Small signal model and efficient parameter extraction technique for deep submicron MOSFETs for RF applications

A new small-signal model for deep submicrometre MOSFETs is proposed for accurately predicting MOS transistor behaviour up to 15 GHz. It is a unified model suitable for both baseband and RF simulation and valid in both the triode and saturation regions. The model is implemented as a macromodel with parasitic elements added to the BSIM3v3 core. The BSIM3v3 model is the ideal basis for RF simulations as it consists of an accurate nonquasistatic model and a capacitance model. The parameter extraction methodology is analogous to the `divide and conquer´ strategy used in computer science. This approach is superior to the traditional method of optimising the entire model to fit the measured S-parameters. It has shown an increase in accuracy and optimisation speed while reducing convergence problems caused by global optimisation. Excellent agreement between measured and simulated S-parameters at different biasing conditions in the range from 50 MHz to 15 GHz has been obtained