Factors and Considerations for Modeling Loss of a GaN-based Inverter

Zhe Yang,Paige Renne Williford,Edward A. Jones,Jianliang Chen,Fei Wang,Sandeep Bala,Jing Xu

DOI: https://doi.org/10.1109/tpel.2020.3012958

IF: 5.967

2021-03-01

IEEE Transactions on Power Electronics

Abstract:The article investigates the impacts of four often-neglected factors on the loss model of a GaN-based full-bridge inverter: parasitic capacitance of the devices, dynamics of junction temperature (T_j) under time-varying power dissipation (P_loss), case temperature estimation, and detailed considerations of the passive components. Procedures to calculate the converter loss considering the above factors are proposed and implemented. A 4.5-kW hard-switching inverter prototype using gallium nitrite (GaN) high-electron-mobility transistors is used to experimentally demonstrate the impact of each factor on the converter loss model. It is found that the accuracy of a converter loss model is mainly affected by the passive components at the light load condition, whereas the thermal and loss models of the active components become the major factors as the output power increases. The results show that after considering the above factors, the converter loss discrepancy between calculation and measurement can be reduced from 30.6 W (28%) to 2.5 W (less than 3%) at heavy load (P_o = 4.5 kW), while at the light load condition (P_o = 500 W), it is reduced from 3.9 W (28%) to 2.6 W (16%). Furthermore, the difference between simulated and measured case temperature of the GaN devices is within 6 °C.

engineering, electrical & electronic

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to improve the accuracy of the loss model of the full - bridge inverter based on gallium nitride (GaN). Specifically, the paper explores the influence of four often - ignored factors on the inverter loss model. These factors include:

1. **Parasitic Capacitance of the Devices**: In practical applications, due to the external parasitic capacitance (\(C_p\)) introduced by filters, heat sinks and printed circuit boards (PCBs), the switching loss is more sensitive to parasitic elements. This influence is usually negligible in silicon - based devices.

2. **Dynamics of Junction Temperature (\(T_j\))**: Due to the low heat capacity of GaN devices, the junction temperature responds rapidly to the time - varying power loss (\(P_{\text{loss}}\)). Especially in inverters, the instantaneous power loss varies widely within a 50/60 Hz line cycle, which leads to significant changes in the junction temperature.

3. **Case Temperature Estimation**: GaN devices usually require a high - density PCB layout to reduce parasitic inductance and improve switching performance, which will complicate the thermal coupling between adjacent devices. Traditional thermal models may not be accurate enough when using lumped thermal resistance to estimate the case temperature.

4. **Detailed Considerations of the Passive Components**: The losses of passive components such as inductors and capacitors cannot be ignored in high - frequency converters. In particular, the core loss may be much higher at low temperatures than at high temperatures, and the loss of capacitors is mainly caused by the equivalent series resistance (ESR), whose value depends on the working conditions.

By proposing and implementing a converter loss calculation method that takes the above factors into account, the paper aims to reduce the difference between calculation and measurement. The experimental results show that under heavy - load conditions (output power \(P_o = 4.5 \text{kW}\)), after considering these factors, the difference in converter losses is reduced from 30.6 W (28%) to 2.5 W (less than 3%); while under light - load conditions (output power \(P_o = 500 \text{W}\)), the difference is reduced from 3.9 W (28%) to 2.6 W (16%). In addition, the difference between the simulated and measured case temperatures of GaN devices is within 6°C.