LDPC Offset BP Decoding with Corrected Row LLR Approximation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The Offset BP-based decoding method for LDPC codes has low precision and inadequate decoding performance due to its approximate arithmetic operations, which are influenced by variables, leading to a high number of arithmetic operations and reduced circuit implementation efficiency.

Innovation Solution

A decoding method that approximates the row arithmetic operation of the Sum-Product decoding method using an Offset BP-based approach by setting the row LLR as the minimum of column LLRs minus an offset, thereby improving precision and reducing arithmetic operations, while maintaining a similar circuit scale to the Offset BP-based method.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If Sum-Product decoding method is used, then decoding performance is high, but the amount of arithmetic operations is large and circuit implementation is difficult

Engineering Contradiction:
Improvedecoding performanceVSAvoidcircuit implementation complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent changes the parameter representation from natural logarithms (Sum-Product) to base-2 logarithms. This transformation allows the use of binary arithmetic operations instead of complex floating-point operations, significantly reducing circuit complexity while maintaining decoding performance. The row LLR and column LLR calculations are reformulated using bitwise operations and lookup tables based on the base-2 logarithmic transformation.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If Offset BP-based decoding method is used, then the amount of arithmetic operations is reduced, but decoding performance deteriorates due to low precision approximate arithmetic operations

Engineering Contradiction:
Improvearithmetic operation efficiencyVSAvoiddecoding performance
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent introduces a correction parameter γ (gamma) that is added to the minimum absolute value of column LLRs during row processing. This correction term compensates for the approximation error inherent in Offset BP methods. By dynamically adjusting γ based on the minimum LLR magnitude, the method maintains high decoding performance while preserving the computational efficiency of approximate arithmetic operations.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent implements a feedback mechanism where the correction parameter γ is determined based on the minimum absolute value of column LLRs from the previous iteration. This feedback loop allows the system to adaptively adjust the correction amount, ensuring that approximation errors are compensated without requiring full Sum-Product decoding complexity. The correction is applied iteratively to improve decoding accuracy.

Inventive Principle:
Principle #23Feedback

3Reliability

If variable offset parameter is used, then decoding performance is improved, but the amount of arithmetic operations increases

Engineering Contradiction:
Improvedecoding performanceVSAvoidarithmetic operation complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent pre-calculates and stores correction parameters γ in lookup tables before the actual decoding process. These pre-computed values are based on the minimum LLR magnitudes and are prepared in advance to avoid real-time complex calculations. During decoding, the system simply retrieves the appropriate γ value from the lookup table and applies it, maintaining computational efficiency while achieving high decoding performance through the variable correction parameter.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2264906B1Modified Offset BP-based LDPC decoding
Publication Date: 2015.09.09 MITSUBISHI ELECTRIC CORP
  • EP2264906B1 patent drawingFigure 1
  • EP2264906B1 patent drawingFigure 2
  • EP2264906B1 patent drawingFigure 3

AI summary

In row calculation, a value which is obtained by subtracting an offset according to a minimum of the absolute values of column LLRs from the minimum of the absolute values of the column LLRs is set as a row LLR corresponding to a column of the column LLRs. An object of the present invention is to provide a decoding method and a decoding device that can approximate a row arithmetic operation based on a Sum-Product decoding method with a high degree of precision by using an Offset BP-based decoding method, thereby improving the performance of decoding of an error correcting code.