优化凿孔极化码在NAND FLASH中的纠错技术研究

Research on Error Correction Technology of Optimized Punctured Polar Code in NAND FLASH

为了改善NANDFLASH(闪存)数据存储的可靠性,提出一种基于优化巴氏参数的凿孔极化码编码方法。使用非均匀存储感知法提高感知精度,计算存储单元重叠区域的相邻边界值和感知电压,并根据得到的感知电压和闪存单元的概率密度函数计算对数似然比(LLR),再将存储单元中各比特的LLR值拟合成高斯分布并得到方差,从而迭代计算每个存储单元位的巴氏参数。并且,根据闪存页容量和元数据信息构造码率匹配的凿孔极化码,使之可以更好地应用于闪存中。仿真结果表明,该方法构造的极化码性能优于蒙特卡洛和高斯近似构造方法,当采用SCL(L=4)译码算法时,在不同的凿孔码字下相比于高斯近似法分别有约2.32、2.1、1.7dB的性能增益,相比于蒙特卡洛法分别有约0.12、0.13、0.15dB的性能增益。研究表明,使用优化巴氏参数法构造的凿孔极化码应用于闪存能够提升纠错性能。当闪存擦除单元的标准差减小时,极化码的纠错性能也会提升。

For improving the reliability of NAND FLASH(flash memory)data storage,a punctured polar code coding method based on optimized bhattacharyya parameters is proposed.The sensing accuracy is improved using the heterogeneous memory sensing meth⁃od,and the adjacent boundary values and sensing voltage of the overlapping region of the storage cell are calculated,then to calculate the log likelihood ratio(Log likelihood ratio,LLR)according to the resulting sensing voltage and the probability density function of the flash cell,and to synthesize the gaussian distribution of the LLR values of each bit in the storage cell and obtain the variance to itera⁃tively calculate the bhattacharyya parameter of each memory cell bit.Construct the punctured polar codes matching the code rate ac⁃cording to the flash page capacity and the metadata information,so that it can be better applied in the flash memory.The simulation re⁃sults show that the performance of the polar code constructed by the method used in this paper is better than the Monte Carlo and Gauss⁃ian Approximation construction methods.When SCL(L=4)decoding algorithm is used,the performance gains under different punc⁃tured codewords are about 2.32dB,2.1dB and 1.7dB respectively compared with Gaussian Approximation method and about 0.12dB,0.13dB and 0.15dB respectively compared with Monte Carlo method.The punctured polar code constructed by the optimized bhattacha⁃ryya parameters method can be applied to the flash memory to improve the error correction performance.When the standard deviation of the flash erase unit is reduced,the error correction performance of the polar code will also be improved.