Accuracy Evaluation and Proposed Dynamic Tuning Procedure of a Compact SiC SPICE Model
Konferenz: PCIM Europe 2024 - International Exhibition and Conference for Power Electronics, Intelligent Motion, Renewable Energy and Energy Management
11.06.2024-13.06.2024 in Nürnberg, Germany
doi:10.30420/566262199
Tagungsband: PCIM Europe 2024
Seiten: 10Sprache: EnglischTyp: PDF
Autoren:
DeBoi, Brian; Nelson, Blake; Curbow, Austin
Inhalt:
The commercialization of wide bandgap technology has increased demand for accurate circuit-level simulation models of devices such as Silicon-Carbide (SiC) MOSFET power modules. These models assist with design challenges such as minimizing overshoot and electromagnetic interference associated with wide bandgap switching speeds. However, it is challenging to create highly accurate models across all the conditions that the device may be operated in, such as varying gate resistance, temperature, operating voltage, and operating current. A major contributor to this problem is that the procedure for characterizing and modeling SiC MOSFETs is simplistic relative to their complexity in switching. In particular, the standard methods for characterizing and modeling the device capacitances are simplified, ignoring dependencies on voltage biases and frequency. However, increasing the model complexity to address these issues is generally not feasible because 1) the models must be efficient to converge and simulate quickly, and 2) the necessary characteristics are often impossible to measure. Rather than increasing the model complexity, this paper builds upon a previously presented compact behavioral model and applies a dynamic tuning procedure to improve alignment with empirically derived datasheet parameters. The procedure is applied to a half-bridge SiC MOSFET power module model and it is demonstrated that the overall dynamic accuracy of the model is increased by 50% across a wide range of double pulse test conditions. While this procedure further divorces the behavior model from physical reality, the tradeoff is acceptable given the purpose of this model: accurately predicting device behavior in application while minimizing computational complexity.