Determining accuracy bounds for simulation-based switching activity estimation

Discrete simulation-based methods for switching density estimation can provide very accurate results, but some unresolved problems such as determining the required number of input patterns to guarantee a final accuracy impede the usefulness of this technique. In this paper, we present novel solutions for the following issues in switching density estimation: 1) derivation of simulation stopping criteria based on run-time information, and 2) a priori estimation of the number of input patterns required to guarantee a final specified accuracy. These problems are solved through the definition of a set of multinomial random variables and functions based on the parameters of these random variables. Experimental results indicate that the proposed techniques deliver a reliable stopping criteria as well as accurate a priori estimation of the number of required input patterns.