Dynamics of multichannel feedforward adaptive systems

The role played by reference-signal compensation in the convergence of multichannel feedforward adaptive systems is addressed. Convergence conditions are established, and effects of compensation errors are analyzed. We also explore, by means of altering the reference-signal compensation filtering from the conventional choice, ways to force systems to converge to arbitrary solutions of possible interest other than the standard Wiener solution. This could be useful, for example, if one wished to cancel noise at a very large number of measurement points, which, conventionally, would require a correspondingly complex, possibly prohibitively large, controller. The scheme developed enables very efficient usage of error signals, such that systems with large numbers of disturbance-cancellation points need employ only a relatively small number of error signals in the actual control-system implementation