Non-Volatile Memory Read Recovery Using Cell-Specific Voltage Offsets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Non-volatile memories face errors in data read due to cell-dependent systematic noise, which page-wide techniques cannot effectively address since the read reference voltage is the same for all cells in a page.
Innovation Solution
A method is introduced where a quantized estimate of the offset in the read reference voltage is made for each cell to correct for systematic noise, adjusting the reference voltage for each cell and processing the combined results through error correction again.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If page-wide techniques are used to handle systematic noise, then the read reference voltage is uniformly applied to all cells in a page, but cell-dependent systematic noise cannot be effectively corrected since the same reference voltage is used for all cells
Solution Approach 1:
The patent segments the page-wide noise correction problem into cell-level corrections by dividing cells into multiple groups based on their physical locations. Each group receives a tailored read reference voltage adjustment based on its specific systematic noise characteristics, thereby resolving the contradiction between uniform treatment and cell-dependent correction needs.
Solution Approach 2:
The patent applies local quality by assigning different read reference voltage offsets to different cell groups based on their specific physical locations and noise characteristics. Instead of uniform page-wide correction, each local region receives customized correction parameters, enabling effective cell-dependent systematic noise correction while maintaining manageable complexity through group-based organization.
2Reliability
If quantized estimate of offset is made for each cell and read reference voltage is adjusted accordingly, then cell-dependent systematic noise is corrected, but the processing complexity and time increase
Solution Approach 1:
The patent reduces processing time by segmenting cells into groups based on physical location and applying group-level offset estimates rather than individual cell-level estimates. This segmentation approach maintains effective noise correction while significantly reducing the computational burden and processing time associated with evaluating each cell separately.
Solution Approach 2:
The patent applies partial action by estimating offset values for representative cells within each group rather than all cells, then applying these estimates to entire groups. This partial estimation approach achieves sufficient correction accuracy while reducing processing time and computational resources compared to full cell-level analysis.
Data Source
AI summary
Various embodiments of the invention pertain to a technique of recovering data from a portion of a non-volatile memory which was not reliably read because the number of read errors exceeded the ability of the ECC process to correct those errors. For each cell in that portion of memory, a quantized estimate is made of the amount of offset in the read reference voltage that is predicted to correct for any systematic noise that may have affected the reading of that cell. For each quantized offset, the read reference voltage is adjusted by that amount and data from the relevant cells is read. The combined results for all the cells are then processed through the ECC again.