Turbo Decoder State Metric Storage for Lower Memory and Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The implementation of MAP-based algorithms in turbo decoders faces challenges with high memory requirements and latency due to the need to store and process extensive state metric information, particularly in integrated circuit implementations for wireless communication systems.
Innovation Solution
The method of embedded state metric storage, where selected state metric information is embedded into the state metrics, reducing the memory requirement by dividing or subtracting the selected metric to create a constant, thus omitting the need to store and calculate the embedded metric, which reduces memory and computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional MAP decoding is implemented with full state metric storage, then decoding accuracy is maintained, but memory requirement increases significantly
Solution Approach 1:
The patent extracts and removes the least significant bits (LSBs) of state metric values from storage. By identifying and discarding the LSB portion of each state metric, the system reduces memory requirements while preserving the most significant bits that contribute most to decoding accuracy. This extraction principle directly addresses the contradiction by removing unnecessary storage burden while maintaining essential decoding functionality.
Solution Approach 2:
The patent changes the precision parameter of state metric storage by truncating bit depth. Instead of storing full-precision state metrics, the system stores truncated versions with reduced bit depth (e.g., removing LSBs). This parameter change allows the system to maintain acceptable decoding accuracy while significantly reducing memory requirements, directly resolving the contradiction between reliability and quantity of storage.
2Reliability
If full precision state metrics are stored and processed, then decoding performance is optimized, but computational complexity increases
Solution Approach 1:
The patent extracts and removes the least significant bits from state metric values before processing. By eliminating LSBs that contribute minimally to decoding performance, the system reduces the computational burden of processing full-precision values while maintaining acceptable decoding performance. This extraction reduces arithmetic operations complexity without significantly degrading reliability.
Solution Approach 2:
The patent applies partial precision processing by retaining only the most significant bits of state metrics rather than processing full precision values. This partial action approach processes a subset of the available precision information, reducing computational complexity while maintaining sufficient decoding performance for practical applications.
3Quantity of substance
If state metric precision is reduced to lower memory usage, then memory requirement decreases, but decoding accuracy degrades
Solution Approach 1:
The patent changes the precision parameter of state metric storage by selectively truncating bit depth. Instead of uniformly reducing precision, the system specifically removes least significant bits while preserving most significant bits. This targeted parameter change achieves memory reduction while minimizing degradation of decoding accuracy, as the MSBs contain the critical information for accurate decoding.
Solution Approach 2:
The patent applies different quality levels to different parts of the state metric data structure. Rather than uniformly reducing precision across all bits, the system preserves high precision for most significant bits and reduces precision for least significant bits. This local quality differentiation ensures that critical decoding information maintains high accuracy while non-critical portions use reduced precision, optimizing the balance between memory usage and decoding performance.
Data Source
AI summary
A method, an embedded state metric storage, is used for MAP (Maximum A Posterior)-based decoder of turbo codes to reduce the memory requirement of state metric storage. For MAP decoder, this method comprises selecting any state metric from the updated state metrics for each recursion direction, forward and reverse, and dividing the state metrics by the selected state metric; the selected state metric value becomes a constant, namely, one. The constant one state metric is embedded into the resulted state metrics. For log-MAP decoder, this method comprises selecting any state metric from the updated state metrics in each direction, forward and reverse, and subtracting the state metrics from the selected state metric; the selected state metric value becomes a constant, zero. The constant zero state metric is embedded into the resulted state metrics. One advantage of the embedded state metric storage during state metric updating and likelihood ratio calculation is to embed the information of the selected state metric into the resulted state metrics. Thus, the selected state metric is not required to be kept in the state metric memory, and calculation of a constant state metric in the resulted state metric can be omitted. Therefore, the latency and the area of implementation in ASIC will be reduced with this method of embedded state metric storage.