ULP LDPC Decoder Thresholding Using Syndrome Weight Lattices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Ultra-Low Power (ULP) LDPC decoders face sub-optimal threshold values due to circles in the generator graph and neglecting temporary Syndrome Weight (SW) during pre-calculated threshold generation, leading to inefficient error correction in data storage devices.
Innovation Solution
The use of soft quantization and lattice interpolation combined with reinforcement learning to dynamically generate threshold values for bit flipping decisions in ULP LDPC decoders, based on Syndrome Weight and distances between lattice points, improving convergence and decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If pre-calculated threshold values are used in ULP LDPC decoders, then decoding speed is improved, but threshold optimality deteriorates due to generator graph circles and neglecting temporary Syndrome Weight
Solution Approach 1:
The patent transforms the static pre-calculated threshold approach into a dynamic adaptive threshold system. The threshold values are no longer fixed but are adjusted based on the current Syndrome Weight (SW) and clock cycle, allowing the decoder to adapt to changing decoding conditions and achieve optimal thresholds for each specific situation while maintaining fast decoding speed.
Solution Approach 2:
The patent changes the parameters used for threshold determination from fixed pre-calculated values to dynamic values based on Syndrome Weight and clock cycle. By introducing SW as a variable parameter and using lattice interpolation to generate thresholds based on current SW and distance from lattice points, the system achieves both speed and accuracy.
2Measurement precision
If reinforcement learning with lattice interpolation is used to generate threshold values, then threshold optimality is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-selecting lattice points and pre-calculating thresholds at these lattice points during the training phase. During actual decoding, the system only needs to perform simple lattice interpolation and distance calculations rather than full reinforcement learning computations, significantly reducing real-time computational complexity while maintaining threshold optimality.
Solution Approach 2:
The patent uses partial action by performing reinforcement learning only at discrete lattice points rather than continuously for all possible Syndrome Weight values. The lattice interpolation method then extends these partial results to cover the entire range of SW values, reducing the computational burden while achieving near-optimal thresholds throughout the full operating range.
3Productivity
If dynamic threshold generation based on Syndrome Weight is implemented, then error correction efficiency is improved, but memory requirements increase for storing threshold tables
Solution Approach 1:
The patent segments the continuous range of Syndrome Weight values into discrete lattice points. By storing threshold values only at these discrete lattice points rather than for all possible SW values, the memory requirements are significantly reduced while still enabling dynamic threshold generation for any SW value through interpolation. This segmentation approach maintains error correction efficiency while minimizing memory usage.
Data Source
AI summary
Advanced ultra-low power error correcting codes are generated using soft quantization and lattice interpolation based on clock and Syndrome Weight. Reinforcement learning may be used to generate threshold values for flipping bits for low density parity check Ultra-Low Power error correction codes. The threshold values can be generated offline and downloaded to a storage device or generated while the storage device is in use.