Windowed MAP Decoding With Alternating Warm-Up Recursions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Windowed MAP decoders perform poorly when puncturing is used to raise the code rate, especially in the 3GPP standard, due to issues with state-metric convergence and divergence, leading to subpar error correction performance compared to un-windowed decoders.
Innovation Solution
The solution involves alternating the direction of warm-up recursions between forward and reverse with each successive Turbo decoder iteration and storing snapshots of state metrics to initialize warm-up phases, allowing for efficient state-metric initialization and convergence, even under high puncturing conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If puncturing is used to raise the code rate in windowed MAP decoders, then code rate increases, but error correction performance deteriorates due to state-metric convergence issues
Solution Approach 1:
The patent applies preliminary action by performing a warm-up phase of recursions before the main decoding window to initialize state metrics properly. This preliminary computation ensures that state metrics converge to correct values even under high puncturing conditions, preventing the performance deterioration that would otherwise occur when raising the code rate through puncturing.
Solution Approach 2:
The patent changes the initialization parameters of state metrics by using values computed from the warm-up recursions rather than default or zero initialization. This parameter change allows the decoder to maintain accurate state-metric convergence even when puncturing removes significant redundancy, thereby preserving error correction performance at higher code rates.
2Reliability
If un-windowed MAP decoders are used to maintain error correction performance, then decoding accuracy improves, but memory consumption increases
Solution Approach 1:
The patent applies segmentation by dividing the decoding process into a warm-up phase and a main windowed decoding phase. The warm-up phase computes initial state metrics that are then used to initialize the windowed decoder, enabling the segmented approach to achieve performance close to un-windowed decoders while maintaining the memory efficiency benefits of windowing.
Solution Approach 2:
By performing preliminary computations in the warm-up phase to generate initialized state metrics, the patent enables the windowed decoder to start with accurate values rather than defaults. This preliminary action allows the use of smaller memory structures characteristic of windowed decoders while achieving error correction performance comparable to un-windowed approaches.
3Ease of operation
If warm-up recursions are performed in a fixed direction, then initialization is simplified, but convergence performance deteriorates under high puncturing
Solution Approach 1:
The patent applies dynamics by making the warm-up recursion direction adaptive rather than fixed. The direction of warm-up recursions is adjusted based on the decoding context and puncturing pattern, allowing the system to dynamically optimize convergence performance while maintaining operational simplicity through automated direction selection.
Solution Approach 2:
The patent employs inversion by performing warm-up recursions in alternating directions (forward and backward) depending on the decoding phase. This alternating approach ensures that state metrics are properly initialized regardless of the puncturing pattern, achieving convergence performance comparable to full un-windowed decoders while retaining the memory efficiency of the windowed approach.
Data Source
AI summary
One aspect provides a method. The method comprises receiving a signal comprising a sequence of encoded symbols, each corresponding to one of a plurality of possible states. For each symbol in the sequence, the method further comprises determining a set of state metrics, each representing a probability that the respective symbol corresponds to the plurality of states. The method further comprises decoding the signal by processing runs of recursions using runs of forward recursions, whereby a later state metric in the sequence is updated based on a preceding state metric, and runs of recursions using runs of reverse recursions, whereby a preceding state metric in the sequence is updated based on a later state metric. The method further comprises outputting the decoded signal to a device. The decoding comprises performing a plurality of repeated iterations over the sequence.