The aim of thesis was to establish a methodology for 210Pb measurements by Accelerator Mass Spectrometry (AMS). The potential application is to measure 210Pb in people who have been exposed to radon. This will better our understanding of radon toxicity, which is not possible now with current radiometric and mass spectrometry techniques. The determination of 210Pb by AMS was done in two major studies 1) Studying Pb chemistry in a Cs+ sputter source used in AMS and 2) Evaluating 204,205 & 208Pb spikes for the quantification of 210Pb by isotope dilution. Pb chemistry was investigated using an 834 SIMS-type and a SO-110 Cs+ sputter source at the IsoTrace Laboratory and A.E Lalonde AMS facility, respectively. Different molecular anions of Pb were studied with the 834 SIMS-type Cs+ sputter source and the strongest molecular anion current of Pb and thus greatest ionization efficiency was achieved form the superhalogen PbF3-. The average 208PbF3- current was unaffected by varying the ratio of the fluorinating compounds (AgF2 and CsF) packed into a target. The average current of 208PbF3- was reproducibly increased by chemically mixing the targets of AgF2, CsF and PbF2 in concentrated HF rather than mechanically mixing them the powders with a stir rod. The count rate of 210Pb reproducibly increased by a factor of 20 when μg quantities of PbF2 were present in mg AMS targets compared to AMS targets that had pg quantities of PbF2. The average current of 208PbF3- for pure PbF2 targets in an SO-100 Cs+ sputter source was reproducibly increased when the Cs+ flux was decreased by a factor of 10. This phase of my work maximized the overall efficiency of PbF3-, to a value of 1.8x10-10 ±8x10-11s-, which was a key first step in the measurement by AMS. Then isotope dilution was tested to quantify 210Pb and the next stage of my work evaluated the use of 204,205 & 208Pb spikes. 210Pb was measured in the +3 charge state by isotope dilution assays using 204,205 & 208Pb spikes. 204Pb+3 reproducibly suffered from the molecular interference from 68Zn3+3, which could not be easily removed without negatively impacting the detection limit for 210Pb. 205Pb+3 continually suffered from 205Tl+3 interference which could be readily be removed but not without negatively affecting the II detection limit for 210Pb. 208Pb+3 suffers from no molecular interferences but if a large amount of 208Pb is needed to swamp the Pb in a sample, this could limit the detection limit for 210Pb as the abundance sensitivity is 210Pb/208Pb=1.3×10-12. A calibration curve is required when 208Pb is used as a spike due to a difference in collection efficiency of a Faraday cup, where 208Pb+3 is detected and the gas ionization chamber, where 210Pb+3 is detected. The quantification of 210Pb with 208Pb as a spike yielded a detection limit of 4.4mBq at the IsoTrace facility. A theoretical detection limit of ≤0.11mBq is expected at the A.E Lalonde AMS facility. The expected detection limit at the A.E Lalonde AMS facility is on par with α-spectroscopy but AMS samples can be counted in less than 1 hour whereas alpha spectrometry samples must be counted for about 1 day.

Les klystrons sont des tubes électroniques dans lesquels un ou plusieurs faisceaux d'électrons sont modulés afin de d'amplifier un signal radiofréquence. Pour ce faire, des cavités passives constituant la ligne d’interaction du klystron échangent de l’énergie électromagnétique avec le faisceau. Il en résulte une modification périodique de la vitesse des électrons, qui sont regroupés en paquets. A la fin de la ligne d’interaction, cette énergie est partiellement transmise à un circuit radiofréquence externe. Afin d’améliorer les performances des klystrons et d’augmenter leur rendement énergétique, une nouvelle architecture, inspirée des RFQ (Radio-Frequency Quadrupole), a été proposée. Il s’agit d’utiliser un nombre plus important de cavités, faiblement couplées au faisceau, et permettant la mise en paquets des électrons de façon très progressive et économe en énergie. Dans l’optique de tester cette nouvelle architecture, le point de départ de ce projet a été la modification de la ligne d’interaction d’un klystron existant. Les éléments d’origines ont été modélisés avec différents programmes de simulation (AJ-Disk, Klys2D et MAGIC2D). Plusieurs architectures ont été ensuite simulées avec ces programmes, afin d’améliorer le rendement énergétique de ce klystron. Une fois que les paramètres électromagnétiques des cavités ont été déterminés, il a été nécessaire de concevoir les pièces constituant cette nouvelle ligne. Le développement de ce nouveau klystron a imposé la création d’un système d’accord des cavités en fréquence, et d’une méthode de titanage, afin d’empêcher la création de phénomène de multipactor. Pour ce faire, deux séries de prototypes ont été conçues, fabriquées et testées. Cette étape a également permis la vérification des méthodes de brasage prévues avec les pièces finales. Les différents éléments de la ligne d’interaction ont été réalisés par un partenaire industriel extérieur au projet. Ces pièces ont été ensuite assemblées à Thales Electron Devices Vélizy, avant que le nouveau klystron soit testé sur un banc dédié. / Klystrons are vacuum tubes that amplifiy microwave signals with the modulation of its beam electrons velocities. The electron beam and the klystron cavities exchange electromagnetic energy. This energy exchanges accelerate or decelerate electrons, and create bunches. The electromagnetic energy carried by these bunches is transferred to an external radio-frequency circuit in the last cavity. In the frame of the improvement of klystrons efficiency, a new design inspired by RFQ (Radio-Frequency Quadrupole) had been proposed. With more cavities, poorly-coupled with the beam, the electron bunching is expected to be smoother and more efficient. The purpose of this project was the test of this new design by replacing the interaction line of an existing klystron. This klystron was simulated with several codes such as AJ-Disk, Klys2D and MAGIC2D. We then used this simulation software to try new designs to improve this klystron efficiency. Once we chose the new cavities electromagnetic properties, we designed the interaction line parts.This klystron development involved the design of a new tuning system for the cavities frequencies, and a new titanium deposition method, in order to prevent multipactor effect. This step needed the design and the fabrication of two series of cavities prototypes to test our new parts and methods. The brazing method was also checked with these cavities. The final interaction line parts were fabricated by a subcontractor, and then assembled and brazed at Thales Electron Devices Vélizy. This new klystron was finally tested on a dedicated conditioning and testing bench.

abstract: The advances of Deep Learning (DL) achieved recently have successfully demonstrated its great potential of surpassing or close to human-level performance across multiple domains. Consequently, there exists a rising demand to deploy state-of-the-art DL algorithms, e.g., Deep Neural Networks (DNN), in real-world applications to release labors from repetitive work. On the one hand, the impressive performance achieved by the DNN normally accompanies with the drawbacks of intensive memory and power usage due to enormous model size and high computation workload, which significantly hampers their deployment on the resource-limited cyber-physical systems or edge devices. Thus, the urgent demand for enhancing the inference efficiency of DNN has also great research interests across various communities. On the other hand, scientists and engineers still have insufficient knowledge about the principles of DNN which makes it mostly be treated as a black-box. Under such circumstance, DNN is like "the sword of Damocles" where its security or fault-tolerance capability is an essential concern which cannot be circumvented. Motivated by the aforementioned concerns, this dissertation comprehensively investigates the emerging efficiency and security issues of DNNs, from both software and hardware design perspectives. From the efficiency perspective, as the foundation technique for efficient inference of target DNN, the model compression via quantization is elaborated. In order to maximize the inference performance boost, the deployment of quantized DNN on the revolutionary Computing-in-Memory based neural accelerator is presented in a cross-layer (device/circuit/system) fashion. From the security perspective, the well known adversarial attack is investigated spanning from its original input attack form (aka. Adversarial example generation) to its parameter attack variant. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2020

No description available.

This thesis deals with utilizing C2H technique for implementation algorithm on FPGA. Several structures of digital filters FIR and IIR are implemented within this work with usage of C2H. For such a comparison is in terms of FPGA resources utilized, the maximum frequency, latency, complexity of implementation and acceleration obtained to Nios II processor itself. Example for image processing using local operators implemented using C2h is also created to display the result on the LCD.

The thermodynamics inside the helium liquefier in the FREIA laboratory still contains many unknowns. The purpose of this project is to develop a theoretical model and implement it in MATLAB, with the help of the CoolProp library. This theoretical model of the FREIA liquefaction cycle aims at finding the unknown parameters not specified in the manual of the manufacturer, starting from the principle of enthalpy conservation. Inspiration was taken from the classical liquefaction cycles of Linde-Hampson, Claude and Collins. We developed a linear mathematical model for cycle components such as turboexpanders and heat exchangers, and a non-linear model for the liquefaction in the phase separator. Liquefaction yields of 10% and 6% were obtained in our model simulations, with and without liquid nitrogen pre-cooling respectively - similar to those in the FREIA liqueuefier within one percentage point. The sensors placed in FREIA showed similar pressure and temperature values, even though not every point could be verified due to the lack of sensors. We observed an increase of more than 50% in yield after adjustments of the heat exchanger design in the model, especially the first one. This constitutes a guideline for possible future improvements of the liquefier. / Termodynamiken bakom heliumförvätskaren i FREIA-laboratoriet innehåller fortfarande många okända aspekter. Detta kandidatarbete syftar till att utveckla en teoretisk modell och implementera den i MATLAB med hjälp av biblioteket CoolProp. Denna modell av FREIA:s förvätskningscykel syftar till att hitta de okända parametrar som inte specificerats av tillverkaren, och baserar sig på principen om entalpins bevarande. Inspiration togs från de klassiska förvätskningscyklerna Linde-Hampson, Claude och Collins. Vi utvecklade en linjär matematisk modell för cykelkomponenter såsom expansionsturbiner och värmeväxlare, och en icke-linjär modell för själva förvätskningen i fasseparatorn. En förvätskningsverkningsgrad på 10% och 6% uppnåddes i våra modellsimuleringar, med respektive utan förkylning med flytande kväve - liknande verkningsgraderna i FREIA- förvätskaren inom en procentenhet. Sensorerna placerade i FREIA visade på liknande tryck och temperaturer, även om bristen på sensorer gjorde att vi inte kunde bekräfta varje punkt. Vi observerade en ökning på mer än 50% i verkningsgrad efter att ha justerat värmeväxlardesignen något, speciellt för den första. Detta kan utgöra riktlinjer för var man framöver kan förbättra den faktiska förvätskaren. / <p>Subject reader/Ämnesgranskare: Roger Ruber</p>

Accelerator Mass Spectrometry (AMS) is a sensitive technique for the analysis of rare isotopes. Optimizing the output of the cesium-sputter ion source is a fundamental method for improving measurement precision, efficiency, and reliability. Several strategies for improving the ion source are discussed and lead to an understanding of the electrodynamics within the ion source to inform further improvement in design and operating parameters. At the Andr´e E. Lalonde Accelerator Mass Spectrometry Laboratory (Lalonde AMS), the High Voltage Engineering Europa (HVEE) SO-110C ion source was modelled using Integrated Engineering Software (IES)’s Lorentz-2E ion trajectory simulation software. Lorentz-2E incorporates the mutual space-charge interaction between the positively charged cesium ion beam and the sputtered negative ion beam. A critical component of this work was the development of the Rijke code. Rijke communicates with Lorentz-2E to initiate, generate, and run varied sequences of simulations, as well as analyze and record the input and output data in formats convenient for timely analysis. This software and its interconnection with Lorentz- 2E is described in extensive detail for a prospective user. Initial simulation work examined the effects of modifying various electrode geometries within the source such as the extraction cone, the target aperture, a simple cratered sample model as well as examining the effects of varying the cesium ion current. The self-repulsion of cesium was found to be important at currents of 250 µA and above. At high enough cesium currents, the expansion of the cesium beam is such that parts of it impinge outside the extents of the sample material. Through both simulation and experiment, it was demonstrated that this effect can be mitigated by either recessing (translating along the axis of symmetry away from the ionizer) the target holding the sample or by adjusting the potential difference between the target and ionizer. Experimentally, at routine settings (6 kV target to ionizer potential, 115 ◦C cesium oven temperature, and 35 keV output energy), a target recess of 1 mm gave the most stable and sustained output of 12C from graphite blanks. While the peak current was less than the unrecessed case, the total measured charge from the recessed target was higher. Cesium currents at these routine settings were found to be below the theoretical space-charge limited maximum. Using 10Be standards, a multi-dimensional experimental study examined the effects of increasing the cesium current, adjusting the target-ionizer potential from 4 to 11 kV, while also examining target recesses of 0 to 4 mm. Multiple combinations of these settings produced enhanced currents of 9Be2+, measured at the high-energy offset Faraday cup, as high as 13.5 µA. This was higher than previously observed, resulting in the most precise measurement of 10Be performed to date at Lalonde AMS. The electrodynamics within the ion source can be characterized as three competing processes: a) a strong locus of positive space charge located at the centre of the sample, depending primarily on the focusing of the cesium beam, which draws negative ions across the axis of symmetry; b) a bulk positive space charge external to the negative ion beam, depending primarily on the magnitude of the cesium current, draws the outer-most negative ions away from the axis; and c) the raw field from the electrode potentials and geometry which is mainly defocusing for negative ions. These effects are mitigated the most when the cesium beam is distributed across the entire sample surface with the additional critical benefit of maximizing the sample material accessed for sputtering. This thesis work has demonstrated that both the mutual and self space-charge interaction of the cesium and negative ion beams were critically important and that the use of the simulation software can inform both improved design and operation settings of the ion source.

The integral image representation of an image is important for a large number of modern image processing algorithms. Integral image representations can reduce computation and increase the operating speed of certain algorithms, improving real-time performance. Due to increasing demand for real-time image processing performance, an integral image architecture capable of accelerating the calculation based on the amount of available resources is presented. Use of the proposed accelerator allows for subsequent stages of a design to have data sooner and execute in parallel. It is shown here how, with some additional resources used in the Field Programmable Gate Array (FPGA), a speed increase is obtained by using a one-dimensional Systolic Array (SA) approach. Additionally, extra guidelines are given for further research in this area.

This thesis reports on a study to determine the viability of using a mobile tablet device as a brain concussion detection tool. The research builds upon the results of a prior method of collecting data for measuring motion sensitivity, where a user presses and releases a force sensor to balance a rising and falling line on a computer display. The motion sensitivity data collected using this force sensor device was shown to have less irregularity in persons with concussion. The MotionScan application, developed for this research, uses the accelerometer of a tablet device to record motor movement of a user while the user tries to control a free-moving ball on the tablet screen to trace a line. Data collection sessions were conducted with 20 participants, where researchers recorded motor performance data for similar tasks using both the MotionScan application and the force sensor device. Researchers analyzed the performance outcomes on the tablet application and force sensor device, and validated that they both record motor movements similarly. Participants were also asked for their feedback on the interface of MotionScan and the data collection process, which was used to improve the usability of MotionScan and data collection processes. The research demonstrates that a tablet device can measure the variability in a person’s motor sensitivity and with more research could be used as a concussion detection tool.

京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第21441号 / 理博第4434号 / 新制||理||1637(附属図書館) / 京都大学大学院理学研究科物理学・宇宙物理学専攻 / (主査)教授 中家 剛, 教授 鶴 剛, 准教授 市川 温子 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM