LDPC Decoding with Selective Variable Node Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sum-product decoding in LDPC systems requires a large number of iterative calculations, leading to increased computational complexity and slow convergence, particularly due to the need for updating prior value ratios for all variable nodes after check node processing is completed.
Innovation Solution
A decoding apparatus and method that alternately execute check node processing and variable node processing, identifying a next-target element within the same column for variable node processing after each row of the parity check matrix is updated, reducing the number of elements subjected to variable node processing and processing in units of fundamental matrices, including zero matrices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If shuffled decoding is used to update prior value ratios for all variable nodes after each check node processing, then convergence speed is improved, but calculation amount increases proportionally to the number of non-zero elements in the column direction
Solution Approach 1:
The patent extracts and processes only the necessary subset of variable nodes (those corresponding to non-zero elements in the current row) rather than updating all variable nodes. This selective extraction reduces the calculation amount while maintaining convergence speed improvements through the shuffled decoding approach.
Solution Approach 2:
Instead of performing complete updates for all variable nodes (excessive action), the patent performs partial updates only for the specific variable nodes that are relevant to the current check node processing. This partial action is sufficient to achieve convergence while avoiding unnecessary calculations.
2Device complexity
If conventional sum-product decoding is used to execute variable node processing after all check node processing is completed, then calculation amount is reduced, but convergence speed decreases
Solution Approach 1:
The patent performs preliminary updates of prior value ratios for specific variable nodes immediately after each check node processing step, rather than waiting for all check node processing to complete. This preliminary action accelerates convergence by making useful updates earlier in the decoding process without requiring complete processing of all check nodes first.
Solution Approach 2:
The patent introduces dynamic, incremental updates to prior value ratios during the decoding process, allowing the system to adapt and converge faster by continuously refining estimates as check node processing progresses, rather than performing static batch updates after all processing is complete.
Data Source
AI summary
A check node processing unit updates an extrinsic value ratio based on a prior value ratio for each row of a parity check matrix with respect to input data. An identifying unit identifies, based on an element of the parity check matrix that can be identified by a row and column associated with the updated extrinsic value ratio, a next-target element in the same column and in a different row. The identifying unit identifies an element to be updated in the next step by the check node processing unit, from among multiple elements included in the same column. A variable node processing unit updates, based on the extrinsic value ratio, a prior value ratio associated with the identified next-target element after the check node processing unit completes the updating of each row. The check node processing unit and the variable node processing unit alternately and iteratively execute their operations.