Abstract
We have already proposed a method for grouping of relevant and equivalent inputs of a logical circuit under test (LCUT) by use of M-sequence correlation. We call this method as input grouping method. Relevant input implies those inputs which give some influence to the output, and equivalent input means those input pairs for which the output pattern does not change when they are exchanged. Input grouping shows the relation between the inputs and output, so this method is one of the method to represent the logical structure of the LCUT. The relevant inputs and equivalent inputs are calculated from the correlation functions between the input and output, when pseudorandom M-sequences are applied to the LCUT.
On the other hand, the input grouping can also be calculated from a truth table. We call this method as truth table method. The authors showed that the input grouping by use of correlation functions stated above, is far advantageous over the truth table method if we admit small percentage of error. This is due to the fact that the number of correlation functions necessary to calculate input grouping can be considerably reducible, if we admit small percentage of error.
In this paper, relations between the input groupings and fault probabilities of a LCUT are investigated for various cases, and undetected fault ratios of the fault positions by use of this input grouping are calculated. Same calculations were made in case where the input grouping by use of a part of correlation functions are used. These simulations were carried out for various kind of actual logical circuits, and it is shown that the undetected fault ratios of fault positions are sufficiently small even when a part of correlation functions are used.
This method of grouping of relevant and equivalent inputs of LCUT would be very useful for a fault diagnosis of actual logic boards.
