Nonvolatile Memory Data Scrambling for Pattern Error Recovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Nonvolatile memory systems face reliability issues due to data pattern-induced errors, which conventional error correction codes and mirroring techniques are insufficient to address, especially in Multi-Level Cell (MLC) designs as cell interactions lead to program disturb and uncorrectable data patterns.
Innovation Solution
Implementing a combination of Redundant Array of Independent Disk (RAID) techniques with data scrambling, where different data transformations are applied to each device to break data patterns and provide redundancy, ensuring that errors induced by specific data patterns are not replicated across devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional error correction codes and mirroring techniques are used, then data redundancy is provided, but data pattern-induced errors cannot be prevented and reliability remains insufficient
Solution Approach 1:
The patent applies asymmetry by using different data transformations (first transformation and second transformation) for different devices storing the same data. This asymmetric treatment ensures that data patterns vary across devices, preventing the propagation of pattern-induced errors while maintaining redundancy for reliability.
2Reliability
If data is stored with redundancy across multiple devices, then reliability is improved, but storage efficiency decreases due to additional redundant bits
Solution Approach 1:
The patent changes the transformation parameters applied to data before storage in different devices. By varying the transformation applied to the same data across devices, the system provides redundancy without requiring additional physical storage capacity, as the same data space is utilized with different transformations applied.
3Stability of the object's composition
If the same data transformation is applied across all devices, then data consistency is maintained, but data pattern errors are replicated across devices
Solution Approach 1:
The patent deliberately introduces asymmetry by applying different transformations to the same data across different devices. This asymmetric approach prevents error propagation while maintaining a form of consistency through the use of reversible transformations that allow data recovery from any device.
4Ease of manufacture
If no data transformation is applied, then storage is simple and direct, but data patterns cause errors that cannot be prevented
Solution Approach 1:
The patent applies preliminary action by transforming data before storage in the memory devices. This pre-transformation step modifies data patterns to prevent program disturb errors from occurring in the first place, rather than attempting to correct errors after they occur, thereby maintaining simplicity while preventing harmful effects.
Data Source
AI summary
A nonvolatile memory array includes two or more devices, each device containing data that is scrambled using a different scrambling scheme. When the same data is provided and stored in both devices, different data patterns occur in each device, so that if one of the patterns causes data pattern induced errors, the original data can be recreated from another copy that does not share the same data pattern.