EXPERIMENTAL VALIDATION OF NUMERICAL HEAT TRANSFER PREDICTIONS FOR SINGLE- AND MULTI-COMPONENT PRINTED CIRCUIT BOARDS IN A FORCED CONVECTION ENVIRONMENT: PART 2 - RESULTS AND DISCUSSION

The need to validate the prediction accuracy of numerical Computational Fluid Dynamic (CFD) tools used for the thermal design of electronic systems has been established in the accompanying Part I publication, and a pragmatic approach of applying experimental data to achieve this was outlined [Rodgers et al., 1999a]. In addition, the experimental characterisation process and numerical modelling approach used to generate the results presented and discussed in this paper were also defined. In this study numerical predictions are assessed against corresponding experimental measurements from three, board-mounted thermal test components (SO16, TSOP 48 and PQFP 208), exposed to a 2 m/s airflow parallel to the Printed Circuit Board (PCB) in-plane. Benchmark criteria used to validate prediction accuracy were based on the parameters of die junction temperature and component-PCB surface temperature gradients. The analysis of surface temperature profiles on the singlecomponent PCBs revealed a tendency of the CFD code to overpredict convective heat transfer immediately downstream of the component. Despite this, best prediction accuracy was obtained on the singlecomponent test PCBs, with worse case discrepancies within 4°C. For the three test cases it was shown that prediction accuracy was not sensitive to flow model selection, whereas a combination of aerodynamic effects and different PCB thermal boundary conditions resulted in high sensitivity on the multi-component PCB. Analysis of the experimental junction temperature data from the multi-component PCB allowed the aerodynamically sensitive regions of the PCB to be identified. Not surprisingly, the greatest prediction discrepancy and variation between the flow models also occurred in these regions. This clearly demonstrated that the applied flow model must be capable of capturing the specific physical phenomena generated by ones application. Overall, the rules governing the application of a laminar or turbulence model are not clear and the results obtained highlight why this is so.