Nonvolatile Memory Data Recovery Read Offset Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Nonvolatile memory systems face challenges in reducing read latency and improving reliability during data recovery read operations due to uncorrectable errors caused by severe memory cell deterioration, which existing methods address inadequately.
Innovation Solution
A data recovery read operation method that involves performing a first read operation on adjacent memory cells, determining offset values based on cell count values and additional information such as program/erase cycles, temperature, and location, and using these values to calibrate the read level for target memory cells, thereby optimizing the read operation through machine learning-derived algorithms or offset tables.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional ECC-based error correction is used for deteriorated memory cells, then error correction capability is maintained, but read latency increases and reliability decreases for severe deterioration cases
Solution Approach 1:
The patent performs preliminary actions by reading adjacent memory cells before reading target memory cells, and by pre-calculating offset values based on cell count values from adjacent cells. This preliminary characterization of degradation patterns enables faster and more accurate error correction during the actual data recovery read operation, reducing latency while improving reliability.
Solution Approach 2:
The patent dynamically adjusts read levels by applying calculated offset values to compensate for degradation. Instead of using fixed read levels, the system changes the read level parameter based on the degradation characteristics inferred from adjacent memory cells, enabling accurate reading of deteriorated memory cells without increasing latency.
2Reliability
If multiple read operations are performed on target memory cells to recover data, then data recovery reliability improves, but read latency increases
Solution Approach 1:
The patent employs self-service by using information from adjacent memory cells to characterize degradation patterns and automatically determine appropriate offset values for reading target memory cells. This self-characterization eliminates the need for extensive trial-and-error read operations, achieving high reliability with minimal read operations and maintaining high read throughput.
Solution Approach 2:
The system uses feedback from reading adjacent memory cells to adjust the reading strategy for target memory cells. The cell count values and degradation patterns observed in adjacent cells provide feedback that guides the selection of offset values and read levels, enabling accurate data recovery in fewer operations and maintaining high productivity.
3Measurement precision
If offset values are determined based on comprehensive degradation information, then read accuracy improves, but system complexity increases
Solution Approach 1:
The patent segments the memory array into adjacent memory cells and target memory cells, and further segments the degradation analysis into discrete cell count value measurements. This segmentation allows the system to determine offset values based on localized degradation patterns rather than requiring comprehensive analysis of the entire memory array, improving read level accuracy while managing complexity.
Solution Approach 2:
The patent introduces cell count values as an intermediary metric that simplifies the relationship between physical degradation and read level adjustments. Instead of directly measuring complex degradation parameters, the system uses cell count values from adjacent cells as an intermediary to infer degradation patterns and determine appropriate offset values, reducing the complexity of the offset determination process.
Data Source
AI summary
A data recover read (DRR) operation method of a nonvolatile memory system includes: performing a first read operation on adjacent memory cells, which are connected to an adjacent wordline adjacent to a target wordline, based on a first specific read level; obtaining a cell count value for the adjacent wordline; determining offset values for a normal read level of target memory cells, which are connected to the target wordline, based on the cell count value for the adjacent wordline; and performing a second read operation on the target memory cells, based on the determined offset values and a result of the first read operation.