Flash Memory Testing With Adaptive Soft-Sensing Noise Matching

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional flash memory decoding methods using the Additive White Gaussian Noise (AWGN) model fail to accurately simulate the complex noise conditions of modern flash memory technologies, leading to suboptimal error bit correction capabilities.

Innovation Solution

A decoding method that utilizes a non-Gaussian N4 noise model to adjust soft sensing step sizes and log likelihood ratios based on statistical description parameters, optimizing the decoder's error bit correction capability by projecting usage conditions into a mesh structure for optimal match combinations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If the Additive White Gaussian Noise (AWGN) model is used to simulate noise in flash memory decoding, then the decoding process can be simplified and performed efficiently, but the error bit correction capability becomes insufficient under complex real-world noise conditions

Engineering Contradiction:

Improvedecoding efficiencyVSAvoiderror bit correction capability

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent changes the noise model parameters from the simplified AWGN model to a more complex non-Gaussian noise model that better represents real flash memory noise characteristics. This includes adjusting statistical parameters to match actual noise distributions, thereby improving error correction capability while maintaining decoding efficiency through optimized parameter selection.

Inventive Principle:

Principle #35Parameter changes

2Quantity of substance

If QLC or PLC technology is adopted to increase storage density, then the number of bits per cell increases to 4 or 5 bits, but potential drift and overlap occur more easily, making the system more susceptible to noise

Engineering Contradiction:

Improvestorage densityVSAvoidpotential drift and overlap

Core Design Contradiction:

Quantity of substanceVSObject-affected harmful factors

Solution Approach 1:

The patent implements feedback mechanisms where decoding results and noise characteristics are continuously monitored and used to adjust the noise model parameters. This adaptive feedback loop allows the system to compensate for potential drift and overlap effects in QLC/PLC technologies, maintaining accurate error correction despite increased storage density.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250279152A1Flash memory testing mechanism capable of optimally match error bit correction capability of decoder with real noise

Publication Date: 2025.09.04 SILICON MOTION INC

AI summary

A decoding method includes: comparing a decoded result with a pre-decoded result of a codeword after decoding the codeword stored by a flash memory, to obtain statistic description parameter(s) in a 3D space for the codeword; determining whether a soft sensing step size currently used is matched to a log likely ratio of a decoder; when it is not matched, changing and adjusting the soft sensing step size to calculate change information of the statistic description parameter(s) caused by a change of the adjusted soft sensing step size to predict a match combination of a soft sensing step size and a corresponding log likely ratio; and using the match combination of predicted soft sensing step size and predicted corresponding log likely ratio to read a next codeword stored by the flash memory.