Accuracy bounds in switching activity estimation

Calculation of switching densities in digital ICs is necessary for both power and reliability estimation. Discrete simulation-based methods are popular since they can produce accurate estimates of switching density. However, a number of unresolved problems such as estimating the required number of input patterns hinder this technique. In this paper, we address the issue of obtaining an a-priori estimate for the number of simulation patterns required for a specified accuracy. By approximating each signal as a binomial random variable and using the normal approximation to the binomial distribution, we obtain bounds on the number of required patterns. Experimental data shows that our model produces a tight bound on the minimum number of required patterns