Monitoring wire bonding via time-frequency analysis of horn vibration

In this paper, a technique is developed to monitor a wire-bonding process. Vibration of the horn of a bonder is picked up by a piezoelectric material patch attaching on the surface of the horn and analyzed by joint time-frequency analysis. With neural network, features extracted from the time-frequency distributions of the horn vibration are then exploited for classifying bonding machine setting and predicting bond quality. The results demonstrate the advantages of the time-frequency analysis of horn vibration in monitoring wire bonding.