Transfer function identification with filtering techniques

This paper addresses the problem of identifying the transfer function of a plant with the aid of filtering techniques. Conditions for obtaining fair estimates of a transfer function are derived. It is shown that, to be able to reduce the effect of noise and other stochastic effects, highly tuned filters are required. It is also found that proper excitation signals are necessary for effective implementation of the proposed scheme. Some methods for designing the required highly tuned filters are proposed. The relationship between the proposed scheme and existing methods is discussed. It is shown that for some specific cases the proposed scheme is, in effect, a computationally efficient implementation of existing methods. However, for some other cases the proposed scheme gives more accurate results. The theoretical predictions are substantiated with computer simulations