On-line detection of intermittent faults in digital-to-analog converters

This paper proposes a method to detect intermittent faults in digital-to-analog converters (DACs), whose effects could be either a temporary incorrect analog output value or a temporary analog value out of specifications. The method employs information theory as the theoretical basis and demonstrates a statistical procedure for intermittent fault detection based only on a DAC's behavioral model, not specific internal circuit design. The method is most useful in on-line test during system operations and can generate digital signatures for fault detection. Simulation results are presented to support the fault detection method, with intermittent fault probabilities ranging from very low values to expected values in system operations.