On Doped SC-LDPC Codes for Streaming

Roman Sokolovskii, Alexandre Graell i Amat, Fredrik Brännström

Abstract

In streaming applications, doping improves the performance of spatially-coupled low-density parity-check (SC-LDPC) codes by creating reduced-degree check nodes in the coupled chain. We formulate a scaling law to predict the bit and block error rate of periodically-doped semi-infinite SC-LDPC code ensembles streamed over the binary erasure channel under sliding window decoding for a given finite component block length. The scaling law assumes that with some probability doping is equivalent to full termination and triggers two decoding waves; otherwise, decoding performs as if the coupled chain had not been doped at all. We approximate that probability and use the derived scaling laws to predict the error rates of SC-LDPC code ensembles in the presence of doping. The proposed scaling law provides accurate error rate predictions. We further use it to show that in streaming applications periodic doping can yield higher rates than periodic full termination for the same error-correcting performance.

Introduction

Spatially-coupled low-density parity-check (SC-LDPC) codes are a cornerstone of modern communications because they approach theoretical capacity limits using practical decoding methods. In streaming scenarios, data is transmitted in a continuous chain. Efficient decoding requires "decoding waves" to propagate from the boundaries of the chain inward.

However, in very long chains, these waves can get "stuck" due to noise, leading to error propagation. To prevent this, researchers use doping—periodically inserting better-protected bits or modified structures to jump-start new decoding waves. While effective, predicting the performance of specific doping patterns at finite lengths (real-world block sizes) previously required massive simulations. This paper introduces a mathematical scaling law to fill this gap.

Theoretical Framework and Methods

The authors focus on Variable Node (VN) doping, where specific bits in the chain are fixed to a known value (typically zero) to assist the decoder. The core innovation is the "Switch" Model, which treats a doping point as a probabilistic switch. With a certain probability ($\psi_d$), the doping triggers two new decoding waves, effectively acting like a full termination. Otherwise, it fails to impact the decoding process.

The success of this "switch" depends on whether the decoder can survive a "dip" in available information—specifically, the number of degree-one check nodes—just before the doping-induced waves take over. This local minimum acts as a critical bottleneck.

Evolution of degree-one check nodes showing the local minimum bottleneck during decoding.

The methodology involves several key steps:

  1. Density and Mean Evolution: Tracking the average behavior of the decoder to identify the local minimum.
  2. Gaussian Approximation: Modeling the success probability of the doping point based on the variance around this local minimum.
  3. Geometric Distribution: For semi-infinite streaming chains, treating the distance between successful doping points as a geometrically distributed random variable to average error rates over effective chain lengths.

Key Results

The proposed scaling law demonstrates an "all-or-nothing" behavior: doping either successfully launches a new pair of decoding waves or fails entirely. This binary nature allows for precise modeling of error rates.

The scaling law's predictions for Bit Error Rate (BER) and Block Error Rate (BLER) show remarkable agreement with heavy numerical simulations. This accuracy confirms that the probabilistic switch model captures the essential physics of the decoding process in doped SC-LDPC ensembles.

Comparison of predicted BER and BLER from the scaling law versus numerical simulations.

Furthermore, the authors demonstrate that periodic doping can achieve higher "design rates" (more data per transmission) than periodic full termination while maintaining the same target reliability. This makes doping a more flexible and efficient tool for streaming applications.

Conclusion and Future Work

The derived scaling law provides a powerful tool for engineers to optimize doping intervals and window sizes without exhaustive simulations. While currently tailored for the Binary Erasure Channel (BEC), future work aims to adapt this model for more complex channels like Additive White Gaussian Noise (AWGN). Additionally, extending the framework to protograph-based SC-LDPC codes and exploring "soft" doping patterns remains a promising direction for enhancing hardware-implemented streaming codes.

Summary updated

Abstract

In streaming applications, doping improves the performance of spatially-coupled low-density parity-check (SC-LDPC) codes by creating reduced-degree check nodes in the coupled chain. We formulate a scaling law to predict the bit and block error rate of periodically-doped semi-infinite SC-LDPC code ensembles streamed over the binary erasure channel under sliding window decoding for a given finite component block length. The scaling law assumes that with some probability doping is equivalent to full termination and triggers two decoding waves; otherwise, decoding performs as if the coupled chain had not been doped at all. We approximate that probability and use the derived scaling laws to predict the error rates of SC-LDPC code ensembles in the presence of doping. The proposed scaling law provides accurate error rate predictions. We further use it to show that in streaming applications periodic doping can yield higher rates than periodic full termination for the same error-correcting performance.
BibTeX
@article{sokolovskii2021doped,
  author = {Roman Sokolovskii and Alexandre Graell i Amat and Fredrik Brännström},
  title = {{On Doped SC-LDPC Codes for Streaming}},
  journal = {IEEE Communications Letters},
  year = {2021},
  doi = {10.1109/lcomm.2021.3069628},
}
Figure