Self-learning machines applied to the testing of semiconductor memories

The concept of self-learning testers is explored with reference to the detection of certain classes of soft failures. An appropriately organized machine acquires testing "know-how", while exercising memories with a virtually endless sequence of pseudo random test vectors. Verification and sorting of failures is straightforward, but imaginative compression of the acquired information is mandatory. The efficacy of the tester is, indeed, predicated on the notion of the "time trace" of a sequence of test vectors.