Multivariate normal distribution based statistical timing analysis using global projection and local expansion

This paper employs general multivariate normal distribution to develop a new efficient statistical timing analysis methodology. The paper presents the theoretical framework of the statistical timing analysis technique by introducing two strategies: the global projection and the local expansion rules. By using the global projection rule, the large problem size occurring due to the delay correlations among a large number of gates across a chip caused by widely varying global sources, is sufficiently reduced. By using the local expansion rule, the arrival time correlations caused by re-convergent paths in circuits are efficiently evaluated. The experimental results show that the computed mean and the standard deviation values of the maximum circuit delay estimated by the proposed method have average errors of 0.2% and 2.9%, respectively.