Adaptive Memory Read Write Systems for Multi-Level Cells
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-level memory cells, such as flash memory cells, face challenges in precisely programming and reading data due to Gaussian distribution of charge levels, leading to retention loss after cycling, which affects the accuracy of read and write operations.
Innovation Solution
Adaptive memory read and write systems that compute estimated means and variances of level distributions using signal samples from pilot cells, employing equations to adjust and refine these values, and utilizing slicer thresholds to facilitate accurate data writing and reading.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multi-level memory cells store varying amounts of charge to indicate multiple logic levels, then storage density is improved, but programming precision deteriorates due to Gaussian distribution of charge levels
Solution Approach 1:
The system performs read operations to measure actual charge levels, uses this feedback to compute updated mean and variance values for each charge level distribution, and adjusts subsequent write operations based on these computed statistics. This closed-loop feedback mechanism compensates for the Gaussian distribution effects and improves programming precision.
Solution Approach 2:
The system dynamically changes the parameters (mean and variance) of charge level distributions based on measured data from read operations. By computing updated statistical parameters and using them to adjust write operations, the system adapts to the actual behavior of memory cells and maintains precision despite Gaussian distribution characteristics.
2Measurement precision
If adaptive read and write systems compute estimated means and variances using signal samples, then read and write accuracy is improved, but system complexity increases
Solution Approach 1:
The system performs preliminary read operations on pilot cells to estimate the mean and variance of charge level distributions before performing actual data storage and retrieval operations. This preliminary characterization allows subsequent operations to use optimized parameters, improving accuracy without requiring complex real-time computation during normal operation.
Solution Approach 2:
The system uses pilot cells as copies or representatives of the actual data storage cells. By measuring and characterizing the pilot cells, the system obtains statistical information that can be applied to the actual cells without requiring complex individual characterization of each cell, thus reducing overall system complexity.
3Measurement precision
If detection thresholds are set at intersections of adjacent level distributions, then read operation accuracy is improved, but retention loss after cycling worsens
Solution Approach 1:
The system dynamically adjusts detection thresholds and charge level parameters based on measured statistics from read operations. Rather than using fixed thresholds, the system adapts thresholds to the actual distribution characteristics of charge levels in the memory cells, compensating for shifts caused by cycling and maintaining both accuracy and retention.
Solution Approach 2:
The system uses feedback from read operations to continuously update the statistical parameters of charge level distributions and adjust detection thresholds accordingly. This feedback mechanism allows the system to compensate for retention loss and distribution shifts over time, maintaining read accuracy despite cycling effects.
Data Source
AI summary
Adaptive memory read and write systems and methods are provided that may compute estimated means and variances of multi-level memory cells to facilitate writing and reading of data to and from the multi-level memory cells are described herein. The systems may include an apparatus comprising multi-level memory cells, and an estimation block configured to compute estimated means and variances of level distributions of the multi-level memory cells by processing signal samples provided by at least a subset of the multi-level memory cells, the estimated means and variances to be used to facilitate writing and/or reading of data to and/or from at least selected ones of the multi-level memory cells, the multi-level memory cells having M-levels where M is an integer greater than 1, and each of the level distributions is associated with a corresponding level of the M-levels.