• Improved channel thermal noise model incorporating the parasitic sourcee/drain resistances of submicron MOSFETs. • Extraction of accurate noise parameters such as S id , R n , NF min , G opt , and B opt. • Application of the proposed model in high-frequency amplifier design. • Analysis of the impact of the parasitic source/drain resistances in circuit noise performance under various bias conditions. For sub-micron metal oxide semiconductor field effect transistors (MOSFETs), parasitic series source/drain resistance has a significant impact on channel thermal noise (S id ) and noise parameters. In this work, we propose an improved analytical channel thermal noise model considering parasitic resistance, based on physical thermal noise models of sub-micron intrinsic MOSFETs. To validate the proposed model, measurements were performed at room temperature (25°C) on nMOSFETs fabricated in a commercial 130-nm (0.13-µm) bulk RF CMOS technology. All RF S-parameter and noise measurements were conducted on-wafer at room temperature, with open/short de-embedding applied to accurately remove pads and interconnect parasitics. The model was calibrated by extracting parameters in a spice with the standard BSIM4 model as a baseline and validated against measured data such as S id , R n , NF min , G opt , and B opt . Furthermore, the proposed model is extended to a circuit-level analysis by deriving the noise figure of a high-frequency amplifier (HFA) using Cadence Virtuoso (Spectre). A good agreement between the measurement and the developed model is observed, particularly under high gate bias (V gs ) conditions where the potential drop at the parasitic resistance becomes apparent. The analysis demonstrates that accurate modeling of parasitic resistance is essential for predicting the accurate noise figure of the HFA in high-current regimes. The improved model predicts the thermal noise of both the extrinsic MOS device and the HFA circuit well, thereby supporting accurate noise simulations for high-frequency circuits that operate under a wide range of gate bias conditions.
Jeong et al. (2026) studied this question.