A New Steady State Analytical Thermal Modelling Method for High Power Density Power Module Design – The Fourier Series Thermal Model for Multi-layer Structures with Various Longitudinal and Lateral Dimensions

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)


The requirements of the thermal modelling methods for the design of power modules include: fast and accurate, easy to integrate with models in other domains, being able to model the geometrical and layout differences among the numerous design candidates. There are three kinds of thermal modelling methods available: the numerical methods, the analytical methods and the lumped-parameter thermal models. Due to the geometry characteristics of power modules which share a relatively simple and regular multi-layer structure, the analytical Fourier Series thermal model (FSTM) stands out among these methods as a suitable method for power modules design. However, the existing steady state FSTM can only model the multi-layer structure with each layer having the same size. Therefore, the structural simplification of power modules by extending the solder layer and the top copper layer to the full size of the Direct Bonded Copper (DBC) substrate has to be made to apply this FSTM. This structural simplification is only applicable for some power modules where the semiconductor chips are relatively far away from the copper track edges, but would cause large errors for the other power modules. In order to make reasonable structural simplification, the heat spreading characteristics of the power modules are studied. Based on the findings, the example SEMIKRON power module is simplified to a multi-layer structure with regular rectangular layers which have various longitudinal and lateral dimensions, with only 1% error caused. To solve the problems of the existing FSTM, a new FSTM which can model the multi-layer structure with various longitudinal and lateral dimensions is proposed by the author. The accuracy and calculation speed are analyzed by comparing to the mature FEM method, proving its fast calculation speed of finishing one simulation in 0.26s and achieve results with only 2% error compared to the FEM analysis which takes 2555s.
Date of Award24 Jun 2021
Original languageEnglish
Awarding Institution
  • The University of Bristol
SupervisorXibo Yuan (Supervisor)

Cite this