Page 9 of 12

Conclusion

In this paper we described the challenges faced by Intel in continuing with functional test as the primary mechanism for screening manufacturing defects, and we examined structural test as an alternative. Three major test quality risks were identified in migrating to structural test:

• Reduced test data volume due to the inefficiencies in loading test patterns from a structural tester.
• The loss of collateral defect coverage provided by functional tests that are applied at speed in the normal functional mode of operation.
• Sub-micron trends indicating that interconnect defects such as bridges and opens will dominate the defect distribution. Simulation results were presented that indicate that an increasing number of defects will result in speed failures rather than hard failures, requiring alternate ways of generating test patterns.

Defect-based test was introduced as an approach to mitigate some of these risks by increasing the effectiveness of ATPG-generated vectors. While this approach is intuitively appealing, it poses formidable challenges. Little hard evidence is available on the effectiveness of such an approach. Fault models that represent defect behavior well, and are tractable from a test generation viewpoint, have to be developed. Tools for the enumeration of likely fault sites and for test generation tools with the new fault models need to be implemented.

The technology development strategy for DBT was presented as an evolutionary cycle that builds on prototype capabilities and uses strategic partnerships with design teams. Silicon data collected from these experiments are used to refine and validate fault models and the tooling collateral as they are developed.

The tooling challenges for defect-based test for large, high-performance designs were discussed. Commercial capabilities that exist today are either insufficient, or cannot be scaled to meet the needs of next-generation microprocessor designs. These challenges span the design flow from logical to physical design, and they will require a concerted effort by the CAD industry to make defect-based test a robust, scalable solution.