Box-Jenkins continuous-time modeling (Q1571080)

The authors consider continuous-time SISO systems where the parameters of the transfer function are to be identified from samples of input and output respectively. The output is disturbed by noise, the input signal is an arbitrary band-limited excitation. The key idea is to assume that the disturbing noise can be written at the sampling instants as a filtered white discrete-time noise and to combine this discrete-time noise model with a continuous-time plant model, which results in a hybrid Box-Jenkins model structure. It has the advantage that (1) asymptotically there is no approximation error over the complete frequency band from DC to Nyquist and (2) it allows general parametric noise models to be identified. The modeling approach is illustrated on real measurements of an electrical circuit.