Retention-Aware Belief-Propagation Decoding for NAND Flash Memory

Abstract

To recover decoding errors caused by data retention noise in nand Flash memory, a novel retention-aware belief-propagation (RA-BP) decoding scheme is proposed. The RA-BP scheme is based on two rounds of BP decoding in which the memory cell's charge-loss effect is systematically compensated. The most prominent feature of the proposed RA-BP scheme is that it recovers the retention-failure decoded data without performing additional memory sensing operations. Instead, the likely retention-failure memory cells are identified by using the available readback voltage signal and the decoded codeword bits. For all such likely retention-failure memory cells, the corresponding log-likelihood ratio information is modified in such a way as to reverse the effect of retention noise, and a second round of BP decoding is performed afresh. Through numerical simulations, it is shown that the proposed RA-BP decoding scheme can improve the error rate performance by more than three orders of magnitude and increase the retention time limit by up to 60% compared with a single round of BP decoding.