LDPC Decoding Based on Hybrid Stochastic Computing

Compared with traditional computing methods, stochastic computing (SC) is simpler in circuit implementation. The Belief Propagation (BP) or Min-Sum (MS) algorithms used in traditional Low Density Parity Check (LDPC) decoding require a large amount of circuits for the calculation of check nodes (CNs). In this paper, decoding is completed on a decoding structure that combines the advantages of the Offset Min-Sum (OMS) algorithm and SC. CN information is calculated in the probability domain, and variable node (VN) information is calculated in the Log Likelihood Ratio (LLR) domain. It is proposed to use the gold sequence combined with a synchronizer to generate a positively correlated sequence, and combine the Noise-Dependent Scaling (NDS) algorithm to obtain a positively correlated random number generator to complete the conversion from LLR to a positively correlated bit stream. The positively correlated bit stream is input into an AND gate to find the minimum value of the input bit streams. Simulation results show that the bit error rate curve of the proposed method is close to the OMS algorithm.