Adaptive Single Phase Decoding of LDPC Codes

A new "memory-aware" Low-Density Parity-Check codes decoding is presented. The proposed algorithm performs a single-phase update in which the variable node elaboration is embedded in the check node phase. This reduces the memory size to the number of bits in the codeword, for a saving of 70-80% depending on the code in use. Despite these rough approximations, simulation results on AWGN channel provide almost the same BER/FER performance of the exact belief propagation (only about 0.2 dB is lost in the transmission SNR) thanks to the introduction of stabilising factors driving the algorithm, which are adaptively updated through the decoding process. Remarkably, the decoding throughput is roughly doubled at high SNR and the computational complexity is also lower than current 2-phase serial implementations.