Adaptive Turbo Decoding for Non-Convergent Bit Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Turbo decoders require high computational power and complexity to achieve optimal performance, especially with increasing iterations and interleaver size, which translates to higher MIPS and power consumption without significant improvements in bit error rate.
Innovation Solution
A channel adaptive iterative turbo decoder system that reduces computational complexity by continuing MAP decoding iterations only for non-convergent data points and using a smaller window for calculations when a predefined number of non-convergent points is reached, employing Log Max*() and MaxLogMax() methods for calculating forward, reverse, and LLR metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the number of iterations and interleaver size are increased to improve turbo code performance, then the bit error rate performance is improved, but the computational power and power consumption increase significantly
Solution Approach 1:
The patent segments the decoding process by identifying and separating converged bits from non-converged bits. Only non-converged bits continue to require computational processing in subsequent iterations, while converged bits are excluded from further calculations. This segmentation dramatically reduces the computational load and power consumption while maintaining the bit error rate performance benefits of multiple iterations.
Solution Approach 2:
The patent applies partial action by performing MAP decoding operations only on the subset of non-converged bits rather than processing all bits in each iteration. The stopping criterion determines when to halt processing for specific bits, allowing the system to achieve sufficient performance with less computational effort than full iterative processing would require.
2Reliability
If the number of iterations and interleaver size are increased to improve turbo code performance, then the bit error rate performance is improved, but the MIPS (million instructions per second) requirement increases
Solution Approach 1:
The patent segments the set of data bits into converged and non-converged subsets. The stopping criterion identifies converged bits, and subsequent MAP decoding iterations process only non-converged bits. This segmentation reduces the number of instructions that must be executed in each iteration, thereby reducing the MIPS requirement while preserving the performance improvements gained from multiple iterations on the critical non-converged bits.
Solution Approach 2:
The patent extracts and removes converged bits from the processing pipeline once they meet the stopping criterion. By taking out these already-sufficiently-decoded bits from further processing, the system eliminates unnecessary computational instructions, reducing the overall MIPS requirement while maintaining decoding performance.
3Reliability
If conventional turbo decoding is performed for all data bits in each iteration, then complete decoding is achieved, but computational complexity and processing time increase unnecessarily for converged bits
Solution Approach 1:
The patent performs preliminary identification of converged bits using a stopping criterion after a predetermined number of iterations or when convergence conditions are met. Once bits are identified as converged, the action of excluding them from further processing is taken in advance, preventing unnecessary computational steps and reducing overall processing time while ensuring decoding completeness for all bits.
Solution Approach 2:
The patent applies partial action by performing decoding operations only on non-converged bits in subsequent iterations. Instead of excessively processing all bits uniformly, the system applies decoding effort selectively only where needed, reducing processing time while maintaining decoding completeness through the stopping criterion that ensures all bits eventually meet convergence requirements.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A channel adaptive iterative turbo decoder for computing with MAP decoders a set of branch metrics for a window of received data, computing the forward and reverse recursive path state metrics and computing from the forward and reverse recursive path state metrics the log likelihood ratio for 1 and 0 and interleaving the decision bits; and identifying those MAP decoder decision bits which are non-convergent, computing a set of branch metrics for the received data, computing from the forward and reverse recursive path state metrics the log likelihood ratio (LLR) for 1 and 0 for each non-converged decision bit and interleaving the non-convergent decision bits.