A new technique for nonlinear transistor modeling is suggested. The technique includes an analytical extraction and a multistage optimization, providing fully automated modeling. An extraction algorithm for the parameters of an internal voltage-controlled current source Ids is shown. A fully automated workflow of small-signal model extraction is presented. An approach for obtaining parameters of nonlinear capacitances is described. A 0.15um GaAs pHEMT nonlinear model was built and verified with measured IV-curve, multi-bias S-parameters and power characteristics.
A golden device selection is a necessary step of model extraction for microwave monolithic integrated circuit (MMIC) elements. A golden device, or golden die, is an element sample on a wafer that represents the centerline process technology. The existing golden device selection algorithms consider several independent DC or AC parameters expressed as real values. But for MMIC elements the microwave characteristics are more important than DC or low-frequency parameters. The proposed in this paper algorithm carry out frequency-dependent characteristics and real-valued parameter simultaneously to select a golden device. The paper presents numerical experiment and experimental application of the proposed algorithm. The proposed algorithm is applied to GaAs pHEMT low-noise amplifier design. But it is potentially applicable to other circuit types (switches, digital, power amplifiers, mixers, oscillators, etc.) with a minor tuning only in the considered characteristics set. The algorithm performance allows operating with a rather large dataset.