24
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Discipline
Örneklenmiş işaret için bulanık mantığa bağlı aradeğerleme algoritması
ÖZET Günümüzün gelişen teknolojileri artık geleneksel elektronik denetim birimlerinden yeteri kadar verim alamamaktadır. Gün geçtikçe ortaya çıkan daha hassas birimler ve kaçınılmaz olan enerjiden tasarruf sağlama zorunluluğu bilim adamlarını bu yönde araştırmalar yapmaya itmiştir. Gitgide mükemmele yaklaşma isteği ve doğanın belki de bir gün aynısının yapay yollarla ortaya çıkarılmaya çalışılması, yapay zeka, yapay sinir ağları, çok değerli mantık ve bunlarla birlikte bulanık mantığın ortaya çıkarılmasına neden olmuştur. Bulanık mantık her gün kullandığımız ve davranışlarımızı yorumladığımız yapıya ulaşmamızı sağlayan matematiksel bir disiplindir. Temelini doğru ve yanlış değerlerin belirlediği bulanık küme kuramı oluşturur. Burada yine geleneksel mantıkta olduğu gibi bir (I) ve sıfır (0) değeri vardır. Ancak bulanık mantık yalnızca bu değerlerle yetinmeyip, bunların ara değerlerini de kullanarak; örneğin bir uzaklığın yalnızca yakın ya da uzak olduğunu belirtmekle kalmayıp, ne kadar yakın ya da ne kadar uzak olduğunu da söyler Bu tezde, iletim veya depolama sırasında örnek değerleri gürültü veya başka nedenlerden dolayı bozulmuş olan bir işaretin tekrar elde edilmesinde, bulanık mantığa dayalı bir aradeğerleme algoritması geliştirilmiştir. Sonuçlar, örnek değerleri belirli bir şekilde seçilmiş olan gürültüsüz ve toplamsal gürültülü işaretler için ayrı ayrı elde edilmiş ve bozulan örnek değerlerin kestirilme performansı belirlenmiştir. Anahtar kelimeler: Bulanık Mantık, Bulanık Küme, Elektronik Kontrol Biçimi, Yapay Zeka, Çokdeğerli Mantık, Sinir Ağları, Aradeğerleme.
Tek kanallı mikroelektrodlarla aksiyon potansiyellerinin kaydı ve analizi
Bazı hücre sınıflarının elektrokimyasal aktivitelerinin sonucu olarak biyoelektriksel potansiyeller üretilir. Elektriksel olarak istiharat ve aksiyon potansiyeli ortaya çıkar. Aksiyon potansiyeli hücre zarının zaman içinde özel iyonlara karşı özellikle sodyum ve potasyuma geçirgenliğinin değişmesiyle ortaya çıkar. Bu çalışmada, serebellar (beyincik) hücrelerinden tek kanallı mikropipetlerle spontan aksiyon potansiyellerinin kaydı için bir donanım gerçekleştirilmiştir. Gerçekleştirilen bu donanımla, anti-psikotik ilaçların bu hücrelere olan etkilerinin incelenmesi amaçlanmıştır. Geliştirilen sistem şunları içerir; 1) solüsyondaki iyonik akımı, teldeki elektron akımına dönüştüren mikroelektrod, 2) hücre içine giren mikropipet maniplatör, 3) biopotansiyel sinyallerini yükselten amplifikatör, 4) 50Hz ve onun harmoniklerini süzen filtre, 5) A/D çevirici ve 6) aksiyon potansiyellerini görüntüleyen ve katdeden bilgisayar. Mikropipet mikromaniplatöre sabitlenir ve mikromaniplatör hareket ettirilerek hücreye girilmeye çalışılır. Hücreye girildiğinde beynin normal çalışması sırasında oluşan aksiyon potansiyeli hem osiloskop ekranında görülür. İstenirse bilgisayara kayıt yapılır. Daha sonra gerçekleştirilen bilgisayar programıyla yapılan kayıtlar filitrelenir ve incelenir. Bu sistem kullanılarak Sprogue-Dowley sıçanlarından (rat) hem in-vivo hem de in- vitro olarak aksiyon potansiyellerinin kaydını yapmak mümkün olmuştur. Anahtar Kelimeler : Aksiyon Potansiyeli, Membran Potansiyeli, Mikropipet,Yüksek Empedans, Gürültü, purkinje Hücreleri, Beyincik VI
Endüksiyon makinalarında durum değişkenleri ve parametre kestirimi
Bir endüksiyon makinesinin parametrelerinin kestirimi (tanısı), performans tahmini, benzetim analizi ve kontrol uygulamaları açısından oldukça önemlidir. Bu çalışmada, sinüzoidal, altı adım ve PWM (Pulse Width Modulation) besleme gerilimleri altında, üç fazlı sincap kafesli bir endüksiyon motorunun stator gerilimleri ve akımları ile rotor hızı ölçümleri kullanılarak, durum değişkenleri ve parametre tanısı için, genişletilmiş Kalman filtreleme algoritması uygulanmıştır. Kullanılan filtreleme algoritmasının diğer istatistiksel filtrelere göre bir üstünlüğü, ölçüm büyüklüklerinin istatistiksel özelliklerine göre kendi parametrelerini ayarlamasıdır. Diğer filtreleme metotlarında, model parametrelerini ayarlamak için kullanılan ön tahmin büyüklükleri ve ölçüm büyüklükleri arasındaki hata yüzdesi, tüm örnekler için sabit olarak kabul edilir veya programcı tarafından ayarlanır. Halbuki Kalman filtreleme algoritması, bu hata yüzdesini en iyi performans elde edilecek şekilde otomatik olarak değiştirir. Ölçüm büyüklüklerinde çok ani değişimlerin söz konusu olduğu altı adım ve PWM beslemeler için, filtre, daha büyük örnekleme frekansı gerektirmektedir. Ancak bu durumda hesaplama süresi artmaktadır. Tüm besleme gerilimleri için, filtreleme algoritması, durum değişkenleri ve parametrelerin tanısında oldukça iyi performans göstermiştir. Anahtar Kelimeler: Endüksiyon Motoru, Endüksiyon Motor Benzetimi, Vektör Kontrolü, Genişletilmiş Kalman Filtrelemesi, Durum Değişkenleri, Parametre Kestirimi
Derin öğrenme teknikleri kullanılarak meyve ve sebzede çeşitli hastalıkların tespit edilmesi
Meyve ve sebze hastalıklarının gıda güvenliği ve sürdürülebilir tarım pratikleri açısından kritik önemi bulunmaktadır. Dolayısıyla hastalıklar ürün verimini düşürmekte, kaliteyi azaltmakta ve böylece küresel gıda arzını tehdit etmektedir. Bu hastalıklar aynı zamanda biyoçeşitliliği de olumsuz etkilemekte, ekosistem dengesini bozmakta ve çiftçilerin geçim kaynaklarını zayıflatmaktadır. Bu çalışmada, derin öğrenme teknikleri kullanılarak meyve ve sebzelerde görülen hastalıkların tespiti yapılmıştır. Bu araştırma kapsamında 12 sınıfa ait 2907 adet RGB görüntüden çevrimiçi bir veri seti elde edilmiştir. Her sınıf için veri genişletme yöntemi ile veri seti 2907'den 17442'e kadar çıkarılmıştır. Meyve ve sebzelerdeki çeşitli hastalıkların tespiti için 10 katmanlı evrişimli derin ağ modeli oluşturulmuş ve ön eğitimli derin ağ mimarileri ( InceptionV3 ve ResNet50) kullanılmıştır. Elde edilen sonuçlar, en başarılı yöntemleri belirlemek için zaman ve başarı oranı açısından karşılaştırılmıştır. Sağlanan analizlerin sonuçları ayrıca tasarlanan bu gerçek zamanlı sistem ile meyve ve sebzelerde hastalık görüntülerini tespit etme ve tahminlerini bilgisayar ekranına aktarmak için gerçekleştirilmiştir.
Enerji-verimli aygıt ve bağ özelliklerine dayalı bluetooth multipikonet formasyonu
ABSTRACT ENERGY-EFFICIENT BLUETOOTH SCATTERNET FORMATION BASED ON DEVICE AND LINK CHARACTERISTICS Canan PAMUK M.S. in Electrical and Electronics Engineering Supervisor: Assist. Prof. Dr. Ezhan Karaşan August 2003 Bluetooth is a promising ad hoc networking technology. Although construction and operation of piconets are well defined in Bluetooth specifications, there is no unique standard for scatternet formation and operation. In this thesis, we propose a distributed and energy-efficient Bluetooth Scatternet Formation algorithm based on Device and Link characteristics (SF-DeviL) that is compatible with Bluetooth specifications. SF-DeviL handles energy efficiency using classes of devices, battery levels and the received signal strengths. SF-DeviL forms scatternets with tree topologies that are robust to battery depletions, where devices are arranged in an hierarchical order in terms of battery power and traffic generation rate. SF-DeviL is dynamic in the sense that the topology is reconfigured when battery levels are depleted, thereby increasing the lifetime of the scatternet. Unlike many of the algorithms in the literature SF-DeviL is also multihop, i.e., there is no requirement for each node to be in the transmission range of all other nodes. Keywords: Bluetooth, scatternet formation, energy efficiency, class of device, received signal strength, RSSI, battery level, multihop, tree topology, distributed. u
31-modu açık sclcndcr tüp yapılı transdüser tasarımı
In this work, design and electromechanical analysis of radially polarized 31-Mode free-flooded ring transducers which have high power capability in deep submergence are explained. 31-Mode free-flooded ring transducers have Helmholtz resonance caused by the water inside, besides the radial resonance. By adjusting the dimensions of the ring, these resonance frequencies can be changed to the preferred values and desired band characteristics can be obtained.Transducer is designed using circuit theory techniques on the electrical equivalent circuit of free-flooded ring. The ring transducer is modeled in ANSYS, a Finite Element Modeling tool. The results obtained by analyzing the electrical equivalent circuit and the ANSYS outputs are compared. Electrical equivalent circuit parameters are updated considering the result obtained in finite element analysis. Effects of mounting end tubes are investigated by finite element analysis. New components related to the end tubes are inserted to the electrical equivalent circuit.
Büyük sinyal analizi ve alçak frekansli geri besleme kullanilarak düşük faz gürültülü osilatör tasarim ve simülasyonu
Spectral purity of oscillators is of great importance in both commercial and military systems. Implementing communication, radar, and Electronic Warfare systems with increasingly higher frequencies, wider bandwidths, greater data rates, and more complex modulation schemes require low phase noise signal sources. Phase noise in signal sources is not a well understood area. Although an- alytical models accomplish to describe the phase noise of known signal sources accurately, a unifying and reproducible model or method that provides a priori information for the design of a low phase noise oscillator is still not established. Due to this lack of methodical approach, mostly empirical design practices that are known to produce good results are widely adopted. Proposed design method is similar. Design and simulation of a low phase noise Dielectric Resonator Oscillator is studied. Noise sources in oscillators are briefly summarized. Phase noise models are compared. Dielectric resonators, which use small, disc-shaped ceramic mate- rials that have high quality factors at microwave and millimeter-wave frequencies, are introduced with a concise theoretical coverage. Eff ect of circuit con guration on phase noise is studied on two di erent FET devices. Common-gate con guration gave best simulation results for both tran- sistors. Parameters of coupling to the resonator are studied based on large signal anal- ysis of the active device. The optimal parameters are described with supporting simulation results. Comparisons with suboptimal designs are provided, results indicate that optimization improves the phase noise on the order of tens of dBs. Low frequency feedback method is investigated. Simulation results showed signi cant improvement in close-in phase noise when such networks are used. A large data set is obtained with input parameters of frequency, device, bias point, and feedback con guration; and optimality of such schemes are discussed based on it. The methods for suppressing both close-in and away from the carrier phase noise are presented in the most generalized way, only to be reproduced for the intended device of operation.
Mikrodenetleyici tabanla ısı kontrol sisteminin tasarımı
Mikrodenetleyici tabanla ve dijital PID kontrol algoritmalı yazılım kullanılarak tasarlanan ısı kontrol sistemi kullanıcıya birçok avantaj ve üstünlük sağlamaktadır. Sanayide veya evlerimizde sıcaklığı kontrol edilen prosesin ısı değişimlerine, dijital tabanlı PID kontrol algoritmasıyla lineerlik kazandırılarak hassasiyeti ve doğruluğu yüksek, kontrolü kolay bir kontrol sistemi elde edilir. Bu yöntemle elde edilen kazanımları örneklersek; endüstride ekonomik ve katma değeri yüksek kaliteli ürünlerin elde edilmesini, evlerimizde kullanılan ısıtma amaçlı sistemlerin kararlı bir sıcaklıkta çalışmasını sağlayarak insan sağlığına olan olumlu ve üstün yanlarını sayabiliriz.Sıcaklık kontrolünün, insan hayatında ve sanayideki birçok proseste önemli bir yeri olduğundan, bu alanda yapılan ve yapılmaya devam edilen çalışmaların değeri her zaman artmaya devam edecektir ve bilimsel anlamda önemini her zaman koruyacaktır.Anahtar kelimeler: NTC sensörlü ısı kontrol sistemi ve tasarımı, PID ısı kontrol sistemi ve tasarımı, mikrodenetleyici tabanlı ısı kontrol sistemi ve tasarımı,
EEG sinyallerindeki epileptiform aktivitenin veri madenciliği süreci ile tespiti
Türkiye'de ve dünyada; biyomedikal ve sinyal işleme konularındaki çalışmalara bakıldığında, hem teorik hem de uygulamalı olarak çok sayıda çalışmanın bulunduğu göze çarpmaktadır. Bu çalışmalar genel olarak; elektronik sistem tasarımları, matematiksel modeller, istatistik metotlarla yapılmış çalışmalar, yapay zekâ tabanlı (bulanık mantık, yapay sinir ağları ve genetik algoritmalar içeren) çalışmalar ve tıbbi yazılımlar (genel olarak mobil uygulamalar ve cihaz otomasyonları) olarak gruplandırılabilir. Bu alandaki çalışmaların özellikle 2000 yılından sonraki artışı dikkat çekicidir. Türkiye'de ve dünyada veri madenciliği alanında çalışmaların çok yeni ve az oluşunun yanı sıra, tıbbi veriler üzerindeki veri madenciliği çalışmalarının yok denecek kadar az olması konunun önemini ortaya koymaktadır. Veri madenciliği süreçlerinin tıbbi veriler üzerinde uygulanması sonucunda analiz süresinin kısaldığı ve yüksek doğruluk oranı içeren sonuçlar elde edildiği görülmektedir. Çalışma bu bakımdan ele alındığında, güncel bir uygulama olarak önem taşımaktadır.Bu çalışmada; elektroensefolagram (EEG) verileri üzerinde, epileptik aktivitelerin olup olmadığının belirlenmesi ve daha sonraki aşamalarda geliştirilecek ilave yazılımlarla otomatik teşhis koymaya yardımcı bir araç geliştirilmesi amaçlanmıştır. Verilerin veri madenciliği süreçleri kapsamında sınıflama ve kümeleme algoritmaları kullanarak tespit edilebilmesi için öncelikli olarak sekiz adet öznitelik parametresi seçilmiş ve belirlenen öznitelik değerleri hesaplanmıştır. Belirlenen öznitelikler için elde edilen sonuçlar, on bir ayrı veri madenciliği algoritmasına tabi tutulmuş ve seçilen bazı algoritmaların yüksek doğruluk oranı ile epileptik aktiviteyi tespit ettiği görülmüştür. Böylelikle farklı veri madenciliği algoritmaları ile elde edilen sonuçların hesaplama sürelerinin ve doğruluk oranlarının kıyaslanması sağlanmıştır. En yüksek doğruluk oranını verdiği tespit edilen algoritmalar yardımı ile bu alanda çalışan uzmanların epileptik aktivite teşhisi koymalarına zemin oluşturmak, ilgililerin teşhis sürecinde karar vermelerini kolaylaştırmak ve konulan teşhislerde doğruluk oranını yükselterek, Türkiye'de nöroloji ve bilgisayar bilimleri literatürüne katkı sağlanması hedeflenmiştir.
Kesir dereceli PID denetleyicinin model referans tabanlı tasarımı
Oransal-İntegral-Türevsel (PID) denetleyici kolay tasarımı ve yeterli kontrol performansı sağlaması nedeniyle otomatik kontrol uygulamalarında sıklıkla tercih edilmektedir. Oransal kazanç (Kp), İntegral kazancı (Ki) ve Türevsel kazanç (Kd) olarak adlandırılan üç temel parametreye sahiptir. Geleneksel PID denetleyicideki integral ve türev operatörleri tamsayı derecelidir. Araştırmacılar kontrol performansını artırması sebebiyle geleneksel PID denetleyicideki tamsayı dereceli integral ve türev operatörleri yerine kesirli dereye sahip integral ve türev operatörlerini kullanarak kesirli dereceli PID denetleyiciyi önermişlerdir. Kesir dereceli PID denetleyici, geleneksel PID denetleyiciye göre ilave olarak kesirli integratör derecesi (λ) ve kesirli türev derecesi (µ) içerir. Bu tez çalışmasında, kesir dereceli PID denetleyicinin zaman alanında referans bir modele göre tasarımına odaklanılmıştır. Referans model olarak Bode'nin ideal transfer fonksiyonu kullanılmıştır. Bode'nin ideal transfer fonksiyon modelinin zaman alanı cevabı ile kontrol edilmek istenen sistemin çıkışı arasındaki hatanın minimize edilmesi ile kesir dereceli PID parametrelerinin elde edilmesi amaçlanmıştır. Hatanın minimizasyonu için çalışmada Genetik Algoritma (GA) optimizasyon yöntemi kullanılmıştır. Çalışma tek girişli çok çıkışlı yapıya sahip ters sarkaç sistemi üzerinde benzetim çalışması şeklinde yürütülmüştür. Ters sarkaç sisteminde araba ve sarkaç için iki ayrı kesir dereceli PID denetleyici kullanılmış ve GA optimizasyonu bu iki denetleyicinin Kp, Ki, Kd, λ ve µ parametreleri ayarlanarak, ters sarkaç sisteminin çıkış cevabı referans modele yaklaştırılabilmiştir.
Elektrookülar nesne kontrolünde grafen tekstil arayüz kullanımı
Study of eye movements (EMs) and measurement of the resulting biopotentials, referred to as electrooculography (EOG), may find increasing use in applications within the domain of activity recognition, context awareness, mobile human-computer interaction (HCI) applications, and personalized medicine provided that the limitations of conventional "wet" electrodes are addressed. To overcome the limitations of conventional electrodes, this work, reports for the first time the use and characterization of graphene-based electroconductive textile electrodes for EOG acquisition using a custom-designed embedded eye tracker. This self-contained wearable device consists of a headband with integrated textile electrodes and a small, pocket-worn, battery-powered hardware with real-time signal processing which can stream data to a remote device over Bluetooth. The feasibility of the developed gel-free, flexible, dry textile electrodes was experimentally authenticated through side-by-side comparison with pre-gelled, wet, silver/silver chloride (Ag/AgCl) electrodes, where the simultaneously and asynchronous recorded signals displayed correlation of up to ~87% and ~91% respectively over durations reaching hundred seconds and repeated on several participants. Additionally, an automatic EM detection algorithm is developed and the performance of the graphene-embedded "all-textile" EM sensor and its application as a control element toward HCI is experimentally demonstrated. The excellent success rate ranging from 85% up to 100% for eleven different EM patterns demonstrates the applicability of the proposed algorithm in wearable EOG-based sensing and HCI applications with graphene textiles. The system-level integration and the holistic design approach presented herein which starts from fundamental materials level up to the architecture and algorithm stage is highlighted and will be instrumental to advance the state-of-the-art in wearable electronic devices based on sensing and processing of electrooculograms.
Otomotiv uygulamaları için 77 GHz SiGe BiCMOS FMCW RADAR
SiGe BiCMOS technologies have been improved such that they offer comparable, sometimes even better RF performances with respect to their counterparts, especially III-V technologies. Further, SiGe BiCMOS technologies can integrate CMOS transistors and SiGe HBTs on the same chip. These advancements make possible to implement IC designs with low-cost and high integration. Also, due to the increasing demand for compact system designs with low-cost, and low-power-consumption, millimeter-wave systems became attractive for both industry and academia. These developments have paved the way for millimeter-wave systems. Therefore, in this thesis work, a 77 GHz SiGe BiCMOS FMCW Monostatic RADAR for automotive applications is presented. The system was designed and implemented with IHP's 0.13 µm SiGe BiCMOS SG13G2 technology which offers HBTs with ft/fmax of 300 GHz/500 GHz. The schematic designs and implementation of each sub-block and the overall system were shown. Then, the simulation results for all the sub-blocks and the system were presented. The total chip area for the system is 5.05 mm2. The total power consumption of the system is 0.32 mW. The system achieves 8.5 dBm output power with only one TX channel. The simulated NF of the RX-channel is 9.1 dB and RX-gain is 18 dB. The simulation results show that the designed system can detect targets up to 180m distance with a maximum velocity of 170km/h with accuracy. Also, the designed system can separately identify two targets with a range resolution of 15 cm. This work is the first demonstration of a 77 GHz long-range-RADAR designed with an EBD.
İnce taneli perakende ürün tanıma sistemi için istatistikyöntemler
In recent years, computer vision has become a major instrument in automating retail processes with emerging smart applications such as shopper assistance, visual product search (e.g., Google Lens), no-checkout stores (e.g., Amazon Go), real-time inventory tracking, out-of-stock detection, and shelf execution. At the core of these applications lies the problem of product recognition, which poses a variety of new challenges in contrast to generic object recognition. Product recognition is a special instance of fine-grained classification. Considering the sheer diversity of packaged goods in a typical hypermarket, we are confronted with up to tens of thousands of classes, which, particularly if under the same product brand, tend to have only minute visual differences in shape, packaging texture, metric size, etc., making them very difficult to discriminate from one another. Another challenge is the limited number of available datasets, which either have only a few training examples per class that are taken under ideal studio conditions, hence requiring cross-dataset generalization, or are captured from the shelf in an actual retail environment and thus suffer from issues like blur, low resolution, occlusions, unexpected backgrounds, etc. Thus, an effective product classification system requires substantially more information in addition to the knowledge obtained from product images alone. In this thesis, we propose statistical methods for a fine-grained retail product recognition. In our first framework, we propose a novel context-aware hybrid classification system for the fine-grained retail product recognition problem. In the second framework, state-of-the-art convolutional neural networks are explored and adapted to fine-grained recognition of products. The third framework, which is the most significant contribution of this thesis, presents a new approach for fine-grained classification of retail products that learns and exploits statistical context information about likely product arrangements on shelves, incorporates visual hierarchies across brands, and returns recognition results as "confidence sets" that are guaranteed to contain the true class at a given confidence level.
26 GHz 5G uygulamaları için 130-nm sige teknolojisiyle dört-kanallı faz-dizili hüzmeleyici ve özgirişim kaldıran tam dubleks alıcı/verici geliştirilmesi
This thesis is on the design of radio-frequency (RF) integrated front-end circuits for next-generation 5G communication systems. The demand for higher data rates and lower latency in 5G networks can only be met using several new technologies including, but not limited to, mm-waves, massive-MIMO, and full-duplex. Use of mm-waves provides more bandwidth that is necessary for high data rates at the cost of increased attenuation in air. Massive-MIMO arrays are required to compensate for this increased path loss by providing beam steering and array gain. Furthermore, full-duplex operation is desirable for improved spectrum efficiency and reduced latency. The difficulty of full-duplex operation is the self-interference (SI) between transmit (TX) and receive (RX) paths. Conventional methods to suppress this interference utilize either bulky circulators, isolators, couplers or two separate antennas. These methods are not suitable for fully-integrated full-duplex massive-MIMO arrays. This thesis presents circuit and system-level solutions to the issues summarized above, in the form of SiGe integrated circuits for 5G applications at 26 GHz. First, a full-duplex RF front-end architecture is proposed that is scalable to massive-MIMO arrays. It is based on blind, RF self-interference cancellation that is applicable to single/shared antenna front-ends. A high-resolution RF vector modulator is developed, which is the key building block that empowers the full-duplex frontend architecture by achieving better than state-of-the-art 10-b monotonic phase control. This vector modulator is combined with linear-in-dB variable gain amplifiers and attenuators to realize a precision self-interference cancellation circuitry. Further, adaptive control of this SI canceler is made possible by including an on-chip low-power IQ downconverter. It correlates copies of transmitted and received signals and provides baseband/dc outputs that can be used to adaptively control the SI canceler. The solution comes at the cost of minimal additional circuitry, yet significantly eases linearity requirements of critical receiver blocks at RF/IF such as mixers and ADCs. Second, to complement the proposed full-duplex front-end architecture and to provide a more complete solution, high-performance beamformer ICs with 5-/6-b phase and 3-/4-b amplitude control capabilities are designed. Single-channel, separate transmitter and receiver beamformers are implemented targeting massive-MIMO mode of operation and their four-channel versions are developed for phased array communication systems. Better than state-of-the-art noise performance is obtained in the RX beamformer channel, with a full-channel noise figure of 3.3 dB.
Düşük enerjili HEVC ve VVC video sıkıştırma donanımları
Video compression standards compress a digital video by reducing and removing redundancy in the digital video using computationally complex algorithms. As spatial and temporal resolutions of videos increase, compression efficiencies of video compression algorithms are also increasing. However, increased compression efficiency comes with increased computational complexity. Therefore, it is necessary to reduce computational complexities of video compression algorithms without reducing their visual quality in order to reduce area and energy consumption of their hardware implementations. In this thesis, we propose a novel technique for reducing amount of computations performed by HEVC intra prediction algorithm. We designed low energy, reconfigurable HEVC intra prediction hardware using the proposed technique. We also designed a low energy FPGA implementation of HEVC intra prediction algorithm using the proposed technique and DSP blocks. We propose a reconfigurable VVC intra prediction hardware architecture. We also propose an efficient VVC intra prediction hardware architecture using DSP blocks. We designed low energy VVC fractional interpolation hardware. We propose a novel approximate absolute difference technique. We designed low energy approximate absolute difference hardware using the proposed technique. We propose a novel approximate constant multiplication technique. We designed approximate constant multiplication hardware using the proposed technique. We quantified computation reductions achieved by the proposed techniques and video quality loss caused by the proposed approximation techniques. The proposed approximate absolute difference technique and approximate constant multiplication technique cause very small PSNR loss. The other proposed techniques cause no PSNR loss. We implemented the proposed hardware architectures in Verilog HDL. We mapped the Verilog RTL codes to Xilinx Virtex 6 or Xilinx Virtex 7 FPGAs and estimated their power consumptions using Xilinx XPower Analyzer tool. The proposed techniques significantly reduced power and energy consumptions of these FPGA implementations.
50 nm altında altın nanoyapıların nanofabrikasyonu için elektron demeti litografi ve yüzeyden kaldırma süreçlerinin optimizasyonu
Since the demonstration of the first integrated circuit in the late 1950s, the microelectronics industry has witnessed a vast transformation with transistor densities doubling roughly every two years as a result of continuous scaling down of device dimensions, referred to as miniaturization. The fundamental concept of miniaturization has not only been employed for the realization of ultra large scale integrated (ULSI) circuits with reduced manufacturing costs, lower power consumption, higher speed and computational power; but also, for developing novel transducer elements and energy storage devices by harnessing the unique physical effects that arise at micro/nanoscales such as higher surface-to-volume ratios. One of the most important technologies in micro/nano device fabrication, if not the single most important, is lithography. The broad range of lithographic techniques ranging from conventional optical lithography methods (e.g. ultraviolet-UV, deep ultraviolet-DUV, extreme ultraviolet-EUV) to unconventional ones (e.g. electron beam lithography, x-ray lithography, ion-beam lithography, stereolithography, scanning probe lithography, nanoimprint lithography, directed self-assembly) can be used to create features with microns to tens of nanometer resolution and below. Among these, electron beam lithography (EBL) stands out as a powerful direct-write tool offering nanometer scale patterning capability and is especially useful in low volume R&D prototyping. However, patterning with EBL requires careful balance of process parameters which need to be considered in conjunction with the pattern transfer technology that can be either etching or lift-off specifically for the case metallic layers. Accordingly, this thesis provides a systematic study to address the gap in process optimization of lift-off process based on EBL patterning of sub-50 nm metallic nanostructures using a lower cost PMMA/PMMA positive tone bilayer resist spin approach. The governing parameters in EBL including exposure dose, bake temperature, develop time, developer solution, substrate effect, proximity effect (PE) are experimentally studied and their effects on nanopatterning are characterized by field emission scanning electron microscopy (FE-SEM) of fabricated nanostructures.
5G (28 GHz) haberleşme sistemleri için elektriksel denge dupleksleyici bant-içi tam dupleks sistem
The evolution of wireless communication systems is witnessed every day by their users. Currently, 4G meets the needs of the users in terms of data-rate and latency. 4G is spreading to the world and dominating the worldwide market. However, ever-increasing user demands for higher data-rate, lower latency, and an increased number of users are generating new challenges that 4G could not manage to meet the demands. To overcome the challenges in 4G, the institutions are studying the next generation (5G) communication systems. With the introduction of 5G, the utilization of various new technologies such as full-duplex becomes necessary to meet the demand for higher data-rate and lower latency. With the in-band full-duplex operation, the spectrum efficiency is doubled theoretically. The improvements in the SiGe BiCMOS technology have made them be a good candidate for low-cost and high-performance transceivers. In this thesis, the design of an electrical balance duplexer based in-band full-duplex system for 5G (28 GHz) communication systems implemented in IHP's 0.13 $\mu$m SiGe BiCMOS technology is presented. The electrical balance duplexer is constructed with a hybrid transformer and an impedance tuner to suppress the leakage from the transmitter to the receiver by reducing the mismatch between the antenna and balance port impedances. The duplexer achieves isolation between the transmitter and receiver sides more than 40 dB for various antenna impedances at 28 GHz. The transmitter gain of the system is more than 30 dB by achieving OP1$_{dB}$ of 12 dBm whereas the receiver gain of 5 dB is obtained. The minimum noise figure of the receiver is 8.2 dB.
Grafen tekstil elektrotlarla giyilebilir yuzey elektromiyografi (sEMG) teknolojileri
Ability to acquire, record, and process muscular biopotentials with wearable health trackers, diagnostic and/or assistive devices through the integration of soft, gel-free surface electrodes will enable seamless monitoring of muscle status in dynamic settings and can facilitate various applications. To this end, one of the fundamental limitations against the development of such systems is due to the drawbacks of clinical Ag/AgCl wet electrodes which have a gel layer that causes discomfort and can lead into skin irritation especially in wearable, portable applications with typically longer monitoring durations. The major objective of this thesis is to develop, explore and evaluate the application of novel and truly wearable graphene textile electrodes specifically in surface electromyography (sEMG) applications. Benchmarking of the textile electrodes with respect to clinical electrodes was performed in terms of their skin electrode impedance (SEI) values and signal-to-noise ratios (SNR) acquired through static experiments. SEI values of textile electrodes were found to be well within the acceptable range and comparison of SNR values showed that graphene textile performance in static settings swings between 65\% and 90\% level of clinical signal quality. Custom-designed wearable platforms consist of muscle specific elastic bands with integrated graphene electrodes and battery-powered, small footprint hardware that can stream data wirelessly over Bluetooth. Feasibility of the developed wearables were shown with two different applications in dynamic conditions where a calf band was designed and used for activity tracking, and an arm band was developed for localized muscle fatigue assessment. The holistic system design concept presented here that is inclusive of fundamental material aspects up to the implementation of high-level user interface will be valuable for the development of wearable EMG platforms.
Fourier özniteliklerinin sinir ağları ile incelenmesi ve örgü ağlarda bağlantı yönlendirmeye uygulanması
Random Fourier features provide one of the most prominent ways to classify large-scale data sets when the classification is nonlinear. However, Fourier features, in its original proposal, are randomly drawn from a certain distribution and are not optimized. In this thesis, we investigate the use of Fourier features by a single hidden layer feedforward neural network (SLFN) and optimize those features (instead of drawing randomly) with several gradient-descent based approaches. The optimized Fourier features are deduced from the radial basis function (RBF kernel), and implemented in the hidden layer of the SLFN which is followed by the output layer. The resulting classification accuracy is compared with the results of SVM with RBF kernel. Particularly, (1) we tune the parameters such as the hidden layer size and RBF kernel bandwidth, and (2) test with ten different classification data sets. The introduced SLFN provides substantial computational gains with similar accuracy figures compared to the ones of SVM. We also test our SLFN for steering in wireless mesh networks and observe promising smart steering capabilities.
Uyarlanabilir beyin-bilgisayar arayüzlerine doğru: Zihinsel durum tanıma için istatistiksel çıkarım
Brain-computer interface (BCI) systems aim to establish direct communication channels between the brain and external devices. The primary motivation is to enable patients with limited or no muscular control, including amyotrophic lateral sclerosis (ALS) and stroke patients, to use computers or other devices by automatically interpreting their intent based on the measured brain electrical activity. Furthermore, enabling healthy individuals to use BCI systems as an additional communication channel in certain human computer interaction systems is also a current topic of interest. Current experimental BCI systems are trained in a supervised fashion and then evaluated during test sessions. With increasing demands for daily and long-term use of BCIs in real-life applications such as in semi-autonomous cars, BCIs have been tested on longer sessions in which researchers have observed considerably lower performance of trained systems. This is believed to be caused by the nonstationary nature of the electroencephalographic (EEG) signals. As a result, semi-supervised adaptation of BCI systems based on test data has emerged as a new research domain. One of the main reasons underlying the nonstationarity of signals involves changes in the users' cognitive states such as the cognitive load, alertness, attention, fatigue, boredom, and motivation. However, dynamically extracting information about such cognitive states from EEG signals and using that to improve the performance of BCI systems is currently an open research problem. In this thesis, we tackle the highly complex problem of estimating the level of alertness and vigilance of users during execution of cognitive tasks. To identify the neural, EEG-based correlates of long-term task and response time consistency, we devise a series of experiments running the sustained attention to response task (SART). After proposing a novel adaptive scoring scheme for vigilance, we provide new evidence on the close relationship between intrinsic resting and task-related brain networks and develop models to predict consistency in tonic performance and response time using neural networks and feature relevance analysis from spatio-spectral features of resting-state EEG signals. Next, focusing on the imminent goal of predicting low and high vigilance intervals, we propose fully automated systems based on convolutional neural networks (CNNs) using phase locking value features as successful pre-trial predictors of phasic vigilance and performance consistency. In all of these contributions, we consider the personal vigilance traits and individual psychophysiological differences for modeling and detecting the extremely alert and drowsy trials in long and monotonous experiments, and enrich the literature with the evidence on spatio-spectro-temporal correlates of vigilant and consistent behavior. We then utilize Bayesian changepoint models for sequential inference and detection of instants at which continuous vigilance levels of users enter a new phase. We demonstrate the success of our online and offline vigilance models in detecting changepoints from both the SART datasets collected in our lab and driving datasets that contain vigilance labels. Finally and as the highlight of this thesis, we hypothesize that the underlying vigilance levels affect users' reaction time and thus the ability to focus and engage in motor imagery BCI paradigms. We then introduce an adaptive alertness-aware MI classification system for motor imagery BCI that uses a series of novel unsupervised learning schemes for labeling trial vigilance levels during training and test sessions, and leads to a method with full adaptation in both feature extraction and training of its classifier parameters. Three different versions of this adaptive classification approach are introduced that are trained differently on trials labeled with low vigilance levels by our various vigilance clustering schemes. We report improvements in the overall test accuracy of adaptive versions with respect to the original, non-adaptive baseline for our own SPIS MI-BCI dataset and the BCI Competition IV Dataset 2a. A number of datasets collected in our BCI laboratory are uploaded to a public repository at https://github.com/mastaneht.
Süper çözünürlüklü yapılandırılmış aydınlatma mikroskopisi için sıkıştırılmış algılama ve öğrenmeye dayalı yöntemler
Using an optical microscope, most viruses, proteins, and small molecules cannot be successfully imaged because of Abbe's diffraction limit. The super-resolution structured illumination microscopy (SIM) technique overcomes this issue and expands the lateral resolution to the half of the diffraction limit. The cost of the SIM technique results from the need to record at least nine raw images to reconstruct a single super-resolution image. This requirement has two consequences: photobleaching and motion artifacts. To alleviate these problems, we need a system that is extremely fast for recording raw images (to observe high dynamic processes) and projects less excitation light onto the sample (to avoid photobleaching). Compressed sensing (CS) can be a candidate for achieving these objectives. First, CS allows us to record an object scene with a photomultiplier tube (PMT) instead of a camera. The acquisition speed of a PMT is much higher than a scientific complementary metal-oxide-semiconductor (sCMOS) camera. Second, the scene in the CS framework is sampled faster (thanks to the higher frame rate of a digital micromirror device - DMD), and also sampled with lower excitation light (because of sampling patterns). Third, the CS framework can recover the scene reliably with few measurements, reducing the overall data collection time further. The main objective of this dissertation is to combine CS and SIM techniques, but we also make various contributions to this framework. The main contributions of this dissertation are (1) proposing a dictionary learning method based on the multi-layer convolutional sparse coding (ML-CSC) model to improve the performance of a CS recovery algorithm; and (2) proposing a method for the combination of CS and SIM and demonstrating the method with simulation-based studies as well as real data collection experiments. In early attempts in the sparse representation theory, some off-the-shelf dictionaries were utilized. However, training dictionaries instead of using a known transform significantly improved signal reconstruction quality. On the other hand, the success of a CS recovery algorithm is directly related to the sparsity level of a signal. The sparsity level of a signal depends on the sparsifying transform or dictionary. With that perspective, we need to learn a sparsifying transform or dictionary that is compatible with a signal of interest. Therefore, we propose a dictionary learning method based on ML-CSC. The method does not depend on any parameters or the success of a CS recovery algorithm involved in the dictionary learning steps although the ancestor of the proposed algorithm depends on some parameters and the recovery algorithm. We also implement the learned dictionaries into a CS recovery algorithm and discuss the performance of the proposed learning algorithm. The other main contribution of this dissertation is to combine the CS framework and the SIM technique. We demonstrate this combination utilizing a simulation-based study. The mathematical foundation of the proposed study is demonstrated. Then, experimental results for both stationary and non-stationary objects are presented. We utilize some CS recovery algorithms presented previously and compare the reconstruction results for the case of the combination of CS and SIM. We propose an optical configuration for the data collection problem with the photomultiplier tube (PMT), and then we discuss the limitations of the DMD in the laboratory. Then, an optical configuration for the combination of CS and SIM is introduced. Using the proposed configuration, an experimental study is performed for both stationary and non-stationary objects. The normalized intensity profiles of the reconstructions and the other conventional microscopy methods for the same object are compared. The proof-of-principle solution for the photobleaching issue is evaluated for the real optical configuration. We also present a CS approach for holography and demonstrate the extraction of depth information from a single hologram. An optical configuration for holographic data collection is first presented. The depths of the variety of digital holograms (include compressive ones) are obtained using the stereo disparity method. The proposed method does not require the phase information of the hologram but two perspectives of the scene, which are easily obtained by dividing the hologram into two parts (two apertures) before the reconstruction. We investigated the effects of gradual and sharp divisions of the holograms for the disparity map calculations, specifically for divisions in the vertical, horizontal, and diagonal directions. After obtaining the depth map from the stereo images, a regular two-dimensional image of the object is merged with the depth information to form 3D visualization of the object.
Derin sinir ağları ile sentetik açıklıklı radar görüntüleme
Synthetic aperture radar (SAR) is a remote sensing imaging modality that has been in use since the 1960s. Conventional image formation in SAR is based on 2D inverse Fourier transform of the reflectivity field of the scene to be imaged. This conventional image formation technique is developed for a clean and complete data collection scenario. However, in reality, the collected data are only a reduced representation of the underlying scene due to hardware limitations and uncertainties in the data collection geometry, and hence suffer from reduction and phase errors. Therefore, many SAR image formation frameworks using regularization have been proposed over the years, in order to account for these limitations. In this dissertation, we have focused on the SAR imaging problem, particularly image formation, phase error correction, and automatic target recognition (ATR), and developed three frameworks. The first framework tackles the SAR image formation problem. In this framework, SAR image formation is formulated as a regularized optimization problem, and using the plug-and-play (PnP) priors framework, we have incorporated deep learning-based priors into our formulation. Our second framework is an extension of the first one, which aims at joint image formation and phase error correction. Experimental results show the effectiveness of these two frameworks and our proposed methods exceed the state-of-the-art image formation and phase error correction performances in the majority of the scenarios considered. The third proposed framework focuses on the ATR problem, and within this framework two ATR approaches are presented which perform the ATR task in the data domain rather than image domain. We have experimentally shown that the ATR task can be successfully performed in the data domain, and with further development, it might be possible to reach state-of-the-art performance. Overall, we have shown that the performance in various SAR imaging tasks can be improved significantly using deep learning tools.
Taşınabilir ölçü aletleri için 8-bit 2 GSPS ayrık zamanlı ardışık yaklaşımlı analog sayısal çevirici tasarımı
A successive approximation register analog to digital convertor (SAR ADC) is a type of ADC that converts a continuous analog signal into a discrete digital representation for all the quantization levels as possible. For this thesis the SAR ADC is consist of a gain stage, a split capacitor array digital to analog convertor (DAC) block, and a SAR block to supply the requirements for high bandwidth real-time oscilloscopes. As soon as the bandwidth requirements extend beyond the sample rate capability of the available ADC, it becomes necessary to find other techniques to utilize available ADCs to meet those extended requirements or design a new generation ADC. Time interleaving (TI) is a common technique to extend the performance of existing designs. The ADCs are used as parallel. Each ADC provides a sample rate at least half the total sample rate required to meet the Nyquist requirement. The time interleaved SAR ADC is separated in 10 sub-ADC blocks, each sub-ADC is consist of 10 SAR ADC and clocked 36° out of phase. Data is stored in the memory behind each ADC, and once the acquisition is completed, the complete 2GS/s representation of the signal could be reconstructed by demuxing the data. The time interleave technique has been used by all the major oscilloscope structures to get the performance up into the GHz range. The real-time oscilloscope needs a fast and high bandwidth ADC. Based on this purpose, an 8-bit 2GS/s time interleaved SAR ADC is designed as usable in high bandwidth portable measurement devices especially real-time oscilloscopes. In this thesis, a reliable and feasible 8-bit 2GS/s SAR ADC with 2GHz bandwidth is presented using 180nm CMOS technology. A comparator consists of a latch and 5 amplifier with 3 of them are offset cancelled. The simulation results of the comparator is available in the relevant section. A fully differential DAC is consist of a sample & hold circuit and it is generated as split capacitor array DAC. The simulation results of the DAC is available in the relevant section. The DAC is driven by the SAR logic. The SAR block consist of a phase shifter with 9 D flip flops, a register array with 9 D flip flops, and one more register array with 8 D flip flops. The simulation results of the SAR logic is available in the relevant section. The SAR ADC processes with the overdrive-recovered amplifiers, offset cancelled comparator, and charge injection rejected split capacitor array DAC. Also the sample and hold procedure is completed by DAC block. The designed TI-SAR ADC is simulated and all the differential non-linearity (DNL), integral non-linearity (INL), signal to noise ratio (SNR), spurious free dynamic range (SFDR), signal to noise and distortion (SINAD), effective number of bits (ENOB) characterizations are determined.
Elektrikli araçlardaki süperkapasitör bankaları için ilk yatırım maliyeti tahmin yaklaşımı
Bu çalışma, elektrikli araçlarda enerji depolama sistemleri kapsamında süperkapasitörlerin kullanımını ele almakta ve özellikle süperkapasitör bankalarının tasarımı ve optimizasyonuna odaklanmaktadır. Fosil yakıtlara bağımlılığın azaltılması ve çevresel kaygıların artması, elektrikli araçları sürdürülebilir ulaşım çözümlerinin ön saflarına taşımıştır. Ancak, lityum-iyon bataryaların sınırlı güç yoğunluğu, uzun şarj süresi ve çevrim ömrü gibi dezavantajları nedeniyle süperkapasitörler tamamlayıcı bir alternatif olarak öne çıkmaktadır. Hızlı şarj/deşarj kabiliyeti, uzun ömür, yüksek güç yoğunluğu ve geniş sıcaklık aralığında çalışma gibi üstün özellikleri sayesinde süperkapasitörler, ani hızlanma ve rejeneratif frenleme gibi uygulamalarda etkili bir çözüm sunmaktadır. Bu bağlamda, süperkapasitör teknolojisinin temel ilkeleri, farklı türleri (EDLC, psödokapasitör, hibrit) ve elektrot malzemeleri (karbon esaslı, geçiş metali oksitleri, iletken polimerler) detaylı olarak incelenmiştir. Ayrıca elektrolitler ve ayırıcıların performansa etkileri değerlendirilmiş ve enerji yoğunluğunu artırmaya yönelik son malzeme geliştirmelerine de yer verilmiştir. Süperkapasitörlerin elektrikli araçlardaki uygulamaları arasında hızlanma desteği, rejeneratif enerji geri kazanımı, soğukta çalıştırma ve batarya ömrünü uzatma gibi konular analiz edilmiştir. Bataryalarla birlikte kullanılan hibrit enerji depolama sistemlerinin (HEDS) avantajları ve farklı yapılandırmaları (pasif, yarı aktif, aktif) açıklanmıştır. Çalışmanın temel bölümlerinden biri süperkapasitör bankalarının boyutlandırılması ve maliyet analizidir. Tasarım kriterleri arasında gerilim, kapasite, ısı yönetimi, ağırlık, hacim ve sistem entegrasyonu gibi faktörler detaylandırılmıştır. Hücre maliyeti, Batarya Yönetim Sistemi (BYS) maliyeti, soğutma ve montaj gibi bileşenlerin maliyetleri değerlendirilmiş ve uzun ömür nedeniyle süperkapasitörlerin toplam yaşam döngüsü maliyetinde avantaj sağlayabileceği belirtilmiştir. Deneysel ve analitik kısımda, belirli bir sabit enerji düzeyinde (2304 J) ve değişen çalışma gerilimlerinde (16 V, 24 V, 32 V) çeşitli modül konfigürasyonları analiz edilmiştir. Gerilim arttıkça kapasitenin düştüğü, seri hücre sayısının ve ESR'nin yükseldiği; paralel hücre sayısı azaldıkça akım taşıma kapasitesinin azaldığı gözlemlenmiştir. Maxwell BCAP serisinden farklı hücrelerin karşılaştırmalı analizi sonucunda BCAP3000 K2 hücresinin düşük ağırlığı, düşük hücre sayısı, yüksek güç yoğunluğu ve düşük sistem karmaşıklığı sayesinde elektrikli araçlar için en uygun seçenek olduğu belirlenmiştir. Sonuç olarak, bu tez çalışması süperkapasitörlerin elektrikli araçlardaki kullanımının teknik ve ekonomik açıdan uygulanabilirliğini kapsamlı şekilde değerlendirmiştir. Süperkapasitörlerin bataryalarla birlikte hibrit sistemlerde kullanılması, araçların performansını, güvenilirliğini ve verimliliğini önemli ölçüde artırma potansiyeli taşımaktadır. Elde edilen veriler, gelecekteki elektrikli araç teknolojileri için yol gösterici niteliktedir.