Nonlinear behavioral modeling of oscillators in VHDL-AMS using Artificial Neural Networks

In this paper an approach to behavioral modeling of microwave oscillators is described. The presented model takes into account start-up, steady state behavior and phase noise. To describe the nonlinearities, an artificial neural network (ANN) is employed. The dynamic behavior of the oscillator is described by a system of differential equations that are solved in VHDL-AMS. As opposed to input-output models of microwave devices, this paper presents a self sustaining oscillation, which starts from a small injected excitation (e.g. noise) and ends in a stable limit cycle. Additionally, the phase noise characteristics of the oscillator in the 1/f² and flat region are emulated.