On generation of high quality tests for defect detection and diagnosis.

摘要

Defect detection and defect diagnosis are two fundamental goals in VLSI testing. In this thesis, we present methods to generate tests of high quality for defect detection and defect diagnosis, and to order test patterns for higher diagnosis resolution when the fail data is truncated.; Defects in deep sub-micron VLSI circuits are known to be predominantly opens and bridges. We propose a uniform test generation method to model interconnect opens and bridges using constrained multiple line stuck-at fault model. A novel feature of the proposed fault model is its flexibility to accommodate increasing levels of accuracy. Additionally the model does not require accurate device level circuit models to achieve desired accuracy.; Redundant logic identification and removal are useful for logic optimization and testability enhancement. Based on our work in the modeling of multiple line stuck-at faults, we present new procedures for identifying redundant stuck-at faults including redundant multiple line stuck-at faults on the branches of fan-out stems. The methods proposed include new procedures to identify stuck-at faults that are simultaneously redundant thus allowing simultaneous removal of logic associated with several redundant faults.; Defect diagnosis plays a crucial role in silicon debug, field return analysis, yield learning and process improvement. We investigate the use of n-distinguishing test sets, which distinguish pairs of single stuck-at faults n times, to enhance the probability of distinguishing unmodeled defects. Independent of the specifics of the diagnosis procedure used, n-distinguishing test sets intrinsically offer enhanced defect diagnosis resolution. In addition, it is necessary for defect diagnosis to consider the situations where the fail data available to the diagnosis tool is truncated. We propose a test pattern ordering algorithm for defect diagnosis with truncated fail data. Higher diagnosis resolution can be achieved with the test set appropriately ordered using the proposed algorithm.