Accurate global & local circuit leakage current analysis based on design of experiment method

The paper introduces an efficient statistical methodology for an accurate estimation of the static leakage current in a circuit, while taking into account both local and global process variations. The methodology is based on Design of Experiment method. It consists in characterizing the standard cell libarary with response surface models or equations for modeling the electrical circuit behavior. Statistical simulations are then performed on the equations for estimating the overall leakage current performances of the circuit. The method has been validated on a 16 bit adder in 65 nm bulk technology. In so doing, we achieved a faster simulation run time (times 4) compared to traditional Monte Carlo analysis, with simulation errors of less than 5%.