Statistics in Semiconductor Test: Going beyond Yield

Semiconductor test has evolved from simply screening individual units to a data-intensive manufacturing operation which enables decisions going far beyond “pass versus fail.” The quantity and complexity of data generated at each test manufacturing step, and indeed in the entire test manufacturing flow, make processing the data into a useful form a daunting task. Many fields have experienced similar explosive growth in data volume and also use statistical methods to understand and predict outcomes. In addition to developing new techniques test should exploit applicable statistical methods from any field such as agriculture or genetics to make test decisions, optimize test flows, and guide what test data should be acquired. Although statistics is a big subject, a small set of methods outlined in this paper are a solid start to the foundation of statistical test.