Decision feedback equalization of data with spectral nulls

An adaptive decision-feedback equalizer (DFE) update algorithm called the row-action projection (RAP) algorithm is presented. It does not suffer from the numerical stability and noise amplification problems of conventional adaptive equalization algorithms, and it offers linear computational complexity and fast tracking capability. Simulation results are presented that compare the algorithm to the conventional least-mean-squares (LMS) and recursive least squares (RLS) algorithms for decision-feedback equalization of a time-varying, bandlimited channel. The RAP algorithm is based on multiple use of the filter state vectors, viewed as rows of the system data matrix. The multiple updates have a geometric interpretation as projections toward the hyperplanes described by the rows of the system data matrix. The algorithm provides improved performance over the conventional LMS and RLS algorithms for DFE equalization of data whose frequency spectrum contains nulls as well improved mean square error (MSE) at the output of the DFE compared to the RLS algorithm for spectrally nulled data corrupted by noise