Various methods are known for the obtaining a model of the type of the transfer function from the database for step response. Implicit requirement of such methods is that the data of the characteristic to be relatively free from noise. In the present work, to obtain a model from step response with noise, the method of orthogonal decomposition applied directly to noisy step response. Although the filtering properties of orthogonal models in the literature traditionally analyzed objects/processes

more »

... with negligible noise level. For the step response it may use the integration by parts, so that the derivative instead of the step response to be held by the orthogonal functions. This way avoids the approximation of the derivative of the disturbance. Good approximating properties of the process is concluded in its robustness to the overlay noise in output data and high reliability in both the time and frequency domain.