Probabilistic Eigenvalue Shaping for Nonlinear Fourier Transform Transmission

Andreas Buchberger, Alexandre Graell i Amat, Vahid Aref, Laurent SchmalenJournal of Lightwave Technology 2017 · DOI · OpenReview

Abstract

We consider a nonlinear Fourier transform (NFT) based transmission scheme, where data are embedded into the imaginary part of the nonlinear discrete spectrum. Inspired by probabilistic amplitude shaping, we propose a probabilistic eigenvalue shaping (PES) scheme as a means to increase the data rate of the system. We exploit the fact that for an NFT-based transmission scheme, the pulses in the time domain are of unequal duration by transmitting them with a dynamic symbol interval and find a capacity-achieving distribution. The PES scheme shapes the information symbols according to the capacity-achieving distribution and transmits them together with the parity symbols at the output of a low-density parity-check encoder, suitably modulated, via time-sharing. We furthermore derive an achievable rate for the proposed PES scheme. We verify our results with simulations of the discrete-time model as well as with split-step Fourier simulations.

Introduction

In modern optical fiber networks, the "Kerr nonlinearity"—where the fiber's refractive index changes with light intensity—is a fundamental limit to data rates. While standard systems try to compensate for this as noise, the Nonlinear Fourier Transform (NFT) treats the fiber as a nonlinear system, mapping signals into a "nonlinear spectral domain" where the channel acts as a simple multiplicative filter.

However, NFT-based transmission faces two major hurdles:

  1. Inefficient Timing: Previous schemes used a fixed symbol interval based on the longest possible pulse, leaving large "dead zones" of silence for shorter pulses.
  2. The Shaping Gap: Standard systems use a uniform distribution of symbols. In the nonlinear domain, this is far from the theoretical capacity. Because different eigenvalues have different "costs" in terms of time and noise sensitivity, a non-uniform (shaped) distribution is required to maximize efficiency.

Methods

The proposed system embeds data into the imaginary part of the discrete nonlinear spectrum (eigenvalues). The Probabilistic Eigenvalue Shaping (PES) scheme introduces three main innovations:

Dynamic Symbol Interval

Instead of waiting for a fixed duration, the transmitter sends the next pulse as soon as the current one has decayed. This turns the communication problem into one with "variable-length symbols." By removing fixed guard bands, the system gains a significant boost in data rate.

Comparison of fixed vs. dynamic symbol intervals, illustrating the reduction in 'dead zones' between pulses.

Capacity Optimization

The capacity-achieving distribution is derived by maximizing the time-scaled mutual information. For this channel, the optimal distribution is discrete and roughly exponential, often favoring the smallest and largest eigenvalues.

PES Architecture

Inspired by Probabilistic Amplitude Shaping (PAS), the system uses a Distribution Matcher (DM) to transform raw bits into a sequence of eigenvalues following the optimal distribution. These are then passed through a systematic Low-Density Parity-Check (LDPC) encoder. Since the optimal distribution is not symmetric, parity bits are mapped to a separate set of eigenvalues and sent via time-sharing.

Key Results

The PES scheme was evaluated using discrete-time Monte-Carlo simulations and high-fidelity Split-Step Fourier (SSF) simulations to account for real-world impairments like fiber loss and noise.

Rate Improvement and Capacity

The PES scheme yields a significant improvement—up to 2x the data rate—compared to unshaped NFT systems with fixed intervals. Using off-the-shelf LDPC codes, the scheme operates within approximately 2 dB of the theoretical constrained capacity.

Mutual information vs. SNR for various modulation orders (M), showing the performance gains of PES (solid lines) over unshaped systems (dashed lines).

Transmission Distance

In physical fiber simulations, the system maintained a Bit Error Rate (BER) below $10^{-5}$ over distances of 2960 km to 3200 km, depending on the modulation order ($M=4$ to $16$). Compared to previous state-of-the-art unshaped systems, PES provided rate gains of 20%, 26%, and 95% for modulation orders of 4, 8, and 16, respectively.

Conclusion and Future Work

Probabilistic Eigenvalue Shaping effectively addresses the timing and shaping inefficiencies of NFT-based transmission. While the gains are substantial, several challenges remain:

  • Synchronization Complexity: Transmitting pulses with variable durations makes timing recovery difficult, as errors can lead to desynchronization.
  • Spectral Efficiency: Future work should explore higher-order solitons or the continuous part of the nonlinear spectrum to further increase efficiency.
  • Computational Overhead: Real-time NFT calculation and distribution matching remain significant hardware challenges for high-speed transceivers.

Summary updated

Abstract

We consider a nonlinear Fourier transform (NFT)-based transmission scheme, where data is embedded into the imaginary part of the nonlinear discrete spectrum. Inspired by probabilistic amplitude shaping, we propose a probabilistic eigenvalue shaping (PES) scheme as a means to increase the data rate of the system. We exploit the fact that for an NFT-based transmission scheme the pulses in the time domain are of unequal duration by transmitting them with a dynamic symbol interval and find a capacity-achieving distribution. The PES scheme shapes the information symbols according to the capacity-achieving distribution and transmits them together with the parity symbols at the output of a low-density parity-check encoder, suitably modulated, via time-sharing. We furthermore derive an achievable rate for the proposed PES scheme. We verify our results with simulations of the discrete-time model as well as with split-step Fourier simulations.
BibTeX
@article{buchberger2017probabilistic,
  author = {Andreas Buchberger and Alexandre Graell i Amat and Vahid Aref and Laurent Schmalen},
  title = {{Probabilistic Eigenvalue Shaping for Nonlinear Fourier Transform Transmission}},
  journal = {Journal of Lightwave Technology},
  year = {2017},
  doi = {10.1109/jlt.2018.2854760},
}
Figure