Asymptotic and finite-length performance of irregular spatially-coupled codes

Abstract

The newest family of low-density parity-check (LDPC) codes, spatially-coupled (SC) codes, is shown to have several desirable characteristics including low implementation complexity and close-to-optimal performance over a range of channels. Furthermore, because of their ribbon-shaped parity-check matrices, window decoding can be used to decode these codes, which leads to low-delay implementations. Researchers have focused on asymptotically regular SC code ensembles and have examined several aspects of the code construction processes. In this paper, we concentrate on irregular SC code ensembles. We evaluate their decoding thresholds over the binary erasure channel and show that their performance is better than their regular SC counterparts. It is also shown that the gap between asymptotic coding thresholds of irregular SC ensembles and the fundamental Shannon limit gets negligibly small. For the sake of a better comparison, we have also evaluated the finite-length error performance of selected regular and irregular SC codes over the additive white Gaussian channel and it is also observed that finite-length error performance of these irregular SC codes outperforms regular SC codes. To further improve the error performance of these codes and to lower the possible error floors, progressive edge growth algorithm has also been considered in the finite-length performance analysis.