A Study of Flicker Noise in MOS Transistor Under Switched Bias Condition

with K, γ, constants, and γ ≈ 1. It is quite well accepted that the sources of low frequency noise are mainly carrier number fluctuations due to random trapping–detrapping of carriers in energy states, named ‘traps’, near the surface of the semiconductor. From some time ago, switched biasing has been proposed as a technique for reducing the flicker noise itself in MOSFET’s [1]. An intuitive explanation of the phenomenon is that periodically turning ‘off’ the transistor’s channel, periodically forces a significant fraction of occupied traps to a known empty state, thus introducing some ‘order’ in the random process. A switched MOSFET flicker noise PSD resembles the plot in Figure 1.b [2,13]. Usual 1/f spectrum is seen at frequencies greater than the switching frequency. At lower frequencies the noise (log scale) increases with a much smaller slope. Finally at an even lower frequency, the slope resembles again the original 1/f spectrum. Several authors proposed models to explain this particular behavior [3,4,12] however, the exact mechanism and the statistics of the switched noise current, are not yet clear. Particularly, reported models [3] predict a plateau at lowest frequencies that do not correctly address experimental results [2,13]. The model presented in [12] is not simulation based as [3,4] and shows a different approach at this problem. The goal of this paper is to discuss in detail flicker noise in a switched MOS transistor. Let us first examine the DC bias case: consider a MOS transistor, and a small channel element of differential area dA = W.dx as in Figure 1.a. Defects inside and at the surface of the oxide generate localized states (traps with energy Et), which may be occupied by carriers from the channel. Electrons (and holes) in the channel may tunnel to, and back from, these traps in a random process thus generating a noise current. N ' A [m-2] will denote the number of occupied traps per unit area in the whole oxide volume above the channel element dA. The relation between the carrier densities in the channel named N ' [m-2], and N ' A is given by the Reimbold’s coefficient r [5][10]. To find the drain current noise, the impact on ID of local N ' fluctuations is integrated along the channel [5,6]. Thus a physics based flicker noise model should begin finding an expression for SN'A (f)(the PSD of N ' A). This paper is organized as follows: in section II, an explicit analytic deduction of SN'A (f) (non-switched transistor) is presented using SRH statistics and the autocorrelation formalism. In section III the study is extended using simulations to examine the switched bias flicker noise. ABSTRACT