Application of multi-frequency test signals to closed loop system identification

The authors assess the performance of dynamic response estimators as a function of closed loop plant topology, random noise and externally applied perturbation signals. Analysis of a general feedback structure has shown that the conventional stochastic frequency response estimator contains the forward and reverse path dynamics each weighted by a real valued spectral window, resulting in causal and anti-causal contributions in the associated impulse response estimates. It was found that signal averaging in conjunction with an external periodic perturbation signal yields response estimates which become progressively independent of noise and closed loop characteristics when successive data segments are added to the estimators.