Measurement of power supply currents was found to be very usefulnfor testing CMOS IC's because of its potential to detect a large classnof manufacturing defects. However, this technique was used mainly fornfault detection and was confined to digital circuits. In this paper, wenpresent a suited methodology for fault diagnosis of analog circuitsnbased on the observation of power supply currents. In the proposedntechnique, fault signature dictionaries are generated from the currentsnin the power supply bus. To obtain signatures rich in information fornefficient diagnosis, the transistors in the circuit are forced tonoperate in all possible regions of operation by using a ramp signal atnthe supply instead of the conventional constant DC signal or groundnvoltage. The signatures are then clustered into different groups using anKohonen neural network classifier. This technique has the potential tondetect and diagnose single and multiple shorts as well as open circuits.nThe theoretical and experimental results of the proposed technique arenverified using a CMOS Operational Transconductance Amplifier (OTA)ncircuit
展开▼
