Statistical test: A new paradigm to improve test effectiveness & efficiency

A fundamentally new concept of test has been steadily developing and gaining acceptance in the semiconductor industry. This concept is to apply the statistical methods that have long been used to analyze yield off-line, and to control fabrication processes on-line, to the making of product test decisions. Whereas traditional semiconductor manufacturing test has compared measurements to predefined limits as they are made in a static sequence, statistical test looks for outliers from patterns of measurements after they accumulate and even alters the intensity and sequence of testing to most efficiently find the defective units to the level required to meet the customer´s quality expectations. Thus statistical test adapts both the test limits and test flow to optimize the tradeoff between outgoing quality and cost of test. Statistical test has been shown to improve both effectiveness, the fraction of defects a test can detect, and efficiency, the cost of testing to find these defects.