Extraction of small-signal model parameters of Si/SiGe heterojunction bipolar transistor using least squares support vector machines

A novel straightforward methodology for extracting bias-dependent small-signal equivalent circuit model parameters (SSECMPs) of silicon/silicon-germanium heterojunction bipolar transistors is presented. The inverse mapping between SSECMPs and scattering (S) parameters is established and fitted using simulated data of the SSECM. Since the problem has large input space, S-parameters at many frequency points, the least squares support vector machines concept is used as regression technique. Physical SSECMPs values are obtained using the proposed methodology. Moreover, an excellent agreement is noted between the S-parameters measurements and their simulated counterpart using the extracted SSECMPs in the frequency range from 40 MHz to 40 GHz at different bias conditions.