburg, Sweden, 2017, pp. 1–3.
[4] S. Okamoto, M. Terayama, M. Yoshida, K. Kasai, T. Hirooka, and M.
be achieved at 17.7971 bit/symbol with the increasing of SNR Nakazawa, “Experimental and numerical comparison of probabilistically
for the linewidth at the FEC threshold of 3.8×10−3 . The final shaped 4096 QAM and a uniformly shaped 1024 QAM in all-Raman
net bit rate is 293.7 Gbit/s considering a 7% FEC overhead, a amplified 160 km transmission,” Opt. Exp., vol. 26, no. 3, pp. 3535–3543,
Feb. 2018.
pilot insertion ratio of 1/27 and redundant symbols from the first [5] S. L. Olsson, J. Cho, S. Chandrasekhar, X. Chen, E. C. Burrows, and P. J.
stage. Winzer, “Record-high 17.3-bit/s/Hz spectral efficiency transmission over
50 km using probabilistically shaped PDM 4096-QAM,” in Proc. Opt.
Fiber Commun. Conf., 2018, pp. 1–3.
D. Computational Complexity Analysis [6] S. L. Olsson, J. Cho, S. Chandrasekhar, X. Chen, P. J. Winzer, and S.
Makovejs, “Probabilistically shaped PDM 4096-QAM transmission over
The complexity of the DSP algorithm is important for prac- up to 200 km of fiber using standard intradyne detection,” Opt. Exp.,
tical application. The low complexity algorithm can effectively vol. 26, no. 4, pp. 4522–4530, Feb. 2018, doi: 10.1364/oe.26.004522.
[7] M. Terayama, S. Okamoto, K. Kasai, M. Yoshida, and M. Nakazawa,
reduce the system power consumption and increase the speed “4096 QAM (72 Gbit/s) single-carrier coherent optical transmission with
of hardware processing. As shown in Table II, the complexity a potential SE of 15.8 bit/s/Hz in all-Raman amplified 160 km fiber link,”
includes four aspects: real multipliers, real adders, comparison, in Proc. Opt. Fiber Commun. Conf. Expo., 2018, pp. 1–3.
[8] Y. Wang, S. Okamoto, K. Kasai, M. Yoshida, and M. Nakazawa, “Single-
and decision operation. The parameters are selected when the channel 200 Gbit/s, 10 Gsymbol/s-1024 QAM injection-locked coherent
three algorithms are based on optimum implementations. Tak- transmission over 160 km with a pilot-assisted adaptive equalizer,” Opt.
ing the linewidth of 100 kHz as an example (N = 100, B = Exp., vol. 26, no. 13, pp. 17015–17024, 2018, doi: 10.1364/oe.26.017015.
[9] X. Chen, S. Chandrasekhar, J. Cho, and P. Winzer, “Transmission of 30-
128), the DP-BPS algorithm requires 12NB+2B real multiplier, GBd polarization-multiplexed probabilistically shaped 4096-QAM over
12NB+2B real adders, 2B comparison, and 2NB decision oper- 50.9-km SSMF,” Opt. Exp., vol. 27, no. 21, pp. 29916–29923, Oct. 2019.
ation. In the 8APSK-PA-CPR scheme the DC-VV algorithm is [10] X. Chen, J. Cho, A. Adamiecki, and P. Winzer, “16384-QAM transmission
at 10 GBd over 25-km SSMF using polarization-multiplexed probabilistic
used to estimate the phase of the 8APSK pilot, which increases constellation shaping,” in Proc. 45th Eur. Conf. Opt. Commun., Dublin,
a small amount of complexity (2 Comparison) compared with Ireland, 2019, pp. 1–4.
QPSK-PA-CPR. The complexity of the 8APSK-PA-CPR algo- [11] T. A. Eriksson, M. Sjödin, P. Johannisson, P. A. Andrekson, and M. Karls-
son, “Comparison of 128-SP-QAM and PM-16QAM in long-haul WDM
rithm is also reduced by 99.99% compared with the DP-BPS transmission,” Opt. Exp., vol. 21, no. 16, pp. 19269–19279, Aug. 2013.
algorithm. [12] J. K. Fischer et al., “Generation, transmission, and detection of 4-D set par-
titioning QAM signals,” J. Lightw. Technol., vol. 33, no. 7, pp. 1445–1451,
Apr. 2015.
V. CONCLUSION [13] J. Lu et al., “Carrier phase recovery for set-partitioning QAM formats,” J.
Lightw. Technol., vol. 36, no. 18, pp. 4129–4137, Sep. 2018.
We demonstrate a scheme of probabilistic shaping 4D ultra- [14] P. J. Winzer, “High-spectral-efficiency optical modulation formats,” J.
high order modulation format with pilot aided carrier phase re- Lightw. Technol., vol. 30, no. 24, pp. 3824–3835, Dec. 2012.
[15] F. Steiner et al., “Experimental verification of rate flexibility and prob-
covery. The four-dimensional set partitioning modulation format abilistic shaping by 4D signaling,” in Proc. Opt. Fiber Commun. Conf.
based on amplitude translation in probabilistic shaping systems Exhibit., San Diego, CA, USA, 2018, pp. 1–3.
is designed to approach the Shannon limit. The 8APSK is used as [16] Q. Guo, W. Peng, C. Yan, and Y. Bai, “Multi-dimensional probabilistic
shaping for higher fibre nonlinearity tolerance,” in Proc. Eur. Conf. Opt.
the pilot format to optimize the performance of the CPR scheme. Commun., Dublin, Ireland, 2019, pp. 1–3.
It is demonstrated that 8APSK-PA-CPR shows great resilience [17] J. Hu, Z. Sun, X. xu, Y. Lu, and Y. Qiao, “Carrier phase recovery exploiting
to phase noise compared with the traditional PA-CPR and BPS recursive blind phase search algorithm for probabilistic shaping 64QAM
systems,” in Proc. 2020 Asia Commun. Photon. Conf. Int. Conf. Inf.
algorithm. The SNR gain of the 8APSK-PA-CPR scheme ob- Photon. Opt. Commun., Beijing, China, 2020, pp. 1–3.
tained by amplitude optimization is ∼0.4 dB compared with the [18] I. Fatadin, D. Ives, and S. J. Savory, “Laser linewidth tolerance for 16-
traditional QPSK-PA-CPR. The steep drop of BPS with different QAM coherent optical systems using QPSK partitioning,” IEEE Photon.
Technol. Lett., vol. 22, no. 9, pp. 631–633, May 2010.
PS factors is overcome by 8APSK-PA-CPR when the SNR ratio [19] Y. Gao et al., “Low-complexity two-stage carrier phase estimation for
is lower than 24 dB. The maximum improvement of the general- 16-QAM systems using QPSK partitioning and maximum likelihood
ized mutual information (GMI) is 1.3 bit/symbol when the SNR detection,” in Proc. Conf. Opt. Fiber Commun., Los Angeles, CA, USA,
2011, pp. 1–3.
is higher than 26 dB. The complexity of the 8APSK-PA-CPR [20] Q. Zhang and C. Shu, “Viterbi and Viterbi algorithm based phase recovery
algorithm is reduced by 99.9% compared with the BPS. We for probabilistically shaped signals,” IEEE J. Lightw. Technol., vol. 39,
demonstrated the performance of PS-4D- SP-1024QAM with no. 5, pp. 1364–1370, Mar. 2021.
[21] J. Lu, S. Fu, M. Tang, M. Xiang, P. Shum, and D. Liu, “Low-complexity
8APSK-PA-CPR under different laser linewidth. The simulation carrier phase estimation for M-ary QAM based on blind phase search
results show that the proposed scheme obtained ∼2.8 dB SNR using simplified measurement,” in Proc. Int. Conf. Opt. Commun. Netw.,
gain compared with PDM-PS-1024QAM. Hangzhou, China, 2016, pp. 1–3.
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.
TIAN et al.: DESIGN OF PROBABILISTIC SHAPING 4D ULTRA HIGH ORDER MODULATION 3697
[22] D. A. A. Mello, F. A. Barbosa, and J. D. Reis, “Interplay of probabilistic Qi Zhang (Member, IEEE) received the Ph.D. degree from the Beijing Univer-
shaping and the blind phase search algorithm,” IEEE J. Lightw. Technol., sity of Posts and Telecommunications, Beijing, China, in 2005. Her research
vol. 36, no. 22, pp. 5096–5105, Nov. 2018. interests include optical communication and satellite communication.
[23] J. Cho, X. Chen, S. Chandrasekhar, and P. Winzer, “On line rates, infor-
mation rates, and spectral efficiencies in probabilistically shaped QAM
systems,” Opt. Exp., vol. 26, no. 8, pp. 9784–9791, 2018.
[24] G. Böcherer, F. Steiner, and P. Schulte, “Bandwidth efficient and rate
matched low-density parity-check coded modulation,” IEEE Trans. Com-
mun., vol. 63, no. 12, pp. 4651–4665, Dec. 2015.
[25] R. Kschischang and S. Pasupathy, “Optimal nonuniform signaling for
Gaussian channels,” IEEE Trans. Inf. Theory, vol. 39, no. 3, pp. 913–929, Ran Gao (Member, IEEE) received the Ph.D. degree in electronic science and
May 1993. technology from the Beijing Institute of Technology, Beijing, China, in 2015. He
[26] G. Böcherer, “Probabilistic signal shaping for bit-metric decoding,” in is currently a Professor with the School of Information and Electronics, Beijing
Proc. IEEE Int. Symp. Inf. Theory, Honolulu, HI, USA, 2014, pp. 431–435. Institute of Technology. His research interests include fiber optical sensors,
[27] M. Morsy-Osman, Q. Zhuge, L. R. Chen, and D. V. Plant, “Feedfor- optical waveguide, and measurement instruments.
ward carrier recovery via pilot-aided transmission for single-carrier sys-
tems with arbitrary M-QAM constellations,” Opt. Exp., vol. 19, no. 24,
pp. 24331–24343, 2011.
[28] H. Cui, X. Xu, S. Liu, Y. Lu, and Y. Qiao, “QPSK pilot-aided carrier
phase recovery algorithm for 800-Gb/s/λ DP-256QAM transmission,” in
Proc. Asia Commun. Photon. Conf./Int. Conf. Inf. Photon. Opt. Commun.,
Beijing, China, 2020, pp. 1–3.
[29] A. Sheikh, A. Graell i Amat, and G. Liva, “Achievable information rates Zhipei Li received the B.S. degree in microelectronics from Harbin Engineering
for coded modulation with hard decision decoding for coherent fiber-optic University, Harbin, China, in 2013, and the Ph.D. degree in electronic science
systems,” J. Lightw. Technol., vol. 35, no. 23, pp. 5069–5078, Dec. 2017. and technology from the Beijing University of Posts and Telecommunications,
Beijing, China, in 2019. He was with Transmission and Access Research
Department of Huawei for one year. He is currently working on postdoctoral
research with the School of Information and Electronics, Beijing Institute of
Feng Tian received the Ph.D. degree from the Beijing University of Posts and Technology, Beijing, China. He is also engaged in research on high-speed
Telecommunications, Beijing, China, in 2013. Her research interests include fiber communication systems including high-baudrate and high-order modu-
high-speed fiber communication systems and digital signal processing. lation format signal generation, impairment compensation for high-bandwidth
optoelectronic devices, beyond 1Tb/s transmission in single channel, and other
coherent digital signal processing techniques.
Tianze Wu received the B.S. degree from Yanbian University, Yanji, China,
in 2018. He is currently working toward the Ph.D. degree with the Beijing
University of Posts and Telecommunications, Beijing, China. His research
interests include high-capacity fiber communication system, coded modulation,
and probabilistic shaping.
Qinghua Tian received the Ph.D. degree from the Beijing University of Posts
and Telecommunications, Beijing, China, in 2013. Her research interests include
optical communications and satellite communications.
Yiqing Ji received the B.S. degree from Xi’an Shiyou University, Xian, China,
in 2020. Her research interests include mutiband fiber communication systems
and digital signal processing.
Chuxuan Wang received the B.S. degree in electronic science and technology Xiangjun Xin received the Ph.D. degree from the Beijing University of Posts
in 2017 from the Beijing University of Posts and Telecommunications, Beijing, and Telecommunications, Beijing, China, in 2004. His research interests include
China, where he is currently working toward the Ph.D. degree. His current high-speed fiber communication systems, broadband optical transmission tech-
research interests include coherent optical communication system, few-mode nologies, and all optical networks.
fiber, and nonlinear compensation.
Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on September 26,2023 at [Link] UTC from IEEE Xplore. Restrictions apply.