LDPC Parity Check Matrix for Layered Decoding Without Shared Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing LDPC parity check matrices, such as those used in MoCA networks, have a lower triangular structure that leads to increased decoding errors and iterations, particularly in the last parity column with a column weight of one, resulting in reduced robustness.
Innovation Solution
A robust LDPC parity check matrix is designed with a systematic and parity portion, where the column weights of the parity portion are uniform, achieved by cyclic shifts in identity submatrices, allowing for simultaneous LLR calculations across check nodes without common variable nodes, enhancing decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a lower triangular structure is used for the LDPC parity check matrix, then the matrix structure is simple and systematic, but decoding errors increase and iterations increase particularly in the last parity column
Solution Approach 1:
The patent transforms the symmetric lower triangular structure into an asymmetric structure by applying cyclic shifts to rows and columns. Specifically, even-weight columns are cyclically shifted down by 2 positions and odd-weight columns are cyclically shifted down by 1 position, creating an asymmetric matrix that eliminates the problematic last column structure while maintaining overall sparsity and systematic properties
Solution Approach 2:
The patent changes the structural parameters of the parity check matrix by introducing cyclic shift operations with specific shift amounts (2 for even columns, 1 for odd columns). This parameter transformation modifies the matrix structure to distribute column weights more uniformly and eliminate the degenerate case of the last column having weight 1, thereby improving decoding performance
2Device complexity
If the last parity column has a column weight of one, then the matrix maintains sparsity, but decoding iterations increase and errors increase
Solution Approach 1:
The patent applies different local transformations to different columns based on their weight properties. Even-weight columns receive a cyclic shift of 2 positions while odd-weight columns receive a cyclic shift of 1 position. This localized differentiation ensures that the last column (which would have weight 1) is transformed to have weight 2, eliminating the problematic local structure while preserving sparsity elsewhere
3Reliability
If variable nodes are connected to multiple check nodes, then the code provides better error correction, but simultaneous LLR calculation becomes impossible in layered decoding
Solution Approach 1:
The patent segments the variable nodes into two distinct groups: even-indexed variable nodes connected to even-weight check nodes, and odd-indexed variable nodes connected to odd-weight check nodes. This segmentation creates two independent subgraphs that can be processed simultaneously in parallel, enabling layered decoding while maintaining the required connectivity for error correction
Data Source
AI summary
An LDPC parity check matrix, includes a systematic portion having a plurality of systematic elements and a parity portion having a plurality of parity elements. The value of each systematic element determines a cyclic shift to be applied to rows of an identity submatrix corresponding to that element. The value of each parity element determines a cyclic shift to be applied to rows of an identity submatrix corresponding to that element. The weights of two or more columns of the parity portion are the same.