An accurate statistical analysis of random dopant induced variability in 140,000 13nm MOSFETs

Abstract

Intrinsic variability of transistor parameters has become a major concern for next generation CMOS technologies[1]. On modern giga-transistor chips, devices occurences more than 60σ from the mean have become important in design. It is thus necessary to have detailed information about the distribution of transistor. parameters such as threshold voltage, including the distant tails, which can be used to accurately predict design margins. Previously published simulation results have tended to analyse only small statistical samples, typically of 200 devices[2],[5]. In order to make detailed predictions of the distribution of random dopant induced threshold voltage fluctuations, a significantly l..