The thermal design of electrical components is critical: exceeding thermal limits can endanger the safety of both equipment and users. In printed circuit boards (PCBs), the main design parameters are the number and size of thermal vias, the size and thickness of copper pads, the thickness of the flame retarded epoxy-resin (FR4) boards, and the number of layers. Understanding and modeling how these parameters influence the temperature of the design is a necessary step in the design process. The finite element method (FEM) involves breaking the domain of a geometry into a sub-domain and solving the system of differential equations governing the problem of interest at each node. Printed circuit boards (PCBs) components range in size from tens of micrometers to tens of centimeters, which presents a difficulty in this process: capturing small components results in large meshes, which are computationally expensive to generate and solve, while increasing the mesh size produces poorly conditioned and degenerated elements, which will likely not converge. Thermal vias have a copper plating of typically 25 um. Regions with a high concentration of thermal vias bounded by copper plating create a relatively homogeneous thermal gradient. Accordingly, the thermal vias and PCB material in these regions may be replaced by homogeneous, anisotropic materials. This thesis proposes combining regions with large concentrations of thermal vias into bodies of anisotropic, thermal conductivity, dimensioned by their abstracted thermal resistance networks. This process performs well theoretically, and has an R2 = 0.99191 compared to modeling with discrete elements. A FEM model of TI reference design PMP21440, an 8 W, 7 V to 0.8 V, 10 A output current, GaN buck converter, is dimensioned by this process, and compared against measurements. The resulting model has an R2 = 0.74568 that is relatively stable across three power loss injections. The results are adequate. A lack of information on the experimental setup make it difficult to isolate error sources. This approach is likely worth further investigation.
| Datum der Bewilligung | 2024 |
|---|
| Originalsprache | Englisch |
|---|
| Betreuer/-in | Mario Jungwirth (Betreuer*in) |
|---|
PCB Thermal Design: Homogenization of Anisotropic Bodies by Thermal Network Abstraction for Computational Efficiency
Janson, J. R. (Autor). 2024
Studienabschlussarbeit: Masterarbeit