Adaptive Equalizer Tap Weight Initialization via Cross-Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Adaptive decision feedback equalizers in digital communication receivers face challenges in converging to desired tap weights under difficult channel conditions, especially due to the initialization of tap weights to zero without knowledge of the channel impulse response.
Innovation Solution
A method that involves cross-correlating a known training sequence with received training data to produce a cross-correlation vector, calculating a threshold value, and iteratively selecting and subtracting correction vectors to improve the accuracy of channel impulse response estimation, reducing noise components and refining the channel estimate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If tap weights are initialized to zero without channel knowledge, then the equalizer is simple to implement, but convergence fails under difficult channel conditions
Solution Approach 1:
The patent applies preliminary action by estimating the channel impulse response before equalization begins. The receiver cross-correlates the received training signal with the known transmitted training signal to obtain an initial channel estimate, then uses this estimate to initialize the tap weights of the feedforward and feedback filters. This preliminary channel knowledge enables reliable convergence under difficult channel conditions while maintaining implementation simplicity through automated estimation procedures.
2Measurement precision
If channel impulse response is estimated using cross-correlation, then initial tap weights can be calculated accurately, but noise components reduce estimation accuracy
Solution Approach 1:
The patent applies the extraction principle by separating and removing noise components from the channel estimate. After obtaining the initial channel estimate through cross-correlation, the method identifies and extracts significant noise components by comparing the estimated channel impulse response against the known training signal characteristics. These extracted noise components are then subtracted from the channel estimate, leaving a cleaner, more accurate channel representation for tap weight initialization.
Solution Approach 2:
The patent employs feedback by using the known transmitted training signal as a reference to evaluate and refine the channel estimate. The receiver continuously compares the estimated channel impulse response with expectations based on the known training sequence, identifying discrepancies that indicate noise contamination. This feedback mechanism guides the iterative noise removal process, progressively improving estimation accuracy by eliminating identified noise components.
3Ease of operation
If adaptive algorithms like LMS are used, then implementation is easy and performance is reasonable, but convergence fails under difficult channel conditions
Solution Approach 1:
The patent combines preliminary action with adaptive algorithms by initializing the LMS algorithm with tap weights derived from channel estimation rather than random or zero values. This preliminary initialization based on cross-correlation and noise removal provides the adaptive algorithm with a head start, enabling it to converge reliably under difficult channel conditions while maintaining the simplicity and ease of operation of the LMS implementation.
Data Source
AI summary
A channel impulse response is determined from a cross-correlation of a received training sequence and a stored version of the transmitted training sequence. The channel impulse response is iteratively cleansed of noise caused by the finiteness of the cross-correlation. Initial values of the tap weights for the taps of an equalizer such as a decision feedback equalizer may be determined based on the cleansed channel impulse response.