COFEST: A Power Based Testing Methdology

This study was done to determine the efficacy of power-based testing. Power-based testing refers to observing power consumption inorder to detect faults in the circuits.

The presence of faults alters the toggling behavior in a circuit when excited by input stimuli. The toggling behavior refers to the charging and discharging of the output capacitors, which is inturn indicative of the power dissipated by the circuit. The presence of faults affects the toggling behavior by blocking the switching activity (For ex. in case of a SA0 fault) and sometimes increases toggling activity. This behavior is seen as a potential oppurtunity for detecting faults. A fault is said to be detected if it causes appreciable differance in the power consumption between a faulty and a fault-free circuit.

The tool is a COcurrent Fault and Energy SimulaTor, i.e., it can simulate 32 faulty machines parallely and estimate the energy consumed by them. For calculating the energy, a unit-capacitance, unit-delay model was chosen. The tool was tested on 5 ISCAS benchmarks.
The tool gave mixed results. It was inferred from the results that the input stimuli is critical for the sucess of this testing scheme. The weak point of the tool in its present form, is the way input vectors are generated. This aspect is being worked upon.

If you want to know more about this testing scheme, you can read the section titled "Approach taken for the project" in the report. Download the report by clicking here