The Method of Predicting Retention Threshold Voltage Distribution for NAND Flash Memory Based on Back-Propagation Neural Network
Kunliang Wang, Gang Du, Zhiyuan Lun, Xiaoyan Liu · 2019
A new retention threshold voltage (Vth) distributions prediction method for 3-D TLC and QLC NAND flash memories based on back-propagation neural network (BpNN) is proposed. The data pre-processing and post-processing techniques are developed to improve the prediction accuracy of BpNN model. Based on our proposed BpNN model, the predicted Vthdistributions after different data retention times (t) and program/erase (P/E) cycles have good agreement with the measurements, which can be used for read voltage optimization (RVO) not only hard-decision decoding algorithm but also soft-decision decoding algorithm. Especially, this BpNN model can be embedded into the SSD controller and help in improving the reliability of NAND flash memory.