Author: Andrew Joseph SellersPublisher:
ISBN:
Category : Power electronics
Languages : en
Pages : 0
Book Description
This dissertation investigates the propagation of information between models of disparate computational complexity and simulation domains with specific focus on the modeling of wide bandgap semiconductors for power electronics applications. First, analytical physics models and technology computer-aided design numerical physics models are presented. These types of physics models are contrasted by ease of generation and computational complexity. Next, processes generating transient simulations from these models are identified. Mixed-mode simulation and behavioral device models are established as two available options. Of these two, behavioral models are identified as the method producing superior computational performance due to their much-reduced simulation time. A comparison of switching performance for two wide bandgap field-effect transistors manufactured with the same process is next presented. Empirical and simulated switching results demonstrate that available models predict the slew rates reasonably well, but fail to accurately capture ringing frequencies. This is attributed to two primary causes; the modeling tool used for this comparison is incapable of producing a sufficiently high-quality fit to ensure accurate prediction and the devices are sensitive to parasitic values beyond the measurement uncertainty of the characterization hardware. To remedy this, a two-fold approach is necessary. First, a new model must be generated which is more capable of predicting steady-state performance. Second, a characterization procedure must be produced which tunes parameters beyond what is possible with empirical characterization. To the first point, a novel model based on the Curtice model is presented. The novel model adapts the Curtice model by adding gate-bias dependence to model parameters and introducing an exponential smoothing function to account for the gradual transition from linear to saturation exhibited by some wide bandgap field-effect transistors. Care is taken to model forward conduction, reverse conduction, and transfer characteristics with high accuracy. Non-linear capacitances are then modeled using a charge-based lookup table demonstrated by previous work in the literature to be effective. Thermal performance is accounted for with both the incorporation of thermal scaling factors and a thermal RC network to account for joule-heating. The proposed model is capable of capturing device steady-state and small-signal performance more precisely than previous models. A tuning and optimization procedure is next presented which is capable of tuning device model parasitic values within uncertainty bounds of characterization data. This method identifies the need for and introduces new model parameters intended to account for dispersive phenomena to a first degree. Pairing this method with the aforementioned model, significant improvements in transient agreement can be achieved for fast-switching devices. A method is also presented which identifies and quantifies the impact of parameters on transient performance. This process can be used to remove model parameters from the tuning set and possibly decouple parameter tuning. The propagation of these fully-tuned device and circuit models to the system level is next discussed. The cases of a buck converter and double pulse test are used as examples of dc switching circuits which may be used for switching characterization and to account for switching losses. Simulation is used to demonstrate that these circuits, when using similar components, produce comparable results. This allows the use of double pulse tests for switching characterization in simulation, thus eliminating the need for quasi-steady-state conditions to be reached in converter simulation. Methods are proposed for the inclusion of this data into system-level models such that simulation time will be minimally impacted. When used in conjunction, the methods presented in this chapter are sufficient to propagate information from the physics level all the way through to the system level. If specific circuits and system components are known, the impact of including a theoretical device can be assessed. This lends itself to advanced design of each type of model by analyzing the interactions predicted by various levels of models. This has serious implications for accelerating the deployment of wide bandgap semiconductor in power electronics by addressing the primary concerns of reliability and ease of implementation. By using these methods, devices, circuits, and systems can each be optimized to fully benefit from the theoretical advantages presented by wide bandgap semiconductor materials.
