Improved monitoring of ultrasonic wire bonding via input electrical impedance

The technology of using input electrical impedance to monitor a process of ultrasonic wire bonding has existed for a few decades. From literature it is seen that the waveforms of "impedance" in these methods were detected only approximately and much information was missing. In this paper, the method of detecting electrical impedance is improved so that the true waveforms of both the real and imaginary part of the input impedance of a wire bonder are detected and used to monitor bond quality and machine operation condition in-situ and real-time. In order to automate the monitoring, 25 features of the waveforms were selected and fed into a 3 layer back propagation artificial neural network which provides condition indicators after proper training. Since the input impedance characterizes the behavior of a dynamic system completely, once accurate measurement is made, it allows accurate prediction of bond strength. Comparison with results obtained from off-line standard shear tests well demonstrates this capability. When trained to identify operation conditions, the proposed system also identifies the drifting of operation parameters quite accurately