Storage Reliability Optimization Of 3D Charge Trap NAND Flash Based On Data Type Characteristics

With the rise of emerging technologies such as artificial intelligence,deep learning and 5G,the demand for digital storage is increasing day by day.In a wide variety of memory markets,NAND flash memory has gradually become the mainstream of the market by virtue of its superior high reliability and large capacity.However,the traditional Two-dimensional(2D)NAND has encountered a bottleneck in the size of the miniaturization process.In order to seek a breakthrough,the market began to turn to Three-dimensional(3D)stacked NAND flash memory.Today,after more than ten years of development,3D NAND flash memory has been rapidly developed from 24 layers of stacking to 200+layers of stacking,and storage density and storage capacity are constantly improving.However,due to the complex three-dimensional stacking structure of 3D NAND flash and different preparation technology from 2D flash,the reliability problem of 3D NAND flash is more complex.How to combine the characteristics of different data types to optimize the reliability design of 3D NAND flash is an important subject to realize high reliability memory system.In the first part of this paper,the cold data in 3D Charge Trap(CT)NAND is systematically studied and the corresponding solution is proposed.Since data retention is the main error source of cold data,the data retention characteristics of 3D NAND are deeply investigated.The experimental results show that charge loss is the main cause of error during data retention in the storage state with more charge,but the data reading operation will lead to charge injection,which can neutralize part of the error bits caused by charge loss.Based on this mechanism,this paper first analyzes the influence of read frequency and read times on the reliability of cold data,and then proposes two kinds of optimization schemes to suppress charge loss:the prewrite scheme which writes the data before the target data is written to improve the reliability;Based on practical application scenarios and combined with the characteristics of digital image data,a high-to-low flipping scheme(HLF)is proposed.Both schemes can effectively reduce the bit error rate of 3D NAND flash memory and extend the erasable life of the memory.The second part of this paper explores the impact of data mode on the reliability of hot data storage.For flash memory,the reliability is not only related to its physical mechanism,but also the data mode of stored data has a certain impact on the reliability of data storage.A large storage state gap between adjacent data will aggravate the reliability error.Therefore,this paper designs two data modes,namely the hierarchical distributed data coding of storage state and the interleaving distributed data coding of storage state.The experimental results show that improving the distribution of states in the stored data has a certain optimization effect on the reliability of data.The third part of this paper combines 3D NAND flash reliability with Low Density Parity Check Code(LDPC)to explore the decoding capability of LDPC in 3D NAND for different types of data.The relationship between the data retention characteristics and the decoding success rate of LDPC is revealed through the chip measurement analysis of the coded data,and an optimization scheme is proposed to reduce the bit error rate by reading multiple times after data retention.Through testing and verification,the scheme not only effectively improves the reliability of flash chip particle storage,but also improves the success rate of LDPC decoding.Based on 3D CT NAND flash memory,the reliability and related physical mechanism are systematically studied in this paper,and a variety of optimization schemes are proposed based on a large number of measured data of flash memory chips.The innovation points of this paper are as follows:combined with the data characteristics of cold data in the actual storage application,a prewrite scheme which can effectively reduce the bit error rate is proposed by analyzing the reliability of the actual data.According to the performance requirements of digital image storage application,an HLF scheme is proposed to improve the storage reliability of image data,which effectively improves the life of the memory chip.Two data types,namely state ladder distributed coding and layer interleaved coding,are designed to explore the reliability of thermal data and the correlation between data modes.Finally,by combining LDPC with 3D NAND flash memory,the chip measurement analysis of LDPC encoded data is carried out to reveal the law of 3D NAND memory’s own data type characteristics affecting LDPC encoding and decoding ability,and its influence on LDPC decoding ability is analyzed combined with reliability optimization scheme.The above research in this paper provides important basic data for improving the reliability of 3D NAND flash memory,and provides a technical optimization path for realizing high reliability flash memory system.