The use of artificial neural networks in nonlinear microwave devices and circuits modeling: An application to telecommunication system design (invited article)

Abstract

The design of telecommunication systems is a hierarchical process involving use of a large set of simulation tools and relying on appropriate modeling of system elements to obtain get‐it‐right‐the‐first‐time fabrication. This paper investigates in detail possible application of neural networks to modeling of large‐signal hard‐nonlinear behavior of power transistors for circuit design purpose, and modeling of nonlinear circuits such as power amplifiers for system design purpose. The problem of finding a good model is discussed through solutions offered by neural networks, with particular interest in wavelet networks trained by BFGS algorithms. Finally, experimental results are reported. ©1999 John Wiley & Sons, Inc. Int J RF and Microwave CAE 9: 198–215, 1999.

Application of artificial neural network models to linear and nonlinear RF circuit modeling

“…Next, we design an efficient training algorithm based on the Powell's optimization scheme, with the initial point determined by the GA. An alternative trainw x ing procedure described in ref. 12 can also be used. The latter is based on the wavelet theory and quasi-Newton Broyden᎐Fletcher᎐Goldfarb᎐ Shannon algorithm.…”

Neural-Based Models of Semiconductor Devices for SPICE Simulator

“…The universal approximation property of ANN [9][10] provides them the ability to learn any arbitrarily nonlinear input-output relationships [11 -14] from corresponding measured or simulated data. Moreover, researches started investigating NN approaches to model transistor DC [15][16][17][18][19] , small signal [20][21] , and large-signal [22][23][24][25][26][27] behaviors. Xiuping et al [28] have proposed an improved microwave active device modeling technique based on the combination of the equivalent circuit and ANN approaches.…”