Fast convergence decoding scheme for regular LDPC codes

We propose a new belief propagation (BP) decoding algorithm for fast convergence decoding of regular LDPC codes. The fast convergence is achieved by grouping and scheduling updates of nodes on a bipartite-graph without additional computations or approximations. For this reason, the proposed algorithm can be applied to a serially constructed decoder, and can reduce the power consumption and decoding delay of the decoder. Simulation results show that the convergence speed of the proposed algorithm depends on the degree, d_v, of the variable-nodes of the bipartite-graph, and the average number of iterations of the decoder can be reduced to (d_v-1)/d_v by using the proposed algorithm without performance degradation or additional computation.