Possibility of VLSI board fault diagnosis using adaptive image restoration to the thermography

At present, VLSI chips have increased their number of input/output pins, therefore standard methods of fault diagnosis, which uses limited input/output pins, has become difficult. This research establishes a new concept diagnosis method which does not use input/output pins but apply the restoration thermography. This paper compares three methods standard method, Histogram method and This method, after this, we confirmed that this method is superior to the other two methods. Consequently, this system has verified the possibility of fault IC diagnosis by thermography restoration using the neural network which was constructed by the little learning data (134). Using this method, a 96.43 % fault diagnosis rate was obtained. As for this reason, this paper´s neural network which inputs the difference image using learning data of IC coordinate (IC position) with gradation value (IC heat generation temperature) satisfactorily performed for the fault IC image restoration. Moreover, by adding the addition emphasis process, the fault diagnosis rate progressed 1.72 % more than the nonaddition method.