Relating random telegraph signal noise in metal-oxide-semiconductor transistors to interface trap energy distribution

Abstract

In this work, we study random telegraph signal ͑RTS͒ noise in metal-oxide-semiconductor field effect transistors when the device is periodically and rapidly cycled between an "on" and an "off" bias state. We derive the effective RTS time constants for this case using Shockley-Read-Hall statistics applied under transient conditions. In this way, we show that the oft-observed reduction in RTS noise under such bias conditions can be explained by a nonuniform ͑e.g., U-shaped͒ distribution in energy of interface traps.

1/f noise simulation in MOSFETs under cyclo-stationary conditions using SPICE simulator

“…However, when bias conditions shift significantly during circuit operation, as in cyclo-stationary operation, noise simulation may not accurately estimate circuit noise. The works of [1][2][3][4][5] presented experimental evidence of flicker noise reduction when the transistor is cycled between strong inversion and accumulation-additional to that expected from noise modulation-and which is not accounted for in the BSIM4 equation model. the premise that the trap density in the device is not energy independent, as proposed by [7], but rather U-shaped [4,10].…”

“…However, in order to determine N t from E f n , a trap distribution in terms of the energy, f (E Fn ), is required. Based on the observations from [4,10], this distribution is assumed U-shaped. It is important to notice that this distribution is an input parameter of the method and should be provided in the model card of the device.…”

“…This behavior, which has been attributed to the distribution of traps in energy [4] has led to the proposition of noise reducing techniques, such as transistor switching [5] and the application of a forward body bias (FBB), also known as forward substrate bias (FSB), during the low transition of the input signal [6]. From a designers perspective, the development of compact models for noise simulation that account for these noise reducing techniques is crucial to properly simulate a circuit.…”

Low-Frequency Noise Phenomena in Switched MOSFETs

“…Instead, the occupancy varies between and . Substituting and (8) and equating and , we may derive an expression for and [41]. If the switching frequency is made very high compared to the RTS corner frequency ( much smaller than and ), and converge to the same value, , and the RTS becomes stationary.…”

“…As will be shown, this means the effective RTS time constants and will change. The derivation below is given in more detail in [28] and [41].…”