Sensitivity of the continualization of poles and its effect in system identification

In a continuous-time parameter identification process which uses a digital data acquisition system, the form in which the discrete-time model of the dynamic system is represented influences the performance of the estimation process. In this paper, it is shown that alternatives to the conventional representation of discrete-time systems, which uses the z operator, possess pole-transformation sensitivity magnitudes converging to unity as the sampling interval approaches zero, while that of the z form becomes unbounded. This fact is shown to influence the accuracy of continuous-time pole estimates and favors the use of alternatives to the widely-used z operator in practical implementations of discrete-time systems, as provided with experiments and simulations on a voice-coil-driven flexible positioner.