Research On LDPC Decoding In NAND Flash Memory Based On Soft-Information Sensing Strategy

In the non-volatile memory, the popular NAND flash memory has become an important storage media in the field of data storage with its good throughput performance, low energy consumption and the advantages of data storage stability. The multi-level cell(MLC) technology, which doubles or triples the memory density to increase the flash memory capacity, has been widely used by flash memory manufacturers. However, as the feature size of NAND flash memory decreases, the threshold voltage is interfered by noises in the flash memory and its stability degrades. The low-density parity check(LDPC) codes, a kind of advanced error correcting code, could provide a feasible error correction solution for the system of NAND flash memory.In the thesis, the LDPC decoding based on soft-information sensing strategy in 2-bit MLC NAND flash memory is studied. Through analyzing the data representation and the main noises in NAND flash memory, a threshold voltage distribution model of MLC NAND flash memory is summarized. The LDPC decoding methods with the different soft-information sensing strategies are applied to correct the errors. The main contents of this article are summarized as follows:1. The physical structure and operation model of NAND flash memory are introduced. Based on the physical features of flash memory cells, the interference noises and sources in NAND flash memory are summarized. The definition and representation of LDPC codes are introduced. Besides, the common LDPC codes decoding methods are also presented.2. The physical characteristics and noises characteristics of MLC NAND flash memory are analyzed. The models of the threshold voltage distribution and the channel noises interference are summarized. Based on these models, the Bit-Flip decoding algorithm with soft-information uniform sensing strategy is simulated. The simulation result shows that the algorithm can effectively reduce the errors caused by the noises in the flash memory model.3. Through the analysis of threshold voltage distribution characteristics, a soft-information non-uniform sensing strategy is presented, which improves the accuracy of LLRs. Asimulation is carried out to show that the proposed strategy can improve the Bit-Flip decoding performance. Based on the non-uniform sensing strategy, the Min-Sum decoding algorithm is applied in the NAND flash memory model, which demonstrates that the Min-Sum decoding algorithm has a better decoding performance than the Bit-Flip decoding algorithm.