Undetected Error Probability in the Short Blocklength Regime: Approaching Finite-Blocklength Bounds With Polar Codes

Alexander Sauter, Ahmet Oguz Kislal, Giuseppe Durisi, Gianluigi Liva, Balázs Matuz, Erik G. Ström

TL;DR

This paper tackles the challenge of minimizing undetected errors in short data packets, crucial for applications like 5G and beyond. The authors introduce novel theoretical bounds and practical coding schemes using polar codes with Cyclic Redundancy Checks (CRCs), demonstrating that a threshold-based decoding approach generally outperforms CRC-based methods for reliable communication over clean channels, but the opposite holds true in the presence of noisy channel information.

What Problem Does This Paper Address?

In modern communication systems, especially those aiming for ultra-reliable low-latency communication (URLLC) like 5G, data packets are often short. While error correction codes are used to combat transmission errors, two types of errors are critical: detected errors (where the decoder flags a failure) and undetected errors (where the decoder outputs an incorrect message without realizing it). Undetected errors can be particularly harmful in mission-critical applications. For long packets, adding a CRC code is an effective way to detect errors with minimal impact on the overall data rate. However, for short packets, this rate penalty becomes significant. This paper investigates how to achieve a good trade-off between total error probability (TEP) and undetected error probability (UEP) in the short blocklength regime, focusing on polar codes, which are known for their excellent performance in this area.

Methods

The authors propose two main contributions:

  1. Novel Achievability Bounds: They develop two new theoretical bounds for finite blocklengths that specifically address the TEP-UEP trade-off.
  2. Bound 1 (CRC-based): This bound considers an outer error detection code (like a CRC) that works in conjunction with the inner error correction code.
  3. Bound 2 (Threshold-based): This bound is inspired by Forney's optimal erasure decoder and uses a threshold test on the generalized information density. This approach aims to be more computationally tractable for achieving bounds.
  4. CRC-Aided (CA) Polar Code Schemes: They propose and analyze two practical coding schemes that combine polar codes with CRCs:
  5. Scheme 1 (CRC-based): This scheme reserves a portion of the CRC parity bits for error detection, while the remaining bits are used for error correction.
  6. Scheme 2 (Threshold-based): This scheme adapts Forney's optimal threshold test to the output of a successive cancellation list (SCL) decoder, which is a common decoding method for polar codes.

The paper also introduces a saddlepoint approximation to efficiently evaluate the pairwise error probabilities needed for these bounds, which is crucial for the low error rates required in URLLC.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and simulations over synthetic channel models.

Key Insights and Ideas

  • Thresholding vs. CRC for Short Packets: For clean channels (like the AWGN channel), threshold-based decoding (Scheme 2) generally outperforms CRC-based error detection (Scheme 1) for short blocklengths. This is because CRC-based methods incur a rate penalty that is more detrimental when packets are short.
  • Sensitivity to Channel State Information (CSI): The advantage of threshold-based methods diminishes significantly when the channel state information at the receiver is noisy or inaccurate. In such "mismatched" scenarios, CRC-based error detection proves to be more robust and preferable.
  • Bounds as Performance Predictors: The newly derived bounds are shown to be more accurate than traditional Forney bounds for short blocklengths and the target error probabilities, providing a better benchmark for practical coding schemes.
  • Trade-off Tuning: The parameters within both the bounds and the practical schemes (like the threshold value or the split of CRC bits) allow for fine-tuning the trade-off between TEP and UEP.

Key Results

  • Bi-AWGN Channel:
    • Threshold-based CA polar codes (Algorithm B) achieve better performance (lower SNR for target error rates) than CRC-based CA polar codes (Algorithm A) for short blocklengths (e.g., n=64).
  • The gap between threshold-based and CRC-based approaches narrows as blocklength increases.
  • Algorithm B can even operate below Forney's theoretical achievability bound.
  • The proposed bounds (especially Theorem 2) provide tighter estimates of achievable performance than Forney's bound.
  • Block-Memoryless Phase-Noise Channel (Imperfect CSI):
    • The performance gap between the two bounds (Theorem 1 and 2) is significantly reduced.
  • CRC-based CA polar codes (Algorithm A) outperform threshold-based ones (Algorithm B) in this mismatched scenario, with Algorithm A showing a clear advantage across all blocklengths.
  • Threshold-based methods are highly sensitive to CSI errors, leading to substantial performance degradation.

Limitations and Future Work

  • Computational Complexity: While the saddlepoint approximation helps, evaluating the bounds can still be computationally intensive.
  • Mismatched Decoding: The paper highlights the significant performance degradation of threshold-based methods with imperfect CSI. Future work could explore more robust thresholding techniques or adaptive strategies for CSI mismatch.
  • Optimization of Parameters: The paper shows the importance of optimizing parameters (like the threshold s in Theorem 2) for performance. Further research could investigate more efficient methods for parameter optimization in practical systems.
  • Generalization: While the paper focuses on polar codes and specific channel models, extending these findings to other code families or more complex channel environments would be valuable.

Summary updated

Abstract

We analyze the trade-off between the undetected error probability (i.e., the probability that the channel decoder outputs an erroneous message without detecting the error) and the total error probability in the short blocklength regime. We address the problem by developing two new finite blocklength achievability bounds, which we use to benchmark the performance of two coding schemes based on polar codes with outer cyclic redundancy check (CRC) codes -- also referred to as CRC-aided (CA) polar codes. The first bound is obtained by considering an outer detection code, whereas the second bound relies on a threshold test applied to the generalized information density. Similarly, in the first CA polar code scheme, we reserve a fraction of the outer CRC parity bits for error detection, whereas in the second scheme, we apply a threshold test (specifically, Forney's optimal rule) to the output of the successive cancellation list decoder. Numerical simulations performed on the binary-input AWGN channel reveal that, in the short-blocklength regime, the threshold-based approach is superior to the CRC-based approach, both in terms of bounds and performance of CA polar code schemes. We also consider the case of decoding with noisy channel-state information, which leads to a mismatched decoding setting. Our results illustrate that, differently from the previous case, in this scenario, the CRC-based approach outperforms the threshold-based approach, which is more sensitive to the mismatch.
BibTeX
@article{sauter2025undetected,
  author = {Alexander Sauter and Ahmet Oguz Kislal and Giuseppe Durisi and Gianluigi Liva and Balázs Matuz and Erik G. Ström},
  title = {{Undetected Error Probability in the Short Blocklength Regime: Approaching Finite-Blocklength Bounds With Polar Codes}},
  journal = {IEEE Trans. Communications},
  year = {2025},
  doi = {10.1109/tcomm.2025.3545648},
}
Figure