A class of mixed transversal and ladder adaptive filters with pure order updates

A class of exact recursive-least-squares (RLS) adaptive filter structures with pure order updates, called mixed transversal and ladder (MITAL) filters, is presented in which the propagation of numerical errors along in time is limited to at most M sampling periods, so the long-term numerical stability can be guaranteed. The adaptive structures can be treated either as transversal forms or as ladder forms to meet the requirements of different applications. The outstanding numerical performances of the adaptive filters are illustrated by computer simulations