On model reduction of discrete time systems (Q1820099)

A well-known method of model reduction of linear discrete-time systems is further elaborated. If the assumption can be made that r of n poles are located close to the origin in the complex z-plane, then precisely these r poles are omitted to form the reduced model of order n-r, whereas the positions of the remaining poles are slightly modified. It is therefore apparent that the approximate transfer function will be a better fit near \(z=1\) than away from that point. In particular, by using connections between polynomial coefficients, the roots and the Schur-Cohn determinants, the range of validity of reduction is established. Furthermore, the lattice realization of the original system compared with that of the reduced one is presented. Finally, some relationships between the transfer function of both systems are obtained.