Van Brandt, Léopold
[UCL]
Flandre, Denis
[UCL]
Raskin, Jean-Pierre
[ELEN]
Electrical noise is defined as a randomly time-fluctuating electrical signal. In large-area MOSFETs, the stochastic component of the drain current is dominated by the flicker ($1/f$) noise in low frequency, and by the white Gaussian thermal noise in high frequency. These two noise sources are nowadays well incorporated into circuit simulators (i.e. SPICE), which allows analog designers to perform noise analysis both in transient and AC. In $28 anometer$ technology and beyond, a granular non-Gaussian low-frequency noise arises from the presence of a few traps within the MOSFET gate oxide: the emph{random telegraph noise} (RTN). Its amplitude is found to increase as the inverse of the gate area. Study and modelling of this noise get thus substantially motivated by the use of advanced CMOS technologies in VLSI design. This work aims to derive an AC compact modelling of the RTN, and to discuss the cointegration of such a model with existing models for 1/f noise. The analytical expression of the RTN power spectral density is exactly known from stochastic process considerations. One goal of this work is to validate this expression trough spectral analysis based on experimental measurement in the time domain. The use of the analytical expression is only possible if accurate estimations of the RTN parameters are available. Thus, we propose a comprehensive methodology for both time-domain and spectral analysis of the RTN. A robust software tool has been implemented in Matlab extsuperscript{ extregistered}. Recommendations for a suitable choice of the measurement parameters such as the sampling frequency and the time window are made, in a large-scale characterization perspective.

Bibliographic reference |
Van Brandt, Léopold. *AC compact modelling of low-frequency noise for 28nm CMOS technology and beyond.* Ecole polytechnique de Louvain, Université catholique de Louvain, 2017. Prom. : Flandre, Denis ; Raskin, Jean-Pierre. |

Permanent URL |
http://hdl.handle.net/2078.1/thesis:10725 |