Segmented LDPC Decoding for Throughput and Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional LDPC decoders face challenges in high-throughput implementations due to the costly overhead of obtaining soft information for error correction, which can lead to reduced bit error rate (RBER) correction strength, especially in high-speed communication and noisy storage systems.
Innovation Solution
The proposed solution involves segmenting the codeword into portions to apply variable weighting and scaling to check nodes based on expected error rates, allowing for differential log likelihood ratio (LLR) assignments and reduced decoding time by focusing on error-free segments, thereby improving RBER correction strength and power efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If soft information (LLR) is obtained for different portions of the codeword to improve decoding performance, then decoding performance is improved, but costly overhead is incurred that reduces throughput
Solution Approach 1:
The codeword is divided into multiple segments, and the check nodes are divided into corresponding groups. Each group of check nodes processes only its associated codeword segment independently. This segmentation enables parallel processing of multiple segments simultaneously, improving throughput while maintaining the ability to apply soft information where needed without processing the entire codeword sequentially.
Solution Approach 2:
The decoder applies soft information (LLR) selectively only to segments that require error correction, rather than processing the entire codeword with full soft information overhead. For segments that are already error-free or have low error rates, the decoder can skip expensive soft information processing, reducing overall overhead while maintaining decoding performance where it matters most.
2Reliability
If RBER diversity information is applied to improve error correction in different codeword portions, then error correction strength is improved, but significant storage space is required for check node operation
Solution Approach 1:
Check nodes are divided into multiple groups, with each group associated with a specific codeword segment. Each check node group stores and processes only the RBER diversity information relevant to its assigned segment, rather than maintaining storage for the entire codeword. This segmentation dramatically reduces the total storage space required while preserving the ability to apply RBER diversity information for improved error correction in each segment.
3Reliability
If the entire codeword is decoded without segmentation, then complete error correction is achieved, but decoding time increases and power consumption rises
Solution Approach 1:
The codeword is segmented into multiple independent processing units, each handled by a dedicated check node group. These segments can be decoded in parallel rather than sequentially, significantly reducing total decoding time. The segmentation maintains error correction completeness by ensuring that all segments are processed, while the parallel execution and selective processing of only necessary segments reduce both time and power consumption.
Solution Approach 2:
The decoder performs preliminary processing of codeword segments that appear to be error-free or have low error rates, identifying them early in the process. These segments can then be skipped or given minimal processing in subsequent decoding stages, reducing overall decoding time and power consumption while still maintaining complete error correction for segments that actually require it.
Data Source
AI summary
A low-density parity-check (LDPC) decoder performs check node computations as N different segments of the check nodes which have connections only to a codeword segment of length C/N bits as well as check nodes that have connections across the entire codeword of length C. The decoder can include a controller or other compute hardware to decode the codeword, including to perform computations for separate segments of C/N bits of the codeword. The system can perform computations including adjustment of the decode computations based on an expected error rate for selected segments of the codeword.