Neural FET small-signal modelling based on mel-frequency cepstral coefficients

In this paper, a new neural approach for field effect transistor (FET) small-signal modelling, based on Mel-frequency cepstral coefficients (MFCCs), is presented. This approach uses the scattering parameters of the FET at discrete frequencies in a certain band to predict the small-signal circuit elements. The proposed approach assumes that the neural network inputs represent a random data sequence. Few MFCCs are extracted from this random data sequence and used to train the neural networks that will relate these coefficients to the circuit elements. Two major objectives can be achieved using this approach; a reduction in the number of neural network inputs, and hence a faster convergence of the neural training algorithm, and a robustness against measurement errors in the testing phase. Experimental results show that the MFCCs are less sensitive to measurement errors than the actual measured scattering parameters.