Reducing Aging Impacts in Digital Sensors via Run-Time Calibration

Hazards or intentional perturbations must be identified in safety- and security-critical applications. Digital sensors have been shown to an appealing approach detect such abnormalities. However, as any sensor technology, digital are prone mis-calibration. In particular, even if the initial calibration is correct, rate of false missed alarms might increase when aged. this paper, we thoroughly study impact aging-induced alarms. Indeed aging relates usage time, a priori model (historical data for environmental variation) predicting unrealistic tracking time with related temperature voltage variation imposes high overhead. Accordingly, propose alternative where not one but two deployed. practice, used deviations, while second reference. respect, only operated seldom, mostly re-calibrate active From dual input (unaged aged sensor), corrective models derived. We account methods, namely simple effective offset correction, adjustment based on machine-learning. conduct extensive characterizations (both pre-silicon simulations post-silicon measurements FPGA) which quantitatively confirm applicability sensitivity sensors.