Abstract
The concept of model-based test was developed in order to reduce the production test effort for data converters (Cherubal and Chatterjee (IEEE Trans Circuits Syst part I 50(3):317---327, 2003); Stenbakken and Souders (1985) Modelling and test point selection for data converter testing. In: ITC, Int Test Conf, pp 813---817; Wegener and Kennedy (IEEE Trans Circuits Syst I 51(1):213---217, 2004); Wrixon and Kennedy (IEEE Trans Instrum Meas IM-48(5):978---985, 1999)). In applying this concept, a vector of model parameters is determined for each device under test (DUT). Typically, this model parameter vector is merely used to calculate the DUT performance characteristic which is then subject to specification-oriented testing. However, each element of the model parameter vector represents an independent error source which contributes to performance degradations; thus, the model parameter vector can be viewed as a signature of the error sources. In this work, analyzing the error source signature is used to devise a model-based methodology for hard-fault detection and diagnosis. We investigate conditions under which hard-faults are detectable/diagnosable in spite of masking effects due to manufacturing process variations. In particular, we show that taking the model parameter vector as the fault signature is optimal as it minimizes the masking effects and thus maximizes detectability/diagnosibility.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have