LDPC Layered Min-Sum Quantization for Fast Message Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current LDPC decoding methods, particularly in Tanner graphs, face inefficiencies in message passing between check nodes and variable nodes, which affect decoding performance and error correction, especially due to limitations in quantization methods used in layered min-sum decoding.
Innovation Solution
A novel quantization method is introduced where values below a threshold are uniformly quantized, and those above are non-uniformly quantized, with a corresponding inverse-quantization defined, and the layered min-sum decoding algorithm is enhanced with a fast mapping and de-mapping method to improve error correction performance while reducing hardware implementation complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If uniform quantization is used for all message values in layered min-sum decoding, then hardware implementation is simpler, but decoding performance and error correction capability deteriorate
Solution Approach 1:
The patent applies different quantization strategies to different ranges of message values: uniform quantization for small magnitude values and non-uniform quantization for large magnitude values. This local differentiation allows the system to optimize for both hardware simplicity (using uniform quantization where sufficient) and decoding performance (using non-uniform quantization where needed), resolving the contradiction between ease of manufacture and reliability.
2Reliability
If non-uniform quantization is used for all message values, then decoding performance improves, but hardware implementation complexity increases
Solution Approach 1:
Instead of applying non-uniform quantization universally, the patent selectively applies it only to message values above a certain threshold magnitude. This local application reduces hardware complexity compared to full non-uniform quantization while maintaining error correction capability for the critical cases where large message values occur.
Solution Approach 2:
The patent segments the message value range into two regions: a lower range handled by simple uniform quantization and an upper range handled by non-uniform quantization. This segmentation allows the system to achieve good decoding performance without the full hardware complexity of universal non-uniform quantization.
3Measurement precision
If more memory is allocated for message storage in Tanner graphs, then decoding accuracy improves, but memory requirements and implementation cost increase
Solution Approach 1:
The patent changes the quantization parameters (uniform vs. non-uniform) based on the magnitude of message values, allowing accurate representation of both small and large values within a limited memory range. This parameter adaptation enables high decoding accuracy without proportionally increasing memory requirements.
Solution Approach 2:
The quantization step size dynamically adapts to the message value magnitude: small step sizes for small values and large step sizes for large values. This dynamic quantization allows the system to maintain precision where needed while reducing memory usage for less critical values.
Data Source
AI summary
A method is disclosed for performing LDPC decoding, specifically layered min-sum decoding using a Tanner graph including check nodes (CN) and variable nodes (VN). Messages passed between nodes are quantized in a non-uniform manner. Values below a threshold are uniformly quantized whereas values above the threshold are non-uniformly quantized. A corresponding inverse-quantization is also defined.