LDPC Parity Check Message Computation With Adaptive Min-Sum Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing LDPC decoding methods face challenges in achieving an ideal tradeoff between decoding performance, data rate, implementation complexity, and energy consumption, particularly in high-speed communication systems like spatial communications.
Innovation Solution
The proposed method involves a novel approach for calculating parity check messages in LDPC decoding, using a method called 'Adapted Offset Min-Sum' (AOMS), which determines correction values based on comparisons with multiple thresholds, and includes on-the-fly scaling of data to manage saturation and improve error rate performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the Min-Sum algorithm is used to reduce computational complexity, then implementation complexity is reduced, but decoding performance deteriorates due to over-evaluation of parity check messages
Solution Approach 1:
The patent modifies the Min-Sum algorithm by introducing a correction value that compensates for the over-evaluation of parity check messages. This correction value is determined based on the difference between the two smallest absolute values of variable messages, effectively adjusting the parameter (parity check message value) to restore decoding performance while maintaining the low computational complexity of the Min-Sum approach
Solution Approach 2:
The patent implements a feedback mechanism where the correction value is calculated based on the local characteristics of the variable messages (the difference between the two smallest absolute values). This feedback loop allows the algorithm to dynamically adjust the parity check messages to compensate for the inherent over-evaluation problem in the Min-Sum algorithm
2Reliability
If BP-SPA algorithm is used to achieve good decoding performance, then reliability is improved, but computational complexity increases due to use of hyperbolic tangent, logarithm and exponential functions
Solution Approach 1:
The patent replaces the complex mathematical functions (hyperbolic tangent, logarithm, exponential) used in BP-SPA with simpler arithmetic operations (minimum, addition, subtraction). This substitution uses computationally 'cheap' operations that can be executed efficiently, achieving comparable decoding performance without the high computational burden of transcendental functions
3Device complexity
If data quantization is applied to reduce hardware complexity, then device complexity is reduced, but error rate performance deteriorates due to saturation of data
Solution Approach 1:
The patent applies preliminary scaling to the variable messages before they are used in the parity check message calculation. This preliminary action prevents saturation from occurring during the decoding process, allowing the use of fixed-point quantization without losing information, thus maintaining error rate performance while enabling hardware complexity reduction
Data Source
AI summary
The disclosure relates to an LDPC decoding method which involves performing iterations until a stop criterion is satisfied. Each iteration involves computing variable messages (αn,m), computing parity check messages (βm,n), and computing a posteriori estimation variables. Computing a parity check message (βm,n) for a parity check node (CNm) involves determining the two smallest values (Mini1, Min2) among the absolute values of the variable messages associated with the parity check node (CNm), comparing a difference between said values with a threshold, determining a correction value according to the result of the comparison, and computing the parity check message according to the correction value.