ML-Based Read Voltage Determination for Memory Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory systems face challenges in accurately reading data due to changes in cell threshold voltages caused by factors like read disturbance, leading to failed bits and increased ECC decoding overhead, with current methods requiring extensive human resources and large lookup tables for optimal read voltage determination.
Innovation Solution
The implementation of machine learning algorithms to determine optimal read voltages for memory systems, using supervised ML training to generate transfer functions that adjust read voltages based on input parameters, reducing ECC decoding overhead and improving read accuracy by dynamically updating read voltage data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional methods are used to determine read voltages with large lookup tables, then read voltage determination is possible, but memory overhead and device complexity increase
Solution Approach 1:
The patent changes the approach from storing fixed lookup tables to using machine learning models that dynamically determine read voltages based on input parameters. The ML model learns optimal read voltage mappings from training data and applies this knowledge without requiring large stored tables, thus reducing memory overhead while maintaining read accuracy.
Solution Approach 2:
The patent replaces the traditional mechanical/memory-based lookup table system with a machine learning-based computational system. Instead of storing pre-computed voltage values in large tables, the system uses trained ML models to predict optimal read voltages, substituting physical storage with intelligent computation.
2Measurement precision
If extensive lookup tables are used for optimal read voltage determination, then read voltage accuracy is maintained, but computation time and processing overhead increase
Solution Approach 1:
The patent performs preliminary machine learning training during the manufacturing or initialization phase to create optimized voltage determination models. This pre-computation allows the system to make fast, accurate read voltage decisions during actual operations without requiring extensive real-time computation or large lookup tables, thus reducing read latency while maintaining precision.
3Ease of manufacture
If traditional read voltage methods are used, then simple implementation is possible, but failed bit count increases due to threshold voltage changes
Solution Approach 1:
The patent introduces dynamic adaptation by using machine learning models that can adjust read voltage decisions based on learned patterns from training data. The system transitions from static, fixed voltage determination to dynamic, adaptive voltage selection that accounts for threshold voltage changes and other factors, reducing failed bit counts while maintaining implementation feasibility.
4Reliability
If machine learning algorithms are implemented to determine read voltages, then read accuracy and reliability improve, but device complexity and implementation difficulty increase
Solution Approach 1:
The patent segments the machine learning functionality into distinct components: a training phase that creates the voltage determination model, and an inference phase that applies the model during read operations. This segmentation allows the complex ML functionality to be implemented in a manageable way, separating the heavy computational training from the lightweight operational use, thus improving reliability while controlling implementation complexity.
Data Source
AI summary
Systems, methods, and apparatus including computer-readable mediums for determining read voltages for memory systems with machine learning (ML) are provided. In one aspect, a memory system includes a memory and a memory controller configured to: obtain a first reading output of memory data using a first read voltage corresponding to a first set of parameters associated with the memory data; if the first reading output fails to pass an Error Correction Code (ECC) test, obtain a second reading output of the memory data using a second read voltage corresponding to a second set of parameters associated with the memory data and including the first set of parameters, the second read voltage being generated using at least one ML algorithm based on the second set of parameters; and if the second reading output passes the ECC test, output the second reading output as a target reading output of the memory data.