We introduce an accurate CMOS IC Reliability Statistics Tool (IC - REST 2) which predicts device failure statistics using not only devices-under-test (DUT) tune-to-fail data but also time-evolution-to-fail data. This greatly reduces DUT sample size and hence IC time-to-market; The tool combines TCAD simulation, short-time tests and novel solutions to the SiO2 and Si-SiO2 defect generation rate equations (in time and space) under static and dynamic operation. The predicted failure statistics are best described by the generalized Γ-distribution (or more accurately the generalized B-distribution) and on a Gumbel plot look more Weibull than Log-normal . |