EYE-HEIGHT/WIDTH PREDICTION USING ARTIFICIAL NEURAL NETWORKS FROM S-PARAMETERS WITH VECTOR FITTING

Artificial neural networks (ANNs) have been used to model microwave and RF devices over the years. Conventionally, S-parameters of microwave/RF designs are used as the inputs of neural network models to predict the electrical properties of the designs. However, using the S-parameters directly as inputs into the ANN results in a large number of inputs which slows down the training and configuration process. In this paper, a new method is proposed to first disassemble the S-parameters into poles and residues using vector fitting, and then the poles and residues are used as the input data during configuration and training of the neural networks. Test cases show that the ANN trained using the proposed method is able to predict the eye-heights and eye-widths of typical interconnect structures with minimal error, while showing significant speed improvement over the conventional method