Testing an integrated circuit for what it should not do (abstract only)
Charles Kapps · 1985
Integrated circuits are designed to perform some intended function. This is true not only of the circuit as a whole, but also of all of the functional parts that make up the circuit. The general philosophy of testing is to provide some sequence of inputs and test to see if the outputs reflect the intended function. Little notice is given to the possibility that the circuit may have produced undesirable side effects. For example, a processor that is given a “load accumulator” instruction may indeed load data into the accumulator exactly as desired, but it may also incorrectly change the contents of some other register. The problem with testing for this condition is that the number of possibilities may be very large and the time required for exhaustive testing may become excessive. Furthermore, it may become difficult to identify the meaningful possibilities to check for. This paper presents a case study for problems of this kind and shows how the problems encountered defy recognition by normal logic simulation and test generation techniques. Extended methods of test generation are proposed for identifying problems of this type.