Improved Mean-Square Error Estimate for the LMS Transversal Equalizer With Narrowband Interference

When the least-mean-square (LMS) algorithm is used to adapt an adaptive transversal equalizer that is subject to strong narrowband interference, a so-called non-Wiener or nonlinear effect takes place. This results in the mean-square error (MSE) performance of the adaptive equalizer being better than that of the fixed Wiener filter of equivalent structure. Reuter and Zeidler proposed a transfer-function-based approach to provide an estimate of the MSE performance of the equalizer in such an environment. We have recently shown that the mean of the LMS weights in this adaptive equalizer problem shifts away from the Wiener filter solution. As a result, we propose an MSE model for the LMS equalizer that is an improvement over the existing Reuter-Zeidler model. The new model uses the same transfer-function-based approach but incorporates the shift in the mean of the weights. Numerical simulations are provided to illustrate the improvement.