Design of radio frequency (RF) power amplifiers (PAs) for wireless communications requires small- and large-signal data collected from the underlying transistors, including scattering parameters (S-Parameters) and load-pull (LP), to determine optimal impedance targets. High speed devices operating with fundamental frequencies above 35 GHz present extreme challenges for measuring the harmonic signals resulting from nonlinear effects. Predictive physics based simulations in conjunction with compact modeling capabilities are promising alternatives to expensive and time-consuming measurements. To date, tools either exist in the electron transport domain or in the behavioral modeling domain and a key goal is to treat these problems simultaneously because they are strongly coupled at millimeter wave frequencies. Accurate physics based simulations of high speed and high power transistors require proper treatment of hot-electron, self-heating, and full-wave effects. The Boltzmann solver called Fermi kinetics transport (FKT) has been shown to capture all of these important physical effects. FKT can approach the accuracy of Monte Carlo methods while maintaining the computational efficiency of deterministic solvers. The latter trait allows simulation of large electronic devices such as the output stages of PAs. Previous work on FKT provided proof of concept results which demonstrated its versatility and accuracy as an electronic device simulation framework.The purpose and contribution of this thesis is the use of FKT as a predictive TCAD tool to generate RF data required for PA design. This work begins with a thorough investigation of the underlying physical equations and their numerical solution for electronic device simulations. Included in this investigation is an analysis of the full-wave discretization technique called Delaunay-Voronoi surface integration (DVSI), a derivation of the FKT device equations and their discretization in energy- and real-space, and a detailed account on the numerical solution of the fully coupled nonlinear system of equations. The detail provided in this work is meant to provide future device engineers and researchers a thorough understanding of the numerical framework for their application and simulation needs. The FKT device simulator is then applied to real device geometries to generate useful data for RF circuit designers. Extensions of the FKT method required for large-signal LP simulations are presented with representative applications. Additionally, compact behavioral models are extracted directly from FKT device simulations, enabling a computationally efficient means for simulated LP data generation. The resulting TCAD tool is a promising simulation capability for high power RF transistor design and characterization. It is anticipated that PA design for 5G applications will be using techniques like these in the near future.
Read
