To improve the quality of test sets for path delay faults, we describe a variation of n-detection test generation based on the concept of m-tuple detection. An m-tuple test set for a set of path delay faults P includes a test for every m-tuple of faults defined over P . Such a test set has the following advantages. (1) Similar to an n-detection test set, it results in several tests for every path delay fault p ∊ P , increasing the likelihood of testing p under worst-case delay conditions. (2) It increases the likelihood of accidentally detecting non-target path delay faults, which are not included in P , more effectively than a conventional n -detection test set. Experimental results demonstrate that m-tuple test generation for m = 2 is manageable in terms of test set size and run time increase relative to conventional test generation.