Pattern-based bit-flipping decoding algorithm of low density Parity-check codes for NAND flash memory

Abstract

Bit-flipping (BF) algorithm is useful low-density parity-check (LDPC) code decoding algorithm for SSD such as NAND flash memory due to its fast decoding speed and low complexity for hardware implementation. However, there have been researches to improve the BF algorithm because of its poor error correcting capability. In this thesis, we propose the pattern-based BF algorithm (PBFA) that adaptively determines the flipping threshold by analyzing the local maximum patterns of the flipping functions. The flipping thresholds are determined by creating the mapping from the local maximum patterns to the thresholds of the genie-aided BF algorithm, which adaptively chooses the threshold in order to minimize the error probability at each iteration. This proposed algorithm outperforms the existing BF algorithm and the other adaptive threshold choosing BF algorithm at the low maximum iteration. In addition, the better error correcting performance is achievable by applying the backtracking strategy to the investigated local maximum patterns causing the decoding failure.