Non-Uniform LLR Quantization for Lower Memory Soft Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wireless communication systems face high memory and data rate costs due to large interleaving sizes required for robust channel coding against fading effects, which can lead to decreased performance when reducing the number of bits used to represent log likelihood ratios (LLRs).
Innovation Solution
Implementing non-uniform quantization of LLRs, where the width of each quantization bin is tailored based on the fidelity of the incoming signal, allowing for adaptive compression that reduces memory and transmission costs while maintaining performance by compressing LLRs more efficiently in regions of higher uncertainty.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If the number of bits used to represent LLRs is decreased to reduce memory cost, then memory requirements are reduced, but the FEC decoder works less efficiently and reliability of reception decreases
Solution Approach 1:
The patent applies local quality by using different quantization step sizes for different ranges of LLR values. Specifically, smaller step sizes (higher precision) are used for LLR values near zero where uncertainty is highest, while larger step sizes (lower precision) are used for LLR values with larger magnitudes where the decision is more certain. This non-uniform quantization approach optimizes memory usage while maintaining reception reliability by concentrating precision where it is most needed.
2Reliability
If the interleaver size is increased to improve robustness against fading effects, then system robustness improves, but the number of softbits to be stored increases leading to larger memory cost
Solution Approach 1:
The patent changes the parameter of quantization precision from uniform to non-uniform based on the magnitude of LLR values. By adapting the quantization step size according to the LLR magnitude, the system can maintain robust performance with smaller interleaver sizes, thereby reducing the memory cost while preserving robustness against fading effects.
3Quantity of substance
If uniform quantization is applied to reduce LLR precision, then memory cost is reduced, but performance loss increases particularly in regions of high uncertainty
Solution Approach 1:
The patent implements local quality by applying different quantization precision to different regions of the LLR distribution. Regions with high uncertainty (LLR values near zero) receive finer quantization with smaller step sizes, while regions with low uncertainty (large |LLR| values) receive coarser quantization with larger step sizes. This resolves the contradiction by maintaining measurement precision where it matters most while reducing overall memory cost.
Data Source
AI summary
A method of processing a signal by non-uniform quantization of log likelihood ratios is disclosed. A method comprising the steps of: receiving a plurality of bits; calculating a log likelihood ratio, known as a LLR, for each bit; providing a LLR value for each bit based on the calculated LLR; quantizing the LLR values into a plurality of quantization bins, each quantization bin having: a width representative of one or more LLR values; and an index value having a bit length; and associating each bit with the index value that corresponds to its LLR value, wherein the width of each quantization bin is non-uniform. This compresses the LLR values in a more efficient manner, requiring lower memory usage and/or lower bandwidth. A chip for a receiver and a communication system comprising one or more receivers are also disclosed.