Modeling uncertainty in RC timing analysis

A method is presented for modeling the effects of uncertainty in RC analysis. Representing uncertain parameters as intervals, interval algebra was used to create a rigorous framework for manipulating these uncertain values. The authors then modified an existing RC analysis algorithm (Crystal´s PR-slope model) to illustrate their approach. Although Crystal was chosen for the experiments, the same techniques may be applied to other timing analysis algorithms. Compared to Monte Carlo simulation, the interval algorithm is much more efficient, operating several thousand times faster for the same quality results. For stages with few transistors, the accuracy of both methods is similar: both come within 10% of the true values.<>