Study Of Polar Codes Decoding And Channel Estimation For MLC NAND Flash Memories

As a non-volatile memory,NAND flash has been widely used in people's lives due to its advantages of large capacity and low cost.With the continuous development of production processes,the storage density of NAND flash has been increased,while which makes the error probability of the read-back data be larger at the same time.Therefore,it is particularly important to improve the data storage performance of multi-level cell NAND flash memory.In recent years,polar code has been proposed as an error-correcting code which can reach the limit of channel capacity.Therefore,the application of polar code to NAND flash memory has broad prospects.This article studies on the NAND flash memory with focuses on the application of polar code and the channel estimation.The main work is summarized as follows:The related algorithms of polar code and the fundamental of NAND flash memory have been studied.Furthermore,the flash memory channel model was constructed by various noises in flash memory,and the high-rate polar code for NAND flash memory is acquired by information bottleneck-based construction and then RQUP-based shortening.The threshold voltage sensing method based on the maximum mutual information(MMI)criterion is studied.Combined with the characteristics of the channel model,a simplified sensing method based on MMI criterion is proposed.This method is mainly based on the observation that there is nearly no aliasing happens between two non-neighbored states.Thereby,the sensing method of the multi-level memory cell NAND flash memory is simplified.The simulation found that the simplified method has lower complexity,and the same performance as the original method when the Program/Erase times below 50 K.Under the different Program/Erase times,the LLR information look-up tables are given by utilizing the six threshold voltages acquired by the simplified MMI-based sensing.In addition,the simulations show the performance of different initial retention time with the fixed Program/Erase time,and compare the performance of different decoding algorithms of Polar codes on the flash memory channel,and analyze the effect of different Program/Erase times on data retention time.A neural network-based channel estimation is proposed.Compared with the conventional RABP-CU method,the proposed channel estimation shows a higher estimation accuracy at PE=5K and much lower complexity.Moreover,in the polar-coded NAND flash memory,the proposed channel estimation can enlarge the retention time and thus extending the lifetime of the flash memory,which is demonstrated by simulations.Long-term data storage will increase the retention process noise,and channel estimation is an effective method to deal with the retention process noise by considering the impact of data retention time.A channel estimation method is proposed based on the trained neural network structure in this paper.Based on the trained neural network structure and the flash channel probability information,the flash memory threshold voltage distribution parameters are calculated,then the simplified MMI-based sensing and the polar decoding are implemented.Simulation results show that compared with the conventional RABP-CU method,the proposed channel estimation shows a higher estimation accuracy at PE=5K and much lower complexity.Moreover,in the polarcoded NAND flash memory,the proposed channel estimation can enlarge the retention time and thus extending the lifetime of the flash memory,which is demonstrated by simulations.