Approximate SPICE modeling of SiC MOSFETs

The recent adaptation of wide bandgap (WBG) semiconductors pushes the SPICE circuit simulation software to the very edge, requiring computationally light and accurate assessment of rapid and oscillatory transients. One of the main limitations in SPICE modeling of WBG semiconductors is the lack of built-in models in the available software, forcing usage of behavioral modeling with heavily non-linear equations. Using such implementation leads to high computational costs and convergence issues. This work presents a general approach to the approximate modeling of WBG semiconductors, leveraging the functionality of modern open-source SPICE software to improve convergence by decoupling non-linear model formulation from the SPICE model formulation and to move the computational cost of non-linear model updates outside the SPICE software. The proposed modeling approach is suitable for further acceleration by FPGA implementation and machine learning surrogate modeling. To investigate the accuracy of the approximate modeling, the approximations are derived for charge and channel equations of two silicon carbide MOSFETs and compared in a double pulse test against the original simulation model and experimental results. It is shown that the approximate implementation maintains the accuracy of the original model in the transient simulation and improves the convergence, leading up to 59% acceleration.