ON THE ORIGIN OF 1/F NOISE IN MOSFETS

Abstract

Often the 1/f noise in MOSFETs is stated to be an ensemble of many RTS with different time constants. The majority of literature on 1/f noise is overlooking the contribution due to mobility fluctuations that are uncorrelated with number fluctuations. Here, we demonstrate that the so-called proofs for Δ N can also be obtained from the empirical relation. The following misunderstandings and controversial topics on the surface and bulk contributions to the low-frequency noise will be addressed: 1) 1/f and RTS noise can have different physical origins. An analysis in time domain shows that the low-frequency noise with RTS is nothing else than a superposition of a two level noise with a Lorentzian spectrum and a Gaussian noise with a pure 1/f spectrum and different bias dependency. 2) It is very unlikely that in a spectrum consisting of one strong two level RTS and a pure 1/f noise, the 1/f noise is a superposition of many RTS with different time constants. 3) The spreading in WLS I /I2 below a critical WL is not a proof for the Δ N origin. 4) The typical shape in the double log plot from sub threshold to strong inversion of S I/ I 2 versus I , is also not a proof for the Δ N origin.