Applying Heat Transfer Coefficient Data to Electronics Cooling

Many electronics cooling applications involve direct air cooling of the components and their heat transfer performance is often estimated, in the design phase, using heat transfer coefficient data from either in-house data or published sources. The agreement between predicted and observed system performance is quite often not acceptable, which means that the cooling system must be developed empirically—an expensive and time-consuming process. The authors believe they have identified one important reason for this state of affairs: the heat transfer coefficient values presented in the literature are being used improperly. Almost all of the published heat transfer coefficient data are from single-active-component experiments, which implicitly define h based on the adiabatic temperature of the component (h_ad), while most users assume h to be defined on the basis of the mean fluid temperature (h_m). This misunderstanding leads to underpredictions of the temperature rise of the components by 20–30 percent or more. This paper reviews the options for defining the heat transfer coefficient, shows how the problem arises, and then describes the steps necessary to properly use the existing heat transfer coefficient data base. There are two options for applying the existing data base to a fully powered array: One can either calculate the adiabatic temperatures of the components from the known heat release distribution in the array or calculate the values of h_m that can be used with T_m, again using the known heat release distribution. The two options represent different ways to apply the same general method, superposition. The superposition method of Sellars, et al., (1956) is adapted to discrete systems and used as a guide to the forms recommended.