On a new outlier rejection technique

This paper presents a novel approach to reject outliers based on parametric measurements, namely quiescent current and IC speed estimates. Unlike the existing multi-parameter or I-DDQ-based techniques which use additional measurements to estimate the expected healthy I-DDQ response, we focus on the variance to reject outliers. We take advantage of the correlation between the two types of measurements and of the fact that speed estimate measurements are more reliable than leakage ones in order to evaluate the acceptable current variation level. For that, we derive approximate variance equations from basic current and delay equations. Using these variance equations, we are able to identify current variations that cannot be justified by normal process variations, based on simple outlier rejection criteria. This approach is developed to be used as a pre-processing tool to other post-processing techniques. Results are promising.