Modeling microwave devices and circuits for telecommunications system design

The design of telecommunication systems is a hierarchical process involving the use of a large set of simulation tools and relying on appropriate modeling of system elements in order to obtain get-it-right-the-first-time fabrication. Even if today's simulators are powerful tools, they can only be as accurate as the mathematical model that are used. As a consequence constructing good models for devices used in circuits and for circuits used in systems is a basic task. 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 first presented so as 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 power of the proposed approach, to achieve a significant step in telecommunications system simulation accuracy by means of models that embed more information and to promote the development of new system simulators able to totally handle these models.