Ternary Quantized Polar Code Decoders: Analysis and Design

TL;DR

This paper investigates the impact of 3-level quantization on polar code decoders, revealing performance issues and proposing two novel techniques to enhance decoding accuracy in low-complexity, energy-efficient applications.

Contribution

It introduces two innovative methods to improve quantized SCL decoding performance, addressing quantization-induced inaccuracies in low-resource environments.

Findings

Quantized log-likelihood ratios cause significant performance loss in SCL decoding.

Two techniques improve decoding accuracy with 3-level quantization.

Simulations confirm the effectiveness of the proposed methods.

Abstract

The performance of short polar codes under successive cancellation (SC) and SC list (SCL) decoding is analyzed for the case where the decoder messages are coarsely quantized. This setting is of particular interest for applications requiring low-complexity energy-efficient transceivers (e.g., internet-of-things or wireless sensor networks). We focus on the extreme case where the decoder messages are quantized with 3 levels. We show how under SCL decoding quantized log-likelihood ratios lead to a large inaccuracy in the calculation of path metrics, resulting in considerable performance losses with respect to an unquantized SCL decoder. We then introduce two novel techniques which improve the performance of SCL decoding with coarse quantization. The first technique consists of a modification of the final decision step of SCL decoding, where the selected codeword is the one maximizing the maximum-likelihood decoding metric within the final list. The second technique relies on statistical knowledge about the reliability of the bit estimates, obtained through a suitably modified density evolution analysis, to improve the list construction phase, yielding a higher probability of having the transmitted codeword in the list. The effectiveness of the two techniques is demonstrated through simulations.