Testing For Analog Circuits Based On Pseudo-Random Current Injection Adopting Distance Classification

The fact that the research of analog system testing and diagnosis has been going slowly is mainly due to the theoretical and technical immaturity as well as an insufficiency of applicable research results. And one of the basic difficulties lies in the complexity of the fault models in the analog system since both the input stimulus and output response are continuous variables, and the parameters in the circuit components are also continuous too.The faults in analog circuits could be categorized as two groups. One is hard fault, which means the stuck-open fault and the bridge fault; another is soft fault, which refers to the component parameters going beyond the preassumed perfomance range. Usually, these faults do not lead to a total failure of equipments. And it is commonly regarded as fault-free when the comparative deviation of component parameters are within±5% of the nominal value, while in practical operations the allowance range could be set flexibly according to different requirements for the circuit performance. Among the many testing methods of analog and mixed signal circuits booming up in recent years, the pseudorandom testing technique is easier to generate testing stimuli and the current-based testing method could magnify the response differences between fault-free circuits and faulty circuits. The advantages of the two methods are combined together in the pseudo-random current injection testing method, which improves the efficiency and correctness of the test, and which is much more operable, thus it is very applicable in built-in-self-testing of analog and mixed signal circuits in system chips. The fault can be diagnosed by Variance Analysis or Distance classification after getting the correlation functions between the input and output of CUT.This paper particularly studied the Variance Analysis and the Distance Classification,and compared their effects in the application of circuit testing. According to the comparison result,the Variance Analysis will not be applicable when the allowance range has to be set flexibly according to specific operation requirements although it can achieve a high rate of fault detection. For example, when the allowance range of some particular component gets bigger, the Variance Analysis could not precisely define the new judgment boundary. While the Distance Classification is more suitable for the case that the component parameter allowance range is usually varied according to specific requirement.Therefore,combining Distance Classification with...