Skip to Main Content
A defects pattern recognition system has been developed for the flip-chip solder joint quality inspection by using laser ultrasound and interferometric techniques. This system extracts error ratio and dominant frequency as features from ultrasound waveforms. It also performs a cluster analysis of those feature vectors by applying probabilistic neural network classification algorithm. The system can automatically classify chips into different clusters and can, therefore, find differences between good and bad chips, as well as classifying the type of defect.