Digital adaptive filter and method of convergence therein

Abstract

Disclosed is a digital adaptive filter and a method of cancelling an echo component generated by a partial return of a transmission signal to a receiving side. The digital adaptive filter comprises: a main memory unit for storing a first echo replica which can be stored or read by using the parallel signal from a main tap unit as an address signal, at least one submemory unit for storing a second echo replica which can be stored or read by using the parallel signal from said subtap unit as an address signal, and an adding circuit for adding the first echo replica and the second echo replica. The second echo replica stored in the at least one submemory unit has a greater precision than the precision of the first echo replica stored in the main memory unit, and thus the accumulated error due to memory division can be reduced. Instead of or accompanied by the greater precision, the contents of the main memory are updated directly based on the error between the pseudo signal and an input signal or based on a step size corresponding to the error, depending on whether the error is relatively large or relatively small, respectively.