A New Trend for FET Small Signal Modeling Using Cepstral Analysis

In this paper, a new technique is proposed for field effect transistor (FET) small-signal modeling using neural networks. This technique is based on the combination of the Mel-frequency Cepstral Coefficients (MFCCs) with the different discrete transforms such as the discrete cosine transform (DCT), the discrete sine transform (DST), and the discrete wavelet transform (DWT) of the inputs to the neural networks. The input data sets to traditional neural systems for FET small-signal modeling are the scattering parameters and the corresponding frequencies in a certain band, and the outputs are the circuit elements. In the proposed approach, these data sets are considered as forming random signals. The MFCCs of the random signals are used to generate a small number of features characterizing the signals. In addition, other vectors are calculated from the DCT, the DST, or the DWT of the random signals and appended to the MFCCs vectors calculated from signals. The new feature vectors are used to train the neural networks. The objective of using these new vectors is to characterize the random input sequences with much more features to be robust against measurement errors. There are two benefits for these approaches: a reduction in the number of neural networks input, hence a faster convergence of the neural training algorithm and robustness against measurement errors in the testing phase. Experimental results show that the techniques based on the discrete transforms are less sensitive to measurement errors than using the traditional and MFCCs methods.