SISO GLDPC Component Decoding for Low-Complexity Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing error correction mechanisms for non-volatile memory devices, such as NAND flash, require high complexity and extra storage space for parity bits, necessitating more efficient error correction codes to enhance data protection with minimal complexity and parity requirements.
Innovation Solution
Implementing soft-input soft-output (SISO) decoding for component codes of generalized low-density parity-check (GLDPC) codes, which reduces complexity by selecting error patterns based on syndrome and channel information, and modifying noisy codewords until successful decoding is achieved.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If strong error-correction codes are used to enhance data reliability, then data protection is improved, but storage space requirements and system complexity increase
Solution Approach 1:
The patent segments the error correction process into component codes (e.g., Hamming codes) that operate on groups of bits. Each component code handles a specific portion of the codeword, allowing parallel processing and reducing the complexity of any single decoding operation while maintaining strong overall error correction capability through the structured combination of multiple simple components.
Solution Approach 2:
The patent employs soft-input soft-output (SISO) decoding that operates on reliability values (log-likelihood ratios) rather than hard binary decisions. This parameter transformation from hard decisions to soft probabilities enables more nuanced error correction, improving reliability while the iterative message-passing algorithm keeps computational complexity manageable through efficient information exchange between variable and check nodes.
2Reliability
If traditional hard-decision decoding is used, then decoding speed is maintained, but error correction performance degrades in noisy channels
Solution Approach 1:
The patent introduces an intermediary soft-decision layer that translates channel observations into reliability values (log-likelihood ratios) before processing. This intermediary representation allows the decoder to incorporate confidence information from the channel, improving error correction performance. The message-passing algorithm then operates on these soft values through check nodes and variable nodes, achieving better reliability while managing complexity through the structured iterative process.
3Reliability
If maximum a posteriori (MAP) algorithms are used for optimal decoding, then error correction performance is maximized, but computational complexity becomes prohibitive
Solution Approach 1:
The patent segments the complex MAP decoding problem into smaller, manageable component codes (such as Hamming codes) that can be decoded using simpler algorithms. By dividing the overall codeword into groups processed by individual component decoders, the system achieves near-MAP performance through the collective action of multiple simple decoders, avoiding the prohibitive complexity of direct MAP implementation.
Solution Approach 2:
The patent implements iterative message-passing decoding that performs partial decoding operations across multiple passes. Rather than attempting exhaustive MAP computation in a single step, the algorithm performs successive refinements, passing messages between check nodes and variable nodes until convergence or a maximum iteration count is reached. This partial iterative action achieves satisfactory performance with manageable complexity.
Data Source
AI summary
Disclosed are devices, systems and methods for improved decoding of a binary linear code. An example method includes receiving a noisy codeword; computing a syndrome based on the noisy codeword; identifying N error patterns that correspond to the syndrome; selecting M error patterns from the N error patterns, wherein M≤N are positive integers, wherein a distance between a codeword corresponding to each of the M error patterns and the noisy codeword is less than a distance between a codeword corresponding to any other error pattern and the noisy codeword, and wherein the distance excludes a Hamming distance; modifying the noisy codeword based on each of the M error patterns one-at-a-time; and decoding the modified noisy codeword one-at-a-time until a successful decoding is achieved.