Adaptive Equalizer Coefficient Updates Without Training Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data-aided LMS methods require periodic training patterns, which decrease data communication speed and can lead to wrong filter coefficient convergence in adaptive equalization for PCS signals.
Innovation Solution
A filter coefficient update method that combines DDLMS with statistical information updates, using gradients to minimize differences between processed received samples and set values, avoiding the need for periodic training patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data-aided LMS method is used for adaptive equalization, then filter coefficient convergence can be achieved, but periodic training patterns are required which decrease data communication speed
Solution Approach 1:
The patent combines DDLMS (Decision-Directed Least Mean Squares) method with statistical information update method into a hybrid adaptive equalization algorithm. The DDLMS provides reliable filter coefficient convergence through temporary determination results, while the statistical information update method eliminates the need for periodic training patterns by using gradient descent on statistical moments, thereby maintaining high data communication speed without sacrificing convergence reliability.
2Reliability
If data-aided LMS method is used for adaptive equalization, then filter coefficient convergence can be achieved, but wrong convergence may occur in PCS signals
Solution Approach 1:
The patent implements a feedback mechanism where the adaptive equalization algorithm continuously monitors the statistical information of the equalized signal and adjusts the filter coefficients accordingly. By using the gradient of statistical moments (such as fourth-order moments for QAM signals) as feedback, the system can detect and correct wrong convergence situations, ensuring that the filter coefficients converge to the correct optimal values rather than local minima, thus improving convergence accuracy for PCS signals.
Data Source
AI summary
In order to suppress wrong convergence of a filter coefficient without reducing data communication speed, this filter coefficient update amount output device is provided with: a first output unit which outputs a first coefficient update amount derived from a difference between a temporary determination result regarding a processed reception sample value that is the reception sample value on which filter processing by a digital filter has been performed and the processed reception sample value; a second output unit which outputs a second coefficient update amount derived from a gradient with respect to the filter coefficient such that the magnitude of a difference between statistical information on the processed reception sample value during some duration of time, and a set value regarding the statistical information is minimized; and a third output unit which outputs the coefficient update amount derived from the first coefficient update amount and the second coefficient update amount.