ADVANCED MODELING APPROACHES FOR MICROWAVE FET DEVICES & SUB-SYSTEMS

Abstract

In this work, both small-signal and large-signal models for microwave filed-effect transistors (FETs) and behavioral models for wireless sub-systems have been investigated. Firstly, one novel technology-independent small-signal model generation method is presented, where the linear polynomial regression is adopted to improve the optimization speed and the accuracy of the model. The proposed approach is straightforward and suitable for wideband application. Secondly, based on the extracted small-signal model, a novel large-signal model with integration-path-dependent technology is presented to handle the dispersion effect. The consistency between the small-signal model and the large-signal model is highly reserved through the path-dependent integration which is creatively introduced to deal with the non-conservative partial derivative data. In addition, one hybrid electromagnetic (EM) and artificial neural network (ANN) model is proposed to improve the robustness of the large-signal model. The parasitic effect of the model is accurately captured by the distributed components. To deal with the voltage drop on the extrinsic resistors, the intrinsic elements of the model are uniformly redistributed by well-trained ANNs. At the same time, attributed to the universal approximation ability, ANNs are employed to represent the nonlinear voltage-current and voltage-charge relationships within the intrinsic part. The proposed models have been successfully applied in Ka-band GaAs MMIC power amplifier designs and validated. Finally, one novel black-box behavioral model for sub-systems is investigated and developed to handle the issues of inaccuracy of high order harmonics in non-50 Ohm environment. Also, this frequency-domain model includes the memory effects for wide bandwidth wireless system applications.