Highly parallel multichannel LS algorithms and application to decision-feedback equalizers

The efficient least-squares (LS) design of multichannel causal finite-impulse-response (FIR) filters characterized by a different number of memory elements per input channel is discussed. A highly concurrent transversal algorithm for these filters, suitable for very large scale integration (VLSI) implementation, is derived. This batch algorithm is used in the efficient adaptive computation of the multichannel filter. The batch and adaptive algorithms are applied to LS autoregressive-moving-average identification and to the design of efficient decision-feedback equalizers. Simulation results are included