Deep Neural Network Read Voltage Estimation for NAND Flash Memory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In solid state memory storage devices like NAND flash, noise from program disturb errors and inter-cell interference causes voltage distribution and level changes over time, leading to inaccuracies in read voltages, which affects data integrity and longevity.
Innovation Solution
A deep neural network is employed to estimate read voltage thresholds using inputs such as ones count, checksum, and samples from a skew normal distribution, improving the accuracy of read voltages applied to memory devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional read voltage estimation methods are used, then the device complexity is low, but the measurement precision of read voltages deteriorates due to noise and interference
Solution Approach 1:
A deep neural network is introduced as an intermediary component between the memory device and the read operation. The DNN processes cell count data and ones count information to estimate optimal read voltages, acting as a mediator that transforms raw measurement data into accurate voltage thresholds despite noise and interference effects
Solution Approach 2:
The patent replaces traditional mechanical or algorithmic voltage estimation methods with a data-driven deep neural network approach. Instead of using fixed algorithms or manual calibration, the system uses machine learning models trained on memory device characteristics to dynamically estimate read voltages, substituting conventional computational mechanics with neural network-based inference
2Reliability
If deep neural network is used to estimate read voltages, then the measurement precision improves, but the device complexity increases
Solution Approach 1:
The deep neural network is integrated directly into the memory device controller, enabling the system to perform self-calibration and self-optimization of read voltages. The DNN uses data from the memory device itself (cell counts, ones counts) to automatically adjust read voltage estimates without requiring external calibration equipment or manual intervention, making the system self-sufficient
Solution Approach 2:
The deep neural network serves multiple functions simultaneously: it estimates read voltages, compensates for program disturb errors, corrects inter-cell interference effects, and adapts to different memory device characteristics. This multi-functional approach improves reliability across various operating conditions without requiring separate systems for each function
3Loss of information
If read voltages are not accurately adjusted, then the device operation is simple, but the loss of information increases due to voltage distribution changes
Solution Approach 1:
The deep neural network performs preliminary estimation of read voltages using cell count data and ones count information before actual read operations. By pre-calculating optimal voltage thresholds based on current memory device state, the system prevents information loss from voltage drift and interference before read operations occur, rather than correcting errors after they happen
Solution Approach 2:
The system implements a feedback mechanism where cell count measurements and ones count data from read operations are fed back to the deep neural network. The DNN uses this feedback to continuously refine its voltage estimates and adapt to changing memory device characteristics, thereby preventing information loss from accumulating over time due to voltage distribution changes
Data Source
AI summary
Devices, systems and methods for improving the performance of a memory device are described. An example method includes obtaining a plurality of cell counts for each of a plurality of read voltages applied to the memory device, generating, based on the plurality of cell counts and the plurality of read voltages, at least one ones count, at least one checksum, and a plurality of samples corresponding to a distribution function of at least one read voltage of the plurality of read voltages, determining an updated value for the at least one read voltage based on an output of a deep neural network whose input comprises the at least one ones count, the at least one checksum, and the plurality of samples, and applying the updated value of the at least one read voltage to the memory device to retrieve information from the memory device.