Modeling microwave devices and circuits for telecommunications system design

This paper investigates in detail the possible application of neural networks to the modeling of large-signal hard-nonlinear behaviour of power transistors for circuit design purpose, and to the modeling of nonlinear circuits such as power amplifiers for system design purpose. System design methodology is presented to identify the main modeling steps. Structures are proposed to embed behavioural information, based on time varying Volterra series for device modeling and on bilateral model for circuit modeling. The problem of finding a good model is then discussed through solutions offered by neural networks, with a particular interest in wavelet networks trained by BFGS algorithms. Finally, two modeling examples are presented and simulation results are compared with classical model-based simulations or with measurements, in order to demonstrate the effectiveness of the proposed approach