A variable observation time method for testing delay faults

Test methodologies for delay faults usually observe output patterns at a single observation time, and the same observation time is used for all faults in the circuit under test. It is shown that use of a single observation time is not advantageous for testing delay faults, and the detection threshold can be dramatically improved by using a testing methodology that allows variable, fault-dependent, and output-dependent observation times. A 'waveform-type' simulation method is used for calculating detection thresholds for definitely detectable faults. Statistical distributions of delay fault detection thresholds are presented for ten benchmark circuits.<
>