Reproduce “High-Cardinality Geometrical Constellation Shaping for the Nonlinear Fibre Channel” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Eric Sillekens, Gabriele Liga, Domaniç Lavery, Polina Bayvel, Robert I. Killey

Analytical mutual-information and generalized-mutual-information gradients enable efficient optimization of high-cardinality geometrically shaped constellations for both AWGN and nonlinear fibre channels. The method scales to 8192-point 2D and 4D formats, bringing an 8192-point 2D constellation within 0.06 bit per 2D symbol of AWGN capacity and extending simulated transmission reach by up to roughly 120 km for a 128-point 4D format.

Key insights

  • Large constellation cardinality improves the ability of geometrical shaping to approximate Gaussian signalling, but the required number of points grows with dimensionality and target spectral efficiency.
  • Four-dimensional formats provide additional geometric degrees of freedom that can yield shaping gains and reach improvements even when optimized primarily for AWGN conditions.
  • Nonlinear-aware shaping is most useful when nonlinear interactions are strong: it can retain part of the shaping gain while reducing the nonlinear penalty toward that of square QAM.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.