Sensitivity of power dissipation to uncertainties in primary input specification

Abstract

To accurately estimate power dissipation, the exact signal probability (probability of a signal being logic ONE) and signal activity (probability of signal switching) of primary inputs are assumed to be known. In general, very accurate specification of primary input signal probability and activity may not be available. This in turn may result in uncertainties in average power estimation. In this paper we present a novel and efficient technique to estimate the sensitivity of average power dissipation to input signals using a symbolic estimation technique. Results for benchmark circuits show that power sensitivities can vary widely for different primary inputs of a circuit. Hence, in order to accurately estimate average power dissipation, the sensitive inputs of a circuit have to be specified accurately. We have also developed a Monte-Carlo based technique to estimate power sensitivity which also acts as a figure of merit for the symbolic technique.