Soft-Decision LLR Generation for Flash Read Margin Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data storage systems, particularly those using flash memories, face challenges in decoding data due to reduced read margins caused by physical conditions and wear, leading to overlapping voltage distributions that single-read inputs cannot adequately address, resulting in insufficient information for accurate data retrieval.
Innovation Solution
The implementation of a soft-decision input generation method using multiple reads to generate log-likelihood ratios (LLRs) for LDPC decoders, which involves multiple voltage threshold reads to provide additional bits of information and improve data decoding reliability, along with an error management module that tracks voltage thresholds and uses look-up tables for calibration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If single-read input is used for decoding, then the decoding process is simple and fast, but the reliability is insufficient when voltage distributions overlap due to wear and physical conditions
Solution Approach 1:
The decoding process is segmented into multiple stages: first attempting hard-decision decoding with single-read input, and if that fails, proceeding to soft-decision decoding with multiple-read input. This segmentation allows the system to handle most cases simply while providing a more complex fallback for difficult cases.
Solution Approach 2:
The system performs multiple reads (excessive action) only when necessary - specifically when hard-decision decoding fails. For normal cases, a single read suffices. This partial application of the complex solution minimizes overhead while ensuring reliability when needed.
2Measurement precision
If multiple reads are performed to generate soft-decision inputs, then sufficient information is obtained for accurate decoding, but the read time and processing complexity increase
Solution Approach 1:
Multiple reads are performed only when hard-decision decoding fails, not for every decoding operation. This selective approach ensures high measurement precision when needed while minimizing time loss for the majority of cases that can be decoded with single-read input.
Solution Approach 2:
The system uses feedback from the hard-decision decoding result to determine whether to proceed to soft-decision decoding. If hard-decision succeeds, the process terminates; if it fails, the feedback triggers additional reads and soft-decision processing.
3Productivity
If hard-decision input is used, then the decoding process is computationally efficient, but it cannot adequately address overlapping voltage distributions caused by wear
Solution Approach 1:
The decoding system is segmented into two pathways: a fast hard-decision pathway for efficient processing of clear cases, and a more reliable soft-decision pathway for handling cases with overlapping voltage distributions. This segmentation maintains high productivity for most cases while ensuring reliability when needed.
Solution Approach 2:
The computationally intensive soft-decision process is applied partially - only when hard-decision fails. This allows the system to maintain high average productivity while having the capability to handle difficult cases that require more thorough processing.
Data Source
AI summary
An error management system for a data storage device can generate soft-decision log-likelihood ratios (LLRs) using multiple reads of memory locations. 0-to-1 and 1-to-0 bit flip count data provided by multiple reads of reference memory locations can be used to generate probability data that is used to generate possible LLR values for decoding target pages. Possible LLR values are stored in one or more look-up tables.