Investigation of behavioral models for MMIC elements

Behavioral models for MMIC elements based on three multidimensional approximation techniques, i.e., artificial neural networks (ANN), radial basis functions method, and inverse distance weighting (IDW) method, are investigated. As a test example, the parametric models for describing a frequency dependence of scattering parameters of GaAs pHEMT in different DC operating points have been constructed. It is shown that IDW model demonstrates the accuracy and calculation speed that are comparable with ANN model. However, IDW method needs much lesser time for the model building. Thus, it can be successfully used for constructing MMIC elements´ behavioral models.