Abstract
Braided convolutional codes (BCCs) are a class of spatially coupled turbo-like codes (SC-TCs) with excellent belief propagation (BP) thresholds. In this paper, we analyze the performance of BCCs in the finite block-length regime. We derive the average weight enumerator function (WEF) and compute the union bound on the performance for the uncoupled BCC ensemble. Our results suggest that the union bound is affected by poor distance properties of a small fraction of codes. By computing the union bound for the expurgated ensemble, we show that the floor improves substantially and very low error rates can be achieved for moderate permutation sizes. Based on the WEF, we also obtain a bound on the minimum distance which indicates that it grows linearly with the permutation size. Finally, we show that the estimated error floor for the uncoupled BCC ensemble is also valid for the coupled ensemble by proving that the minimum distance of the coupled ensemble is lower bounded by the minimum distance of the uncoupled ensemble.
Introduction: The Finite-Length Challenge
Spatially coupled codes, such as SC-LDPC codes, are renowned for "threshold saturation," where practical belief propagation (BP) decoders approach the theoretical limits of optimal maximum a-posteriori (MAP) decoders. Braided Convolutional Codes (BCCs) serve as the turbo-like counterparts to these codes, demonstrating exceptional asymptotic performance.
However, practical communication systems operate at finite block lengths. In this regime, uncoupled BCC ensembles often exhibit unexpectedly high error floors. This research investigates the underlying distance properties of BCCs to understand this phenomenon and explores how techniques like expurgation and spatial coupling can mitigate these performance bottlenecks.
Analytical Methodology
To evaluate finite-length performance, the authors utilize several mathematical frameworks:
- Weight Enumerator Functions (WEF): Average WEFs for the uncoupled BCC ensemble are derived using transition matrices for the component rate-2/3 convolutional encoders.
- Union Bounds: Truncated union bounds for Bit Error Rate (BER) and Frame Error Rate (FER) are calculated for the Additive White Gaussian Noise (AWGN) channel.
- Expurgation: To address the high error floors predicted by standard union bounds, the authors compute an "expurgated" bound. This involves mathematically excluding a small fraction of permutations with poor distance properties.
- Distance Proofs: A theorem is established to prove that codeword weights in a coupled BCC are lower-bounded by those in an uncoupled BCC, ensuring that coupling does not degrade minimum distance.
Key Insights: Distance Properties and Coupling
A primary finding is that the high error floor in uncoupled BCC ensembles is not intrinsic to the code structure but is caused by a tiny fraction of "bad" permutations. By expurgating just 5% of the worst-performing codes, the minimum distance of the remaining ensemble is shown to grow linearly with the permutation size.
Furthermore, the research proves that spatial coupling is strictly beneficial for distance. The minimum distance of a coupled BCC is guaranteed to be at least as large as its uncoupled counterpart. Consequently, the low error floors achieved via expurgation in uncoupled codes are also applicable to coupled ensembles.
Performance Results
The impact of expurgation is significant. For uncoupled BCCs, transitioning from uniformly random permutations to fixed (good) permutations drastically lowers the error floor. For instance, at $E_b/N_0 = 2.5$ dB, the FER improves from $9.5 \times 10^{-5}$ to $6.8 \times 10^{-7}$.
Spatially coupled BCCs (Type-II) demonstrate even more dominant performance. For a permutation size of $N=5000$, coupling provides a gain of nearly 1.5 dB in the waterfall region compared to uncoupled versions. Even under latency constraints (e.g., comparing $N=1000$ coupled vs. $N=5000$ uncoupled), the coupled architecture remains superior.
Limitations and Future Outlook
The theoretical bounds derived in this study assume optimal Maximum Likelihood (ML) decoding. In practice, BCCs are typically decoded using BP algorithms. While ML bounds provide a fundamental limit, BP performance can be influenced by "absorbing sets" or "trapping sets." Future research will focus on characterizing these sets to further optimize BCCs for practical hardware implementations.
Summary updated