Thermal analysis of power hybrid microelectronic packages

Abstract

In this dissertation a simplified nondimensional approach for the thermal analysis of power hybrid circuits is presented. The new technique uses only the metallization and the substrate as layers and represents everything below the substrate by an external thermal resistance (expressed as an equivalent convective heat transfer coefficient, h). In this study, the impact on thermal management of thick film metallization and copper cladding on alumina, aluminum nitride, and beryllia ceramic substrates is compared. The thermal conductivity of the substrate material, the thickness of the copper layer, the thermal resistance of the heat sink system, the size of the device, and the spacing between two heat dissipating devices are considered. The model results show that increasing the thickness of the copper layer can significantly decrease the device temperatures on alumina but may increase temperatures on high thermal conductivity substrates. Moreover, the model results show that increasing the thickness of the copper layer requires that the devices be placed farther apart to prevent thermal interaction. The results also demonstrate that the external heat sink resistance can have a significant impact on the heat flow paths and temperatures in the substrate. As the external resistance increases, the spacing required to prevent thermal interaction also increases.

In addition to the above, a series of experiments were conducted on various hybrid circuits samples for a low and high heat sink external resistance, i.e., large and small convective heat transfer coefficients, respectively. These samples were constructed using thick film resistors as heat sources on alumina and beryllia substrates. The temperature rise was measured using infrared thermal imaging technique. These experimental results were compared to results predicted by the thermal model. In general, the model underpredicts or overpredicts the experimental temperature rise by 0-2 ·C and the agreement is within the experimental uncertainty of ±2°C.