Optimization Algorithms for Dynamic Tuning of Wide Bandgap Semiconductor Device Models

Abstract

Circuit and device parasitics have an outsized effect on the switching voltage and current waveforms of wide bandgap semiconductors. The variation of these parasitic components greatly hinders the ability to develop simulation models of wide bandgap semiconductors that accurately predict transient device performance. As a solution, the concept of dynamic tuning has become prevalent in the modeling and simulation of wide bandgap semiconductor-based power electronics. This paper presents dynamic tuning applied to two different behavioral models of the same 100 V gallium nitride (GaN) device. Although the models are of the same device, they are disparate in their prediction capability of the device’s empirically measured static characteristics. The different static characteristics also lead to a marked discrepancy in their transient prediction capabilities. Through dynamic tuning, the error between empirically mea-sured and simulated transient characteristics is improved for both models. This paper thus shows two important results. First, the frequency dependence of the parasitic components within a circuit can be accounted for, to a first order, through dynamically tuning a constant lumped element model of the parasitics. Second, dynamic tuning can be successfully, albeit not as effectively, applied to accurately predict transient behavior even for device models that do not precisely match the data sheet.