Dynamic properties testing of solders and modeling of electronic packages subjected to drop impact

The solder joints reliability of electronic packages during board level drop impact is a great concern for semiconductor manufacturers. Many researchers have adopted numerical simulation to investigate the drop performance of electronic package because it is fast and cost-effective. However, the solder balls, which are recognized as vital parts for the integrity of solder joints and the overall function of package, were always assumed to be elastic or simple elastoplastic in previous studies, this is not sufficient to capture the true mechanical behavior of solder balls during drop impact. To obtain more accurate results in numerical simulation, the strain rate dependent material properties of solders are further studied in this paper and measured using Split Hopkinson Pressure Bar (SHPB) technique. The measured stress-strain data at different strain rates are then incorporated in Johnson-Cook model which can adequately represent the rate-sensitive deformation response during high-rate loading. The accuracy of this material model is evaluated by performing FEM simulation of the SHPB test and comparing the finite element simulation with test results. The drop reliability of a fine pitch BGA package is simulated using validated input acceleration (Input-G) method with Johnson-Cook model for solder balls. It is found that the failure mode and critical solder ball location predicted by this modeling correlate well with test. Comparing with traditional elastic or elastoplastic model for solder balls, this material model has better correlation with experimental measurement of dynamic strain, PCB center deflection. Therefore, the stress/strain distribution in solder ball is more precise. At last, an impact life prediction model based on maximum peeling stress of critical solder joints is proposed for board level drop test to estimate the number of drops to failure. With this model, the drop performance of new packages can be quantified, and further enhanced through modelin- - g.