A neural-based CAD tool for RFmicrowave modeling

Abstract

The dramatic development of the commercial markets for wireless communication products leads to an increasing need for accurate and fast models of RF and microwave components and circuits. The traditional modeling approaches have the disadvantage of being either expensive or time-consuming. Although the basic artificial neural network as a fast and accurate modeling approach has been applied in diverse situations, the use of knowledge-aided neural networks is quite new. In this thesis, we focus on the development of a neural-based computer aided design (CAD) tool for the general Multi-Layer Perceptrons (MLP) neural network, the Knowledge-Based Neural Network (KBNN), and the Prior Knowledge Input (PKI) neural network. KBNN and PKI were used, for the first time in this thesis, in the modeling of a mixer and multistage amplifiers. Since in the RF and microwave field, the training data are usually obtained from measurements or simulations, which are either expensive in data generation or CPU time consuming, such applications of knowledge-aided neural networks (KBNN and PKI) have been proved to be capable of reducing the need for a large number of training data, and improving the accuracy and efficiency as well.