New test paradigms for yield and manufacturability - Invited address

Summary form only given, as follows. Test holds the key to success in the rapidly changing world of process technology and design complexity as well as the fashionable area of Design-for-Manufacturability (DFM). As test chips become prohibitively expensive and less statistically valid, it is only through statistical analysis of volume product test data that we can assess the improvements in parametric variation and in defectivity necessary in both the design and the manufacturing processes to meet the yield and supply chain targets for complex ICs. Volume test data is also key to the implementation of adaptive testing. Probabilistic decisionmaking can be applied to the test flows to reduce test costs and improve test quality by specifically targeting the parameters and defects that are likely to cause failures and reduce unnecessary testing and burn-in of defect-free die. Some key paradigms in the test world have to change, however, if test is to keep up with these challenges or test will continue to be relegated to the "non-valueadded" category that has been a long-standing barrier to investment in test equipment and software.