Offshore wind power is attracting increasing levels of research and investment. The use of HVDC transmission and the development of DC grids are topics with similar high levels of interest that go hand in hand with the development of large scale, far from shore wind farms. Despite increased capital cost of some components, DC power transmission can have significant advantages over AC transmission, in particular in the offshore environment. These advantages are well established for large scale, long distance point to point transmission. This thesis assesses the suitability of a multi-terminal DC power collection network, with short cables and relatively small amounts of power, addresses a number of the technical challenges in realising such a network and shows methods for overall system cost reduction. Technical and modelling challenges result from the interaction between power electronic DC-DC converters and the cables in a DC transmission network. In particular, the propagation of the ripple current in bipole DC transmission cables constructed with a metallic sheath and armour is examined in detail. The finite element method is used to predict the response of the cable to the ripple current produced by the converters. These results are used along with wave propagation theory to demonstrate that cable design plays a crucial role in the behaviour of the DC system. The frequency dependent cable models are then integrated with time domain DC-DC converter models. The work in the thesis is, broadly, in two parts. First, it is demonstrated that care and accuracy are required in modelling the cables in the DC transmission system and appropriate models are implemented and validated. Second, these models are combined with DC-DC converter models and used to demonstrate the practicality of the DC grid, make design recommendations and assess its suitability when compared with alternative approaches (e.g. AC collection and/or transmission).

Increasing the power density of power electronic converters while reducing or maintaining the same cost, offers a higher potential to meet the current trend inrelation to various power electronic applications. High power density converters can be achieved by increasing the switching frequency, due to which the bulkiest parts, such as transformer, inductors and the capacitor's size in the convertercircuit can be drastically reduced. In this regard, highly integrated planar magnetics are considered as an effective approach compared to the conventional wire wound transformers in modern switch mode power supplies (SMPS). However, as the operating frequency of the transformers increase from several hundred kHz to MHz, numerous problems arise such as skin and proximity effects due to the induced eddy currents in the windings, leakage inductance and unbalanced magnetic flux distribution. In addition to this, the core losses whichare functional dependent on frequency gets elevated as the operating frequency increases. Therefore, this thesis provides an insight towards the problems related to the high frequency magnetics and proposes a solution with regards to different aspects in relation to designing high power density, energy efficient transformers.The first part of the thesis concentrates on the investigation of high power density and highly energy efficient coreless printed circuit board (PCB) step-down transformers useful for stringent height DC-DC converter applications, where the core losses are being completely eliminated. These transformers also maintain the advantages offered by existing core based transformers such as, high coupling coefficient, sufficient input impedance, high energy efficiency and wide frequencyband width with the assistance of a resonant technique. In this regard, several coreless PCB step down transformers of different turn’s ratio for power transfer applications have been designed and evaluated. The designed multilayered coreless PCB transformers for telecom and PoE applications of 8,15 and 30W show that the volume reduction of approximately 40 - 90% is possible when compared to its existing core based counterparts while maintaining the energy efficiency of the transformers in the range of 90 - 97%. The estimation of EMI emissions from the designed transformers for the given power transfer application proves that the amount of radiated EMI from a multilayered transformer is lessthan that of the two layered transformer because of the decreased radius for thesame amount of inductance.The design guidelines for the multilayered coreless PCB step-down transformer for the given power transfer application has been proposed. The designed transformer of 10mm radius has been characterized up to the power level of 50Wand possesses a record power density of 107W/cm3 with a peak energy efficiency of 96%. In addition to this, the design guidelines of the signal transformer fordriving the high side MOSFET in double ended converter topologies have been proposed. The measured power consumption of the high side gate drive circuitvitogether with the designed signal transformer is 0.37W. Both these signal andpower transformers have been successfully implemented in a resonant converter topology in the switching frequency range of 2.4 – 2.75MHz for the maximum load power of 34.5W resulting in the peak energy efficiency of converter as 86.5%.This thesis also investigates the indirect effect of the dielectric laminate on the magnetic field intensity and current density distribution in the planar power transformers with the assistance of finite element analysis (FEA). The significanceof the high frequency dielectric laminate compared to FR-4 laminate in terms of energy efficiency of planar power transformers in MHz frequency region is also explored.The investigations were also conducted on different winding strategies such as conventional solid winding and the parallel winding strategies, which play an important role in the design and development of a high frequency transformer and suggested a better choice in the case of transformers operating in the MHz frequency region.In the second part of the thesis, a novel planar power transformer with hybrid core structure has been designed and evaluated in the MHz frequency region. The design guidelines of the energy efficient high frequency planar power transformerfor the given power transfer application have been proposed. The designed corebased planar transformer has been characterized up to the power level of 50W and possess a power density of 47W/cm3 with maximum energy efficiency of 97%. This transformer has been evaluated successfully in the resonant converter topology within the switching frequency range of 3 – 4.5MHz. The peak energy efficiency ofthe converter is reported to be 92% and the converter has been tested for the maximum power level of 45W, which is suitable for consumer applications such as laptop adapters. In addition to this, a record power density transformer has been designed with a custom made pot core and has been characterized in thefrequency range of 1 - 10MHz. The power density of this custom core transformer operating at 6.78MHz frequency is 67W/cm3 and with the peak energy efficiency of 98%.In conclusion, the research in this dissertation proposed a solution for obtaining high power density converters by designing the highly integrated, high frequency(1 - 10MHz) coreless and core based planar magnetics with energy efficiencies inthe range of 92 - 97%. This solution together with the latest semiconductor GaN/SiC switching devices provides an excellent choice to meet the requirements of the next generation ultra flat low profile switch mode power supplies (SMPS).

This thesis analyzes the general performance and design requirements of photovoltaic(PV) systems, and specifically how they relate to the design of a system intended to supply power to a rotating satellite. The PV array geometry was discussed, different DC-DC converter topologies were analyzed, and optimum array geometry and converter topologies were determined. The potential reference quantities for use in control of the system are examined. Due to its comparably greater linearity with respect to changes in apparent load and its relative insensitivity to insolation changes, voltage was determined to be the best reference quantity for use in stable tracking of the maximum power operating point of photovoltaic modules. The preceding work is used to design and model a photovoltaic system for a rotating satellite ensuring the supply of the maximum available power as well as stable operation. Simulations of the system are performed at rotational velocities up to 300 rev/min and its behavior is analyzed to demonstrate the validity of the preceding work. It was concluded that: ● parallel connected photovoltaic panels provide greater efficiency than series connected panels. ● Buck, Boost, and Cuk Converter architectures are best suited to PV applications ● PV Voltage is the best reference quantity for use in stable control of PV systems.

Este trabalho apresenta um conversor CC - CC baseado em Capacitores Chaveados de 6 fases e tempos intercalados com o objetivo de coletar energia fotovoltaica projetado em tecnologia CMOS de 130 nm para ser usado em aplicações em Internet das Coisas e Nós Sensores. Ele rastreia o máximo ponto de entrega de energia de um painel fotovoltaico policristalino de 3 cm x 3 cm através de modulação da frequência de chaveamento com o objetivo de carregar baterias. A razão da tensão de circuito aberto foi a estratégia de rastreio escolhida. O conversor foi projetado em uma tecnologia CMOS de 130 nm e alcança uma eficiência de 90 % para potencias de entrada maiores do que 30 mW e pode operar com tensões que vão de 1.25 até 1.8 V, resultando em saídas que vão de 2.5 até 3.6, respectivamente. Os circuitos periféricos também incluem uma proteção contra sobre tensão na saída de 3.6 V e circuitos para controle, que consomem um total máximo de potência estática de 850 A em 3.3 V de alimentação. O layout completo ocupa uma área de 300 x 700 m2 de silício. Os únicos componentes não integrados são 6x100 nF capacitores.

Power electronics have a significant role in modern electrical devices, for instance, hybrid electric vehicles. Power electronics are the technology in between the source and the load circuits and can convert the power from dc to ac or from dc to ac. There are also many types of dc-dc converters, like such as boost and buck converters, which exhibit switching ripple behavior. A boost converter increases the output voltage (with respect to the input voltage) and reduces the output current. A buck converter decreases the output voltage and increases the output current. Many models are used to predict the behavior of the boost and buck converters. The detailed (DET), state-space averaged (SSA), and generalized averaging method (GAM) models are capable of predicting the average behavior of dc-dc converters. For DET and GAM models, the rippling behavior can also be predicted. These models differ in terms of required run time, existence of constant equilibrium points, and accuracy. The DET model has a long run time and does not have constant equilibrium, but it is very accurate. The SSA technique is a mathematical and time-invariant model that capable of describing the behavior of a dc-dc boost converters. It can derive the small signal ac equations of a switching converter and is used to illustrate the average behavior of any linear or nonlinear system in converters. The SSA does not take extensive runtime simulation and has constant equilibrium points, and can be applied to continuous, discrete and sample data systems. The GAM model can predict the average and ripple behavior in power electronic systems and has constant equilibrium and fast run time. However, it has a numerical stability issue. The integrator stabilized multifrequency averaging (ISMFA) model is employed to solve the stability issue in the GAM model, but it is a complicated dynamic method and has restrictions in its process. In the present study, a simplified but stable GAM model is introduced to predict the average and ripple behavior of boost dc-dc converters and to overcome the limitations of other methods. In this work, the stabilized GAM model has been used for a dc-dc boost converters. The stability of the proposed model is analyzed. The performance of the improved GAM model is compared with the DET, SSA, and GAM models. The results show that the stabilized GAM model is stable with the additional poles created by the GAM assignable by parameter choice. The new GAM model predicts the same results as the existing GAM method without the underlying stability concerns. The stabilized GAM model exhibits constant ii equilibrium point and requires significantly lower run times than the DET model, but it is also able to predict the ripple performance of the converter. The stabilized GAM model does not take a long run time, is less complicated, has fewer restrictions, has constant equilibrium and internal stability, and has more straightforward implementation than other models, like the ISMFA model. It represents a suitable alternative to DET models when high accuracy simulations are desired without long simulation run times.

A system to improve efficiency performance of a DC-DC converter is simulated and built. The proposed system combines multiple DC-DC converters in parallel and implements a digital control scheme and load-share controller. A model of the system is developed in MATLAB Simulink and the model demonstrates the improved converter’s efficiency particularly at low load conditions. This simulation is then designed into a hardware system running three DC-DC converters in parallel, controlled by a microcontroller and a load-share controller. The hardware also confirms the simulation results, although some hardware refinements are evident as simulation results are superior. The system is designed to be scalable in the number of converters and the total output power, as well as being DC-DC converter topology-independent. Simulation results show the system maintaining better than 88 % efficiency over almost 90 % of the load range of the system. This system could be implemented where dynamic loads typically occur, such as in electric vehicle charging.

This thesis explores high voltage converter circuits for MEMS applications using micromachined devices. A novel MEMS based tunable DC-DC converter has been developed. Conventional high voltage converters based on charge pumps are unable to convert voltages to higher than few tens of volts due to power handling limitations of the CMOS components. In order to overcome this limitation a high voltage circuit has been proposed, which when integrated with micromachined switches will generate output voltages in the range of 100 Volts. The converter is based on a two phase switched capacitor circuit, and allows regulation of voltage conversion ratio. Three prototype circuits have been built for proof of concept. A test program has been written for synchronized CPLD based control of the switched capacitors. Individual capacitor fabrication technology is explored using two methods - Porous Silicon and DRIE processing. A micromachined capacitor bank has also been fabricated in silicon using a novel process sequence which provides for critical real estate savings and integration benefits. It enables on-chip integration of numerous microcapacitors, without losing customized configurability of the capacitor bank. The technique utilizes polyimide to facilitate lithography on a highly contoured surface. Plain capacitors have been fabricated on silicon with oxide-nitride-oxide stack being used as the dielectric to provide a building block for further fabrication of a variety of capacitors.

A la recherche de la modularité et de la généricité, notre approche consiste à partir de l'idée qu'une cellule élémentaire peut être conçue, intégrée et réalisée en grand nombre. Ensuite elle va être associée en grand nombre pour pouvoir, au final, couvrir n'importe quelle application de conversion DC-DC en électronique de puissance. Ce travail de thèse a pour but tout d'abord d'introduire notre approche de manière consistante. Ensuite, un bilan de l'état de l'art autour de la même direction et aussi la présentation détaillée de notre approche sont abordés. Le coeur de la thèse se concentre sur l'étude du couplage magnétique entre les cellules élémentaires via la mise en parallèle des enroulements entre les transformateurs de puissance. Ce couplage avec notre commande aléatoire sert à la fois la réduction du filtrage et la possibilité de l'intégration des composants passifs. Enfin, le travail est bouclé par la validation expérimentale d'un réseau de 12 convertisseurs de 10W faisant une conversion DC-DC de 120W.

Dans toute alimentation cohabitent des composants à semi-conducteurs, des circuits de commande et les composants passifs des filtres, inductances et condensateurs. Si les actifs sont réalisés sur silicium, les passifs sont actuellement sous forme discrète. L'étude menée porte sur la conception d'une inductance intégrée sur silicium ainsi que sur la modélisation et la commande des composants de puissance à semi-conducteurs d'un redresseur synchrone, dans le but de réaliser des convertisseurs DC-DC entièrement sur silicium pour de futures applications d'alimentations basse tension de circuits électroniques (téléphones cellulaires, ordinateurs portables, etc...). Ainsi, des outils de choix de dimensionnement du convertisseur et de prédétermination du rendement global sont présentés permettant de dégager des enseignements importants concernant les gammes d'utilisation de tels convertisseurs, ainsi que la recherche d'optimums de performances.

Cette thèse se place dans le contexte d'alimentation des systèmes électroniques nomades fonctionnant sous faible tension et faible courant. Le dimensionnement et la réalisation d'éléments passifs magnétiques intégrés, pour le stockage d'énergie à l'échelle de la commutation y est abordé pour tenir compte de la taille, la position de l'alimentation de puissance par rapport au système à alimenter ainsi que de son rendement et sa densité de puissance. De plus, face à la demande de systèmes rapides, à faible encombrement et finement régulés, les fréquences de commutation tendent naturellement à augmenter. L'intégration monolithique des composants des convertisseurs de puissance est une des solutions plausibles. Toutefois, la réalisation de tels composants pose actuellement un réel défi technologique. De nombreuses équipes travaillent depuis plus de trente ans sur le problème de l'intégration de composants magnétiques sur silicium. Ces réalisations font appel aux techniques de réalisation microélectroniques, tel la croissance électrolytique, pour le dépôt des matériaux avec des épaisseurs contrôlées. L'objectif aujourd'hui est de réaliser une inductance de 12H capable de supporter 1W avec des tensions de l'ordre du volt, dans une gamme de fréquence allant de 1MHz à 10MHz. Pour ces raisons de compacité et pour garantir une plage fréquentielle de fonctionnement suffisante, nous avons travaillé avec des structures inductives planaires spirales, similaires par la forme à celles utilisées dans le domaine de l'intégration pour les radiofréquences (RF). L'étude des phénomènes physiques liés au caractère inductif et résistif de spirales planaires à partir d'équations et d'une définition géométrique normalisée du composant a abouti vers la mise en place d'outils d'étude numérique permettant d'analyser les liens entre les paramètres géométriques et les paramètres électriques du composant. Il est alors apparu que les dimensions géométriques nécessaires à l'accompl issement du cahier des charges précédent (notamment l'objectif de très faible résistance), se heurtaient aux limites des possibilités de réalisation technologiques (facteur de forme des conducteurs, résolution spatiale). Dans le but, de dépasser ces limites, nous nous sommes intéressés à l'association série/parallèle d'inductances magnétiquement couplées. Les simulations ont démontrées qu'une marge de manoeuvre existait, permettant une optimisation des performances électriques des composants. Un procédé technologique basé sur les techniques MEMS de résine épaisse et de croissance électrolytique de cuivre, a été développé pour valider les modélisations précédentes. Ce procédé permet de réaliser de manière générique des inductances planaires classiques ainsi qu'un empilement de plusieurs bobines connectées entre elles ou non, sur un substrat silicium. Le procédé a également été mis au point sur substrat Pyrex® pour supprimer les courants induits dans le substrat, sources de perturbations pour le comportement fréquentiel de l'inductance et de la résistance. Un banc de caractérisation impédance métrique a également été conçu afin de déterminer les limites du fonctionnement fréquentiel des composants réalisés qui ont permis de valider les performances des composants Ce travail ouvre de nombreuses perspectives dans la réalisation de composants magnétiques intégrés, assurant des fonctions électroniques connues et nouvelles, notamment en démontrant la faisabilité de transformateurs avec différents rapports de transformation ainsi que de fonctions de filtrage performant.