Abstract :
One of the purposes of the paper is to discuss a number of properties of periodic test signals relevant to the identification of linear dynamic systems. The emphasis is on multifrequency signals (MFS). These are periodic signals having their power, or the major part of it, distributed in a specified way over a limited number of selected harmonics. The reason for this emphasis is that, different from the spectrum of other periodic test signals such as maximum length binary sequences, the spectrum of MFS can to a certain extent be freely chosen. This freedom may be used to incorporate in the chosen spectrum a priori knowledge about the system and about the spectrum of the noise in the observations of the response. A further purpose is to describe aspects of the use of MFS for the identification of linear dynamic systems. This concerns in the first place the estimation of the Fourier coefficients of the steady-state response to MFS from noise disturbed observations. Furthermore, an example is presented of the use of these Fourier coefficient estimates for the identification
