 | Research Summary
Knowledge regarding the cause and the location of failures is essential to
the improvement of the design and manufacturing process. Once the circuit
under test is determined to be faulty, diagnosis is performed to locate
the fault on the circuit. The design and manufacturing process may be
improved by the rapid identification of the cause of the failure.
Diagnosis utilizes the observed responses of the circuit under test; the
subsequent analysis of the gathered data helps pinpoint the cause of
failure. Typically, locating the cause of the failure necessitates the
comparison of the observed failure data to the simulated failures of the
modeled faults.
The main challenges of the diagnosis process can be divided into three
different groups. First, since the simulated responses of the modeled
faults are used during the diagnosis, the storage of this diagnostic
information constitutes a problem. Second, the transfer of test response
data from the circuit outputs to the diagnosis tool may necessitate a high
bandwidth. Third, since modeled faults do not always represent actual
defects, it may not be possible to find a match between the observed
failures and the behavior of modeled faults. The diagnosis of the
un-modeled faults subsequently requires an intensive research.
Our research focuses on the aforementioned challenges of the diagnosis
process and we develop methods to deal with the challenges. Essentially,
the methods are proposed to reduce the size of the diagnostic data and the
bandwidth required for the transfer of the test responses while preserving
the diagnostic resolution. Furthermore, the methods for the diagnosis of
the un-modeled faults with simple fault models are developed.
|
