Generating Diverse Test Sets for Multiple Fault Detections based on Fault Cone Partitioning ∗ Stelios Neophytou, Maria K. Michael, Kyriakos Christou KIOS Research Center of Intelligent Systems and Networks and Department of Electrical and Computer Engineering, University of Cyprus, Cyprus {sneophytou,mmichael,christou}@ucy.ac.cy Abstract Testing modeled faults multiple times has been shown to increase the likelihood of a test set to detect non-modeled faults, either static or dynamic, when compared to a single detect test set. Test sets that guarantee detecting every modeled fault with at least n different tests are known as n-detect test sets. Moreover, recent investigations examine how different the various tests for a fault should be, in order to further increase their ability in detecting defects. This work proposes a new test generation methodology for multiple-detect (including n-detect) test sets that increases their diversity in terms of the various fault propagation paths excited by the different tests. Specifically, the various tests per modeled fault are guaranteed to propagate the fault effect via different propagation paths. The proposed method can be applied to any linear, to the circuit size, static or dynamic fault model for multiple fault detections, such as the stuck-at or transition delay fault models, and avoids any path or path segment enumeration. Experimental results show increased numbers of propagation paths and non-modeled fault coverages when compared to traditional n-detect test sets. 1 Introduction The on-going increase in the complexity of the modern VLSI microchips demands more sophisticated post- manufacturing testing methodologies and/or procedures. One approach is to develop complex fault models to imitate defect behavior at either the logic or layout level of abstraction. This is a hard process, mainly because of the vari- ety of the potential defects arising due to technology shrinking. Moreover, detailed layout information is typically not available until the fabrication phase, giving limited information to test engineers. A less demanding approach employs multiple detections per fault, using well established fault models of linear complexity (e.g. stuck-at and transition fault models). The rationale behind detecting a fault more than one time is to achieve higher quality in terms of defect coverage, by generating a number of different tests for each modeled fault. Test sets that detect each fault multiple times or with at least n different tests have been shown to give high non-modeled fault coverage [1, 2, 3, 4, 5, 6, 7, 8]. Most of these ATPG methods for multiple-detect or n-detect test sets concentrate, mainly, on reducing the test generation effort while keeping the overall test set size small. Another important issue addressed in the literature is the diversity of the different generated tests per fault. A test set with diverse tests per targeted modeled fault has been shown to increase the defect and non-modeled fault coverage of the test set [9, 10, 11, 12, 13]. Diversity, i.e., how different are the tests that detect a fault, has been defined in various manners. [13] proposed a definition for sufficiently different tests, in terms of how different certain primary input signal values are with respect to the already generated tests for a fault. The work in [9, 10, 13] introduce measures that quantify the difference between tests detecting the same fault, either based on the internal signal values that excite the fault [9, 13] or propagate the fault to a primary output [10]. In essence, the motivation behind the approach followed by all these methods is to introduce sufficient randomness to input signal values when distinguishing the tests. However, primary inputs can be categorize in two groups: (a) those affecting the targeted fault (either activate the fault or propagate the fault to some primary output) and (b) those which do not affect the targeted fault. While differentiating the values in both these groups (without reducing the n-detect fault coverage) has contribution in the defect coverage it is more important to find tests that activate the targeted fault with different internal signal values and/or propagate the fault effect to some primary output via different propagation paths. The work in [11, 12] considers the physical neighborhood of the fault site and enforces that the different ∗ This work has been partially supported by the Cyprus Research Promotion Foundation. 2009 24th IEEE International Symposium on Defect and Fault Tolerance in VLSI Systems 1550-5774/09 $26.00 © 2009 IEEE DOI 10.1109/DFT.2009.24 401 2009 24th IEEE International Symposium on Defect and Fault Tolerance in VLSI Systems 1550-5774/09 $26.00 © 2009 IEEE DOI 10.1109/DFT.2009.24 401 2009 24th IEEE International Symposium on Defect and Fault Tolerance in VLSI Systems 1550-5774/09 $26.00 © 2009 IEEE DOI 10.1109/DFT.2009.24 401