Huffman-Coded Sphere Shaping for Extended-Reach Single-Span Links

Pavel Skvortcov, Ian Phillips, W. Forysiak, Toshiaki Koike‐Akino, Keisuke Kojima, Kieran Parsons, David S. MillarIEEE Journal of Selected Topics in Quantum Electronics 2021 · DOI · arXiv

Abstract

Huffman-coded sphere shaping (HCSS) is an algorithm for finite-length probabilistic constellation shaping, which provides nearly optimal energy efficiency at low implementation complexity. In this paper, we experimentally study the nonlinear performance of HCSS employing dual-polarization 64-ary quadrature amplitude modulation (DP-64QAM) in an extended reach single-span link comprising 200 km of standard single mode fiber (SSMF). We investigate the effects of shaping sequence length, dimensionality of symbol mapping, and shaping rate. We determine that the naïve approach of Maxwell-Boltzmann distribution matching - which is optimal in the additive white Gaussian noise channel - provides a maximum achievable information rate gain of 0.18 bits/4D-symbol in the infinite length regime. Conversely, HCSS can achieve a gain of 0.37 bits/4D-symbol using amplitude sequence lengths of 32, which may be implemented without multiplications, using integer comparison and addition operations only. Coded system performance, with a net data rate of approximately 425 Gb/s for both shaped and uniform inputs, is also analyzed.

Introduction

In modern optical fiber communications, researchers strive to maximize data throughput by employing Probabilistic Constellation Shaping (PCS). The core concept involves using low-energy signal points more frequently than high-energy ones to improve energy efficiency, analogous to how Huffman coding uses shorter codes for common characters in text compression.

However, two major hurdles remain for practical implementation:

  1. Complexity: Many shaping algorithms require intensive calculations, such as numerous multiplications or massive look-up tables, making real-time hardware implementation difficult.
  2. Nonlinearity: Fiber optics are limited not just by random noise (AWGN) but also by the "Kerr effect," where the fiber's refractive index changes with signal power. The standard "ideal" Maxwell-Boltzmann (MB) distribution, optimized for linear noise, often performs poorly in nonlinear fiber channels because it fails to account for these power-dependent interactions.

This paper addresses these issues by evaluating Huffman-Coded Sphere Shaping (HCSS)—a low-complexity, finite-length shaping method—in a real-world, nonlinear, 200 km single-span fiber link.

Methods

The researchers integrated HCSS within a Probabilistic Amplitude Shaping (PAS) framework. The system architecture for the transmitter and receiver is shown below.

Figure 1: Block diagrams of the PAS-based transmitter and receiver architectures using HCSS.

Technical Components

  • HCSS Algorithm: Input bits are mapped to amplitude sequences using Huffman coding to select a "composition" (set of amplitudes) and multiset ranking to pick a specific permutation. For short sequences, this process uses only integer additions and comparisons, eliminating power-hungry multiplications.
  • Symbol Mapping: The study compared 1D, 2D, and 4D mapping strategies. In 4D mapping, a shaped sequence is distributed across the four dimensions of the optical field (in-phase and quadrature components of two polarizations).
  • Experimental Setup: The testbed utilized a 200 km single-span standard single-mode fiber (SSMF) link with 9 channels of DP-64QAM at 56 GBd, achieving net data rates of approximately 425 Gb/s. Low-Density Parity-Check (LDPC) codes with rates of 0.64 and 0.72 were used for coded performance evaluation.

Key Insights and Results

The Nonlinear Advantage of Short Sequences

In linear channels, longer shaping sequences are typically more efficient. However, this study found that in nonlinear fiber, shorter sequences (e.g., $L=32$) performed better. Short sequences help "decorrelate" the signal, reducing the high-power interactions that trigger distortion.

HCSS vs. Maxwell-Boltzmann (MB) Shaping

While the MB distribution is the theoretical optimum for simple noise, it underperforms in nonlinear links. Experimental results showed that MB shaping provided only a 0.18 bit gain, whereas HCSS achieved a gain of 0.37 bits/4D-symbol at a sequence length of $L=32$. Furthermore, HCSS demonstrated superior resilience to nonlinearities; while MB shaping suffered a significant SNR penalty at optimum power, HCSS improved the launch power margin by 1.3 dB.

Figure 5: Power penalty vs. sequence length for HCSS compared to the MB shaping baseline.

Dimensionality and Complexity

Moving from 1D to 4D mapping significantly improved nonlinear tolerance by spreading shaping over more dimensions, reducing the probability of high-peak power bursts. Crucially, HCSS with $L=32$ requires only about 100 kbits of memory for look-up tables, making it highly feasible for integration into actual transceiver hardware using only integer operations.

Conclusion

This research demonstrates that HCSS provides a significant performance boost for high-speed optical links by outperforming "ideal" MB distributions in nonlinear channels. By leveraging short shaping sequences and high-dimensional mapping, HCSS achieves substantial information rate gains with extremely low computational complexity. Future work may explore these gains in multi-span long-haul systems and further optimize the interaction between the shaper and FEC hardware.

Summary updated

Abstract

Huffman-coded sphere shaping (HCSS) is an algorithm for finite-length probabilistic constellation shaping, which provides nearly optimal energy efficiency at low implementation complexity. In this paper, we experimentally study the nonlinear performance of HCSS employing dual-polarization 64-ary quadrature amplitude modulation (DP-64QAM) in an extended reach single-span link comprising 200 km of standard single mode fiber (SSMF). We investigate the effects of shaping sequence length, dimensionality of symbol mapping, and shaping rate. We determine that the naïve approach of Maxwell-Boltzmann distribution matching - which is optimal in the additive white Gaussian noise channel - provides a maximum achievable information rate gain of 0.18 bits/4D-symbol in the infinite length regime. Conversely, HCSS can achieve a gain of 0.37 bits/4Dsymbol using amplitude sequence lengths of 32, which may be implemented without multiplications, using integer comparison and addition operations only. Coded system performance, with a net data rate of approximately 425 Gb/s for both shaped and uniform inputs, is also analyzed.
BibTeX
@article{skvortcov2021huffman,
  author = {Pavel Skvortcov and Ian Phillips and W. Forysiak and Toshiaki Koike‐Akino and Keisuke Kojima and Kieran Parsons and David S. Millar},
  title = {{Huffman-Coded Sphere Shaping for Extended-Reach Single-Span Links}},
  journal = {IEEE Journal of Selected Topics in Quantum Electronics},
  year = {2021},
  doi = {10.1109/jstqe.2021.3055476},
}

Loading related content...

Figure