SSD State Prediction for Read Voltage Shift
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Solid state storage devices face increased error bits and reduced read speed due to shifting threshold voltage distributions in non-volatile memory cells over time, leading to ineffective error correction and prolonged decoding processes.
Innovation Solution
A state prediction method is implemented, where state parameters are collected before the device leaves the factory and used to create prediction functions via machine learning algorithms, allowing for quick prediction of current and future read voltage sets and decoding processes, enabling accurate data retrieval and reducing retry times.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If conventional read voltage sets are used for reading non-volatile memory, then the device structure remains simple, but read speed decreases and error bits increase due to threshold voltage shifting
Solution Approach 1:
The patent applies preliminary action by pre-collecting state parameters before the device leaves the factory and pre-creating prediction functions based on machine learning algorithms. This allows the system to predict future read voltage sets and decoding processes in advance, enabling fast read operations without requiring complex real-time voltage adjustment mechanisms during operation.
Solution Approach 2:
The patent implements parameter changes by dynamically adjusting read voltage sets based on predicted threshold voltage shifts. The prediction functions use collected state parameters (such as program/erase cycle counts, temperature, and error rates) to determine optimal read voltage adjustments, allowing the system to adapt to memory degradation while maintaining simple device architecture.
2Reliability
If retry read operations are performed multiple times to correct errors, then data accuracy improves, but time consumption increases significantly
Solution Approach 1:
The patent applies feedback by continuously monitoring state parameters during program and erase operations, and using this feedback to update prediction functions. The system feeds back error information from previous read operations to refine voltage shift predictions, enabling more accurate predictions of optimal read voltages and reducing the need for multiple retry operations.
Solution Approach 2:
The patent replaces the mechanical retry process with a predictive computational approach. Instead of physically performing multiple read retries with different voltage sets, the system uses machine learning prediction functions to calculate the optimal read voltage set in advance, substituting computational prediction for iterative physical retry operations.
3Measurement precision
If threshold voltage distributions shift due to repeated writes and erasures, then manufacturing simplicity is maintained, but measurement precision of read data deteriorates
Solution Approach 1:
The patent implements self-service by enabling the memory device to automatically monitor its own state parameters and perform self-diagnosis of threshold voltage shifts. The prediction functions allow the device to self-adjust read voltages based on its own operational history and degradation patterns, eliminating the need for external calibration equipment or complex control systems.
Solution Approach 2:
The patent applies preliminary action by pre-collecting state parameters during manufacturing and operation, and pre-training prediction functions to anticipate future threshold voltage distributions. This allows the system to maintain high measurement precision without adding complex real-time adjustment mechanisms, as the prediction models are prepared in advance based on accumulated operational data.
Data Source
AI summary
A solid state storage includes a non-volatile memory and a controlling circuit. The non-volatile memory includes a first block. The controlling circuit is connected with the non-volatile memory. The controlling circuit includes a function storage circuit. The function storage circuit stores plural prediction functions. According to plural state parameters corresponding to the first block and a first prediction function of the plural prediction functions, the controlling circuit predicts a read voltage shift of the first block.