Trellis Equalization Using Nonlinear Models for Satellite ISI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing signal equalization techniques fail to effectively mitigate non-linear inter-symbol interference in satellite communication channels, particularly in downlink channels using bandwidth-efficient modulation, due to the complexity of accurately modeling non-linear distortions introduced by high power amplifiers and additive white Gaussian noise.
Innovation Solution
A trellis-based iterative equalizer is developed that incorporates Volterra series decomposition and the BCJR algorithm, using a soft-in soft-out system to compute branch metrics and generate alpha, beta, and sigma state metrics, allowing for the estimation of sequences and correction of non-linear distortions, while employing different non-linear models such as memory polynomial, Wiener, Hammerstein, and lookup table models for improved performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single non-linear model is used for all trellis computations, then device complexity is reduced, but manufacturing precision deteriorates due to inability to accurately model non-linear distortions
Solution Approach 1:
The patent divides the single non-linear model into multiple distinct non-linear models (first, second, and third non-linear models) that are applied to different computational stages of the trellis algorithm. Specifically, the first non-linear model is used for branch metric computation, while the second and third non-linear models are used for alpha and beta state metric computations respectively. This segmentation allows each model to be optimized for its specific computational role, improving overall modeling accuracy without requiring a single overly complex model.
2Manufacturing precision
If different non-linear models are used for branch metrics, alpha state metrics, and beta state metrics, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The patent applies the principle of local quality by assigning different non-linear models to different local computational stages of the trellis algorithm. The first non-linear model is specifically tailored for branch metric computation, while the second and third non-linear models are optimized for alpha and beta state metric computations respectively. This localized optimization allows each model to be specifically designed for its computational context, improving sequence estimation accuracy while managing complexity through targeted rather than universal model application.
3Reliability
If iterative detection and decoding is performed, then reliability is improved, but loss of time increases due to multiple iterations
Solution Approach 1:
The patent implements preliminary action by pre-computing and storing non-linear model parameters and characteristics before the actual iterative detection and decoding process. The distinct non-linear models are prepared in advance with their respective parameters optimized for specific computational stages. This preliminary preparation reduces the computational burden during iterative operations, allowing faster convergence and reducing the time loss associated with multiple iterations while maintaining improved reliability.
Data Source
Figure 1~2
Figure 3A~3B
Figure 3C~3D
AI summary
A method includes: generating a trellis; generating one or more predicted symbols using a first non-linear model; computing and saving two or more branch metrics using a priori log-likelihood ratio (LLR) information, a channel observation, and the one or more predicted symbols; if alpha forward recursion has not yet completed, generating alpha forward recursion state metrics using a second non¬ linear model; if beta backward recursion has not yet completed, generating beta backward recursion state metrics using a third non-linear model; if sigma forward recursion has not yet completed, generating sigma forward recursion state metrics using the branch metrics, the alpha state metrics, and the beta backward recursion state metrics; generating extrinsic information comprising a difference of a posteriori LLR information and the a priori LLR information; computing and feeding back the a priori LLR information; and calculating the a posteriori LLR information.