Finite Element Based Solder Joint Fatigue Life Predictions for MLPQ Packages

Abstract

The solder joint fatigue life of flip chips and ball grid array packages has been extensively studied and published. However, there are only a few published papers available on the solder joint reliability of a non-leaded, surface mounted package such as the MLPQ (Micro Leadframe Quad Package). In this paper, a non linear Darveaux's solder joint fatigue model has been used to predict the fatigue life of the MLPQ package under accelerated temperature cycling conditions. The primary concern is the modeling method that can closely approximate the real solder fatigue life. Since void formation in the solder joint is one of the many critical factors that can compromise reliability in the end system they inhabit and it has been an uncontrolled factor in the assembly process, void should be considered in the physical modeling of the solder joint. Another factor that well known to have great effect on the solder fatigue life is whether the exposed pad solder down to the PCB or not. Hence, both the effects of void in solder joint and exposed pad soldered down to PCB will be the main focus in the present study. The percentage deviation of the cycles to first failure and characteristic fatigue life obtained from FE models against the experimental data are computed. It was shown that all the predicted absolute values were too optimistic with over-estimated longer lives and the smallest divergence from the real solder joint fatigue life is within 50 to 100% that was achieved by modeling with void in the solder joint and without exposed pad soldered down to PCB. Both void in the solder joint and without exposed pad soldered down to PCB were proven to reduce the fatigue life. The findings also show that a big void laid at the crack propagation path of the solder joint has resulted in a larger drop of fatigue life than without exposed pad soldered down to PCB. From the identified model with solder joint fatigue life closest to the actual, further effort has been done by physically removing the solder joint fillet and the results of FE analyses were again compared with the experimental results. The predicted absolute values were conservative and the divergence were now successfully minimized to solely 20 to 60%. The results indicate that modeling of the solder joint without solder fillet has helped to save the computational time due to reduced number of non linear elements, facilitates the ease of model building besides produced a well accepted prediction that is conservative and closer to experimental data. In summary, the damage model without solder fillet that incorporated with two defective features of void in the solder joint and without exposed pad soldered down to PCB was a preferred choice in an anticipation of the closest approximation to the experimental tested solder joint fatigue life of MLPQs. After all, the modeling approach for the MLPQs is ultimately established that applicable for the solder joint life predictions.