A new approach to mixed-signal diagnosis

The authors present a unique approach to diagnosing mixed-signal circuits down to the failing component or node. This approach uses an innovative artificial intelligence technique called model-based reasoning. Previous attempts to automate mixed-signal diagnosis have used component reliability data, node voltage comparisons, or fault dictionaries. Systems built on these approaches are inaccurate, slow, and very costly to develop and maintain. In contrast, the model-based reasoning approach is based on a thorough understanding of the electronic behavior of circuit components. A model-based system that requires minimal setup time and accurately diagnoses mixed-signal circuits has been built. It is concluded that model-based diagnosis resolves traditionally hard diagnostic problems, including feedback loops, unavailability of current measurements, and infinite symptom-fault relationships. Model-based diagnostic systems are easily configured for new circuits and can be easily integrated with external instrumentation and testers