NAND Memory Health Mitigation Using Decoder Statistics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing non-volatile memory technologies face challenges in proactive management of memory health due to variations in NAND behavior under different conditions and process non-uniformity, leading to inefficiencies in error correction and flash management.
Innovation Solution
Implementing decoder statistic-based memory health mitigation by tracking decoding metrics such as syndrome weight and bit error rate to proactively adjust read reference voltages, trigger soft bit reads, or perform direct look-ahead reading, and manage data refresh based on decoding statistics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If exception-based mitigation schemes are used to handle decoding failures, then decoding failures can be addressed when they occur, but the scheme reacts rather than proactively preventing issues and may trigger over-mitigation due to process non-uniformity
Solution Approach 1:
The patent implements preliminary action by monitoring decoder statistics (such as syndrome weight distributions) before decoding failures occur. The system proactively identifies blocks with deteriorating error correction performance and applies mitigation techniques (like data refresh or read retry) before actual decoding failures happen, transitioning from reactive to proactive memory health management
Solution Approach 2:
The patent employs feedback mechanisms by continuously monitoring decoder statistics and using this information to adjust flash management decisions. The system feeds back decoder performance metrics to the flash management logic, which then dynamically adjusts read parameters or triggers data refresh operations based on the observed decoder health, creating a closed-loop control system
2Reliability
If traditional flash management approaches are used, then standard error correction can be maintained, but process non-uniformity between blocks and word lines causes variation in decoder performance that reduces mitigation effectiveness
Solution Approach 1:
The patent applies local quality by treating different blocks and word lines individually based on their specific decoder statistics. Instead of uniform flash management, the system monitors and manages each block's health independently, applying mitigation only where needed. This allows the system to adapt to local process variations and non-uniformity in decoder performance across different memory regions
3Reliability
If proactive memory health management is implemented using decoder statistics, then the likelihood of decoding failures can be reduced, but additional monitoring and analysis capabilities are required
Solution Approach 1:
The patent implements self-service by having the decoder itself generate the monitoring data needed for health management. The decoder's existing syndrome weight calculations, which are already performed during normal operation, are repurposed as health metrics. This eliminates the need for separate monitoring hardware or complex analysis infrastructure, as the system uses its own operational data for proactive health management
Data Source
AI summary
Technology is disclosed herein for memory health monitoring and mitigation based on decoding statistics. Decoding a frame results in a decoding metric (syndrome weight, fail bit count) for that frame. The system tracks a statistic for different sets of frames. The statistic for a set is based on the decoding metrics for that set. The frames may be assigned to sets based on read reference voltages used to read frames or the physical location of the memory cells that store the frames. Memory health mitigation may be performed based on the decoding statistics. One example mitigation is to modify the read reference voltages for the set. Another example mitigation is to trigger reading at soft bit reference levels for a block. Another example mitigation is to trigger direct look ahead reading for a block. Still another example mitigation is to add a block to list of candidates for data refresh.