Research And Application On Error Corretion Codes In MLC Storage Channels

With the continuous development of big data,artificial intelligence and cloud storage technology,the information area is subject to new risks and challenges,driving the need for high integration and low power consumption in storage technology.Non-Volatile Memory technology has become a lot hot research topic due to its advantages such as low power consumption,high capacity,fast read/write sepped and small size.With the increasing storage density and the development of multi-level storage and 3D stacking technology,the reliability of storage system is decreasing with the increasing storage capacity.The traditional storage technology can no longer meet the requirements of the current new business.Therefore,inproving the reliability of the new storage media is an urgent problem to be sloved.Error Correcting Code（ECC）is an effective way to ensure the reliability of flash data.In Mulit-Level Cell（MLC）flash memory,the information on the same Word Line（WL）is stored in two separate pages,and different error types lead to the asymmetric characteristics of BER of the two logical pages.Traditional Low-density Parity Code（LDPC）code error correction uses the same code rate for error correction,ignoring this asymmetric feature,which leads to waste of system verification redundancy and reduces storage efficiency.This thesis investigates the coding and decoding methods of LDPC codes based on the asymmetric characteristics of MLC flash memory.The main research working is as follows:（1）Summarize the physical structure of MLC flash memory and provide the basic operations of MLC storage.Based on this,study the noise types of MLC,analyze the mechanism of noise generation,and model the MLC storage channel.（2）A bilayer LDPC encoding and decoding algorithm suitable for flash memory channels is proposed for the asymmetric characteristics of MLC storage channels.The algorithm uses the bilayer parity matrix to code and decode the MLC flash’s LSB pages to improve its decoding performance.Simulation shows that the Bilayer LDPC decoding algorithm can improve the decoding performance by 1～2 orders of magnitude compared with the conventional BP D.and increase the programming/erasing（P/E）cycle by at least 5000 times at a Bit Error Rate of 10^-4.（3）The extra checksum bit in the bilayer LDPC decoding algorithm brings new problems to the flash memory system,such as increasd read/write complexity and increased code rate of MLC pages.To address the above problems,a perturbation decoding algorithm based on genetic algorithm theory is proposed in this thesis,the method introduces perturbation noise on the basis of the bilayer LDPC decoding algorithm and optimizeds nosie by genetic algorithm to improve the decoding performance while reducing the extra redundancy length.Simulation show that the algorithm improves the decoding performance by 1～2 orders of magnitude compared with the bilayer LDPC decoding algorithm and reduces the extra redundancy length by 2/3.