A variability-based analysis technique revealing physical mechanisms of MOSFET low-frequency noise

This paper presents a technique for statistical analysis of MOSFET low-frequency noise (LFN) based on the autocorrelation coefficient of numerous LFN power spectral density (PSD) spectra. This correlation analysis reveals information about physical mechanisms behind 1/ƒ noise that is difficult to obtain otherwise. The methodology also enables quantification and validation of old and new statistical LFN models. Numerous PSD spectra from contemporary mixed-signal CMOS technologies were analysed; the results reinforce the idea that the 1/ƒ noise in MOSFETs has a trapping-related origin.