8-Bit Multiplier Simulation Experiments Investigating the Use of Power Supply Transient Signals for the Detection of CMOS Defects

Transient Signal Analysis is a digital device testing method that is based on the analysis of voltage transients at multiple test points. In this paper, the power supply transient signals from simulation experiments on an 8-bit multiplier are analyzed at multiple test points in both the time and frequency domain. Linear regression analysis is used to separate and identify the signal variations introduced by defects and the variations caused by changes in fabrication process parameters. Defects were introduced into the simulation model by adding material (shorts) or removing material (opens) from the layout. Process parameter fluctuations were modeled by randomly varying transistor and circuit parameters individually and in groups over the range of +/-25% of the nominal parameters. The results of the analysis show that it is possible to distinguish between defect-free devices with injected process variation and defective devices. Transient Signal Analysis (TSA) is a parametric approach to testing digital integrated circuits [1][2]. In TSA, defect detection is accomplished by analyzing the transient signals measured at multiple test points of a device. The approach offers two distinct advantages over other logic and parametric testing methods. First, device coupling mechanisms (i.e. power supply) permit the detection of defects at test points that are not directly affected by the defect. Consequently, error observability is greatly enhanced in TSA since they need not be propagated to primary outputs. Second, by cross-correlating the data sampled from multiple test points, false detects caused by mistaking signal variations resulting from process drift as signal variations resulting from defects, are reduced. In fact, all useful parametric test methods must address this problem. The proposed technique works because the effects of process drift tend to be global, changing circuit parameters uniformly across the entire die. Hence, the corresponding change in the transient response of the device produces signal variations that are correlated at all test points on the die. In contrast, signal variations caused by a defect tend to be regional with larger amplitudes at test points closer to the defect site. The RC attenuation effects of the device coupling mechanisms reduce the amplitude of the variation as a function of distance from the defect site. This results in a reduction in the cross-correlation profile of the device. A simple statistical method is presented that detects the absence of correlation in one or more test point signals of defective devices while attenuating the signal variations that are …