Abstract: A finite impulse response (FIR) filter characterization is performed using a modified least means squares approach that updates values of FIR filter taps as groups of taps, periodically. For the taps updated for a current input signal sample, future convolution values for future input signal samples are computed and also a partial estimate of the signal response is computed. The expected response is then computed by summing the partial estimate signal response and other partial estimates of the signal response which are computed from future convolution, missed tap update convolution and normalizes error signal values which have been previously computed at past input signal samples and stored in memory.