As designing technique of the digital system and manufacturing process of integrated circuit are improving, the kinds of faults caused by transistor defect have became a main effect to digital circuit, which bring a strong challenge to the traditional testing methods. A novel fault diagnosis method based on current signal includes IDDQ and IDDT. These new methods adopt the circuit information contained in the power supply line current to realize the fault diagnosis, and they are complementary for traditional testing methods. But the theory of the testing method based on power supply current signal face many problems, it need further research.Firstly, we research the technique of diagnosing digital circuit based on current signal, including logic gate circuit testing, combinational logic circuit testing, and have verified that the method combined IDDQ and IDDT can effectively increase the fault coverage. When we research the technique of diagnosing combinational logic circuit, Fault model is improved to make it coincide with the actual condition. BP neural net work and SVMs are proposed to locate the fault, and the results is analysed. In this paper, we propose the intelligence method combined IDDQ(multi threshold IDDQ) and IDDT to test complex fault in digital circuit. Through the analysis of the simulation results, we prove that the method proposed is an effective way to diagnose complex fault. 94% fault coverage is achieved.Secondly, aiming at the existing problems under VDSM (Very Deep Sub-Micron) technology, the method combined delta-IDDQ and IDDT is presented to test CMOS SRAM and the 0-1 algorithm is also improved for the SRAM test. Compared to the March algorithm, the improved 0-1 Algorithm obviously reduces the test cost. By diagnosing memory circuit with four cells, 100% fault coverage is achieved corresponding to bridge fault, open fault and coupling fault. The test results certify that the new method has high fault coverage and can diagnose some faults which traditional methods with logic testing can not detect. |