LDPC code-dependent quantization for the NAND flash channel

Abstract

With the improvement of scaling technologies and the evolution of storage strategies, the storage density of NAND flash memory is gradually growing. However, the increase in storage density increases error probability, which reduces the lifespan of NAND flash memory. Error correction coding technology is a powerful tool for ensuring data reliability in NAND flash memory, but how to maximize the error correction ability through quantization in the reading process is an important problem that must be addressed. In view of the shortcomings of the maximum mutual information (MMI) quantization, this study proposes to utilize the density evolution (DE) algorithm to perform quantization on the basis of the features of low-density parity-check codes. To meet the conditions of the DE algorithm, this research also proposes to implement a bit flipper in the NAND controller for symmetrizing the asymmetric input distribution of the decoder. Then, the constant ratio method is chosen to further reduce the complexity of searching for the best read-voltage thresholds. Through numerical simulation and verification, the proposed algorithm is proven to provide better decoding performance than the MMI quantization and significantly improve the lifetime of the NAND flash memory.