Ultra narrow-band filters and the use of two loops in a cascade configuration dominate current clock conditioners based on phase-locked-loop (PLL) schemes. Since a PLL exhibits a low-pass transfer function with respect to the reference clock, the noise performance at very close-to-carrier offset frequencies is still determined by the input signal. Although better cleaning can be achieved with extremely narrow loops, an ultra low cut-off frequency could not be selected since the stability of the configuration deteriorates as the filter bandwidth is reduced. This fact suggests that a full-spectrum clock conditioning is not possible using traditional PLL architectures, and an alternative scheme is necessary to attenuate the very-close-to-carrier phase noise (PN). In addition, ultra-narrow loop filters can compromise on-chip integration because of the large size capacitors needed when chosen as passive. Input signal attenuation with relaxed bandwidth requirements becomes the main aspect that a comprehensive clock cleaner must address to effectively regenerate a reference signal. This dissertation describes the Band-Reject Nested-PLL (BRN-PLL) scheme, a modified PLL-based architecture that provides an effective signal cleaning procedure by introducing a notch in the input transfer function through inner and outer loops and a high-pass filter (HPF). This modified response attenuates the reference-signal PN and reduces the size of the loop-filter capacitors substantially. Ultra narrow loops are no longer required because the notch size is related to the system bandwidth. The associated transfer function for the constitutive blocks (phase detectors and local oscillators) show that the output close-to-carrier and far-from-carrier PN sections are mainly dominated by the noise from the inner-PLL phase detector (PD) and local oscillator (LO) located in the outer loop, respectively. The inner-PLL PD transfer function maintains a low-pass characteristic with a passband gain inversely proportional to the PD gain becoming the main contribution around the carrier signal. On the other hand, the PN around the transition frequency is determined mainly by the reference and the inner-PLL LO. Their noise contributions to the output will depend on the associated passband local maxima, which is located at the BRN-PLL transition frequency. Hence, in this region, the inner-PLL LO is selected so that its effect can be held below that of the outer-PLL PD. The BRN-PLL can use a high-Q MEMS-based VCO to further improve the transition region of the output PN profile and an LC-VCO as outer-PLL LO to reduce the noise floor of the output signal. In particular, two tuning mechanisms are explored for the MEMS-VCO: series tuning using varactors and phase shifting of a resonator operating in nonlinear regime. Both schemes are implemented to generate a tunable oscillator with no PN-performance degradation.

The aim of our work was to provide a robust method for evaluating imaging performance of positron emission tomography (PET) systems and particularly to estimate the modulation transfer function (MTF) using the line spread function (LSF) method. A novel plane source was prepared using thin layer chromatography (TLC) of a fluorine-18-fluorodeoxyglucose (18F-FDG) solution. The source was placed within a phantom, and imaged using the whole body (WB) two dimensional (2D) and three dimensional (3D) standard imaging protocols in a hybrid PET/CT scanner. Modulation transfer function was evaluated by determining the LSF, for various reconstruction methods and filters. The proposed MTF measurement method was validated against the conventional method, based on point spread function (PSF). Higher MTF values were obtained with 3D scanning protocol and 3D iterative reconstruction algorithm. All MTF obtained using 3D reconstruction algorithms showed better preservation of higher frequencies than the 2D algorithms. They also exhibited better contrast and resolution. MTF derived from LSF were more precise compared with those obtained from PSF since their reproducibility was better in all cases, providing a mean standard deviation of 0.0043, in contrary to the PSF method which gave 0.0405. By using our novel plane source we developed a method for measuring the image registration shift between the CT and the PET images in fusion. The method gave results that are comparable to the manufacturers’ method for image registration and has the advance that is applicable to all scanner models. Through the MTF it is also feasible to compare commercially available scanners in terms of image quality. In conclusion, the proposed method is novel and easy to implement for characterization of the signal transfer properties and image quality of PET/CT systems. It provides an easy way to evaluate the frequency response of each kernel available. The proposed method requires cheap and easily accessible materials, available to the medical physicist in the nuclear medicine department. Furthermore, it is robust to aliasing and since this method is based on the LSF, is more resilient to noise due to greater data averaging than conventional PSF-integration techniques. In this study was used to assess the image quality. But it can also be used as a method of quality control and stability of system performance over time. / Σκοπός της παρούσας μελέτης ήταν να εισαγάγει μια νέα μέθοδο για την αξιολόγηση των χαρακτηριστικών της ποιότητας εικόνας στην τομογραφία εκπομπής ποζιτρονίων (PET). Ειδικότερα, έγινε εκτίμηση της συνάρτησης διαμόρφωσης μεταφοράς (MTF) σε PET εικόνες με χρήση της συνάρτησης διασποράς γραμμής (LSF). Κατασκευάστηκε μία πρότυπη επίπεδη πηγή, Λεπτού Χρωματογραφικού Χάρτη (TLC), υψηλής ομοιογένειας. Οι καμπύλες της MTF υπολογίστηκαν από την ανακατασκευή των εγκάρσιων τομών της επίπεδης πηγής. Η ανακατασκευή έγινε για όλους τους διαθέσιμους αλγόριθμους ανακατασκευής και φίλτρα καθώς και για τους δύο τρόπους σάρωσης (2D και 3D). Τα αποτελέσματα συγκρίθηκαν με την MTF που προκύπτει μέσω της συνάρτησης διασποράς σημείου (PSF). Με την χρήση της επίπεδης πηγής αναπτύχθηκε μέθοδος με την οποία μπορεί να ελεγχθεί η σύμπτωση της εικόνας του CT και της εικόνας του PET κατά την σύντηξη. Η μέθοδος συγκρίθηκε με τον τρόπο ελέγχου που καθορίζει ο κατασκευαστής του συστήματος. Επιπλέον, πραγματοποιήθηκε σύγκριση μεταξύ δύο εμπορικών συστημάτων από διαφορετικούς κατασκευαστές. Η σύγκριση πραγματοποιήθηκε για τα διαθέσιμα πρωτόκολλα σάρωσης και τους προτεινόμενους αλγόριθμους ανακατασκευής κάθε συστήματος. Η σάρωση του αντικειμένου σε 3D και η χρήση τρισδιάστατων αλγορίθμων ανασύστασης είχε ως αποτέλεσμα την δημιουργία εικόνων με υψηλότερες τιμές MTF. Όλες οι καμπύλες MTF που προέκυψαν από τρισδιάστατους αλγόριθμούς ανασύστασης έδειξαν καλύτερη απόκριση στις υψηλότερες συχνότητες από τους δισδιάστατους αλγόριθμους. Οι εικόνες που προέρχονται από τρισδιάστατους αλγόριθμους έχουν καλύτερη αντίθεση (contrast) και διακριτική ικανότητα (resolution). Οι καμπύλες MTF που προέρχονται από την LSF έχουν καλύτερη ακρίβεια σε σχέση με τις καμπύλες MTF που προέρχονται από την PSF και η επαναληψιμότητα τους ήταν καλύτερη σε όλες τις περιπτώσεις. Η μέση τιμή της τυπικής απόκλισης των καμπυλών MTF που προέρχονται από την LSF βρέθηκε ίση με 0,0043, ενώ η μέση τιμή της τυπικής απόκλισης των καμπυλών MTF που προέρχονται από την χρήση της PSF ήταν 0,0405. Με την προτεινόμενη μέθοδο αξιολόγησης είναι δυνατόν να συγκριθούν δύο ή περισσότερα εμπορικά διαθέσιμα συστήματα ως προς τα ποιοτικά χαρακτηριστικά της εικόνας τους. Τέλος, η μέτρηση της απόκλιση της εικόνας του CT και του PET στην εικόνα σύντηξης με την μέθοδο που αναπτύχθηκε ήταν συγκρίσιμη με την τιμή που μετρήθηκε εφαρμόζοντας την μέθοδο του κατασκευαστή. Συμπερασματικά, η παρούσα εργασία έδειξε ότι η χρήση της MTF, που προέρχεται από την LSF, μπορεί να χαρακτηρίσει πλήρως την ποιότητα εικόνας PET. Με τη προτεινόμενη μέθοδο αξιολόγησης της ποιότητας εικόνας, η διερεύνηση των διαφόρων στοιχείων της αλυσίδας απεικόνισης μπορεί να γίνει πιο απλή και να βελτιωθεί περαιτέρω και η συνολική απόδοση συστημάτων PET. Η προτεινόμενη μέθοδος αξιολόγησης απαιτεί υλικά που είναι συνηθισμένα στο κλινικό περιβάλλον, μπορεί να εφαρμοστεί πειραματικά και να χρησιμοποιηθεί στην κλινική πράξη. Επιπλέον, επειδή η μέθοδος βασίζεται στην LSF, είναι πιο ανθεκτική στο θόρυβο από τις συμβατικές τεχνικές που κάνουν χρήση της συνάρτησης διασποράς σημείου. Σε αυτή τη μελέτη χρησιμοποιήθηκε για την εκτίμηση της ποιότητας εικόνας. Μπορεί όμως να χρησιμοποιηθεί και ως μέθοδος ελέγχου ποιότητας και σταθερότητας της απόδοσης του συστήματος στον χρόνο.

Text in Afrikaans / Summaries in Afrikaans and English / In Afrikaans, soos in NederJands en Duits, word saamgestelde woorde aanmekaar geskryf. Nuwe woorde word dus voortdurend geskep deur woorde aanmekaar te haak Dit bemoeilik die proses van woordafkapping tydens teksprosessering, wat deesdae deur rekenaars gedoen word, aangesien die verwysingsbron gedurig verander. Daar bestaan verskeie afkappingsalgoritmes en tegnieke, maar die resultate is onbevredigend. Afrikaanse woorde met korrekte lettergreepverdeling is net die elektroniese weergawe van die handwoordeboek van die Afrikaanse Taal (HAT) onttrek. 'n Neutrale netwerk ( vorentoevoer-terugpropagering) is met sowat. 5 000 van hierdie woorde afgerig. Die neurale netwerk is verfyn deur 'n gcskikte afrigtingsalgoritme en oorfragfunksie vir die probleem asook die optimale aantal verborge lae en aantal neurone in elke laag te bepaal. Die neurale netwerk is met 5 000 nuwe woorde getoets en dit het 97,56% van moontlike posisies korrek as of geldige of ongeldige afkappingsposisies geklassifiseer. Verder is 510 woorde uit tydskrifartikels met die neurale netwerk getoets en 98,75% van moontlike posisies is korrek geklassifiseer. / In Afrikaans, like in Dutch and German, compound words are written as one word. New words are therefore created by simply joining words. Word hyphenation during typesetting by computer is a problem, because the source of reference changes all the time. Several algorithms and techniques for hyphenation exist, but results are not satisfactory. Afrikaans words with correct syllabification were extracted from the electronic version of the Handwoordeboek van die Afrikaans Taal (HAT). A neural network (feedforward backpropagation) was trained with about 5 000 of these words. The neural network was refined by heuristically finding a suitable training algorithm and transfer function for the problem as well as determining the optimal number of layers and number of neurons in each layer. The neural network was tested with 5 000 words not the training data. It classified 97,56% of possible points in these words correctly as either valid or invalid hyphenation points. Furthermore, 510 words from articles in a magazine were tested with the neural network and 98,75% of possible positions were classified correctly. / Computing / M.Sc. (Operasionele Navorsing)

Recently, 3D printing has been increasingly used to create physical models of the vocal tract with geometries obtained from magnetic resonance imaging. These printed models allow measuring the vocal tract transfer function, which is not reliably possible in vivo for the vocal tract of living humans. The transfer functions enable the detailed examination of the acoustic effects of specific articulatory strategies in speaking and singing, and the validation of acoustic plane-wave models for realistic vocal tract geometries in articulatory speech synthesis. To measure the acoustic transfer function of 3D-printed models, two techniques have been described: (1) excitation of the models with a broadband sound source at the glottis and measurement of the sound pressure radiated from the lips, and (2) excitation of the models with an external source in front of the lips and measurement of the sound pressure inside the models at the glottal end. The former method is more frequently used and more intuitive due to its similarity to speech production. However, the latter method avoids the intricate problem of constructing a suitable broadband glottal source and is therefore more effective. It has been shown to yield a transfer function similar, but not exactly equal to the volume velocity transfer function between the glottis and the lips, which is usually used to characterize vocal tract acoustics. Here, we revisit this method and show both, theoretically and experimentally, how it can be extended to yield the precise volume velocity transfer function of the vocal tract.

Contact microphones, especially throat microphones, have been developed to be used in environments with high background noises to improve the speech intelligibility in communication. They pick up vibrations from the surface they are attached to and are therefore less sensitive to sound and noise from the air. Comparison of the speech intelligibility with other types of microphones have previously been done by letting test persons examine the communication devices in question. This study examines the possibility to make use of the STI-method instead and therefore make a comparison faster, more cost-efficient and customizable. The thought is that if the relationship between speech signals and vibrations were known, it could be used to transform the STI test signal to vibrations and then use the STImethod to estimate the speech intelligibility for the chosen contact microphone. This study, containing 22 men and women, evaluated the vibrations at the most suitable locations on the head for contact microphones and compared it with the speech signal of the same test person. Frequency responses were calculated for all locations of the head and a more detailed study showed that the frequency response of the neck may be approximated as a second order lowpass filter with a cut-off frequency of about 300 Hz that attenuates speech signals with higher frequencies. Experiments were also done to measure the STI value of a throat microphone with the known relationship. However, the results pointed out several problems that needs to be addressed before a STI method can be performed successfully. The results from this study may also be used to deeper study the relationships between different vibrations resulting from speech signals and suggestions on how the performance of contact microphones may be improved are given.

Made available in DSpace on 2015-03-03T15:08:47Z (GMT). No. of bitstreams: 1 FernandoCM_TESE.pdf: 4237848 bytes, checksum: 00e7cb467e7c1cded76198b12c9e6e82 (MD5) Previous issue date: 2013-08-23 / This work has as main objective to find mathematical models based on linear parametric estimation techniques applied to the problem of calculating the grow of gas in oil wells. In particular we focus on achieving grow models applied to the case of wells that produce by plunger-lift technique on oil rigs, in which case, there are high peaks in the grow values that hinder their direct measurement by instruments. For this, we have developed estimators based on recursive least squares and make an analysis of statistical measures such as autocorrelation, cross-correlation, variogram and the cumulative periodogram, which are calculated recursively as data are obtained in real time from the plant in operation; the values obtained for these measures tell us how accurate the used model is and how it can be changed to better fit the measured values. The models have been tested in a pilot plant which emulates the process gas production in oil wells / Este trabalho teve como objetivo principal encontrar modelos matem?ticos baseados em t?cnicas de estima??o param?trica lineares aplicado ao problema do c?lculo da vaz?o de g?s em po?os de petr?leo. Em particular se concentrou na obten??o de modelos de vaz?o aplicados ao caso de po?os que produzem pela t?cnica de plunger-lift em plataformas de petr?leo, pois nesse caso, ocorrem picos elevados nos valores da vaz?o que dificultam sua medi- ??o direta atrav?s de instrumentos. Para isso, desenvolveram-se estimadores baseados em m?nimos quadrados recursivos e fez-se uma an?lise das medidas estat?sticas tais como autocorrela??o, correla??o cruzada, variograma e o periodograma acumulado, que s?o calculadas recursivamente ? medida que dados s?o obtidos em tempo real da planta em opera??o; os valores obtidos para essas medidas indicaram o qu?o exato ? o modelo utilizado e de que forma ele pode ser alterado para melhor se adequar aos valores medidos. Os modelos obtidos foram testados em uma planta piloto que emula o processo de produ??o de g?s em po?os de petr?leo

Made available in DSpace on 2014-12-17T14:55:43Z (GMT). No. of bitstreams: 1 AlineFGS_DISSERT.pdf: 1151209 bytes, checksum: 4ae192dd7519e3e14977154f7b70d951 (MD5) Previous issue date: 2010-06-14 / Conselho Nacional de Desenvolvimento Cient?fico e Tecnol?gico / In this work, a performance analysis of transmission schemes employing turbo trellis coded modulation. In general, the performance analysis of such schemes is guided by evaluating the error probability of these schemes. The exact evaluation of this probability is very complex and inefficient from the computational point of view, a widely used alternative is the use of union bound of error probability, because of its easy implementation and computational produce bounds that converge quickly. Since it is the union bound, it should use to expurge some elements of distance spectrum to obtain a tight bound. The main contribution of this work is that the listing proposal is carried out from the puncturing at the level of symbol rather than bit-level as in most works of literature. The main reason for using the symbol level puncturing lies in the fact that the enummerating function of the turbo scheme is obtained directly from complex sequences of signals through the trellis and not indirectly from the binary sequences that require further binary to complex mapping, as proposed by previous works. Thus, algorithms can be applied through matrix from the adjacency matrix, which is obtained by calculating the distances of the complex sequences of the trellis. This work also presents two matrix algorithms for state reduction and the evaluation of the transfer function of this. The results presented in comparisons of the bounds obtained using the proposed technique with some turbo codes of the literature corroborate the proposition of this paper that the expurgated bounds obtained are quite tight and matrix algorithms are easily implemented in any programming software language / Neste trabalho ? feita uma an?lise de desempenho de esquemas de transmiss?o empregando modula??o codificada turbo em treli?a. Em geral, a an?lise de desempenho de tais esquemas ? guiada pelo c?lculo da probabilidade de erro destes esquemas. O c?lculo exato desta probabilidade ? muito complexo e ineficiente sob o ponto de vista computacional, uma alternativa muito utilizada ? o emprego de limitante da uni?o da probabilidade de erro, por ser de f?cil implementa??o computacional e produzir limitantes que convergem rapidamente. Por se tratar do limitante da uni?o, este deve utilizar de expurgo de alguns elementos do espectro de dist?ncias do c?digo para a obten??o de um limitante apertado. A principal contribui??o deste trabalho ? que a enumera??o proposta ? realizada a partir da perfura??o a n?vel de s?mbolo e n?o a n?vel de bit como na maioria dos trabalhos da literatura. O principal motivo do uso da perfura??o a n?vel de s?mbolo reside no fato que a fun??o enumeradora do esquema turbo ? obtida diretamente das seq??ncias complexas de sinais atrav?s da treli?a e n?o de forma indireta a partir da seq??ncias bin?rias que exigem posterior mapeando bin?rio para complexo, como proposto por trabalhos anteriores. Assim, podem ser aplicados algoritmos completamente matriciais a partir da matriz adjac?ncia, que ? obtida a partir do c?lculo das dist?ncias das seq??ncias complexas da treli?a e n?o das seq??ncias bin?rias. Neste trabalho tamb?m s?o apresentados dois algoritmos matriciais de redu??o de estados do codificador bem como do c?lculo da fun??o de transfer?ncia deste. Os resultados apresentados em forma de compara??es dos limitantes obtidos utilizando a t?cnica proposta com alguns c?digos turbo da literatura corroboram com a proposi??o deste trabalho que os limitantes expurgados obtidos s?o apertados e os algoritmos completamente matriciais s?o facilmente implementados em qualquer software de programa??o simb?lica

Made available in DSpace on 2016-08-17T14:53:10Z (GMT). No. of bitstreams: 1 Bernardo Nogueira Neto.pdf: 936161 bytes, checksum: 8cae25f84e9e07f78244f10040d97213 (MD5) Previous issue date: 2010-06-04 / In this dissertation, we explore the application of the Theory of System Identification to develop a mathematical model that represents the behavior of the electrical current rectifier used to power the traction motors of a locomotive diesel-electric, based on mathematical manipulation of its temperature data. It was developed an ARX model and, using the aid of the computational system identification toolbox of MATLAB ®, simulations were made for estimation and validation of the order of the model that best describes the dynamic response of the physical system. The input and output data of the rectifier are obtained from the direct measurement of the output current of main generator, which is the input, and from the measurement of the temperature on the rectifier, which corresponds to the output. These data are used in developing the mathematical model of the system under study. The goal is to obtain a model that best represents the behavior of the main engine rectifier, allowing control actions to optimize the relation between supplied and required power of locomotive, so that downtime events resulting from self-protection mechanisms against high temperatures during operation of the rectifier in severe operations are minimized. As a result, we have a more advanced and efficient model that allows a strategy to monitoring the equipment at the various points of operation, considering their variation and ensuring a stable and safety locomotive's operation. / Nesta dissertação, aborda-se a aplicação da Teoria de Identificação de Sistemas para escolha de um modelo polinomial, que represente o comportamento do retificador de corrente elétrica, utilizado para suprimento dos motores de tração de uma locomotiva dieselelétrica, a partir da manipulação matemática de seus dados de temperatura. Desenvolve-se um modelo ARX e, com auxílio da ferramenta computacional do toolbox de identificação do MATLAB®, são feitas simulações para estimação e validação da ordem do modelo que melhor descreva a resposta dinâmica do sistema físico. Os dados de entrada e saída do retificador são obtidos a partir da medição direta da corrente de saída do gerador principal, que corresponde à entrada, e da medição da temperatura na bancada retificadora, que corresponde à saída. Estes dados são utilizados na elaboração do modelo matemático do sistema em estudo. Busca-se obter um modelo matemático que represente o comportamento do retificador principal da locomotiva, possibilitando ações de controle para uma melhor otimização entre a potência fornecida e a requerida para tração, de tal forma que os eventos de downtime, decorrentes da atuação da proteção contra altas temperaturas do retificador durante operações em regimes severos, sejam minimizados. Como resultado, tem-se um método mais avançado, eficaz e que possibilita uma estratégia de monitoramento nos diversos pontos de operação do equipamento, considerando suas variações e garantindo uma operação estável e segura do sistema.

Actuellement, la presbytie peut être corrigée chirurgicalement à l’aide d’implants intraoculaires réfractifs ou diffractifs multifocaux (chirurgie endoculaire invasive et irréversible) ou en intracornéen avec une correction multifocale réfractive (correction laser irréversible, ou insertion d’un implant dans le stroma). L’objectif de ce travail est de développer un nouvel implant permettant de corriger la presbytie, qui allie l’innocuité et la réversibilité d’une correction intracornéenne, à l’efficacité du diffractif. Le design des profils optiques bifocaux a été permis grâce au développement d’outils de simulation optique. Les efficacités de diffraction sont calculées à partir de la propagation du champ électrique par spectre angulaire. La qualité optique est déterminée d’après les simulations de Fonction de Transfert de Modulation obtenues sous Zemax. Des simulations de rendu d’images permettent de visualiser les effets de différents profils envisagés. Les paramètres critiques du design optique sont déterminés. Le choix du matériau dépend des contraintes de biocompatibilité de l’implant et des techniques de fabrication. La solution retenue est un hydrogel à forte teneur en eau, couplé à une nouvelle architecture de l’implant. L’hydrogel est obtenu par polymérisation radicalaire de macromonomères difonctionnels de poly(éthylène glycol) de masses molaires de l’ordre de 8000 g.mol‐1 qui conduisent à des propriétés mécaniques et une perméabilité aux nutriments compatibles avec l’application. La réalisation, la stérilisation et la caractérisation optique de prototypes ont abouti à la preuve du concept d’un implant bifocal diffractif intracornéen / Presbyopia can be corrected with surgery by means of refractive or diffractive multifocal intraocular lenses (which imply an irreversible and invasive endocular surgery) or by intracorneal multifocal refractive correction (irreversible laser correction, or insertion of an intrastromal implant). This work aims at developing a new implant to correct presbyopia, which takes advantage of both the harmlessness and the reversibility of an intracorneal correction, and the efficiency of diffractive optics. The design of the bifocal optical profiles was based on the development of optical simulation tools. The diffractive efficiencies are calculated from the distribution of the electric field with the method of angular spectrum. The optical quality is determined according to the simulations of Modulation Transfer Function obtained with Zemax. Images simulations show the effects of the different profiles studied. The critical parameters of the optical design are also determined. The choice of the material depends on several constraints such as biocompatibility and techniques of manufacturing. The adopted solution relies on the used of an hydrogel with high water content and the design of a new implant architecture. The hydrogel is obtained by radical polymerization of difunctional macromonomers of poly(ethylene glycol) with molar masses around 8000 g.mol‐1, allowing mechanical properties and permeability to nutriments compatible with the application. The realization, the sterilization and the characterization of prototypes showed the proof of the concept of a diffractive bifocal intracorneal implant

Les modulateurs electro-optiques (EOM) sont des composants importants dans les systemes de telecommunication. Malheureusement, ils ne sont pas parfaitement stables dans Ie temps et leur fonction de transfert (TF) a tendance a se decaler durant Ie temps d'operation. La derive de la TF d'un EOM peut etre expliquee par des effets differents tels que les changements de la temperature ambiante, de la polarisation ou de l'efficacite de couplage optique. La these presente une methode de me sure de la derive de la TF de I'EOM (a grace d'estimation du comportement non-linearite du modulateur, NLl), et propose deux systemes instrumentaux pour ameliorer son fonctionnement: Ie premier pour contraler la temperature de l'EOM, il est possible de forcement reduire cette derive, Ie deuxieme pour compenser la derive (la stabilisation du point de fonctionnement autour du point quadrature est obtenue jusqu'au 0,22% ou la fluctuation de phase est de 0,44°). Ces techniques ont ete aussi appliquees pour ameliorer Ie comportement d'un oscillateur optoelectronique haute frequence. Nous avons realise les experiences par asservissement de temperature, par compensation de la derive. Pour une stabilite de I'OEO a court terme, la stabilisation de la frequence est amelioree jusqu'a 28% en utilisant un contrale de temperature seule etjusqu'a 71 % avec les deux processus. Pour une stabilite de l'OEO a long terme, en deux cas sans contra Ie, I'OEO fonctionne pendant 3 heures et pendant 1,6 heure (puis I'OEO ne marche plus). En utilisant une compensation de la derive de la TF de I'EOM, l'OEO fonctionne bien meme apres 7 heures et 8 heures. / Electro-optic modulators (EOM) are important components in telecommunication systems. Unfortunately, there are some physical effects like ambient temperature variations, polarization and inherent photorefractive coefficient which are due to the drift of EOM transfer function (TF). The thesis presents a method for measuring the drift of the TF of EOM (based on a nonlinearity behavior of modulator - NLl), and propose two instrumentation systems to improve their operations: one to control the temperature of EOM, it is possible to significantly reduce the drift, the second to compensate the drift (the bias point stabilization around the quadrature point is obtained up to 0.22% or the phase fluctuation is 0.44°). These techniques have been applied to improve the behavior of an optoelectronic oscillator high frequency. Four experiments are carried out by controlling the EOM temperature and by compensating the EOM drift. For a short-term stability of OEO, the stabilization of the frequency is obtained with 28% with using a temperature control, up to 71 % with the two techniques. For a long-term stability of OEO, in cases without any control, the OEO has worked for 3 hours and for 1.6 hours (then, the OEO does not work anymore). By using the drift compensation of the EOM FT, the OEO has correctly worked even more than 7 hours and 8 hours.