Flash Memory Read Retry Using LLR Estimation After Decode Failure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solid state storage systems face challenges in error recovery due to insufficient information from single-read operations, leading to inefficiencies in decoding and increased bit error rates, particularly in flash memory devices where hard decision decoding is insufficient for data recovery.
Innovation Solution
The method involves processing multiple read values to determine read parameters such as log likelihood ratios (LLRs) and center read reference voltages, using signal counts from various regions to customize read parameters for improved decoding and error correction, employing hardware acceleration, firmware, and state machines to enhance error recovery processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple read operations are performed to obtain sufficient information for error recovery, then decoding accuracy is improved, but read latency and system complexity increase
Solution Approach 1:
The patent performs preliminary read operations to gather signal information before decoding is attempted. By pre-obtaining multiple read values and calculating LLRs in advance, the system prepares soft information that can be immediately used for error correction, reducing the need for repeated read cycles and thereby lowering overall latency.
Solution Approach 2:
The patent dynamically adjusts the number of read operations based on channel conditions and error patterns. Rather than always performing a fixed number of reads, the system adapts the read strategy to match actual memory state and error characteristics, optimizing the balance between decoding accuracy and latency.
2Measurement precision
If multiple read operations are performed to obtain sufficient information for error recovery, then decoding accuracy is improved, but system complexity increases
Solution Approach 1:
The patent introduces log likelihood ratios (LLRs) as an intermediary representation that bridges multiple read operations and the decoding process. By converting raw read values into LLRs, the system creates a standardized soft information format that simplifies the integration of multiple reads and makes the data more suitable for iterative decoding algorithms, thereby managing complexity.
Solution Approach 2:
The patent replaces hard decision decoding (mechanical binary decisions) with soft decision decoding using LLRs. This substitution allows the decoder to work with probabilistic information from multiple reads, improving accuracy while the mathematical framework of LLRs provides a systematic way to handle the complexity of combining multiple read operations.
3Device complexity
If hard decision decoding is used for simplicity, then system complexity is reduced, but bit error rate increases
Solution Approach 1:
The patent changes the parameter representation from hard binary decisions to soft probabilistic values (LLRs). By transforming the decoding parameter from discrete 0/1 values to continuous likelihood ratios, the system enables more nuanced error correction while maintaining a manageable system architecture through standardized LLR calculation procedures.
Data Source
AI summary
Read parameter estimation techniques are provided that obtain information from multiple read operations to customize read parameters for data recovery. One method comprises performing the following steps, in response to a decoding failure of a page of a memory or a codeword of the memory: obtaining at least three read values of the page or codeword; and processing the at least three read values to determine read parameters comprising: (i) a log likelihood ratio, and/or (ii) a center read reference voltage, wherein the determination is based on a signal count of a number of bits falling in particular regions of multiple regions of the memory and wherein the determined read parameters are used for a decoding of the page or codeword following the decoding failure and/or a subsequent read operation following a successful decoding of the page or codeword.