GRNN-Based Scattering Parameter Modeling Investigation for HBT at Different Temperature

Abstract

In this paper, the scattering parameter (S-parameter) modeling method for heterojunction bipolar transistor (HBT) at different temperatures is investigated. S-parameters of HBT at different temperatures are randomly divided into training and testing sets, which are modeled by radial basis function (RBF) neural network and general regression neural network (GRNN), respectively. Then, the fitting results of the two models in predicting S-parameter are displayed. The experimental results show that the fitting results of RBF neural network are good, but the fitting errors of some data are existed. Meanwhile, most of the data predicted by GRNN can achieve ideal fitting. In addition, the error curve of RBF neural network is more volatile, while for GRNN it has a small fluctuation range, which can predict the S-parameters more stably. Finally, the mean square error (MSE) of RBF and GRNN neural network prediction model are $8.6178\times 10 ^{-4}$ and $3.1041\times 10 ^{-4}$ , respectively. It is proved that GRNN has a better modeling effect for HBT S-parameter at different temperatures. Therefore, the proposed modeling method can accurately characterize the S-parameter of HBT at different temperatures.

Keywords

• Scattering parameter
• HBT modeling
• RBF neural network
• GRNN