Decoding of LDPC Codes Using Singularity of SNR Mismatch Over AWGN Channel

Abstract

In this paper, we propose a new decoding algorithm that guarantees superior decoding performances for low-density parity-check (LDPC) codes without knowledge of channel parameters. Although obtaining correct channel parameters is very difficult in low-latency and energy-efficient communication systems, most studies on LDPC codes assume perfect knowledge of channel parameters. While receivers can obtain correct channel parameters by utilizing numerous pilot signals, increased overhead makes them infeasible in practical systems. In addition, using incorrect channel parameters for decoding usually degrades decoding performances. These challenges highlight the need for a new decoder that performs reliably without correct channel parameters. Meanwhile, the previous studies on repeat-accumulate (RA) codes and LDPC codes have observed that the signal-to-noise ratio (SNR) mismatch sometimes enhances decoding performances when the maximum number of decoding iterations is small. Inspired by these observations, we investigate singularity of the SNR mismatch in LDPC codes from numerous frame error rate (FER) simulations. By utilizing this singularity, we propose a new decoding algorithm for the energy-efficient communication system, where correct channel parameters are not available at receivers. The proposed decoder shows a performance gain of 0.25 dB at FER = 10(-6) compared to the conventional belief propagation (BP) algorithm.