S-functions behavioral model order reduction based on narrowband modulated large-signal network analyzer measurements

Abstract

In this paper we report for the first time on order reduction applied to S-functions behavioral models. The most dominant model parameters are selected based on the relative uncertainty of their estimated values evaluated against a threshold value. The selection procedure is performed on the same measurement data that is used to extract the model and obtained using a large-signal network analyzer. High level of model order reduction, achieved without any substantial loss of the prediction accuracy, is demonstrated on S-functions extracted for a packaged pHEMT device.