O trabalho ressalta a importância do tema Qualidade da Energia (QE) e apresenta alguns dos diversos distúrbios responsáveis pelo seu comprometimento. Tais distúrbios, quando presentes em um dado sistema elétrico, podem causar sérios danos tanto aos equipamentos de medição e controle pertencentes ao fornecedor de energia, como também aos equipamentos mais sensíveis pertencentes aos usuários deste sistema. Os fenômenos mais freqüentes no estudo da QE foram destacados, bem como suas definições. Conhecidos estes fenômenos, implementou-se um algoritmo computacional, utilizando como ferramenta a transformada rápida de Fourier janelada (Windowed Fast Fourier Transform - WFFT) - TRFJ, para identificar e classificar estas perturbações em um dado sistema elétrico de distribuição. Para que o resultado fosse o mais próximo possível de uma situação real, utilizou-se para análise, um sistema elétrico real de distribuição da CPFL (Companhia Paulista de Força e Luz). Sobre o referido sistema elétrico foram simulados casos de afundamentos, elevações e oscilações de tensão, criando assim, um banco de dados para, posteriormente, podermos testar e validar o algoritmo computacional implementado, na identificação destes fenômenos. Este algoritmo permite variarmos tanto o tamanho da janela, quanto a freqüência amostral do sinal. Neste estudo em particular, utilizamos tamanhos de janelas de 1 ciclo e 1/2 ciclo, e freqüência amostral do sinal, de 7,68 kHz (inicialmente utilizada para obtenção do banco de dados), e 0,769 kHz, aproximadamente. Como resultados, obtivemos que o tamanho da janela de dados para esta implementação, não apresentou diferenças significativas na análise, quando comparados com os dados iniciais. Entretanto, observa-se um comportamento contrário com a variação da taxa amostral, ou seja, quanto menor a freqüência amostral empregada, maior a perda de informações importantes em relação ao sinal inicial. / The present work at lines the importance of Power Quality (PQ) and it presents some of the many disturbances related to it and its implications. Such disturbances, as presented in an electrical system, can cause serious damages in the measurements, in the control of power utility equipments, as well as in sensitive equipments. The most frequent phenomena in the study of PQ had been emphasized, as well as its definitions. Once these phenomena are known, a computational algorithm was implemented, using the Windowed Fast Fourier Transform - WFFT as a tool to identify and classify these disturbances, considering electrical distribution system. In order to have the best situation for the test, a real electrical distribution system from CPFL (Companhia Paulista de Força e Luz) utility was simulated. Some cases of voltage sag, voltage swell and oscillatory transient were simulated, creating a data base to test and validate the computational algorithm implemented. This algorithm allows changing the window length as the sample rate. For the proposed study, 1 cycle and 1/2 cycle as window length was analyzed with sample rate of 7,68 kHz (initially used for obtaining the data base), and 0,769 kHz. It was observed that the window length studied did not influence significantly the proposed analysis concerning PQ. However, concerning the sample rate, it was observed a deterioration of the analysis with the 0,769 kHz rate.

Le présent travail de thèse a pour but d'automatiser la détection des défauts de tissage, à l'aide d'un système informatisé fondé sur le traitement d'image. La structure du tissu, de par sa fabrication, est périodique par répétition de l'armure (dessin d'entrecroisement). L'apparition d'un défaut dans le tissu entraîne la destruction de cette périodicité. de ce fait, le défaut de tissage peut être détecté par la surveillance en continu de la structure du tissu pendant le tissage. Dans notre travail, les techniques de transformée rapide de Fourier et la corrélation croisée sont d'abord mises en œuvre, afin d'examiner les caractéristiques de régularité de la structure de l'image du tissu dans le domaine fréquentiel. Dans un deuxième temps, afin d'améliorer l'efficacité de la technique et d'éviter le problème d'erreur de détection, une opération de seuillage a été implémentée en utilisant un filtre de sélection de niveau. Au moyen de ce filtre, la technique mise en place est capable de détecter uniquement les défauts réels et de mettre en évidence leurs dimensions exactes. L'ensemble de cette procédure a été implémentée au moyen du progiciel Matlab ou Scilab. Puis l'ensemble des procédures a été mis en œuvre sur un tissu simulé, afin de comprendre le comportement du spectre de fréquences, de déterminer et d'optimiser les paramètres de détection les plus importants. Dans toutes ces procédures, le niveau de bruit a bien sûr été pris en compte. L'efficacité des procédures ayant été testée sur des tissus simulés dans un premier temps, elles ont été appliquées aux tissus sortant de machines à tisser. Ainsi, des échantillons contenant divers défauts ainsi que des échantillons écrus, unis ou de diverses couleurs ont pu être testés. Les techniques proposées ont aussi dans ces divers cas montré leur efficacité. En conclusion, un prototype fondé sur un dispositif de vision pour l'inspection des tissus sur métier à tisser en temps réel est proposé. / The purpose of this thesis is to automate the online detection of weaving defects by a computerized system based on image processing software. Obviously, fabric inspection has an importance to prevent risk of delivering inferior quality product. Until recently, the visual defect detection is still under taken offline and manually by humans with many drawbacks such as tiredness, boredom, and, inattentiveness. Fortunately, the continuous development in computer technology introduces the online automated fabric inspection as an effective alternative. Because the defect-free fabric has a periodic regular structure, the occurrence of a defect in the fabric breaks the regular structure. Therefore, the fabric defects can be detected by monitoring fabric structure. In our work, Fast Fourier Transform and Cross-correlation techniques, i.e. linear operations, are first implemented to examine the structure regularity features of the fabric image in the frequency domain. To improve the efficiency of the technique and overcome the problem of detection errors, further thresholding operation is implemented using a level selection filter. Through this filter, the technique is able to detect only the actual or real defects and highlight their exact dimensions. A software package such as Matlab or Scilab is used for this procedure. It is implemented firstly on a simulated plain fabric to determine the most important parameters during the process of defect detection and then to optimize each of them even considering noise. To verify the success of the technique, it is implemented on real plain fabric samples with different colours containing various defects. Finally, a vision-based fabric inspection prototype that could be accomplished on-loom to inspect the fabric under construction with 100% coverage is proposed.

The Decimator, an SED Systems Ltd. product, is a PCI slot card that performs both time and frequency domain measurements of given input signals. It is essentially a more economical version of a bench spectrum analyzer or oscilloscope, with a PC interface. Several issues limit the speed and accuracy of the results of the Decimator, and the study of these issues is the focus of this thesis. These issues, including but not limited to, are as follows: 1) Imbalances between the received In-phase and Quadrature-phase channels; 2) The FFT and Windowing functions are performed by a microcontroller, but it is desired that they be migrated to an FPGA. While solutions to improve the first issue is being implemented and verified, the second issue is not one of simply reducing a source of error. The second issue requires a cost-benefit analysis on the migration of these signal processing algorithms from an ARM microcontroller to a Xilinx FPGA.

The Decimator, an SED Systems Ltd. product, is a PCI slot card that performs both time and frequency domain measurements of given input signals. It is essentially a more economical version of a bench spectrum analyzer or oscilloscope, with a PC interface. Several issues limit the speed and accuracy of the results of the Decimator, and the study of these issues is the focus of this thesis. These issues, including but not limited to, are as follows: 1) Imbalances between the received In-phase and Quadrature-phase channels; 2) The FFT and Windowing functions are performed by a microcontroller, but it is desired that they be migrated to an FPGA. While solutions to improve the first issue is being implemented and verified, the second issue is not one of simply reducing a source of error. The second issue requires a cost-benefit analysis on the migration of these signal processing algorithms from an ARM microcontroller to a Xilinx FPGA.

In this thesis, the conventional low voltage power line communication methods is investigated in the axis of automated meter reading applications and the classification and experimental verification of common noise types for low voltage power line communication network. The investigated system provides the real time transmission of electricity consumption data recorded by electricity meters, initially to a local computer via a low voltage line through a low speed PLC (Power Line Carrier) environment and subsequently to a corporate network through a high speed data transmission medium. The automated meter system provides a more effective tracking and data acquisition, a more detailed and vigorous knowledge about consumer behavior for subscriber assessment in electricity distribution in association with a brand new management and system supervision concept in electricity distribution control and management technology. The theoretical studies are experimentally verified for the Turkish low voltage power infrastructure through laboratory experiments performed in METU Electrical and Electronics Engineering Department, Electrical Machines and Drives Laboratory and R&amp / D Laboratories of MAKEL facilities in Hadimk&ouml / y. The single phase voltage of the mains line between the phase and neutral is monitored to exhibit the disturbing effects of various noise sources. The resulting voltage spectrum is logged by using digital data acquisition devices in time and frequency domain. The waveforms are converted to frequency domain using the Fast Fourier Transform (FFT) functions of the MATLAB. The experimental results are compared to the theoretical findings obtained through literature survey.

We present an efficient method to compute the electrostatic fields, torques and forces in dipolar systems, which is based on the fast Fourier transform for nonequispaced data (NFFT). We consider 3d-periodic, 2d-periodic, 1d-periodic as well as 0d-periodic (open) boundary conditions. The method is based on the corresponding Ewald formulas, which immediately lead to an efficient algorithm only in the 3d-periodic case. In the other cases we apply the NFFT based fast summation in order to approximate the contributions of the nonperiodic dimensions in Fourier space. This is done by regularizing or periodizing the involved functions, which depend on the distances of the particles regarding the nonperiodic dimensions. The final algorithm enables a unified treatment of all types of periodic boundary conditions, for which only the precomputation step has to be adjusted.

In the past, we could rely on technology scaling and new micro-architectural techniques to improve the performance of processors. Nowadays, both of these methods are reaching their limits. The primary concern in future architectures with billions of transistors on a chip and limited power budgets is power/energy efficiency. Full-custom design of application-specific cores can yield up to two orders of magnitude better power efficiency over conventional general-purpose cores. However, a tremendous design effort is required in integrating a new accelerator for each new application. In this dissertation, we present the design of specialized compute fabrics that maintain the efficiency of full custom hardware while providing enough flexibility to execute a whole class of coarse-grain operations. The broad vision is to develop integrated and specialized hardware/software solutions that are co-optimized and co-designed across all layers ranging from the basic hardware foundations all the way to the application programming support through standard linear algebra libraries. We try to address these issues specifically in the context of dense linear algebra applications. In the process, we pursue the main questions that architects will face while designing such accelerators. How broad is this class of applications that the accelerator can support? What are the limiting factors that prevent utilization of these accelerators on the chip? What is the maximum achievable performance/efficiency? Answering these questions requires expertise and careful codesign of the algorithms and the architecture to select the best possible components, datapaths, and data movement patterns resulting in a more efficient hardware-software codesign. In some cases, codesign reduces complexities that are imposed on the algorithm side due to the initial limitations in the architectures. We design a specialized Linear Algebra Processor (LAP) architecture and discuss the details of mapping of matrix-matrix multiplication onto it. We further verify the flexibility of our design for computing a broad class of linear algebra kernels. We conclude that this architecture can perform a broad range of matrix-matrix operations as complex as matrix factorizations, and even Fast Fourier Transforms (FFTs), while maintaining its ASIC level efficiency. We present a power-performance model that compares state-of-the-art CPUs and GPUs with our design. Our power-performance model reveals sources of inefficiencies in CPUs and GPUs. We demonstrate how to overcome such inefficiencies in the process of designing our LAP. As we progress through this dissertation, we introduce modifications of the original matrix-matrix multiplication engine to facilitate the mapping of more complex operations. We observe the resulting performance and efficiencies on the modified engine using our power estimation methodology. When compared to other conventional architectures for linear algebra applications and FFT, our LAP is over an order of magnitude better in terms of power efficiency. Based on our estimations, up to 55 and 25 GFLOPS/W single- and double-precision efficiencies are achievable on a single chip in standard 45nm technology. / text

Η μετάδοση δεδομένων µέσω ασύρματων δικτύων αποτελεί ένα από τα σημαντικότερα αντικείμενα μελέτης στον χώρο των τηλεπικοινωνιών τα τελευταία χρόνια. Λόγω του μεγάλου όγκου δεδομένων που απαιτείται για μετάδοση μέσα σε πολύ μικρό χρονικό διάστημα, απαιτείται η εφαρμογή όσο το δυνατόν αποδοτικότερης κωδικοποίησης και διαμόρφωσης, σκοπεύοντας παράλληλα στην επίτευξη μικρής κατανάλωσης ισχύος και ταυτόχρονα υψηλού throughput. Η παρούσα διπλωματική εργασία πραγματεύεται τη μελέτη, σχεδίαση και την υλοποίηση της αρχιτεκτονικής ενός OFDM συστήματος, βασισμένο στις προδιαγραφές που ορίζει το πρότυπο IEEE802.11. Το πρότυπο αυτό επιλέχθηκε, καθώς τυγχάνει ευρείας αποδοχής, όσον αφορά τη μετάδοση δεδομένων σε ασύρματα δίκτυα. Στα πλαίσια της διπλωματικής παρουσιάζονται εξομοιώσεις και μετρήσεις για την εύρεση των κατάλληλων χαρακτηριστικών και του τύπου των φίλτρων που χρησιμοποιούνται σε ένα OFDM σύστημα, ενώ εισάγονται δύο νέες αρχιτεκτονικές του τμήματος κατασκευής του OFDM συμβόλου, καθώς και τα αποτελέσματα της σύνθεσης αυτών. Ιδιαίτερη βαρύτητα δίνεται στη σχεδίαση και την υλοποίηση των αρχιτεκτονικών αυτών με σκοπό την ενσωμάτωσή τους σε συσκευές FPGA. / Transmission of data through wireless networks is one of the most important aspects in the study telecommunications systems. The large volume of data that needs to be transmitted in a very small time interval has resulted in the need to make the system more efficient while increasing the throughput. This has been accomplished through more efficient coding, reduction of the power consumption and through the use of modulation. This masters thesis deals with the study, design and implementation of the typical OFDM system architecture, based on the standard IEEE802.11. This standard was chosen because of its wide acceptance and use regarding transmission of data in wireless networks. The thesis details the adjustments and measurements needed for determining suitable characteristics and the types of filters required in an OFDM system. Additionally two new OFDM system architectures arte introduced aiming at reducing the overall power consumption and the complexity of the formulae used for symbol construction. Particular importance is given in the designing and the implementation of these architectures regarding their incorporation in FPGA devices.

The computation of the Coulomb potentials and forces in charged particle systems under 3d-periodic boundary conditions is possible in an efficient way by utilizing the Ewald summation formulas and applying the fast Fourier transform (FFT). In this paper we consider the particle-particle NFFT (P2NFFT) approach, which is based on the fast Fourier transform for nonequispaced data (NFFT) and compare the error behaviors regarding different window functions, which are used in order to approximate the given continuous charge distribution by a mesh based charge density. While typically B-splines are applied in the scope of particle mesh methods, we consider for the first time also an approximation by Bessel functions. We show how the resulting root mean square errors in the forces can be predicted precisely and efficiently. The results show that if the parameters are tuned appropriately the Bessel window function can keep up with the B-spline window and is in many cases even the better choice with respect to computational costs.

The objective of the proposed research is to develop structural changes to the design and application of electromagnetic (EM) sensing models to more efficiently and accurately invert EM measurements to extract parameters for applications such as landmine detection. Two different acquisition modalities are addressed in this research: ground-penetrating radar (GPR) and electromagnetic induction (EMI) sensors. The models needed for practical three-dimensional (3D) spatial imaging typically become impractically large, with up to seven dimensions of parameters that need to be extracted. These parameters include, but are not limited to target type, 3D location, and 3D orientation. The new special structures for these models exploit properties such as shift invariance and tensor representation, which can be combined with strategic inversion techniques, including the Fast Fourier Transform and semidefinite programming. The structures dramatically reduce the amount of computation and can eliminate the need to store up to five dimensions of parameters while still accurately estimating them.