Experimental validation of numerical heat transfer predictions for singleand multi-component printed circuit boards in natural convection environments

Increasing power densities and changing component design have increased the need for accurate prediction of temperature effects that may affect system performance or reliability. To highlight these aspects early in the product development cycle, designers resort to using computational fluid dynamics (CFD) based numerical predictive tools. However, users acknowledge that these predictions require experimental verification which is now readily available during the early design phase. Therefore, a need exists to establish well-defined benchmark test cases to help establish confidence in both modelling methodology and numerical tools. This paper presents such information for three package types (SO16, TSOP48, and PQFP208) which are evaluated on single and multi-component PCBs. Benchmark criteria are based on the prediction of steady state component junction temperature and associated component-PCB surface temperature gradients, which are both compared with experimental measurements. While the detailed numerical models typically predicted junction temperature to within 4/spl deg/C, discrepancies as great as 9/spl deg/C were also recorded. The sensitivity of prediction accuracy was assessed against discretization level and both the thermal conductivity and geometry of package materials. Hence it was considered important that all experimental and numerical modelling details be provided for reference.