Flash Memory TVSO Adjustment via Neural Network Reliability Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Solid State Drives (SSDs) face issues with read errors due to incorrect Threshold-Voltage-Shift Offset (TVSO) values when the physical structure of flash memory devices ages prematurely or during low-power states, leading to bandwidth, latency, and Quality of Service (QoS) impacts, and require lengthy calibration processes after firmware updates.
Innovation Solution
A method involving a reliability-state classification neural network (CNN) model is used to predict the reliability state of flash memory devices by training on Flash Characterization Testing Error (FCT-ERROR) values and TVSOmin values, allowing for real-time adjustment of TVSO values and compensating for offline-retention time and temperature effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional threshold-voltage-shift read instructions with fixed TVSO values are used, then the flash memory device can operate with simple control logic, but read errors occur when the physical structure ages prematurely or during low-power states
Solution Approach 1:
The patent implements dynamic adjustment of TVSO values based on monitored physical characteristics of the flash memory device. The system transitions from fixed TVSO values to dynamically selected TVSO values from multiple sets, allowing the read operation to adapt to aging and low-power state conditions while maintaining manageable control complexity through automated monitoring and selection processes
Solution Approach 2:
The system continuously monitors physical characteristics such as retention time, temperature, and program/erase cycle counts, then uses this feedback to select appropriate TVSO values. This closed-loop feedback mechanism ensures read accuracy is maintained under varying conditions without requiring complex control logic, as the selection process is automated based on monitored parameters
2Reliability
If flash characterization testing is performed to identify optimal TVSO values for each reliability state, then read accuracy improves, but the process is time-consuming and complex
Solution Approach 1:
The patent performs comprehensive flash characterization testing and identifies multiple sets of optimal TVSO values for different reliability states during the manufacturing or initialization phase. These pre-characterized TVSO sets are stored in lookup tables, allowing the system to quickly select appropriate values during operation without performing time-consuming calibration tests each time the device state changes
Solution Approach 2:
The system dynamically selects from pre-characterized TVSO sets based on current physical characteristics, avoiding the need to reperform characterization testing. The lookup tables enable rapid selection of optimal TVSO values without the time penalty of repeated calibration processes
3Reliability
If the firmware monitors physical characteristics and performs look-up operations to determine TVSO values, then read errors are reduced, but bandwidth and latency are impacted during firmware updates and calibration
Solution Approach 1:
The system performs firmware updates and TVSO calibration operations in the background or during low-activity periods, preparing lookup tables and updating firmware images without blocking normal read operations. This allows the system to maintain high bandwidth and low latency during updates by avoiding foreground calibration processes that would otherwise halt productivity
Solution Approach 2:
The monitoring of physical characteristics and selection of TVSO values occurs continuously in the background without interrupting normal read operations. The system maintains productive read operations while simultaneously updating firmware and recalibrating TVSO values, ensuring continuous useful action without significant impact on bandwidth or latency
4Reliability
If ECC error recovery processes are used when read errors occur, then data can be recovered, but bandwidth, latency, and Quality of Service are significantly impacted
Solution Approach 1:
The system proactively monitors physical characteristics and adjusts TVSO values before read errors occur, cushioning against potential failures by maintaining optimal read parameters in advance. This preventive approach avoids the need for error recovery processes by ensuring reads are performed with correct TVSO values, thereby maintaining high bandwidth and low latency without triggering ECC recovery operations
Data Source
AI summary
A method and apparatus for determining when actual wear of a flash memory device differs from a reliability state. Configuration files of a reliability-state classification neural network model are stored. The operation of a flash memory device is monitored to identify current physical characteristic values. A read of the flash memory device is performed to determine a number of errors. A neural network operation is performed using as input a set of threshold voltage shift offset values currently being used to perform reads of the flash memory device and the calculated number of errors, to identify a predicted reliability state. The identified current physical characteristic values are compared to corresponding tags associated with the predicted reliability state and a flag or other indication is stored when the comparison indicates that the identified current physical characteristic values do not correspond to the respective tags associated with the predicted reliability state.