ED Sciences Physiques et de l'Ingénieur
5G mmW integrated Bidirectional TRX for hybrid and digital beamforming system
by Lucien PAQUIEN (Laboratoire de l'Intégration du Matériau au Système)
The defense will take place at 10h00 - M256, Grenoble INP - Phelma, 3 Parv. Louis Néel, 38000 Grenoble
in front of the jury composed of
- Nathalie DELTIMPLE - Professeure - Bordeaux INP - Directeur de these
- Sylvain BOURDEL - Professeur - Grenoble INP - Rapporteur
- Hervé BARTHELEMY - Professeur - Université de Toulon - Rapporteur
- Florence PODEVIN - Professeure - Grenoble INP - Examinateur
- François RIVET - Maître de conférences - Bordeaux INP - Examinateur
- Didier BELOT - Ingénieur - STMicroelectronics - CoDirecteur de these
- Antoine FRAPPE - Professeur - Junia - Examinateur
The increasing demand for data rate for mobile telecommunications has led to the use of beamforming systems in order to notably limit the impact of free space propagation losses (FSPL) over the link budget, due to the elevation of the operating frequency. In order to be able to direct a directional beam concentrating the majority of the gain of the antenna array towards a given user, a large number of integrated radio frequency front-ends (RFFE) is necessary. Conventionally, 5G RFFEs generally consist of a low noise amplifier (LNA), and a power amplifier (PA). The latter are physically dissociated, and are alternatively addressed using a commuted element, in order to operate in time division duplexing (TDD). In this case, not only does the switched element involve losses and a significant silicon surface requirement, but also the RFFEs are only used half the time (due to TDD). Also, this large silicon area required must then be multiplied by the number of elements that constitutes the beamforming system. In addition, the spacing between each antenna constituting the antenna array being proportional to the wavelength, the latter could therefore reach higher operating frequencies if the RFFEs are miniaturized. In this work, a solution allowing the elimination of the need for a commuted element, as well as the merging of the LNA and PA is proposed, inducing a strong reduction in the silicon surface area required for the same operation that conventional architectures, using the GF 22nm CMOS FD-SOI technology. Although the design of millimeter functions (mmW) will be discussed, the frequency conversion aspect as well as the study of baseband functions will also be covered, including the design of a RF passive mixer, two reconfigurable second- and fourth-order active-RC low-pass filters, a variable gain amplifier (VGA), a 50Ω analog buffer, a double pole double throw (DPDT) switch, as well as a generation chain of quadrature signals, done from the combination of a hybrid coupler (HCPLR), and an external off-chip local oscillator (LO). The complete system will be simulated to demonstrate the relevancy of these structures regarding performances and required silicon surface, and axis for improvement will also be listed.
Numerical simulations of confined Brownian motion.
by Elodie MILLAN (Laboratoire Ondes et Matière d'Aquitaine)
The defense will take place at 14h00 - Aphithéatre B 351 Cours de la Libération, Bâtiment A29, 33405 Talence
in front of the jury composed of
- Thomas SALEZ - Chargé de recherche - Université de Bordeaux - Directeur de these
- Micheline ABBAS - Maîtresse de conférences - Université de Toulouse III - Rapporteur
- Hendrik MEYER - Directeur de recherche - Université de Strasbourg - Rapporteur
- Salima RAFAI - Directrice de recherche - Université de Grenoble Alpes - Examinateur
- Thomas BICKEL - Professeur - Université de Bordeaux - Examinateur
Brownian motion is the erratic movement of microscopic particles immersed in a fluid due to the thermal agitation of the surrounding fluid molecules. It is possible to describe the Brownian motion using Langevin's equation. However, close to a wall, a particle moves more slowly because of the hydrodynamic no-slip condition at the wall. As a result, the particle's mobilities and diffusion coefficients, both parallel and perpendicular to the wall, are locally impacted by the confinement and lead to the emergence of a so-called multiplicative noise. Consequently, when confined, Brownian motion is no longer Gaussian. Besides, the latter effect is difficult to observe at all time. During my thesis, I developed numerical simulations, optimized to study efficiently, on broad spatial and temporal windows, Brownian motion confined between rigid walls. In this manuscript, I present in detail the algorithm and the set of optimisation methods for reducing the computation time. I also present the methods for analysing Brownian motion and apply them to the confined case in order to characterize qualitatively and quantitatively the non-Gaussian features of the displacements of a Brownian particle. This work has rendered possible to confirm the theoretical predictions, in particular at long times, which are inaccessible experimentally.