QAM LLR Compression for Lower-Memory Soft Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High symbol rates in next-generation networks lead to inter-symbol interference (ISI), causing errors in signal decoding due to the inability of some network components to support such rates without introducing significant distortion, particularly in wireless and optical channels.
Innovation Solution
The method involves compressing log likelihood ratios (LLRs) for higher-order square-QAM constellations by exploiting the piecewise linear relationship between the magnitude components of LLRs for bits corresponding to the same dimension of a symbol, reducing the number of bits required for storage and processing, thereby reducing memory requirements and computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If higher symbol rates are implemented to provide data rates in excess of 100 Gbps, then data transmission speed is improved, but inter-symbol interference increases causing decoding errors
Solution Approach 1:
The patent changes the representation parameters of soft output information by compressing LLRs into a compact format that captures essential decision information. This parameter transformation allows the system to maintain decoding accuracy at high symbol rates by efficiently representing signal decisions despite increased interference conditions.
2Reliability
If full precision LLRs are stored and processed for all bits in higher-order square-QAM constellations, then decoding accuracy is maintained, but memory requirements and computational complexity increase significantly
Solution Approach 1:
The patent extracts only the essential decision information from full precision LLRs by identifying and retaining the most significant bits that carry decision reliability information. This extraction process removes redundant precision while preserving the critical information needed for accurate decoding, thereby reducing memory and computational requirements.
Solution Approach 2:
The patent segments the LLR representation into distinct components: decision bits that indicate the most likely symbol value and confidence bits that indicate reliability. This segmentation allows selective processing and storage of different information types, reducing overall complexity while maintaining decoding performance.
3Device complexity
If compressed LLR representation is used to reduce memory requirements, then device complexity is reduced, but information precision may be lost
Solution Approach 1:
The patent transforms the LLR parameter representation from full precision floating-point or high-precision fixed-point format into a compact integer format that preserves decision information. This parameter change maintains the essential soft information needed for iterative decoding while using significantly fewer bits per LLR value.
Data Source
AI summary
It is possible to compress log likelihood ratios (LLRs) by exploiting the mapping symmetry between bits in the same symbol. For example, two LLRs corresponding to the same dimension of a square Quadrature Amplitude Modulation (QAM) symbol can be compressed into a single compressed LLR that excludes the magnitude bits of one of the LLRs because the magnitude component of LLRs for bits corresponding to the same dimension of a square QAM symbol exhibit a piecewise linear relationship with one another. Similar techniques can be used to exploit piecewise linear relationships between a subset of constellation points in a non-square QAM constellation.