This work describes a Fast novel thermal analysis simulation tool for integrated circuits (FANTASTIC), which is fully automated and relies on an enhanced version of the Multi-Point Moment Matching algorithm. The tool provides a novel equivalent network suitable for use in SPICE-like circuit simulators to perform efficient thermal and electrothermal analyses. FANTASTIC requires much less CPU time and memory storage compared to commercial simulators. The thermal behavior of a state-of-the-art four-finger GaAs HBT is investigated as a case-study.

This paper presents an efficient circuit-based approach for the nonlinear dynamic electrothermal simulation of power devices and systems subject to radical self-heating. The strategy relies on the synthesis of a nonlinear compact thermal network extracted from a finite-element model by a novel model-order reduction method requiring a computational time orders of magnitude lower than conventional techniques. Unlike commonly employed approaches, the proposed network allows reconstructing the whole time evolution of the temperature field in all the points of the domain with high accuracy. Electrothermal simulations are enabled in a commercial SPICE-like simulator by coupling such a network with subcircuits that describe the electrical device behavior by accounting for the temperature dependence of the key physical parameters. As a case study, the dynamic electrothermal analysis of a packaged silicon carbide power MOSFET undergoing a short-circuit test is performed, showcasing the performance of the approach and highlighting the need of including the thermal nonlinearities to achieve reliable results.

This paper presents a computationally efficient 3-D simulation approach for the dynamic electrothermal analysis of SiC power MOSFETs. The strategy relies on a circuit representation of the whole device, where the electrothermal feedback is enabled through an equivalent electrical network, and the elementary device cell is described by a novel behavioral model accounting for the non-intuitive temperature dependences of key physical parameters.

A novel matrix reduction method for the efficient and automatic construction of boundary condition independent dynamic compact thermal models having a chosen accuracy, is proposed. The method is implemented in a code which allows constructing boundary condition independent dynamic compact thermal models of any multi-die package modeled within a 3-D commercial mesher. The proposed approach has many advantages with respect to previous approaches in terms of robustness, efficiency, and applicability. The method is validated through the analysis of a dual-flat no-leads 12-leads package.

A novel nonlinear model order reduction method is proposed for constructing one-port dynamic compact models of nonlinear heat diffusion problems for ultra-thin chip stacking technology. The method leads to models of small state-space dimensions, which allow accurately reconstructing the whole time evolution of the temperature field due to an arbitrary power waveform of practical interest. The approach is also efficient, since the computational time/memory requirements for constructing each dynamic compact model is about one order of magnitude lower than that corresponding to a single 3-D finite element method transient simulation of a nonlinear problem.