A Self-Trained, Low-Complexity Method for Detecting Faults in Analog Circuits
Vassilios D. Vassios, Argyrios T. Hatzopoulos, Ioannis G. Intzes, Kyriakos Tsiakmakis, Dimitrios Papakostas · IEEE Transactions on Instrumentation and Measurement · 2025
This work presents a low-complexity fault detection technique for analog circuits. In this technique, the transient in the power supply current is measured and the decomposition to its mean and rms components is utilized. The classifier signatures are produced from specific points in the rms and mean transients of the power supply current. These points create a reliable detection region/library using the standard deviation in which all other circuits under test are compared to. The main feature of the algorithm is that it is self-trained. With every circuit that is tested, the classification boundaries are updated with the use of recursive calculations of the mean and standard deviation. This eliminates the need for a large number of simulations and/or training runs. The algorithm was tested in simulations and in real circuits with the use of microcontroller-based automated test equipment (ATE) to demonstrate the simplicity and speed of the method. The algorithm was tested on different test bench analog circuits and the results are presented.