Generation of compact test sets with high defect coverage

Abstract

Abstract-Multi-detect (N-detect) testing suffers from the drawback that its test length grows linearly with N. We present a new method to generate compact test sets that provide high defect coverage. The proposed technique makes judicious use of a new pattern-quality metric based on the concept of output deviations. We select the most effective patterns from a large N-detect pattern repository, and guarantee a small test set as well as complete stuck-at coverage. Simulation results for benchmark circuits show that with a compact, 1-detect stuck-at test set, the proposed method provides considerably higher transition-fault coverage and coverage ramp-up compared to another recently-published method. Moreover, in all cases, the proposed method either outperforms or is as effective as the competing approach in terms of bridging-fault coverage and the surrogate BCE+ metric. In many cases, higher transition-fault coverage is obtained than much larger N-detect test sets for several values of N. Finally, our results provide the insight that, instead of using N-detect testing with as large N as possible, it is more efficient to combine the output deviations metric with multi-detect testing to get highquality, compact test sets.

Defect aware X-filling for low-power scan testing

“…This metric is an advanced version of the quality metric proposed in [9] as it evaluates the defect coverage potential of a test vector using concurrently both its first and its second test response (we consider LOC scheme). Thus the proposed technique targets both timing-dependent and timing-independent defects at the same time.…”

“…After the selection of vector v, the rest vectors of set CS(t) are discarded from set CS and the weights wo 0 (k,0) for all k∈MS 0 (v,0), wo 0 (k,1) for all k∈MS 0 (v,1), wo 1 (k,0) for all k∈MS 1 (v,0) and wo 1 (k,1) for all k∈MS 1 (v,1), are divided by a constant factor F 2 (these outputs are expected to detect many defects, after the application of test vector v, and thus they are considered as less effective for the selection of the next vectors). As proposed in [9], the value of F 2 was set equal to 8. Then, the new weights WT(v) are calculated for all remaining vectors v, and the next vector is selected.…”

Defect Coverage-Driven Window-Based Test Compression

“…This is also the case for the more efficient output deviation-based metric that was proposed in [6]. Moreover, both metrics in [6] and [18] evaluate test vectors for either timingindependent or timing-dependent defects, whereas the proposed metric improves the detection of both timing-related and timing-independent defects at the same time.…”

Generation of Compact Stuck-At Test Sets Targeting Unmodeled Defects

“…In addition, it evaluates test patterns for both timingdependent and timing-independent unmodeled defects at the same time. Thus it is more effective than the metric proposed in an earlier version of this paper presented in [9], which generates different test sets to target each kind of these defects.…”