Abstract
Interposers are essential for integrating multiple dies in staked-die products. However, the traditional testing mechanism is not appropriate for interposers. Therefore, in this work, we proposed a contactless testing methodology for pre-bond interposers. Also, to improve the production yield, surface defects and inner defects will be both examined. Our testing mechanism dedicates to detecting defective interposers by thermal images. Critical features will be extracted from thermal images and will be used to construct machine learning algorithms to determine whether the interposer is defective automatically. Experimental results show that our contactless testing mechanism can efficiently improve the production yield of the interposers from average 72.13% to average 96.25%.