CFD-based iterative methodology for modeling natural convection in microelectronic packages

In order to predict and manage the thermal behavior of microelectronic packages cooled principally by natural convection, a two-way coupling method was developed between a accurate conduction model and a computational fluid dynamics (CFD) simulation. An iterative solution loop was performed by applying as a boundary condition the local convection coefficients obtained from the CFD to the conduction model, and the temperature field at the solid-fluid interface obtained from the solid conduction model to CFD simulation. For the first iteration, the conduction heat transfer was solved by considering an initial guess of a uniform convection coefficient (e.g. from empirical formulas) for each solid-fluid interface. The procedure was repeated until the convergence of the solution was reached. The comparison between the total convection coefficients including the radiation obtained with the CFD iterative procedure and those from the conventional empirical methods showed important differences, thus demonstrating the usefulness of the CFD approach to obtain accurate thermal results. The numerical results were compared to measurements from a test vehicle, carried out under the natural convection Jedec JESD51 standards in a still air chamber. Radiation plays an important role in heat transfer in this setting because of the large temperature difference between the walls of the box and the test vehicle. The junction temperatures obtained with the numerical simulation for different operating powers were in good agreement with the experimental measurements, with an error of less than 1°C, showing that the proposed methodology allows the accurate simulation of the natural heat transfer for microelectronic packages mounted on PCB in the horizontal orientation.