부산대학교 도서관

We are witnessing the growth and needs for solutions allowing the assembly of optical components, together on their sub-mounts. The solder quality, the yield and the throughput are often more important than the ultimate accuracy that the equipment can provide. We foresee a technological limitation: It is not efficient to work with fiducials in this field, it is always better to align the (sub)device, using one or several of its active features. In this paper, we aim to show how a smart teaching of the vision software allows reliable and accurate recognition. We also emphasis that a high-quality machine produces in-situ and real time information that can prevent destructive control in many cases. Those two topics are explored with post-bonding sub-micron accuracy need mindset, and we show how this target allows to understand better the production yield topic.