NAND Memory Soft Decoding with Adaptive LLR Shifts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current soft decoding methods in NAND memory systems suffer from low decoding efficiency, particularly when reading data from memory cells with overlapping threshold voltage distributions, leading to increased decoding delays and errors.
Innovation Solution
A method involving multiple shifts of log-likelihood ratios (LLR) and use of lookup tables to improve soft decoding efficiency by adjusting reference read voltages and performing soft decoding operations after each shift, including positive, negative, and cross-shifts, with the option to switch to different lookup tables when decoding fails.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If soft decoding is performed using conventional methods, then data can be read from memory, but decoding efficiency is low and decoding delays increase
Solution Approach 1:
The patent applies dynamics by making the log-likelihood ratio (LLR) values adjustable through multiple shifting operations (positive shift, negative shift, cross-shift) rather than using fixed LLR values. This dynamic adjustment allows the system to adapt to different reading conditions and improve decoding efficiency when conventional soft decoding fails, directly addressing the low decoding efficiency problem while managing decoding time through targeted adjustments rather than exhaustive searches.
Solution Approach 2:
The patent changes the parameter of LLR values by performing multiple shifts (positive, negative, cross-shifts) on the initial LLR values obtained from the first lookup table. This parameter transformation enables the system to explore different decoding possibilities and improve decoding efficiency when the initial soft decoding fails, resolving the contradiction between decoding efficiency and time loss by making the decoding process adaptive rather than static.
2Productivity
If multiple shifts of log-likelihood ratios are performed, then decoding efficiency improves, but device complexity increases
Solution Approach 1:
The patent segments the decoding process into distinct phases: initial soft decoding with first LLR values, conditional shifting operations (positive shift, negative shift, cross-shift), and selective lookup table switching. This segmentation allows the system to manage complexity by breaking down the decoding task into manageable steps, each with specific purposes, rather than implementing a monolithic complex decoding algorithm.
Solution Approach 2:
The patent performs preliminary actions by pre-defining multiple shifting operations and lookup tables before the actual decoding process. The system prepares the decoding framework in advance with predetermined shift values and alternative lookup tables, so that when decoding occurs, the complexity is managed through pre-planned options rather than real-time complex calculations, improving efficiency while controlling complexity.
3Reliability
If conventional soft decoding is used, then the decoding process is simple, but data read accuracy decreases for overlapping threshold voltage distributions
Solution Approach 1:
The patent implements feedback by continuously monitoring the results of soft decoding operations and using this information to determine whether shifting operations or lookup table switches are needed. The system feeds back the decoding outcomes to guide subsequent operations, allowing it to adapt to overlapping threshold voltage distributions and improve data read accuracy while managing complexity through condition-based responses rather than always-executing complex procedures.
Data Source
AI summary
In examples, a method of controlling a memory system comprises obtaining a first soft-bit data corresponding to a hard-bit data read from a memory and a first lookup table, where the first lookup table comprises a first log-likelihood ratio determined based on a first reference read voltage of the memory. The method comprises performing a first soft decoding operation according to the first log-likelihood ratio and the first soft-bit data. The method comprises performing at least one shift to the first log-likelihood ratio and performing a second soft decoding operation according to a log-likelihood ratio after each shift and the first soft-bit data when the first soft decoding operation is determined to have failed to decode.