Non-Binary LDPC Parity Node Control With Limited Candidate Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of decoding non-binary LDPC codes is high due to the need for numerous multiplications in elementary parity nodes, which can be reduced but remains significant even with existing algorithms like EMS that filter and sort incoming messages.
Innovation Solution
A method for controlling elementary parity nodes that receives sorted input lists and generates an output list with reduced calculations by selecting and processing a limited number of candidate elements, using a predetermined criterion to determine the next elements for the output list, thereby reducing the number of operations and comparisons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If elementary parity nodes perform all possible computing operations between incoming messages to ensure accurate decoding, then decoding reliability is improved, but computational complexity increases significantly
Solution Approach 1:
The patent applies partial action by selecting only a limited number of candidate elements (e.g., top 1, 2, or 3) from the sorted input lists rather than performing operations on all elements. This partial processing maintains adequate decoding reliability while dramatically reducing the number of computing operations required in elementary parity nodes
Solution Approach 2:
The patent applies preliminary action by sorting the incoming message lists in descending order before processing. This pre-sorting arrangement ensures that the most significant candidate elements are positioned at the beginning of the lists, allowing the algorithm to select top candidates without exhaustive search, thus reducing computational complexity while maintaining reliability
2Productivity
If the number of computing operations in elementary parity nodes is reduced to improve processing speed, then productivity is improved, but decoding precision may deteriorate
Solution Approach 1:
The patent changes the parameter of candidate selection by introducing a configurable number of candidates to process (e.g., selecting top 1, 2, or 3 elements). This parameter adjustment allows optimization between speed and precision: higher candidate counts improve precision while lower counts increase speed, enabling flexible trade-off management
Solution Approach 2:
The patent applies local quality by concentrating computational resources on the most significant candidate elements (those with highest values in sorted lists) rather than distributing resources uniformly across all elements. This localized processing ensures that critical computations are performed with high precision while less critical operations are reduced or omitted
3Measurement precision
If all candidate elements are processed to maintain output accuracy, then measurement precision is improved, but the number of operations increases
Solution Approach 1:
The patent processes only a partial set of candidate elements (the top N elements from sorted input lists) rather than all possible candidates. This partial processing maintains sufficient output accuracy for practical applications while significantly reducing the number of operations and processing time required
Solution Approach 2:
The patent performs preliminary sorting of input lists in descending order before selecting candidates for processing. This pre-arrangement ensures that the most accurate candidates are identified and selected first, allowing the algorithm to achieve adequate output precision with fewer operations by stopping once the required number of candidates is obtained
Data Source
AI summary
A method for controlling an elementary parity node of a decoder for decoding non-binary LDPC codes or a code decoder using at least one non-binary parity constraint, and to the corresponding elementary parity node. The elementary parity node receives first and second input lists (U1, U2) having nm elements sorted in ascending or descending order, nm being greater than 1, and gives an output list (Uout) of nm′ elements sorted in said ascending or descending order, nm′ being greater than 1, each element of the output list (Uout) being the result of a computing operation φ between an element of the first input list (U1) and an element of the second input list (U2). A limited number of candidates is selected for each element of the output list to be generated so as to reduce the number of operations to be carried out in the elementary parity node.