Machine Learning based Scan Chain Diagnosis for Double Faults

The importance of scan chain diagnosis for improving yield is increasing as the complexity of semiconductor circuits increases. The conventional scan chain diagnosis methods provide sufficient diagnostic accuracy for a single fault and double faults. However, if the backward fault in double faults masks the other fault during the load, it is difficult to obtain sufficiency accuracy because fails in good chains are masked. To solve this problem, machine learning based scan chain diagnosis for double faults is proposed. Experimental results show that the proposed method achieves improved diagnostic accuracy for double faults.