Thermal performance of an air-cooled plastic ball grid array package in natural and forced convection

Thermal performance of a three chip plastic ball grid array (PBGA) package is experimentally investigated under natural and forced convection conditions using thermal test dice. This is done to validate a mechanistic predictive methodology for PBGA package thermal analysis. A four layer, wire-bonded, 35 mm square PBGA with dice encapsulated in a single mold compound block is used as the test vehicle. The package is attached to a 1.524 mm thick, four-layer FR4 PWB with two internal copper planes. The following ranges of parameters are covered in this study: ambient temperature (T_a), room temperature, 105°C, and 125°C for natural convection, and 0.25-2.5 m/s free stream velocities for forced convection. The room temperature experiments are performed in a wind tunnel and the high temperature experiments in an oven. In addition to die temperatures, temperatures at several locations on both package and PWB are also measured. Experimental data is also obtained to verify a linear superposition method for predicting die temperatures when multiple heat sources are present. Also, simulations of the experimental set-up using methods of computational fluid dynamics (CFD) are carried out using a mechanistic approach, proposed and implemented earlier, to predict PBGA package thermal performance. From a comparison between experimental data and CFD analyses for natural and forced convection, it is concluded that the mechanistic approach predicts die temperatures to within 12% of experimental measurements. The linear superposition method predicts the experimental data to within 10% and in most cases within 5%