Method Based On The Current Digital Integrated Circuit Failure Diagnosis

With the development of micro-electronics technology, the requirement of integrated circuit fault diagnosis is increasingly imminent. Based on the technology of IDDQ and IDDT, this paper studies the method of fault diagnosis in circuit by applying theory of wavelet analysis and classification in pattern recognition.The traditional methods of voltage testing have been mature in basic researches. And it has been widely used in practice. But this method can't detect some kinds of faults under some high performance ICs. For improving the fault coverage, a novel fault diagnosis technology including IDDQ and IDDT has recently developed rapidly. These new methods can adopt the circuit information contained in the power supply line current to realize the fault diagnosis. In this paper, the ability of integrated circuit fault diagnosis which is executed with IDDQ or IDDT testing technology is verified through Matlab and Pspice simulation on circuits with defects based on summarizing and improving the advantages and disadvantages of all the available methods.Concerning the IDDQ testing, the bridge fault of integrated circuit is detected by analyzing the fault information of the quiescent power supply current. Simulation result shows that the detected fault coverage is limited by IDDQ testing. In IDDT testing, some kinds of faults which can't be detected by conventional logic testing and IDDQ testing can be correctly detected. However, the faults should be accurately located by signal processing. In fault diagnosis, features are extracted from the circuit outputs together with fault classes by wavelet analysis, and then these faults can be located based on nearest-neighbor algorithm and connective pattern recognition algorithm. Considering these simulation results,we can conclude that these algorithms are effective for integrated circuit fault diagnosis and the simulation result of connective pattern recognition algorithm is more accurate than that of nearest-neighbor algorithm. Finally, application of neural networks to analog fault diagnosis is studied using wavelet transform as preprocessors based on IDDT information of the circuit.From the simulation result, we can see that our neural network can classify faulty components.