Wafer-Scale VLSI Testing

1 Introduction In the manufacturing of VLSI integrated circuits (ICs), it would be desirable that the ICs undergo an accurate test directly on the wafer, before they are cut and packaged. This would save the cost of bounding and packaging the faulty ICs, which may be a relatively large fraction. However, accurate testing on the wafer can hardly be supported by the consolidated technology of testing machines. For this reason, the test of ICs on the wafer is usually limited to providing power, ground and a few signals, and to checking some basic electrical properties. This results in a test with limited coverage, which is unable to identify all faulty ICs. The ICs which pass the preliminary test (which may include a relatively large fraction of faulty ICs) must undergo a more accurate test and are possibly discarded after they have been bounded and packaged, an this may contribute heavily to the overall cost of the manufacturing process. Another drawback is that the test on the wafer proceeds sequentially, since the test machine must step over individual IC, and this implies that the time needed to test the entire wafer is quite long. A different approach to the functional manufacturing test of ICs on the wafer can be based upon concepts inherited from the V\VWHPOHYHOOGLDJQRVLVVWKHRU\ [1], which was introduced to diagnose systems composed by units able to perform mutual test. A test involves two units and, in principle, proceeds as follows: the WHVWLQJJXQLW provides a test input sequence to the WHVWHGGXQLW, which returns an output sequence. The testing unit compares the latter sequence with the expected output sequence and defines a WHVWWRXWFRPH, which is binary: 0 if the test passes and 1 otherwise. The set of all test outcomes (called V\QGURPH) is input to a controlling computer (called GLDJQRVHU), which executes a GLDJQRVLVV DOJRULWKP aimed at identifying the faulty units. The diagnosis algorithm partitions the set of all units into set. of units declared good, set) of units declared faulty, and set 6 of units declared VXVSHFW, i.e. those units which the algorithm is unable to classify as either good or faulty. The diagnosis is said to be FRUUHFW if units assigned to set. are actually fault-free and those assigned to set) are actually faulty. The diagnosis is said to be FRPSOHWH if every unit is classified as either good or faulty, i.e. set 6 is empty. A well-known diagnosis …