Analog and mixed-signal test and fault diagnosis play an essential role in circuit design, device production, and instrumentation maintenance. The driving forces for this research consist of economic factors such as time/cost consideration, and the fact that analog test and diagnosis lags far behind digital test. The benefits include correcting design flaws, reducing time-to-market, increasing manufacturing yield, and reducing the system cost. Fault diagnosis has three tasks: fault detection to find the faulty systems, fault location to identify the faulty parameters, and parameter evaluation to calculate deviations. The difficulties for analog and mixed-signal test and fault diagnosis are coming from ambiguities, increased complexity, reduced accessibility, lack of effective fault models, and increasing test cost.
In this dissertation, above problems are explored. A verification technique based on the ambiguity group locating technique is developed to address the ambiguity problem. Deviations can be accurately computed and fault location is computationally efficient.
To decrease complexity and increase accessibility, a large scale system is decomposed into smaller subsystems. A restriction on accessibility in traditional decomposition is removed, so that some specific inaccessible nodes can be computed for their nodal voltages.
An ideal switch is used to model catastrophic faults. To locate multiple analog catastrophic faults, an analog stuck fault location approach is designed eliminating repetitive simulation requirement among traditional stuck fault location techniques.
The significance of the dissertation research is that efficient and systematic solutions are provided for analog test and multiple fault diagnosis, which are applicable to the general background analog systems.