Multivariate Regression Polynomial: A Versatile and Efficient Method for DC Modeling of Different Transistors (MOSFET, MESFET, HBT, HEMT and G4FET)

Abstract

This work presents multivariate regression polynomial as a versatile and efficient method for DC modeling of modern transistors with very different underlying physics including MOSFET (metal-oxide-semiconductor field-effect transistor), MESFET (metal–semiconductor field-effect transistor), HBT (heterojunction bipolar transistor), HEMT (High-electron-mobility transistor) and a novel silicon-on-insulator four-gate transistors (G4FET). A set of available data from analytic solution, TCAD simulation, and experimental measurements for different operating conditions is used to empirically determine the parameters of this model and a different set of test data is used to verify its predictive accuracy. The developed model expresses the drain current as a single multivariate regression polynomial with its validity spanning across different possible operating regions as long as the chosen independent variables lie within the range of training data set. The continuity of the resulting polynomial and its first and second order derivatives make it particularly suitable for implementation in a circuit simulator. The model also provides a method for further simplification based on prior knowledge of the underlying physical mechanism and shows excellent predictive capability for different kinds of devices. This can be very useful for modeling deep-submicron emerging devices for which any closed-form analytical solution is not yet available.