Evaluating thermal characterization accuracy using CFD codes-a package level benchmark study of IcePak/sup TM/ and Flotherm/sup R/

A benchmark thermal study was conducted using the IcePak/sup TM/ (ver. 2.01) and Flotherm/sup R/ (ver. 1.4) computational fluid dynamics (CFD) software packages. Both CFD applications were used to simulate the thermal performance of a 20-lead heatsink small outline package (20HSOP) in steady state natural convection environment. Five 20HSOP devices were tested while mounted on a Semiconductor Equipment and Materials (SEMI) standard FR4 printed circuit board absent of any thermal enhancements and containing the minimum amount of copper traces needed for electrical connections. Power dissipations were varied to span a regime of radiation and natural convection heat transfer environments. Due to the symmetric nature of the problem, only half of the package/PCB was modeled. Each of the models utilized identical material thermal characteristics, dimensions, and boundary conditions. Simulation results were validated using experimental data at each device power dissipation. The simulated results obtained from both CFD tools agreed well with measured data (within 10%). However, in the majority of the experiments, IcePak computed junction and case temperatures which were slightly more accurate than those calculated by the Flotherm software, while also maintaining decreased run times. Model details are discussed along with the perceived advantages and disadvantages of the two CFD software packages.