Title Promoter Affiliations Abstract
"Spherical Wave Based Macromodelling for EMC/EMI System Analysis" "Georges Gielen" "ESAT - MICAS, Microelectronics and Sensors, Electrical Engineering Technology (ESAT), Bruges Campus, ESAT- TELEMIC, Telecommunications and Microwaves" "Nowadays, full-wave electromagnetic simulation tools are widely used in antenna design, and their employment in the assessment of electronic designs with respect to ElectroMagnetic Compatibility (EMC) has increased significantly over the past years, owing to an exponential increase in the integration of functionality and the clock rates of new designs. However, several developers are still rather reserved about their engagement to embrace the use of full-wave simulations. This trend can be attributed to four factors: (i) the investment cost of a full-wave simulation package, (ii) the expertise of R&D engineers, (iii) the limited availability of component models due to conflicts with confidentiality, and (iv) the computational cost of full-wave simulations. Due to the inherent complexity in today’s electronic designs and their clock rates, often dedicated computing servers are indispensable to manage the run-time and computational resources of full-wave simulations. In this thesis, we therefore aim to develop a macromodelling technique, focussed at system-level EMC analysis, which is independent of any full-wave solver, which has a very low computational cost, and which bears close resemblance to the topology of circuit simulators.We obtain these goals by employing a Generalized Scattering matrix (GS-matrix) formulation based on a Spherical Wave Expansion (SWE) of the electromagnetic fields radiated by a Device Under Test (DUT). To derive efficiently such models, a new technique based on a reduction to Chebyshev polynomials is described, which allows one to compute models in optimal time with very high accuracy. In the scope of system-level EMC analysis, two new truncation criteria are formulated, either applicable to fields sampled in the near-field region of a DUT or to fields sampled in the far-field region of a DUT. Based on these truncation criteria, it is observed that in order to accurately model radiated fields close to a DUT, high-order SWEs are indispensable. Models which have been obtained using this approach are gathered in a common model library, which is accessible to a custom GS-matrix based simulation engine. This simulation tool is designed analogous to a circuit simulation topology, and thus allows its users to compute full-wave interactions between multiple DUTs by ‘plug-and-play’ with models available in a component library. Consequently, the presented framework also fits into existing circuit solvers, and significantly increases the accuracy of these solvers by taking into account full-wave phenomena. The additional background computations which form the core of the GS-matrix based simulation tool are optimized by observing that the necessary operators acting on SWEs are sparse. Additionally, it has been observed that the number of spherical waves needed to accurately compute interactions between multiple GS-matrix instances only forms a subpart of the number of spherical waves needed to accurately represent the fields radiated by a DUT in its near-field region. In combination with numerous additional optimizations, full-wave simulations run in the order of seconds on a simple home laptop. To further show the computational efficiency of the proposed simulation tool, support has been added for a cylindrical scan of a DUT by an antenna. This setup, often corresponding to more than one thousand full-wave simulations, only takes up about one minute on a simple laptop.In using the developed simulation tool, we have encountered that in certain cases multiple reflections between DUTs can compromise its accuracy. However, deriving the scattering parameters for incident spherical waves is not straightforward due to the lack of support for spherical wave excitations in most full-wave simulation tools. Therefore, we have derived an alternative approach based on plane wave illuminations. An appropriate linear combination of plane wave illuminations allows us to mimic, to a sufficient degree of accuracy, arbitrary spherical wave excitations. We have validated this approach by comparing scattered field patterns for arbitrary incident plane waves to full-wave simulations. Subsequently, we have applied the scattering parameters for incident spherical waves in simulations where multiple reflections between DUTs play a significant role. Finally, the simulation tool is extended to support an infinite perfectly conducting plane underneath DUTs. Depending on the distance of the DUTs above this conducting plane, an approach based on mirroring fields sampled on a hemispherical surface or an alternative ‘image theory’ is applied. We have, however, encountered inaccuracies in the computations when image theory is applied. These inaccuracies are attributed to forward scattering of fields reflected by the conducting plane. Nevertheless, using the knowledge of the scattering parameters for incident spherical waves, the deviation due to forward scattering of the reflected fields is tackled in a straightforward manner, and we have subsequently observed a very good correspondence between the GS-matrix simulation tool and a full-wave simulation.We have thus, in summary, developed a numerically stable and efficient circuit based topology that can be employed to study a wide range of system-level EMC problems. Several realistic applications are discussed to validate the proposed methodology."
"System design and image reconstruction for unconventional geometry PET scanner." "Johan Nuyts" "Nuclear Medicine & Molecular Imaging" "Nowadays, tomographic imaging is one of the mast fundamental tools to early and accurate diagnosis and evaluation of the therapy effectiveness of an important numbers of diseases, such as cardiac or oncological. In this thesis project, the development of effective reconstruction algorithms will be explored, in order to obtain high quality and accurate images in a reasonable period of time (seconds). Although this thesis project objective is the PET image reconstruction, the mathematical methods can be easily applied to other techniques like Single-Photon Emission Computed Tomography (SPECT) or Computed Tomography (CT scan). The objective of Positron Emission Tomography (PET) is to generate images of a radioactive tracer inside certain parts of the human body. This tracer is a beta+ (positron), and its interaction with the electrons from the human body generate two photons of 5llkeV that propagate in opposite directions and that can be detected in coincidence in a PET detector. The advantages of this technique are the fact that it is not invasive and offers inner body metabolic information a useful tool to diagnosis and follow-up of the disease of interest. Data acquired in PET systems need to be processed in a reconstruction process, involving analytical algorithms as well as statistical iterative ones. The most common algorithm used in PET image reconstruction is the Maximum Likelihood Expectation Maximization (MLEM). This algorithm needs to wait until the data acquisition has ended to begin the reconstruction process, and also needs a complete matrix modeling of the system, that makes the processing slow. The majority of emission photons are absorbed or scattered from their original trajectory due to media attenuation. This leads to a considerably quantity of events lost or deviating from their theoretical collinearity, implying the addition of undesired noise and artifacts to the image. These undesired effects have to be corrected for by using a reconstruction algorithms that accounts for these effects via prior information as CT scans or via using Time of Flight (TOF) information as MLPA algorithm does. The implementation of scatter and attenuation correction and the normalization in the sensitivity are essential in PET detectors in order to improve image quality. PET detectors crystals and systems used to present systematic and random errors that also have to be corrected. Depending on the scanner geometry certain lines-of-response (LOR) efficiencies will be affected, producing subsequent count-rate variability that leads on a inaccurate image. This undesired effect can be corrected using normalization algorithms that exploits geometry symmetry prior knowledge to estimate the correction factors in a fast and efficient way. The most important parameters for characterizing a PET image are the spatial resolution, the sensitivity and the quantification of real activity. In a few words, spatial resolution allows the localization of small lesions and identifications of individual lesions even if they are close to each other, a high sensitivity means that good image quality can be obtained even with low tracer doses, and activity quantification is the capacity of recovering the real activity in regions of interest in the image. Monte Carlo simulations are an important tool to evaluate and optimize the correction methods in image PET. These simulations are generally used to study the photon trajectories and their interactions with the tissues in the human body. By defining the simulated detector characteristics based on calibration measurements, very realistic simulations are obtained. New designed PETs systems are being designed, such as open-geometry systems or systems dedicated to imaging a particular organ. For several reasons, these systems are often unable to perform transmission measurements. In spite of these difficulties, such dedicated PET systems have many advantages, including better spatial resolution, a reduced production cost and optimal performance for particular imaging tasks, enabling dose and/or scan time reduction. In our project, we aim to obtain a cardiac PET image, building a dedicated PET system that consist on four detector plates with TOF information. This system will have to deal with the limited angle in tomography, so an accurate implementation of normalization for the detector sensitivities and correction for attenuation and scatter will be decisive. New norrnalization algorithms will have to be designed to correct the artifacts in the image due to its certain geometry, making new assumptions that differs from the ring-normalization that commonly appears in the literature. Also, the fact of not having CT scan or MR to give further information for this corrections will make the TOF information decisive to correct attenuation and minimizing the loss of resolution. Also, the scatter correction will be studied and take into account for the heart tissue structure. At the end of this project, we aim to obtain a prototype that can offer real measurements from a cardiac phantom, and to explore new ways to correct image in limited angle TOF-PET systems"