Power Modeling for CMOS Circuits by RBF Network

Low power VLSI circuit design requires fast and accurate power estimation. This paper presents a novel radial-basis function neural network (RBF-NN) power macromodeling technique, which makes a first attempt to apply RBF network in power estimation of CMOS circuits. In contrast to previous modeling approaches, it dose not require empirically construct specialized analytical equations for the power macromodel, and achieves better accuracy by taking into account the statistics of not only the primary inputs, but also the primary outputs. It does not perform any simulation during estimation and yields power estimates within seconds. In experiments with the ISCAS-85 circuits, the average absolute relative error of the macromodel was below 5.0% for most of the circuits. The root-mean-square (RMS) error is about 1 - 3%.