Abstract
Capacitance-coupling and mutual inductance between the neighboring wires of global interconnects give rise to crosstalk effects, which are one of the biggest signal integrity problems in DSM circuits today. Previous models for crosstalk-induced faults assume that all the wires of a bus act together to induce crosstalk effects on a single wire. Based on recent simulation results of Sirisaengtaksin and Gupta, we use a more general model of crosstalk-induced faults that allows tradeoffs between efficiency of tests and quality of tests. We construct provably optimal test patterns that covers all crosstalk-induced faults under this general model. Our test patterns admit simple generators for at-speed testing in BIST. The test methodology proposed here can result in huge savings in test cost. In particular, the required linear number of test patterns can be reduced to a constant number of test patterns by accepting only a few percent of error, thus allowing our test methodology to scale with bus width and technology.