Statistical Timing Analysis using Levelized Covariance Propagation

Variability in process parameters is making accurate timing analysis of nano-scale integrated circuits an extremely challenging task. In this paper, we propose a new algorithm for statistical timing analysis using Levelized Covariance Propagation (LCP). The algorithm simultaneously considers the impact of random placement of dopants (which makes every transistor in a die independent in terms of threshold voltage) and the spatial correlation of the process parameters such as channel length, transistor width and oxide thickness due to the intra-die variations. It also considers the signal correlation due to reconvergent paths in the circuit. Results on several benchmark circuits in 70nm technology show an average of 0.21% and 1.07% errors in mean and the standard deviation, respectively, in timing analysis using the proposed technique compared to the Monte-Carlo analysis.