Improving the solder joint reliability of VQFN packages

2005 
Solder joint reliability during temperature cycling on board (TCOB) is a critical issue not only for BGA but also for quad flat non-leaded (QFN) type packages. On the latter, only a few studies are available in the literature. In this work a parameter study based on Robert Darveaux's model for BGA solder fatigue life prediction was done using full 3D FEA modeling with ANSYSreg. The target was to identify the most sensitive parameters. Subsequently Darveaux's model is modified by calibrating experimentally derived model constants to TCOB test results of different leaded VQFN (very thin QFN) packages. The simulations predict highest plastic work in the solder at the lead side of a corner lead. This crack location corresponds well with experimental findings. The most remarkable result is the very high sensitivity of predicted solder fatigue life on the CTE of the PCB: just 2 ppm/K variation of CTE result in 100% effect on fatigue life. Formulating parameter effects in positive direction only, other parameters with high effect (50-100%) on solder fatigue life are a high die/die-pad length ratio, a large solder contact area, and of course a reduced temperature cycling profile from -40/125degC to 0/100degC. Moderate effects of 20%-50% can be achieved with stiffness reductions on both sides of the solder joint, i.e. thinner die, thinner PCB, lower Young's modulus of the mold compound, reduced number or thickness of Cu-layers in the PCB. An increased solder joint height and the existence of a solder fillet have also a positive effect. Finally a correlation of experimental data and simulation results with TCOB test results of ten different VQFN-packages is performed to improve the quantitative prediction accuracy of the simulation methodology. Here after "recalibrating" the constants of Darveaux's fatigue model all simulation and test results agree within a 2times accuracy range, which is considered acceptable
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    6
    References
    8
    Citations
    NaN
    KQI
    []