Neural approach for temperature dependent modeling of microwave FET noise wave temperatures

The main contribution of this paper is development of an efficient approach for extraction of the microwave FET noise wave temperatures. The approach is based on an artificial neural network aimed at calculating the noise wave temperatures from the given ambient temperature and frequency. As the artificial neural networks give almost immediate responses, the proposed approach can be used as a good alternative to the time-consuming optimization procedures in microwave simulators that are usually used for determination of the noise models´ parameters. The accuracy of the presented approach is validated by comparison of the transistor noise parameters obtained using the extracted noise wave temperatures with the measured noise parameters.