9A: Embedded Tutorial: Test with Variations - How Much Can Be Solved in the Design Process? [Breaker page]

Increased process variations as well as other design parametric variations have given many new challenges in test from various perspectives. Some of these challenges are test-specific and some of them can be design-specific. Following the old saying that ??if a problem can be resolved in design, it should not be resolved in test,?? it is important for the test people to first understand what variations-induced problems can be, or are expected to be resolved by models and analyses. Then, the remaining variations-induced problems may have no choice but be resolved after the first silicon becomes available, where then test needs to get involved. One can think that a problem contains a test aspect if solving it requires the involvement of test patterns, silicon measurement, and/or test equipment. The test side of the variations-induced problems can be divided into two parts: (1) There are new problems to solve after the 1st silicon and before considering mass production. Here, we would be trying to resolve problems left in the design and hence, can be viewed as an extension to the design closure effort. (2) Variations also affect the developments of test strategies for mass production. Developing an effective mass production test strategy would be based on how the design process is carried out and how phase (1) is accomplished. This embedded tutorial brings together two leading experts in design tools and modeling to discuss the issues induced by variations. The two speakers will describe how variations are characterized, modeled and analyzed in the design process, address the limitations of these methods, and point out the potential problems that may require the involvement of test to resolve them.