Read Retry Table Clustering for Non-Volatile Memory Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As the density of non-volatile memory devices increases, they become more prone to errors due to process variations and noise, leading to reduced reliability and durability, which existing error correction codes struggle to address effectively, especially as the number of program/erase cycles increases.
Innovation Solution
A read retry table (RRT) apparatus is used to collect data from non-volatile memory cells, perform clustering based on error correction capability, and dynamically adjust read reference voltages to improve data integrity by identifying and correcting errors through a machine learning mechanism.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If the density of non-volatile memory devices is increased, then the storage capacity is improved, but the error rate increases due to process variations and noise
Solution Approach 1:
The patent performs preliminary clustering analysis on memory cell data before actual read operations to pre-determine optimal read reference voltages for different word line groups. This preliminary action enables the system to proactively compensate for process variations and noise effects that will occur during operation, thereby maintaining low error rates even as storage density increases
Solution Approach 2:
The patent dynamically adjusts read reference voltage levels based on clustered data patterns and error characteristics. By changing the voltage parameters adaptively according to the specific word line group and observed error trends, the system compensates for increased noise and process variations inherent in high-density memory devices, thus reducing error rates while maintaining high storage capacity
2Reliability
If existing error correction codes are used, then some errors are corrected, but they struggle to address errors effectively as the number of program/erase cycles increases
Solution Approach 1:
The patent implements a feedback mechanism where read retry table data and error patterns are continuously collected and analyzed. The clustering results and error statistics from previous operations feed into subsequent read operations, allowing the system to adaptively refine voltage adjustments and error correction strategies as program/erase cycles accumulate, thereby maintaining effective error correction throughout the device lifespan
Solution Approach 2:
The system performs preliminary error analysis and clustering on data patterns before degradation becomes severe. By identifying error trends early in the program/erase cycle lifecycle and pre-adjusting read parameters, the system proactively compensates for wear-related errors before they exceed the correction capability of traditional ECC codes
3Reliability
If a read retry table apparatus with machine learning mechanism is implemented, then error correction capability is enhanced, but the device complexity increases
Solution Approach 1:
The patent segments the memory device into multiple word line groups and applies clustering separately to each segment. This segmentation allows the complex machine learning operations to be distributed and managed in smaller, independent units, reducing the overall system complexity while maintaining enhanced error correction capabilities across the entire device
Solution Approach 2:
The patent uses clustering to identify and copy successful error correction patterns and voltage adjustment parameters from one word line group to similar groups. By replicating proven correction strategies rather than performing full machine learning analysis on every group, the system reduces computational complexity while maintaining high error correction effectiveness
Data Source
AI summary
A read retry table (RRT) apparatus is coupled to a plurality of memory dies via a data path. The apparatus is configured to collect data from a plurality of memory cells coupled to a plurality of word lines in the plurality of memory dies via the data path; perform a first clustering on the plurality of word lines based on an error correction capability of error correction circuitry for collected data; perform a second clustering on an outlier of the first clustering; and generate or update an RRT based on values obtained from the first clustering and the second clustering.