Extending static models by using time series to identify the dynamical behavior

Abstract

We use a simple, static model of an amplifier and augment the model by adding a nonlinear dynamical part in which the dynamics are identified using principles of time series analysis. The static part of the model is a polynomial nonlinearity in the input voltage, and is implemented using a built-in system amplifier model in ADS. The dynamic nonlinear part of the model is implemented using an artificial neural network. This new model is fast to simulate and extends the simple, single frequency system amplifier model to cover a wide bandwidth, maintaining good large-signal predictions.