A Unified Expression for Upper Bounds on the BLER of Spinal Codes over Fading Channels

Aimin Li, Xiaomeng Chen, Shaohua Wu, Gary C. F. Lee, Sumei Sun

Introduction

In the evolving landscape of wireless communications, particularly for 5G and future 6G networks, the demand for high-reliability and low-latency links is paramount. Spinal codes, introduced in 2011, have emerged as a powerful class of rateless codes capable of adapting their coding rate dynamically to fluctuating channel conditions. Unlike fixed-rate codes, Spinal codes can continue transmitting symbols until the receiver successfully decodes the message, making them ideal for unpredictable fading environments.

The Challenge of Fading Channels

Wireless channels are subject to "fading"—signal attenuation caused by multi-path interference and user mobility. While Spinal codes demonstrate exceptional performance in empirical simulations, a rigorous theoretical understanding of their Block Error Rate (BLER) in the Finite Block Length (FBL) regime has been lacking. Most existing analytical frameworks either assume static noise (AWGN) or infinite message lengths, which do not accurately reflect the constraints of real-world low-latency systems like satellite links or industrial IoT. This paper addresses this gap by providing a unified mathematical framework to calculate tight upper bounds on the error probability of Spinal codes across various fading models, including Rayleigh, Nakagami-m, and Rician.

Methodology: A Two-Pronged Bounding Approach

The researchers developed a comprehensive methodology to bound the BLER, moving beyond traditional benchmarks.

Spinal Code Architecture

The core of a Spinal code involves hashing message segments into a sequence of "spines," which then seed a pseudo-random number generator (RNG) to produce constellation symbols. This structure is essential for the code's rateless property.

Figure 1: The architectural flow of Spinal codes, from message segmentation and hashing to symbol mapping across multiple passes.

Refined Upper Bound

The study introduces two primary bounds:

  1. Gallager-based Bound: An extension of the classic random coding bound adapted for fading complexities.
  2. Refined Upper Bound (Theorem 3): A more precise approach that leverages the specific sequential hashing structure of Spinal codes. The authors mathematically prove that this refined bound is strictly tighter than the Gallager-based benchmark.

The Unified Framework

To create a versatile tool for researchers, the authors employed Craig’s formula and Riemann sums to transform complex probability integrals into a generalized expression. This allows for the evaluation of BLER by simply plugging in parameters specific to a fading model (e.g., the $m$ factor for Nakagami-m or the $K$ factor for Rician).

Figure 2: The unified framework utilizes Craig’s formula and Riemann sums to provide a generalized expression for BLER across different fading models.

Key Insights: The Importance of the "Tail"

A significant theoretical contribution of this work is the proof of the Tail Transmission Pattern (TTP) optimality. In Spinal codes, the decoding process is sequential; the successful decoding of early segments is often contingent on the clarity of later segments. Through optimization analysis and a greedy search algorithm, the authors demonstrated that prioritizing the transmission of "tail" symbols—those at the end of the message—acts as a "reliability anchor." This strategy significantly enhances the decoder's ability to resolve the entire sequence, maximizing overall system reliability.

Furthermore, the paper proves that for Maximum Likelihood (ML) decoding, the phase of the fading coefficient is invariant; only the magnitude distribution affects the BLER. This insight simplifies the mathematical modeling of complex fading environments.

Results and Validation

The theoretical bounds were validated using extensive Monte Carlo Simulations (MCS) with $10^6$ iterations per data point. Key findings include:

  • 90% Error Reduction: The Refined Bound reduced the "Relative Error" (the gap between theory and simulation) by approximately 90% compared to standard Gallager-based methods.
  • Robust Tracking: The derived bounds closely track simulation curves across a wide range of Signal-to-Noise Ratios (SNR), from 0 to 20 dB, across Rayleigh, Nakagami-m ($m=2$), and Rician ($K=1$) channels.
  • Practical Utility: The paper provides ready-to-use closed-form expressions, eliminating the need for researchers to run computationally expensive simulations to estimate performance in standard fading scenarios.

Limitations and Future Work

While the framework is robust, it assumes Maximum Likelihood (ML) decoding, which, while optimal for reliability, is computationally expensive for long messages. Practical implementations often use "Beam Search" decoders; future research could extend these bounds to account for such sub-optimal decoding. Additionally, the current model assumes perfect Channel State Information (CSI). Incorporating estimation errors and addressing the slight "error floor" observed at very high SNR levels remain open areas for further mathematical refinement.

Summary updated

BibTeX
@article{li2026unified,
  author = {Aimin Li and Xiaomeng Chen and Shaohua Wu and Gary C. F. Lee and Sumei Sun},
  title = {{A Unified Expression for Upper Bounds on the BLER of Spinal Codes over Fading Channels}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2026},
  doi = {10.1109/twc.2025.3593377},
}
Figure