Probabilistic Shaping for Nonlinearity Tolerance

TL;DR

This tutorial surveys probabilistic shaping as a means to improve nonlinearity tolerance in optical fiber channels, using a first-order time-domain perturbation model and a linear filter perspective to connect shaping gains with NLIN and channel memory. It analyzes both linear shaping gains from MB-like distributions and nonlinear shaping gains arising from amplitude modulation induced noise, showing that memory and block length critically influence finite-length shaping performance. The paper introduces extended models (windowed moments, finite-memory EGN, and a linear NLIN filter) and evaluates practical shaping approaches (CCDM/ESS), mapping strategies, and sequence-selection PAS, including sign-bit integration, to demonstrate measurable gains and trade-offs. Collectively, these insights guide the design of nonlinearity-tolerant PAS systems and motivate future work on data-driven sequence selection and CPR-aware shaping.

Abstract

Optimizing the input probability distribution of a discrete-time channel is a standard step in the information-theoretic analysis of digital communication systems. Nevertheless, many practical communication systems transmit uniformly and independently distributed symbols drawn from regular constellation sets. The introduction of the probabilistic amplitude shaping architecture has renewed interest in using optimized probability distributions, i.e., probabilistic shaping. Traditionally, probabilistic shaping has been employed to reduce the transmit power required for a given information rate over additive noise channels. While this translates into substantive performance gains for optical fiber communication systems, the interaction of shaping and fiber nonlinearity has posed intriguing questions. At first glance, probabilistic shaping seems to exacerbate nonlinear interference noise (NLIN) due to larger higher-order standardized moments. Therefore, the optimization of shaping distributions must differ from those used for linear channels. Secondly, finite-length effects related to the memory of the nonlinear fiber channel have been observed. This suggests that the marginal input-symbol distribution is not the only consideration. This paper provides a tutorial-style discussion of probabilistic shaping for optical fiber communication. Since the distinguishing property of the channel is the signal-dependent NLIN, we speak of probabilistic shaping for nonlinearity tolerance. Our analysis builds on the first-order time-domain perturbation approximation of the nonlinear fiber channel and revisits the notion of linear and nonlinear shaping gain. We largely focus on probabilistic amplitude shaping with popular shaping methods. The concept of shaping via sequence selection is given special consideration, as it inherently optimizes a multivariate distribution for shaped constellations.

Figures (23)

• Figure 1: Top row: Unshaped (left) and shaped (right) two-dimensional constellation with 256 equi-probable signal points. Bottom row: induced one-dimensional distribution. From fischer:2005.
• Figure 2: Achievable rates per two dimensions for different uniform and MB-shaped constellations and the AWGN channel. Dashed lines: $C_\circ$ and $C_{\hbox{\tiny$\square$}}$ are the rates for continuous Gaussian and uniform signaling, respectively. Solid lines: QAM uniform and QAM with MB shaping.
• Figure 3: SNR gain (left) and reach gain (right) for MB-shaping of a 64QAM constellation for several achievable rates. IGN model at optimum transmit power.
• Figure 4: Ratio $\mu_n^{\mathrm{sh}}/\mu_n^{\mathrm{uni}}$, $n=4,6$, of the standardized moments for shaped and uniform 16QAM constellations, where the shaped constellations were obtained in pan2016probabilistic from optimizing the mutual information for fiber links of length (# of spans)$\times80$ km. The EGN model was used for the optimization. Also shown are the moment ratios for the shaped constellation that would have been obtained with the GN model.
• Figure 5: Rate loss (left) and ratio $\mu_n^{\mathrm{sh}}/\mu_n^{\mathrm{uni}}$, $n=4,6$ (right) for ESS with 8PAM and $R_{\mathrm{sh}}=2.5$ bits per amplitude versus shaping block length. Standardized moments are for 64QAM obtained from pairing two 8PAM ESS output blocks.