Rate-Loss Reduction Of Sc-Ldpc Codes By Optimizing Reliable Variable Nodes Via Expected Graph Evolution

The outstanding decoding performance of spatially coupled low-density parity-check (SC-LDPC) codes comes from wave-like propagation of reliable messages. The reliable messages are triggered by shortened (known) variable nodes in some consecutive reliable positions. However, at the cost of the improvement, shortened variable nodes cause rate-loss of SC-LDPC codes. To reduce the rate-loss, additional variable nodes (so called reliable variable nodes) can be added to the reliable positions instead of shortened variable nodes. Density evolution (DE) is an efficient method to design degree distribution of the reliable variable nodes. However, degree distributions obtained by DE show degraded performance in finite-length code performance. In this paper, we generalize the expected graph evolution and use the analysis tool in optimizing degree distribution which shows the minimum rate-loss without finite-length performance degradation. From the well-designed degree distribution, rate loss reduction by 60% can he achieved without finite-length performance degradation.