Committee:
     PRESIDENT: SEGOVIA VARGAS, DANIEL
     SECRETARI: BROQUETAS IBARS, ANTONI
     VOCAL: BOLOMEY, JEAN CHARLES
Thesis abstract: The use of microwave imaging techniques in medical applications has gained significant interest because of its non-ionizing character, low cost, and body penetrability. Different techniques and systems for microwave imaging have been successfully developed and tested for the diagnosis and monitoring of different parts of the human body to detect certain diseases by means of determining their complex permittivity distribution representing the electrical biological condition. This thesis proposes to open a new avenue adding to the previous biological condition imaging capabilities some trace of its biological activity. With this goal, a novel microwave technique oriented to locate and detect the low-frequency signals approaching the functional activity of a certain biological organ. These electrical signals, responsible for the different biological phenomena, are normally identified as action or membrane potential. Due to their low-frequency character (f ≃ 1 kHz), they stay mostly confined inside the human body and different kinds of contact electrodes are normally employed to measure them. In this work, we propose a novel technique to extract (from the inside of the human body) and wirelessly monitor these low-frequency signals (generated by a number of basic microorganisms as cells or bacteria mimicking the activity of the organ under study) based on the use of an illuminating microwave signal. Differently from the low-frequencies, the microwave signals are able to propagate inside (in and out) of the human body, being focused (with a Ultra-Wideband (UWB) - 0.5 to 2.5 GHz- multi-probe geometry) on a specific zone, and eventually becoming affected (modulated) by the low-frequency functional biological signals. Once extracted (de-modulated) from the interrogative microwave signal, the information of the action potential signal representing the functional activity, is spatially located inside the region of interest using a microwave differential imaging technique. In this work it has been specifically studied, the capability of monitoring localized action potential signals in the brain representing a Parkinson’s disease (PD), characterized by electrical signals generated in the subthalamic nucleus (STN) and the globus pallidus (GPi) with specific action potential patterns. Three main points have been developed: * Design of an optimized imaging geometry consisting of four (sequentially Tx and Rx) UWB Extended Gap Ridge Horn (EGRH) antennas disposed on two orthogonal sets.* Design of an UWB-Pulse Amplitude Modulation (PAM) monitoring system producing a nsec interrogation pulse * Modeling and validation of the UWB Microwave Functional Brain Activity Extraction for Parkinson’s Disease Monitoring concept.

Committee:
     PRESIDENT: AGUADO AGELET, FERNANDO
     SECRETARI: GOMEZ MONTENEGRO, CARLOS
     VOCAL: TALONE, MARCO
Thesis abstract: Natural disasters suppose a risk to life and assets, and can benefit from better early detection and monitoring systems. Current monitoring systems include in-situ instruments located along the Earth's land, such as the coast lines, or in the oceans. However, the monitoring needs are not met with the current solutions, due to the latency when retrieving the data. Other sensors are located in land, the latency may be negligible provided they are connected to terrestrial networks, but the warning is given when the disaster is already happening in populated areas. Aside from in-situ instruments, also satellite Earth Observation (EO) payloads monitor the extent and risk areas affected by natural disasters. These payloads generate large amounts of data, and it is often impossible to download it through a Ground Station, imposing duty cycle limitations in the executions.The appearance of the 5G paradigm, and the definition of the Non-Terrestrial Networks, has boosted space-to-Earth communications using Internet of Things (IoT) technologies. These systems can contribute to natural disasters early detection and monitoring by placing IoT ground nodes on the Earth's surface, since they are inexpensive to deploy, and can contain different types of sensors depending on the variables to be monitored. Moreover, the ground nodes can provide constant readings, and forward them to a satellite if an anomalous reading is detected, to wake-up the EO payload to perform a continuous monitoring. This paradigm is one of the main contribution of this thesis, called the On-Demand Satellite Payload Execution strategy.The feasibility of this monitoring paradigm is first analyzed by presenting the monitoring needs, defining the requirements in terms of density of ground nodes, update frequency of the measurements, and types of sensors required. After that, a more in depth work related to IoT space-to-Earth communications is conducted, assuming that the Long Range (LoRa) modulation is used. The physical layer is first studied by computing a link budget, considering different payload architectures, assessing the communication limitations for each of them. Then, the effect that ionosphere scintillation has on the overall throughput is evaluated by a set of tests. Following, the Media Access Control (MAC) layer is studied by identifying the optimum protocols for IoT space-to-Earth communications. Then, the maximum density of ground nodes for each protocol, and for different antenna footprint sizes is computed, based on the monitoring requirements defined previously for natural disasters monitoring. Moreover, a set of proof-of-concept experiments are conducted to simulate the scenarios identified in the MAC layer theoretical study. This encompasses the comparison of two different protocols, identifying the limitations of each one. Finally, the design and software development of a Software Defined Radio (SDR)-based LoRa payload is presented. Another contribution of this thesis is related to the antenna design for small satellites and CubeSat payloads. Starting with two different antennas used for satellite communications. The first one being an IoT LoRa patch antenna for the RITA payload. Then, the COMET B2 Probe Inter-Satellite Link (ISL) antenna is presented, this study comprises first a sensitivity study based on simulations on the optimum antenna geometry. Then, an additional analysis is done simulating the optimum position of the final Commercial Off The Shelf (COTS) antenna used for the mission. Afterwards, the requirements definition, theoretical design, simulations, prototyping, and flight model testing and results are presented for the 3Cat-4 L-Band Helix Antenna and for the FSSCat Nadir Antenna. Both of this antenna are for a Global Navigation Satellite Systems Reflectometer (GNSS-R) and Microwave Radiometer (MWR) payload. Finally, a sensitivity analysis of the optimal positioning the L5 Nadir Antenna, for a GNSS-R payload is presented.

Committee:
     PRESIDENT: ZHU, ANDING
     SECRETARI: COLLADO GOMEZ, JUAN CARLOS
     VOCAL: BECERRA GONZÁLEZ, JUAN ANTONIO
Thesis abstract: With the increasing demand for linearity and efficiency in radio frequency (RF) power amplifiers(PAs). This thesis focuses on the development and implementation of digital predistortion (DPD)techniques for linearizing various high-efficiency PA architectures, including envelope tracking (ET)PAs, outphasing PAs, and Pseudo-Doherty load-modulated balanced amplifiers (PD-LMBAs). Themain objective is to achieve an optimal efficiency-linearity trade-off while addressing the challengesposed by modern communication standards, such as dynamic signal bandwidth and frequencyallocation in 5G New Radio (NR). The thesis introduces novel shaping functions and DPD algorithmstailored for multiple-input-single-output (MISO) PAs. For ET PAs, novel envelope optimizationtechniques are presented for better linearizing the ET PA when operating with 200 MHz widebandsignals for handsets. A novel shaping function extraction method is developed for outphasing PAs,enabling the operation of current mode outphasing (CMOP) PAs with wideband signals. For PD-LMBAs, a frequency-domain shaping function and an artificial neural network (ANN)-based modelare adopted to address the challenging nonlinearity, along with a hardware-in-the-loop pruningmethod to reduce the computational complexity.To cope with the dynamic signal bandwidth and frequency location requirements in mobileapplications, an elastic DPD is proposed. It incorporates frequency-dependent polynomials andincremental-bandwidth techniques, along with a modified doubly orthogonal matching pursuit(DOMP) algorithm, to adapt to varying operating conditions and non-flat frequency responses. Theobjective is to create low complexity DPD models suitable for the implementation in handsets. Thethesis also explores the implementation of DPD algorithms on both field-programmable gate array(FPGA) and graphics processing unit (GPU) platforms. An FPGA implementation of the ET PAlinearizer based on the envelope leakage cancellation and generalized memory polynomial ispresented as a case study. GPU-based implementations of conventional GMP and NN-based DPDmodels are also evaluated.The contributions of this thesis advance the state-of-the-art in DPD techniques for high-efficiencyRF PAs, enabling the realization of highly linear and efficient wireless communication systems. Theproposed methods and implementations provide valuable insights for future research anddevelopment in the field of PA linearization.

Committee:
     PRESIDENT: ESPINOSA, HUGO
     SECRETARI: BROQUETAS IBARS, ANTONI
     VOCAL: RAMÍREZ ARROYAVE, GERMÁN AUGUSTO
Thesis abstract: This doctoral thesis is made up of three research papers that have been published in high-ranking Q1 journals. The papers delveinto the innovative integration of a MIMO radar system, signal processing, and machine learning (ML) in applications related tothe monitoring of human dysfunctional breathing, detection of stress levels in drivers, and detection of pain through facialexpressions. The first study, titled “In-Cabin MIMO Radar System for Human Dysfunctional Breathing Detection,”introduces anovel approach using MIMO radar systems for remote, wireless health monitoring. It specifically focuses on the detection ofdysfunctional breathing patterns in humans. The system showcases its capability to remotely identify and track patterns such ashyperventilation and dominant thoracic breathing syndrome. This technology could replace traditional methods that often requirecumbersome and intrusive portable sensors, and is non-invasive and non-obstructive in the context of vehicle cabs.In the secondresearch, Supervised Machine Learning-Assisted Driving Stress Monitoring MIMO Radar System explores the intersection ofMIMO radar systems and machine learning (ML) in the context of vehicle safety. This study demonstrates how supervised MLalgorithms can be trained to interpret physiological signals captured by a MIMO radar system to accurately assess driver stresslevels. The results demonstrate a sleep detection accuracy of 90\%, alertness of 96%, and anxiety of 85%. This research couldcontribute significantly to the development of smarter and safer vehicles that can adapt to the driver’s emotional state, potentiallyreducing the risk of stress-related accidents and improving the overall road safety.The third research, 3D pain facial expressionrecognition using an ML-MIMO radar profiler, represents a significant advance in pain assessment techniques. This researchsuccessfully applies MIMO radar technology to recognize 3D-based facial expressions as indicative of pain. This non-invasiveand objective method compares the measurements of the MIMO radar system with those of the facial action coding system,revealing an accuracy of 92%. This technology is useful in medical settings where self-evaluation is not possible or is not reliable,such as in intensive care units or pediatric care.These three innovative studies using MIMO radar systems and machine learningfor health monitoring, vehicle safety, and pain assessment offer significant improvements over traditional techniques,demonstrating the versatility and potential of these technologies, and opening new avenues for future research.

Committee:
     PRESIDENT: BOAG, AMIR
     SECRETARI: HELDRING, ALEXANDER
     VOCAL: VIPIANA, FRANCESCA
Thesis abstract: The objective of this thesis is to increase the computational efficiency of integral equation based methods for solving theelectromagnetic scattering problem by incorporating random computations. Randomization allows to ease the computationalburden by replacing costly exact operations by inexpensive approximate ones. Besides, it minimizes the need for inter-processorcommunications, leading to highly scalable algorithms. We select electromagnetic scattering in clusters of plasmonicnanoparticles as our case of study. Randomized methods are suitable for these kind of problems since their collective behaviorcan often be efficiently characterized by statistical means.The electromagnetic scattering problem in the integral form is discretized by the Method of Moments, which consists onexpanding the unkown of the problem as a linear combination of basis functions and projecting both members of the equationonto a set of weighting functions. The result is a linear system of algebraic equations that approximates the original functionalequation. This linear system is usually solved by iterative methods. Also, computing and manipulating the system matrix isunfeasible for most realistic problems. This is resolved by compression methods that take advantage of the reduced number ofdegrees of freedom of the electromagnetic interaction between distant subdomains of the scattering object.In this work, we propose a method for solving the electromagnetic scattering problem based on the Born-Neumann series. Thisformalism expresses the inverse of the scattering operator as a power series of the operator itself. Due to the structure of thisexpansion, Monte Carlo methods can be employed to accelerate the computation of the solution of the scattering problem. Thisthesis presents contributions on the following two topics:First, the relation of the Born-Neumann series with semiiteratiive methods, Krylov subspace methods and conformal mapping.Both Krylov and semiiterative methods approximate the inverse of a linear operator as a linear combination of powers of theoperator itself. The difference lies on the way of finding the coefficients of such linear combination: whereas Krylov subspacemethods actively find these coefficients by minimizing a certain residual, semiiterative methods compute them by applying a fixedtransformation based on assumptions about the operator spectrum. We analyze the relation between both families of methodsand study their connection with conformal maps.Second, the convergence properties of the Ulam-Neumann Monte Carlo algorithm. This method approximates the solution of alinear equation by performing random sampling on the terms of the Born-Neumann expansion. We study the convergenceproperties of this method when submatrices are sampled instead of single matrix elements, and propose an alterantive samplingscheme based on independently approximating each term of the Born-Neumann series that converges on a wider range ofsituations.In the last part of the work, we implement and test a randomized compression method for the impedance matrix obtained by theMethod of Moments. The inherent parallelism of this method allows to exploit the dense linear algebra capabilities of highthroughput processors as GPUs.

Committee:
     PRESIDENT: CAPARRA, GIANLUCA
     SECRETARI: BARTZOUDIS, NIKOLAOS
     VOCAL: FONT BACH, JOSEP ORIOL
Thesis abstract: The rapid evolution in satellite navigation technology (GNSS) requires advanced prototyping tools for exploring new signals and developing innovative systems. Prototyping is essential in the design and development process, as it allows researchers to test and refine their ideas before implementing them on a large scale.Prototyping using commercial GNSS receivers poses several challenges. Currently, these receivers are primarily based on application-specific integrated circuits (ASICs), which are characterized by low power consumption, compact dimensions, and low cost, but offer limited flexibility. Although some commercial devices incorporate software-defined radio (SDR) techniques, they often contain proprietary code that restricts reconfiguration through an application programming interface (API) established by the manufacturer.GNSS receivers based on free and open-source software have become very valuable resources in the field of research and development, especially in satellite navigation. These receivers are highly valued for their adaptability and flexibility, allowing researchers to tailor the software to specific experimental needs or develop new signal processing algorithms. However, software-defined receivers tend to be less energy-efficient compared to hardware-based receivers, as they operate on general-purpose processors, which are not optimized for low power consumption.This thesis focuses on the design and development of a low-cost architecture for prototyping experimental GNSS receivers, based on System-on-Chip Field Programmable Gate Arrays (SoC FPGAs). This architecture overcomes the limitations of commercial GNSS receivers in terms of adaptability, flexibility, and reprogramming capacity, and offers improved energy efficiency compared to software-based receivers that rely on general-purpose processors. The strategy consists of combining the versatility of software-defined radio with the intensive parallelism and optimized energy consumption of programmable logic devices, providing the best of both worlds. This fusion allows the development of compact, portable GNSS receivers, thus facilitating the prototyping of embedded devices suitable for field testing. In addition, the GNSS processing core is based on a free and open-source software implementation, which provides detailed access to the signal processing chain and allows unrestricted exploration and modification of the algorithms used.This thesis also presents a design methodology for the development of new prototypes and new GNSS signal processing algorithms based on the proposed SoC FPGA architecture. This methodology places special emphasis on code reuse, a key aspect for reducing development costs and time.The practical applications of this architecture have been demonstrated through three prototypes: a GNSS receiver for low Earth orbit (LEO), a GNSS signal repeater, and a high-sensitivity GNSS receiver.The innovative approach presented in this thesis facilitates the development of experimental prototypes of flexible and portable GNSS receivers and signal generators, suitable for both laboratory experiments and field testing.

Committee:
     PRESIDENT: ADELANTADO FREIXER, FERRAN
     SECRETARI: PEREZ ROMERO, JORGE
     VOCAL: KOUTLIA, KATERINA
Thesis abstract: The growing number of Wi-Fi users and the emergence of bandwidth-intensive services have necessitated an increase in Access Point (AP) density, resulting in more complex network configuration, optimization, and management tasks. Concurrently, advancements in data monitoring and analytics technologies in wireless networks offer opportunities to extract valuable insights into network and user behavior, facilitating more efficient network management. In this thesis, we propose different Machine Learning based techniques to enhance Wi-Fi network management, focusing on three aspects: user connectivity prediction, Wi-Fi traffic prediction, and Wi-Fi traffic classification. The first aspect of our work focuses on predicting the next Access Point (AP) a user will connect to in a Wi-Fi network. We propose a methodology based on historical information of the AP to which a user has been connected, extracting connectivity patterns at different time scales (hourly, daily, weekly). Predictions are done using techniques based on Neural Networks and Random Forest algorithms. This approach is evaluated using real data from a university campus Wi-Fi network. Predicting the next AP of users allows for proactive network reconfiguration, enhancing the efficiency of techniques like Pairwise Master Key caching and Opportunistic Key Caching, which reduce re-authentication times. Additionally, this prediction helps to identify the geographical region of the User Equipment (UE) and can be used for commercial purposes, such as targeted advertising, by customizing messages based on locations of the users. Secondly, we propose a methodology for predicting the aggregated traffic at access points (APs) by leveraging spatial and temporal correlations from neighboring APs to enhance prediction accuracy. Using real measurements, we evaluate various Deep Learning methods, including Convolutional Neural Network (CNN), Simple Recurrent Neural Network (SRNN), Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and Transformer, and present a hybrid approach combining CNN for spatial processing and RNN for temporal prediction. This hybrid method improves accuracy with minimal additional training time and negligible impact on prediction time. Accurate traffic forecasting at each AP enables better load distribution and can inform resource management techniques such as admission control, congestion control, and load balancing. Additionally, predicting low traffic periods can aid in energy-saving strategies by allowing APs with minimal traffic to be switched off during specific times. Finally, traffic classification is essential for enhancing network performance by allowing better resource allocation and prioritization of services with stringent latency requirements. The increasing demand for Virtual Reality (VR) services poses a significant challenge for Wi-Fi networks to meet strict latency needs, crucial for VR to ensure immediate response and avoid user discomfort. To improve VR Quality of Service (QoS), distinguishing interactive VR traffic from Non-VR traffic is key. We propose a machine learning-based method to identify interactive VR traffic in a Cloud-Edge VR environment by analyzing downlink and uplink data correlations and extracting features from single-user traffic characteristics. Six classification techniques (i.e., Logistic Regression, Support Vector Machines, k-Nearest Neighbors, Decision Trees, Random Forest, and Naive Bayes) are compared. The result of the classification is used for the prioritization of VR traffic over Non VR traffic. We evaluate our method using datasets from various VR applications and Wi-Fi network simulations. Our results show a significant reduction in VR traffic delays with minimal impact on Non-VR service latency.

Committee:
     PRESIDENT: HERNÀNDEZ FERRÀS, JOAN MANEL
     SECRETARI: KOUBYCHINE MERKULOV, YOURI ALEXANDROVICH
     VOCAL: RUMOLO, GIOVANNI
Thesis abstract: The Future Circular Hadron Collider (FCC-hh) is a proposed hadron-hadron collider, with a center-of-mass collision energy of 100 TeV in an approximately 100-km long circular tunnel, aiming to push the frontiers of particle physics. One of the key elements in such a high-energy accelerator is the beam screen, which shields the cold superconducting magnets from the synchrotron radiation emitted by the proton beams at such high energies, ensures ultra-high vacuum conditions, and provides a low impedance to the circulating beams. Beam impedance depends on the electromagnetic interaction between the fields generated by the proton beam and the beam screen. In the current baseline design, the beam screen is made of 1 mm thick stainless steel plates coated with 0.3 mm of copper (Cu), but its surface impedance may not be low enough to guarantee a stable beam operation. The use of High-Temperature Superconductor (HTS) is thus being considered to improve the beam screen performance, particularly, those based on REBa2Cu3O7−x coated conductors (REBCO-CCs being RE = Y, Gd, Eu). It is therefore essential to estimate the surface impedance of these materials in the range of frequencies (2.1 kHz to 3 GHz), DC magnetic fields (1 to 16 T), and temperature (40-60 K) relevant for their use in the FCC-hh. The experimental characterization at the exact FCC-hh operating conditions is extremely challenging, and the surface impedance has to be inferred at conditions that are feasible with the current experimental capabilities. In this thesis, we have addressed the full characterization of REBCO-CC as possible coatings for the beam screen. The first part presents a system of dielectric resonators to characterize the surface impedance of REBCO-CC materials at conditions as similar as possible to those foreseen in the FCC-hh. Results on representative REBCO-CC samples show the validity of the approach developed, providing surface impedance characterization up to 16 T, in a wide temperature range (4-100 K) and a frequency range between 6 and 15 GHz. The second part evaluates the impact of using REBCO-CC to beam stability by performing calculations of the resistive wall beam impedance using a combination of finite integration techniques and beam coupling impedance theory for beam pipes of general cross-section. To this end, hybrid beam screen models combining REBCO-CC and Cu have been considered. The position of REBCO-CC and the percentage of coverage have been systematically studied to determine the optimum configuration that lowers the resistive wall impedance compared to that presented by Cu, while maintaining the magnetic field homogeneity seen by the beam within the specifications. We demonstrate in this study that it is possible to find a REBCO-CC coverage that improves the performance of the nominal beam screen.

Committee:
     PRESIDENT NO PRESENCIAL: AIGNER, ROBERT
     SECRETARI: VALENZUELA GONZALEZ, JOSE LUIS
     VOCAL NO PRESENCIAL: VILLANUEVA TORRIJO, LUIS GUILLERMO
Thesis abstract: The emergence of smartphones not only changed the way people uses its phone for, but it also changed the traffic amount that networks need to carry, increasing the demand of higher data rates. The overall result was the appearance of 4G networks, and nowadays, the current development of 5G, implying the need for more frequency bands, and the application of new techniques such as Carrier Aggregation (CA), MIMO antennas, and so on. All these market driven necessities suppose a great challenge for the radiofrequency (RF) industry, which have been facing the necessity of miniaturization and band coexistence on its devices since the beginning of mobile communications.Microwave filters based on Bulk Acoustic Wave (BAW) resonators, have been able to this day to overcome these limitations. These devices consist in a thin piezoelectric layer comprised by two metal electrodes, and an acoustic confinement method, which can be simply made of air or a Bragg reflector. The use of electroacoustic technology enables to reduce the filter size up to five orders of magnitude, allowing the integration of multiple filters in handsets. This thesis focuses on modelling some of different physical phenomena at the resonator level that affect the performance of the filters.The first part of this thesis is the one regarding the spurious response of BAW resonators. This response is originated by acoustic waves traveling in the lateral dimension of the resonator. These waves couple electromechanically, degrading the filter response. BAW filters have been capable of overcoming this limitation suppressing them by the use of different electrode geometries (Apodization), or by surrounding the electrode by a decreased, or increased frame (Border Ring). The nature of these waves is studied through the thesis and several equivalent models are proposed in order to accurately predict them, helping to the design of the correspondent suppression structures. One of this thesis contributions regarding the lateral spurious resonances, consists in making use of a modified Mason model to determine the origin of the additional spurious resonances generated by the Border Ring. These resonances can be attributed to an acoustic mode, different from the fundamental, propagating across the resonator stack. By adding nonlinear sources to that model, the second harmonic (H2) emissions and the impact of the spurious resonances in them, is also studied.Finally, a new equivalent model based in the Transmission Line Matrix (TLM) method is proposed for the acoustic cavity of a BAW resonator. This new approach is able to model resonators with different electrode geometries, in a much faster way than traditionally used methods like the Finite Elements Method (FEM). In addition, by determining different propagation regions, it can be used to model both the apodization and the Border Ring at the same time.The second family of contributions are the ones regarding to the thermoelastic behavior of the BAW resonators. A solid heats up when compressed and vice versa. In a harmonic oscillation, when heat is able to flow through the solid regions, this flow from hotter to colder regions generates a relaxation of the acoustic wave. This is the thermoelastic damping. A thermo-electro-mechanical Mason model is used for quantifying this source of losses on BAW resonators. The model is compared with experimental data taken at cryogenic temperatures and an analysis of losses of the broadband spurious resonances of the Bragg reflector has been performed.