A method to select correct stimuli levels for S-functions behavioral model extraction

S-functions and other frequency-domain behavioral models based on describing functions rely on the linearization of the response of a nonlinear RF or microwave component around a set of large-signal operating points. To assure the validity of the linearization principle, and thus the validity of the model, one has to select appropriate stimuli levels during the experiments which are performed to extract the model parameters. In this work we propose a formal procedure to identify the correct small-signal power levels using a system identification approach, i.e. taking the uncertainty of the measured data into account. The method is based on the analysis of the cost function of an errors-in-variables (EIV) estimator obtained for simple models.