Noise-rejection model based on charge-transfer equation for digital CMOS circuits

Noise-rejection curve (NRC) is given by minimum noise amplitude that causes failure of digital cell as a function of noise width. It is more accurate noise failure criteria than traditional static noise margin (NM) and it becomes widely used in modern crosstalk aware designs. However, the accuracy of the NRC-based analysis is often insufficient because of inconsistencies between characterization, data representation, and modeling of the NRC. In part, it can be attributed to ambiguous definition of noise width and height. This paper presents a new noise-rejection model and describes special characterization of digital circuits to achieve high accuracy. This new model is founded on a charge-transfer equation and constrains an integral measure of the noise, rather that its amplitude and width. It provides a compact data representation and is independent on NM.