A method for the estimation of defect detection probability of analog/RF defect-oriented tests
The goals of this paper are (1) to improve the effectiveness of IDDQ testing, and (2) to find the characteristic current signature for every defect class. Current signature analysis was performed for the IDDQ data collected on the Murphy test chip. A "total variance" method is proposed to reduce the test escape of IDDQ testing. Compared to the other three IDDQ testing methods, it has the lowest test escape and the highest yield loss. However, there are still 7.5% nonfunctional CUTs which could not be detected by any IDDQ testing method. This result shows that it is not possible to replace Boolean tests by IDDQ testing. The distributions of six current signature types over six different classes are analyzed. The results show that "big-step" is the dominant signature type among all defect classes.