This paper studies the typical failure modes and failure mechanisms of non-wetting in an FCBGA（flip chip ball grid array） assembly.We have identified that the residual lead and tin oxide layer on the surface of the die bumps as the primary contributor to non-wetting between die bumps and substrate bumps during the chipattach reflow process.Experiments with bump reflow parameters revealed that an optimized reflow dwell time and H_2 flow rate in the reflow oven can significantly reduce the amount of lead and tin oxides on the surface of the die bumps,thereby reducing the non-wetting failure rate by about 90%.Both failure analysis results and mass production data validate the non-wetting failure mechanisms identified by this study.As a result of the reflow process optimization,the failure rate associated with non-wetting is significantly reduced,which further saves manufacturing cost and increases capacity utilization.
A novel framework of hyper-heuristic algorithm was proposed to improve the adaption of evolutionary algorithms( EAs)in optimization. The algorithm could be changed during the evolutionary progress according to their performances. In addition,a large number of elite individuals were employed in the algorithm and the elite individuals helped algorithm achieve a better performance,while such number of elite individuals stagnated the global convergence in conventional single algorithm. The time complexity was analyzed to demonstrate the novel framework did not increase the time complexity. The simulation results indicate that the proposed framework outperforms any single algorithm that composes the framework.
