Small signal S-parameter estimation of BJTs using artificial neural networks

The RF/microwave circuit design and analysis, from inception to culmination extensively utilizes and depends on and scattering (S-) parameters. S-parameter data can either be obtained through measurements or simulations or by evaluating closed form expressions. Measurement of S-parameter, while most accurate, is not always possible or convenient. In the recent past, artificial neural networks (ANN) have emerged as an efficient tool for small signal S-parameter estimation. S-parameter estimation of a variety of active microwave devices except BJT has been done using ANN. The novelty of this work is that not only it supplements the prior work by estimating small signal S-parameter of BJT, but does so using equivalent circuit based approach, as compared to the data based modeling approach adopted in prior work. The advantages using this approach have been elucidated in this paper. Results were verified by comparing the S-parameters predicted by ANN against PSPICE output. The results displayed excellent conformity throughout the frequency range (200 MHz to 3 GHz) for intrinsic silicon npn BJTs.