A Technique for Constructing RTS Noise Model Based on Statistical Analysis

Low frequency noise such as random telegraph signal (R TS) noise is more and more important as device shrinks down. A significant part of the high frequency phase noise is generated due to up-converted low frequency noise. RF CMOS circuit performance and the device reliability will be limited and negatively impacted. Therefore, a noise model which accurately predicts the noise characteristics of deep-submicron devices is crucial for the low noise RFIC design. In this paper, the RTS noise parameters, i.e., pulse amplitude, pulse width and pulse delay, are obtained from the Schottky diode under different biases, and a technique for constructing the RTS noise model based on the statistical analysis of these noise data is introduced. From the comparisons of the noise data distributions, it is shown that the difference between the modeled RTS and measured RTS is quite small, and this model is useful in generating noise data because it is closely matched to the measured noise.