基于多元LDPC码的MLC Nand Flash差错控制技术研究

发布时间：2018-05-31 18:16

本文选题：MLC + Nand　；参考：《杭州电子科技大学》2016年硕士论文

【摘要】：数字存储产业属于新兴的高科技产业。以高密度存储技术为代表,Nand Flash技术的高速发展已经引起了产业内的多轮革命性更新换代。基于其结构简单、读写速度快、稳定性高、能耗低等一系列优点,Nand Flash技术以绝对的优势迅速得到普及并成为存储领域的应用主流。随着制造工艺的不断升级,原始的单层单元(Single Level Cell,SLC)存储技术已逐渐发展到存储密度更大的多层单元(Multi-Level Cell,MLC)存储技术。近几年,Nand Flash的存储密度扩展了2倍以上。然而存储密度的增加也带来了新的挑战,即错误概率显著增加。随着误码率的成倍上升,BCH和RS码越来越难以满足Nand Flash对差错控制技术的要求。因此,急需寻找一种纠错能力更加强大的纠错码。低密度奇偶校验(Low Density Parity Check,LDPC)码有接近香农容限的优良特性,可以应用于Nand Flash差错控制领域。目前,已经有学者对用于Nand Flash纠错的二元LDPC码进行了研究。但一般情况下,适用于Nand Flash纠错的二元LDPC码通常很长,并且纠错性能也没有多元LDPC码优越。多元LDPC码有很多优点,但其编码、译码复杂度相对较高,因此,寻找可快速编码的编码算法和译码性能好且复杂度较低的译码算法尤为重要。本文重点对多元LDPC码字的编码、译码算法展开深入研究,并构造出一类适用于MLC Nand Flash差错控制技术的多元LDPC码字。本文组织结构如下:首先,简要介绍MLC Nand Flash的结构特点和基本操作;然后阐述了LDPC码的相关理论知识,重点研究多元LDPC码编码、译码理论,并采用一种无四环且可快速编码的码字构造方法构造出一系列适用于MLC Nand Flash纠错的高码率且可快速编码的多元准循环LDPC(Quasi-Cyclic LDPC,QC LDPC)码字;接着提出一种新的译码算法,即基于平均幅度和停止准则的加权符号翻转(Average Probability and Stopping Criterion Weighted Symbol Flipping,APSCWSF)译码算法,其纠错性能良好且运算量较软判决算法低很多。该算法把与校验节点相关联所有符号节点的平均概率记为权重,获得了更加可靠的翻转函数,提高符号的翻转效率进而达到改善译码性能的目的。最后,通过系统仿真与分析充分验证所构造多元QC LDPC码字及所提出编码、译码算法应用于MLC Nand Flash纠错领域的可行性和优越性。
[Abstract]:The digital storage industry belongs to the emerging high-tech industry. The rapid development of Nand Flash technology, represented by high density storage technology, has caused many revolutionary changes in the industry. Based on its simple structure, fast reading and writing speed, high stability, low energy consumption and a series of advantages, Nand Flash technology has been popularized rapidly and become the mainstream of storage applications. With the continuous upgrading of manufacturing process, the original single Level cell (SLC) storage technology has gradually developed to the multi-level cell / MLC storage technology with higher storage density. The storage density of Flash has more than doubled in recent years. However, the increase in storage density also brings a new challenge, that is, a significant increase in error probability. With the increase of bit error rate (BER), it is more and more difficult for BCH and RS codes to meet the requirements of Nand Flash for error control technology. Therefore, it is urgent to find a more powerful error-correcting code. Low Density Parity check Density Parity (LDPC) codes with low density parity check have good properties close to Shannon tolerance and can be used in the field of Nand Flash error control. At present, some scholars have studied binary LDPC codes for Nand Flash error correction. However, in general, binary LDPC codes suitable for Nand Flash error correction are usually very long, and the error correction performance is not as good as multivariate LDPC codes. Multivariate LDPC codes have many advantages, but their coding complexity is relatively high. Therefore, it is very important to find fast coding algorithms and decoding algorithms with good decoding performance and low complexity. This paper focuses on the coding and decoding algorithms of multivariate LDPC codewords, and constructs a class of multivariate LDPC codewords suitable for MLC Nand Flash error control techniques. The organization and structure of this paper are as follows: firstly, the structure characteristics and basic operations of MLC Nand Flash are briefly introduced, then the relevant theoretical knowledge of LDPC codes is expounded, and the coding and decoding theory of multivariate LDPC codes is studied. A series of multivariate quasi-cyclic LDPC(Quasi-Cyclic LDPCQC code words with high bit rate and fast coding are constructed by a fast coding method with no four loops, and a new decoding algorithm is proposed. That is, the weighted symbol flipping average Probability and Stopping Criterion Weighted Symbol averaging / APSCWSF decoding algorithm based on the average amplitude and stop criterion has good error correction performance and much lower computational complexity than the soft decision algorithm. The algorithm takes the average probability of all the symbol nodes associated with the check node as the weight and obtains a more reliable flipping function to improve the symbol turnover efficiency and improve the decoding performance. Finally, the feasibility and superiority of constructing multivariate QC LDPC codewords and the proposed coding and decoding algorithm in the field of MLC Nand Flash error correction are fully verified by system simulation and analysis.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN911.22

【相似文献】