Flexible Channel Coding Approach for Short-Length Codewords

Mikel Hernaez, Pedro M. Crespo, Javier Del Ser

Abstract

This letter introduces a channel coding design framework for short-length codewords which can achieve lower error floor than previous approaches. The proposed code is based on combining convolutional coding with a q-ary linear combination and unequal energy allocation. Simulation results suggest that for very low Bit Error Rates (BER) the proposed system will exhibit lower error floors than previous approaches, with a small performance penalty at mid-range BERs. On the other hand, when selecting an error floor higher than the previous approaches, the loss in performance at mid-range BERs is negligible.

Introduction

In modern digital communications, the demand for low-latency applications—such as industrial IoT and real-time video streaming—has shifted focus toward short-length codewords (hundreds to thousands of bits). While capacity-approaching codes like LDPC and Turbo codes excel with long data blocks, they often encounter a significant "error floor" when applied to short sequences. Traditionally, addressing this requires a complete redesign of the code's parity-check matrix. This paper proposes a flexible, tunable framework that allows performance optimization through a single energy allocation parameter, $\Lambda$, without a full system overhaul.

Proposed Architecture

The proposed system utilizes a hybrid architecture that integrates convolutional codes with non-binary linear combinations. The source data $\mathbf{U}$ is split into two streams, $\mathbf{U}^1$ and $\mathbf{U}^2$, which are processed by parallel convolutional encoders. The resulting codewords are interleaved ($\Pi_1, \Pi_2$) and then combined using a $q$-ary linear combination over a non-binary Galois Field ($GF(2^q)$).

Block diagram of the proposed flexible encoder architecture.

The "secret sauce" of this design is the unequal energy allocation parameter, $\Lambda$. By adjusting the ratio of energy assigned to the convolutional coded symbols versus the linearly combined symbols, engineers can tune the code's performance. A higher $\Lambda$ typically improves the "waterfall" region (the initial steep drop in errors), while a lower $\Lambda$ can be used to suppress the error floor, making the system adaptable to various application requirements.

Decoding and Analysis

The receiver employs an iterative decoder based on the Sum-Product Algorithm (SPA) operating on a factor graph. To optimize the linear combination coefficients, the authors utilize EXIT (EXtrinsic Information Transfer) charts. These charts visualize the exchange of information between decoder components, allowing for the identification of optimal coefficients that ensure efficient convergence and predict when the decoder might get "stuck."

Performance Evaluation

The framework was evaluated using Monte Carlo simulations over an Additive White Gaussian Noise (AWGN) channel. The proposed code was benchmarked against state-of-the-art short-length codes, including Progressive Edge-Growth (PEG), PEG-ACSE, and PEG-ACE irregular LDPC codes, which are specifically known for their low error floors.

BER performance comparison for different values of q and Lambda against PEG-based codes.

As shown in the simulation results, the proposed system can achieve significantly lower error floors than specialized PEG-ACE codes at high Signal-to-Noise Ratios (SNR). While there is a minor performance penalty (less than 0.5 dB at $10^{-4}$ BER) to achieve this flexibility, the ability to adapt the same hardware to different reliability requirements—from broadcasting to high-reliability links—is a major advantage.

Conclusion

This flexible coding framework provides a practical solution for short-length codeword applications. By leveraging non-binary mixing and tunable energy allocation, it offers a superior trade-off between waterfall performance and error floor suppression compared to fixed-structure codes. Future work will explore extending this framework to more complex channel models and higher-order modulation schemes.

Summary updated

Abstract

This letter introduces a channel coding design framework for short-length codewords which can achieve lower error floor than previous approaches. The proposed code is based on combining convolutional coding with a q-ary linear combination and unequal energy allocation. Simulation results suggest that for very low Bit Error Rates (BER) the proposed system will exhibit lower error floors than previous approaches, with a small performance penalty at mid-range BERs. On the other hand, when selecting an error floor higher than the previous approaches, the loss in performance at mid-range BERs is negligible.
BibTeX
@article{hernaez2012flexible,
  author = {Mikel Hernaez and Pedro M. Crespo and Javier Del Ser},
  title = {{Flexible Channel Coding Approach for Short-Length Codewords}},
  journal = {IEEE Communications Letters},
  year = {2012},
  url = {https://openreview.net/forum?id=uxBqRT0wM6},
}
Figure