When Oscillators Stop Oscillating

Abstract

Diagnosing analog systems, i.e . systems for which physical quantities vary over time in a continuous range is , in itself, a difficult problem. Analog electronic circuits, especially those with feedback loops, raise new difficulties. Oscillators, a special case of such circuits, are even more difficult to diagnose. This is because most faults lead to the same type of symptom: the oscillator just stops oscillating, the different electric signals remaining hopelessly flat! This paper shows how a program implementing a model-based diagnosis theory can be used to diagnose oscillators. We frrst present CATS, an ATMS-Iike domainindependent conflict rccognthon engine, adapted for diagnosing continuous timevarying physical systems. Its original feature is that it works on data that arc arrays of numeric intervals so as to deal with values that change with time ·and are inaccurate. ll1cn we show how to use this program for diagnosing oscillators, pointing out the advantage of using external stimulation. Experimental results arc given through the example of an astable multivibrator.