PAD modeling by using artificial neural network

Abstract

An approach for the PAD modeling technique for microwave on wafer measurement based on a combination of the conventional equivalent circuit model and artificial neural network (ANN) is presented in this paper. The PAD capacitances are determined from S parameters of different size of PAD test structure based on EM (electromagnetic) simulation and described as functions of the dimensions of the PAD structure by using sub-ANN. Good agreement is obtained between ANN-based modeling and EM simulated results up to 40 GHz. The de-embedding procedure for PHEMT device utilizing the ANN based PAD model is demonstrated.