While probabilistic constellation shaping (PCS) enables rate and reach adaption with finer granularity [1] (Cho and Winzer, 2009), it imposes signal processing challenges at the receiver. Since the distribution of PCS-quadrature amplitude modulation (QAM) signals tends to be Gaussian, conventional blind polarization demultiplexing algorithms are not suitable for them [2] (Johnson et al., 1998). It is known that independently and identically distributed (iid) Gaussian signals, when mixed, cannot be recovered/separated from their mixture. For PCS-QAM signals, there are algorithms such as [3] and [4] Dris et al. (2019) and Athuraliya et al. (2004) which are designed by extending conventional blind algorithms used for uniform QAM signals. In these algorithms, an initialization point is obtained by processing only a part of the mixed signal, which have non-Gaussian statistics. In this article, we propose an alternative method wherein we add temporal correlations at the transmitter, which are subsequently exploited at the receiver in order to separate the polarizations. We will refer to the proposed method as frequency domain (FD) joint diagonalization (JD) probability aware-multi modulus algorithm (pr-MMA), and it is suited to channels with moderate polarization mode dispersion (PMD) effects. Furthermore, we extend our previously proposed JD-MMA [5] (Bajaj et al., 2022) by replacing the standard MMA with a pr-MMA, improving its performance. Both FDJD-pr-MMA and JD-pr-MMA are evaluated for a diverse range of PCS (entropy $\mathcal {H}$) of 64-QAM over a first-order PMD channel that is simulated in a proof-of-concept setup. A MMA initialized with a memoryless constant modulus algorithm (CMA) is used as a benchmark. We show that at a differential group delay (DGD) of 10% of symbol period T$_{\text{symb}}$ and 18 dB SNR/pol., JD-pr-MMA successfully demultiplexes the PCS signals, while CMA-MMA fails drastically. Furthermore, we demonstrate that the newly proposed FDJD-pr-MMA is robust against moderate PMD effects by evaluating it over a DGD of up to 40% of T$_{\text{symb}}$. Our results show that the proposed FDJD-pr-MMA successfully equalizes PMD channels with a DGD up to 20% of T$_{\text{symb}}$.

@en

We propose a novel method for blind polarization-demultiplexing of probabilistically shaped signals for coherent receivers. The method is capable of separating signals with (quasi) Gaussian distributions by exploiting temporal correlations added to the transmit signals. The proposed method is evaluated in challenging mixing scenarios.

@en

We present a simple, efficient “direct learning” approach to train Volterra series-based pre-distortion filters using neural networks. We show its superior performance over conventional training methods using a 64-QAM 64 GBaud simulated transmitter with varying transmitter nonlinearity and noisy conditions.@en

High-symbol-rate coherent optical transceivers suffer more from the critical responses of transceiver components at high frequency, especially when applying a higher order modulation format. Recently, we proposed in [20] a neural network (NN)-based digital pre-distortion (DPD) technique trained to mitigate the transceiver response of a 128~GBaud optical coherent transmission system. In this paper, we further detail this work and assess the NN-based DPD by training it using either a direct learning architecture (DLA) or an indirect learning architecture (ILA), and compare performance against a Volterra series-based DPD and a linear DPD. Furthermore, we willfully increase the transmitter nonlinearity and compare the performance of the three DPDs considered. The proposed NN-based DPD trained using DLA performs the best among the three contenders, providing more than 1~dB signal-to-noise ratio (SNR) gains for uniform 64-quadrature amplitude modulation (QAM) and PCS-256-QAM signals at the output of a conventional coherent receiver DSP. Finally, the NN-based DPD enables achieving a record 1.61~Tb/s net rate transmission on a single channel after 80~km of standard single mode fiber (SSMF).

@en

We demonstrate a record 54.5 Tb/s WDM transmission at 11.35 bit/s/Hz over 48 km of field-deployed SMF connecting business and academic parks enabled by a novel joint I-Q Neural Network-based transmitter digital pre-distortion technique.

@en

We propose an efficient neural-network-based equalization jointly compensating fiber and transceiver nonlinearities for high-symbol-rate coherent short-reach links. Providing about 0.9 dB extra SNR gain, it allows achieving experimentally the record single-channel 1.48 Tbps net rate over 240 km G.652 fiber.

@en

We propose a novel digital pre-distortion (DPD) based on neural networks for high-baudrate optical coherent transmitters. We demonstrate experimentally that it outperforms an optimized linear DPD giving a 1.2 dB SNR gain in a 128GBaud PCS-256QAM single-channel transmission over 80km of standard single-mode fiber resulting in a record 1.61 Tb/s net data rate.@en

Nonlinear frequency division multiplexing (NFDM) techniques encode information in the so-called nonlinear spectrum which is obtained from the nonlinear Fourier transform (NFT) of a signal. NFDM techniques so far have been applied to the nonlinear Schrödinger equation (NLSE) that models signal propagation in a lossless fiber. Conventionally, the true lossy NLSE is approximated by a lossless NLSE using the path-average approach which makes the propagation model suitable for NFDM. The error of the path-average approximation depends strongly on signal power, bandwidth and the span length. It can degrade the performance of NFDM systems and imposes challenges on designing high data rate NFDM systems. Previously, we proposed the idea of using dispersion decreasing fiber (DDF) for NFDM systems. These DDFs can be modeled by a NLSE with varying-parameters that can be solved with a specialized NFT without approximation errors. We have shown in simulations that complete nonlinearity mitigation can be achieved in lossy fibers by designing an NFDM system with DDF if a properly adapted NFT is used. We reported performance gains by avoiding the aforementioned path-average error in an NFDM system by modulating the discrete part of the nonlinear spectrum. In this paper, we extend the proposed idea to the modulation of continuous spectrum. We compare the performance of NFDM systems designed with dispersion decreasing fiber to that of systems designed with a standard fiber with the path-average model. Next to the conventional path-average model, we furthermore compare the proposed system with an optimized path-average model in which amplifier locations can be adapted. We quantify the improvement in the performance of NFDM systems that use DDF through numerical simulations.@en

The path-average approximation penalizes NFDM transmission over lumped amplified fiber links.We investigate suitably tapered lossy fibers to overcome the approximation error induced by the path average, making the NFDM transmission exact. Error vector magnitude gains up to 4.8 dB are observed.@en