Thermal modelling of power semiconductor modules in power electronic applications

Thesis

The power module is an important building block of a power electronic converter used in high power applications. It provides the placement of power semiconductor devices. One of the most important design constraints of power module is maximum junction temperature of semiconductor devices. Therefore, thermal modelling of the power module is important because it gives the maximum junction temperature of the devices. There are three different approaches to do the thermal modelling of the power module: analytical, finite element analysis and thermal network. In this work, the steady-state and transient thermal modelling of a rectangular N-layer structure with an arbitrary number of heat sources on the top surface is obtained by a Fourier series solution based on the separation of variables method. Although a typical power module does not exactly resemble an N-layer rectangular structure but it can be closely approximated as the latter. As the structure of power modules can be closely approximated as a rectangular N-layer structure. Various simplified structures are analyzed to understand the effects of structural approximation on the temperature field.

The steady-state model (RNLF method) is compared with the finite-element method simulation, and an excellent matching (approximately maximum 1.00\% temperature error) is found in the centres of the semiconductor dies for this case. Experimental temperature measurements taken at the surface of a commercial SiC power module are also presented demonstrating agreement in the centres of the dies to within 3.5\%. The transient model is compared with a finite-element model and an excellent matching (less than 2\% error) in the transient region is found in the centres of semiconductor dies. However, in the sub-transient region (this is the region where the rate of die temperature rise is much higher ), the error is higher (approximately 15\% error) because of finite thermal conductivity of the semiconductor material (SiC is used here). An experimental validation is performed using a high frame-rate thermal camera. After the model outputs have been compensated to match the limited frequency response of the camera, a good agreement is found between model and experiment (less than 6\% error). For the points where the rate of rise of temperature is relatively slow, this compensation is not necessary.

Authors: Choudhury, K

• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Subjects: Electrical and Electronic Engineering