We proposed a new methodology capable of accurately modeling the partial correlations and geometric dependency in the local random fluctuations of various electrical parameters. This method incorporates principal factor analysis (PFA) into the conventional SPICE-based compact modeling of the mismatch variation which is only focused on the dependency of the variation on device’s geometry and biases. PFA enables one to model correlations among the fluctuations by determining the dominant factors and their weights contributed to each of electrical parameter variations. This new methodology enables a better prediction in both digital and analog circuit performance spread because the Monte Carlo model will cover the comprehensive range of transistor variation without creating unrealistic cases.
Journal: TechConnect Briefs
Volume: Technical Proceedings of the 2005 Workshop on Compact Modeling
Published: May 8, 2005
Pages: 163 - 166
Industry sectors: Advanced Materials & Manufacturing | Sensors, MEMS, Electronics
Topics: Nanoparticle Synthesis & Applications