On the noise measurements and modeling for on wafer HEMTs up to 26.5 GHz

Abstract

In conclusion, we put forward a four-channel fiber ring-down pressure sensor with temperature compensation based on neural networks. Using time-division multiplexing principle, this sensor consists of four fiber ring-down loops whose structure is characterized by a combination of series and parallel connection. Its feasibility has been proved by experiment, which shows its good performance both for time-division multiplexing detection and for removing disturbance of temperature. In addition, considering multimode fiber dispersion in optical fiber, we have also discussed the maximum number of multiplexed channels available in this multiplexing system. The idea of the four-channel fiber ring-down pressure sensor with neural networks makes ring-down detection technology more practical and efficient and can be used in other fields, such as physiological processes involving several biological parameters.

GaN HEMT noise modeling based on 50‐Ω noise factor

“…Thus, the extraction of the noise model is reduced to the determination of only the temperature associated to the intrinsic output resistance T ds , as the other resistance temperatures are selected to be equal to the ambient temperature, coherently with their physics‐based contribution to thermal noise. The reported experimental results show that, in agreement with previously published studies for GaAs and Si transistors , the 50‐Ω noise factor F 50 of the tested GaN HEMT exhibits a behavior as a function of the squared frequency that can be approximated by a first‐order fitting. This observation is of great importance because the most appropriate value of T ds can be successfully determined to enable the noise model to mimic the detected behavior of F 50 .…”

Microwave noise modeling of FinFETs

“…Finally, a first order fitting is used for obtaining a smoother F 50 as a function of the square of the frequency (see Fig. 2) [4,6,8,10]. Fig.…”

“…This device has 60 nm gate length and 45.6 lm gate width, which consists of 50 fingers covering 6 fins each. The procedure for obtaining the NF 50 of the device under test starts by measuring with an E8975A NFA (Noise Figure Analyzer) the NF 50 of the whole system consisting of the transistor, the input and the output stages (i.e., adapters, bias tees, cables, and probe tips) [8]. Then the S-parameters of the DUT, the input and the output stages are measured by using an E8364A PNA (Precision Network Analyzer).…”

“…In particular, the determination of the noise parameters has been simplified by exploiting the noise model based on expanding the small signal equivalent circuit with two equivalent noise temperatures, which are associated to the intrinsic input and output resistances [5,9]. Furthermore, Felgentreff et al have presented a resistor noise model taking into account also the equivalent noise temperature associated to the feedback resistance [8,[10][11][12]. In the present paper, we extend this resistor noise model to the FinFET case by accounting also for the equivalent noise temperature associated to the substrate resistance.…”

“…Consequently, several studies have been addressed to overcome these drawbacks by extracting the noise parameters from the measurements of the NF 50 , which represents the noise figure with a 50 X source impedance, without the need of any tuner [4][5][6][7][8]. In particular, the determination of the noise parameters has been simplified by exploiting the noise model based on expanding the small signal equivalent circuit with two equivalent noise temperatures, which are associated to the intrinsic input and output resistances [5,9].…”