Symbol Judgement Using Adjacent-Sequence LLR Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing symbol judgement methods face challenges in supporting soft decision error correction codes in coherent optical transmission systems due to noise enhancement and increased operation scale caused by band narrowing and device nonlinearity, particularly in methods like MAP estimation that require sequence estimation for all possible candidate sequences.
Innovation Solution
A symbol judgement apparatus and method that generates bit log-likelihood ratios for each bit, selects adjacent symbols, generates candidate sequences, reflects channel response, and performs posterior probability maximum estimation to calculate log-likelihood ratios, allowing for soft decision while limiting the number of states and reducing noise enhancement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If MAP estimation by sequence estimation is used to compensate for band narrowing and device nonlinearity, then noise enhancement is suppressed, but the operation amount increases exponentially with multivalued degree and candidate symbol length
Solution Approach 1:
The patent segments the sequence estimation process by limiting candidate symbol sequences to only those adjacent to the reception symbol. Instead of evaluating all possible candidate sequences, the method divides the search space to include only neighboring symbols in the constellation diagram, thereby reducing the exponential operation amount while maintaining noise suppression capability
Solution Approach 2:
The patent applies partial action by performing sequence estimation not for all possible candidate sequences but only for a limited set of adjacent candidate sequences. This partial estimation approach provides sufficient compensation for band narrowing and device nonlinearity effects without requiring exhaustive search through all possible sequences
2Device complexity
If sequence estimation is limited to candidate symbol sequences to reduce operation amount, then operation scale is reduced, but soft decision error correction code support becomes difficult
Solution Approach 1:
The patent changes the parameter representation from hard decision symbols to soft decision log-likelihood ratios. By calculating and outputting log-likelihood ratios for the limited adjacent candidate sequences rather than hard decisions, the method enables soft decision error correction code support while maintaining the reduced operation amount benefit of limited sequence estimation
3Device complexity
If soft decision is performed without sequence estimation to reduce operation amount, then operation scale is reduced, but noise enhancement occurs when band narrowing and device nonlinearity influence is large
Solution Approach 1:
The patent segments the estimation process to include sequence estimation specifically for adjacent candidate sequences when band narrowing and device nonlinearity effects are significant. This selective application of sequence estimation to only neighboring symbols provides noise suppression where needed while avoiding exhaustive estimation, thereby improving signal quality without excessive operation increase
Solution Approach 2:
The patent applies partial sequence estimation action by performing estimation only for adjacent candidate sequences rather than all possible sequences. This partial estimation provides sufficient noise suppression for band narrowing and device nonlinearity compensation while keeping the operation amount manageable
Data Source
AI summary
The bit log-likelihood ratio generation unit generates a bit log-likelihood ratio of the reception symbol for each bit. The symbol selection unit selects a plurality of adjacent symbols of the reception symbol. A candidate sequence generation unit generates a plurality of candidate sequences by combining adjacent symbols of time-series reception symbols. The posterior probability maximum estimation unit calculates the log-likelihood ratio of the adjacent symbol to maximize the posterior probability using the output obtained by whitening time-series data of branched reception symbols and the candidate sequences reflecting the channel response. A likelihood ratio selection unit selects the bit log likelihood to be corrected based on the bit converted from the adjacent symbol. An addition unit performs weighted addition of the selected bit log-likelihood ratio and a correction value based on a difference between the bit log-likelihood ratio and the log-likelihood ratio.