Memory Read Threshold Prediction Using DNNs for Aging Cells
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Non-volatile memory devices face challenges in maintaining data integrity due to voltage level drops caused by program disturb and inter-cell interference, leading to sub-optimal read voltage thresholds that degrade over time.
Innovation Solution
Utilizing a set of first deep neural networks (DNNs) to estimate cell voltage distribution parameters and a second DNN to integrate these estimates, generating optimized read voltage thresholds that account for memory cell degradation, thereby improving data retrieval accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional fixed read voltage thresholds are used, then initial read accuracy is achieved, but data retrieval accuracy degrades over time due to voltage drop from program disturb and inter-cell interference
Solution Approach 1:
The patent implements dynamic read voltage threshold adjustment by training deep neural networks to predict optimal thresholds based on program/erase cycle counts and voltage distribution characteristics. The thresholds evolve over time to compensate for voltage drop, transforming the static threshold system into a dynamic one that adapts to aging effects and maintains data retrieval accuracy throughout the memory device lifespan
Solution Approach 2:
The system uses feedback from voltage distribution sampling and DNN prediction to continuously optimize read thresholds. The DNN models learn from historical voltage distribution data and P/E cycle information, generating feedback-driven threshold adjustments that compensate for degradation mechanisms, thereby resolving the contradiction between initial accuracy and long-term reliability
2Reliability
If deep neural networks are trained to predict optimal read voltages, then data retrieval accuracy is improved throughout the memory lifespan, but computational complexity and training time increase
Solution Approach 1:
The patent segments the DNN training process into multiple stages: initial training with comprehensive voltage distribution data, fine-tuning with targeted samples, and validation phases. This segmentation reduces overall computational complexity by breaking down the large-scale training problem into manageable sub-tasks that can be executed efficiently
Solution Approach 2:
The system employs partial training approaches where DNNs are trained on selectively sampled voltage distribution data rather than exhaustive datasets. By using representative samples and progressive training strategies, the patent achieves sufficient model accuracy with reduced computational resources, balancing reliability improvement against complexity constraints
Data Source
AI summary
Devices, systems and methods for improving performance of a memory device are described. An example method includes obtaining a plurality of samples corresponding to a probability distribution for each of a plurality of cell voltage distributions of the memory device, each of the plurality of cell voltage distributions corresponding to a read voltage, determining, based on the samples obtained for the plurality of cell voltage distributions, a number of first deep neural networks (DNNs), estimating, for each of the first DNNs, one or more parameters of the corresponding probability distribution based on the plurality of samples, training, based on the samples and the corresponding one or more parameters, each of the first DNNs, and training, based on the samples and the one or more parameters from each of the first DNNs, a second DNN to enable generation of an updated read voltage value for retrieving information from the memory device.