<p> Wireless LAN mesh networks (WMNs) are a cost effective way of deploying wireless LAN (WLAN) coverage over extended areas. As WMNs become more populated, scalability issues may arise due to the co-channel interference which is inherent in publicly available RF (radio frequency) channels. This co-channel interference can severely degrade network capacity and link reliability and may eventually make it impossible to operate with the frequency channels for which the network was originally designed. In this thesis, this problem is addressed by selectively installing supplementary free-space optical (FSO) links when RF link performance has deteriorated. The frequency assignment problem is solved using a heuristic technique based on a genetic algorithm. In order to determine the quality of the results, the proposed algorithm is compared with a lower bound solution obtained using an Integer Linear Programming (ILP) formulation.</p> <p> Another advantage of FSO links is that they may reduce node power consumption compared with conventional RF links. This may be an important consideration in cases where power consumption at the nodes is important, such as in solar powered mesh networks. Power consumption estimates of RF and FSO links are obtained and compared for different data rates. This data is then used along with historical solar insolation data to estimate the solar panel and battery sizes required to guarantee a given node outage probability. The results show that no extra provisioning is required for replacing the deployed wireless nodes with new FSO links.</p> / Thesis / Master of Applied Science (MASc)

<p>With the emergence of 5G/mm-Wave communication, there is a growing need for novel front-end electromechanical devices in filtering and carrier generation applications. CMOS-MEMS resonators fabricated using state-of-the-art Integrated Circuit (IC) manufacturing processes provide a significant advantage for power, area and cost savings. In this work, a comprehensive physics-based compact model capable of capturing the non-linear behaviour and other non-idealities has been developed for MEMS resonators seamlessly integrated in CMOS. As the first large signal model for CMOS-embedded resonators, it enables holistic design of MEMS components with advanced CMOS circuits as well as system-level performance evaluation within the framework of modern IC design tools. Global Foundries 14nm FinFET (GF14LPP) Resonant Body Transistors (fRBT) operating at 11.8 GHz are demonstrated and benchmarked against this large-signal electromechanical model. </p> <p><br></p> <p>Additionally, there is a growing interest in CMOS-integrable ferroelectric materials such as Hafnium Dioxide (HfO2) and Aluminum Scandium Nitride (AlScN) for next-generation memory and computation, as well as electromechanical transduction in CMOS-MEMS devices. This work also explores the performance of 700 MHz Ferroelectric Capacitor-based resonators in the Texas Instruments HPE035 process under high-power operating conditions. Identification of previously unreported characteristics, together with the first nonlinear large signal model for integrated ferroelectric resonators, provides insights on the design of frequency references and acoustic filters using ferroelectric transducers. </p> <p><br></p> <p>Extending the range of unreleased CMOS-MEMS resonators to lower frequency using novel design, we also investigate embedded transducers in chip-scale devices for physical sensing. We have simulated and modeled the transducer coupling for low-frequency propagating modes and benchmarked their projected performance against state-of-the-art conventional MEMS sensors. A new approach to phononic crystal (PnC) Interdigitated Transducers (IDTs) is presented emulating the acoustic dispersion in conventional ICs. Unloaded quality factors up to 15,000 have been measured in $\sim$80 MHz resonators, demonstrating their capacity for resonant rotation sensing. We present a unique methodology to amplify and collimate acoustic waves using CMOS-design-rule-compliant Graded Index (GRIN) Phononic IDTs. Ultimately, the CMOS-MEMS techniques presented in this work for both RF applications and physical sensing can facilitate additional functionality in standard CMOS and emerging 3D heterogeneously integrated (3DHI) ICs with minor or no modifications to manufacturing and packaging. This enables new paradigms in next-generation communications, internet of things (IoT), and hardware security.</p>

This thesis addresses the problem of target positioning and localization using Radio Frequency (RF) based measurements and using a variety of modulation including Time of Arrival (ToA), Phase of arrival (PoA) and Received Strength Indicator of RF signals (RSSI). Starting from finding the planar coordinates of a device from a collection of ranging measurements using weighted least square (WLS) methods, we explore the dependency of the solution uncertainty from the geometric configuration of anchors and then develop solutions that compensate for the effects of geometry and reduce the positioning uncertainty to a value close to the Cramer–Rao Lower Bound (CRLB), a measure which is then used in the proceeding chapters for developing optimal anchor configurations for positioning problem with guaranteed estimation uncertainties. The findings in the positioning part are also used to address the limitations of initializing Ultra-Wideband (UWB) anchors through a random trajectory. This is done by studying the dual of the positioning problem addressed in the first part, that is incorporating CRLB as a measure of optimality to design a trajectory that minimizes the uncertainty of anchor initialization. We finally close the positioning part of the thesis by studying the range and bearing measurements provided by radar sensors for people tracking and positioning in indoor environments. Taking into account the target dynamics, in the second part of the thesis we present observabilty analysis and localization for non-holonomic robots, using a combination of onboard sensors and range-only anchors. By using a discrete-time formulation of the system’s kinematics, we identify the geometric conditions that make the system globally observable and thereby derive the observability-based filter (ObF) to outperform the limitations of the classic Bayesian filters. We then use the implications of this analysis to design an active control and optimal path-planning strategy with guaranteed maximum observability. We close this part of the thesis by investigating localization in presence of intermittent measurements and discuss how the observability of a trajectory can be quantified by the condition number of the system matrix, a subject related to the maneuvers executed by the robot and to the sampling time used to collect the measurements. Eventually, in the last part of this thesis, we address the localization in presence of offset and ambiguities in measurements. First, we show that, while using range-only measurements corrupted with offset, the trajectories can be observed and the offset can be estimated in a finite number of steps. Next, we present an approach to resolve the ambiguity of rang-only measurements obtained from RSSI values at the Ultra-High Frequency (UHF) band by proposing an optimization algorithm that merges RFID and odometry data to reconstruct the entire robot trajectory. Finally, we present a solution to resolve the ambiguity of the RFID signal phase and reconstruct the robot trajectory through sensor fusion and using UHF-RFID passive tags.

<p dir="ltr"> Conventional RF switching technologies struggle to simultaneously achieve high-power handling, low loss, high isolation, broadband operation, quick reconfiguration, high linearity, and low cost, which are desirable for many applications, including communications, radar, and sensors. Moreover, they require electrical bias networks, which degrade performance and, in many cases, inhibit wideband applications, including DC operation. On the other hand, plasma (photoconductive) switches use an optical bias to generate free charge carriers. Recently these switches have begun to not only rival conventional technologies in terms of performance metrics such as switching speeds and loss but have exceeded what is possible in terms of power handling. This work details the strides made in placing solid-state plasma technologies at the forefront of advanced, high-power switching applications including a novel high-power tuner and an absorptive/reflective SPnT switch. In various form factors, SSP has achieved analog control of loss as low as 0.09 dB and isolation as high as 54 dB, linearity of 68.8 dBm (IP3), 110 GHz instantaneous bandwidth, including DC, switching speeds as low as 3.5 us, 100+ W power handling, and 30+ W hot switching. In addition, comprehensive physics modeling has been developed to enable seamless design validation before fabrication commences. This thesis discusses the achievements and design considerations for creating optimized plasma switches and proposes a path for future applications.</p>

CMOS technology is particularly attractive for commercialization of ultra wideband (UWB) radios due to its low power and low cost. In addition to CMOS implementation, UWB radios would also significantly benefit from a radio architecture that enables digital communications. In addition to the normal challenges of CMOS RFIC design, there are two major technical challenges for the implementation of CMOS digital UWB radios. The first is building RF and analog circuitry covering wide bandwidth over several GHz. The second is sampling and digitizing high frequency signals in the UWB frequency range of 3 GHz to 10 GHz, which is not feasible for existing CMOS analog-to-digital converters. In this dissertation, we investigate the two technical challenges at the circuit level and the system level. We propose a systematic approach at the circuit level for optimal transistor sizing and biasing conditions that result in optimal noise and power matching over a wide bandwidth. We also propose a general scheme for wideband matching. To verify our methods, we design two single-stage low noise amplifiers (LNAs) in TSMC 0.18µm CMOS technology. Measurement results from fabricated chips indicate that the proposed LNAs could achieve as high as 16 dB power gain and as low as 2.2 dB noise figure with only 6.4 mA current dissipation under a supply voltage of 1.2 V. At the system level, we propose a unique frequency domain receiver architecture. The receiver samples frequency components of a received signal rather than the traditional approach of sampling a received signal at discrete instances in time. The frequency domain sampling leads to a simple RF front-end architecture that directly samples an RF signal without the need to downconvert it into a baseband signal. Further, our approach significantly reduces the sampling rate to the pulse repetition rate. We investigate a simple, low-power implementation of the frequency domain sampler with 1-bit ADCs. Simulation results show that the proposed frequency-domain UWB receiver significantly outperforms a conventional analog correlator. A digital UWB receiver can be implemented efficiently in CMOS with the proposed LNA as an RF front-end, followed by the frequency domain sampler. / Ph. D.

While the toolset known as Machine Learning (ML) is not new, several of the tools available within the toolset have seen revitalization with improved hardware, and have been applied across several domains in the last two decades. Deep Neural Network (DNN) applications have contributed to significant research within Radio Frequency (RF) problems over the last decade, spurred by results in image and audio processing. Machine Learning (ML), and Deep Learning (DL) specifically, are driven by access to relevant data during the training phase of the application due to the learned feature sets that are derived from vast amounts of similar data. Despite this critical reliance on data, the literature provides insufficient answers on how to quantify the data training needs of an application in order to achieve a desired performance. This dissertation first aims to create a practical definition that bounds the problem space of Radio Frequency Machine Learning (RFML), which we take to mean the application of Machine Learning (ML) as close to the sampled baseband signal directly after digitization as is possible, while allowing for preprocessing when reasonably defined and justified. After constraining the problem to the Radio Frequency Machine Learning (RFML) domain space, an understanding of what kinds of Machine Learning (ML) have been applied as well as the techniques that have shown benefits will be reviewed from the literature. With the problem space defined and the trends in the literature examined, the next goal aims at providing a better understanding for the concept of data quality through quantification. This quantification helps explain how the quality of data: affects Machine Learning (ML) systems with regard to final performance, drives required data observation quantity within that space, and impacts can be generalized and contrasted. With the understanding of how data quality and quantity can affect the performance of a system in the Radio Frequency Machine Learning (RFML) space, an examination of the data generation techniques and realizations from conceptual through real-time hardware implementations are discussed. Consequently, the results of this dissertation provide a foundation for estimating the investment required to realize a performance goal within a Deep Learning (DL) framework as well as a rough order of magnitude for common goals within the Radio Frequency Machine Learning (RFML) problem space. / Doctor of Philosophy / Machine Learning (ML) is a powerful toolset capable of solving difficult problems across many domains. A fundamental part of this toolset is the representative data used to train a system. Unlike the domains of image or audio processing, for which datasets are constantly being developed thanks to usage agreements with entities such as Facebook, Google, and Amazon, the field of Machine Learning (ML) within the Radio Frequency (RF) domain, or Radio Frequency Machine Learning (RFML), does not have access to such crowdsourcing means of creating labeled datasets. Therefore data within the Radio Frequency Machine Learning (RFML) problem space must be intentionally cultivated to address the target problem. This dissertation explains the problem space of Radio Frequency Machine Learning (RFML) and then quantifies the effect of quality on data used during the training of Radio Frequency Machine Learning (RFML) systems. Taking this one step further, the work then goes on to provide a means of estimating data quantity needs to achieve high levels of performance based on the current Deep Learning (DL) approach to solve the problem, which in turn can be used as guidance to better refine the approach when the real-world data quantity requirements exceed practical acquisition levels. Finally, the problem of data generation is examined and provides context for the difficulties associated with procuring high quality data for problems in the Radio Frequency Machine Learning (RFML) space.

With the increase in spectrum congestion, intelligent spectrum sensing systems have become more important than ever before. In the field of Radio Frequency Machine Learning (RFML), techniques like deep neural networks and reinforcement learning have been used to develop more complex spectrum sensing systems that are not reliant on expert features. Architectures like Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) have shown great promise for applications like automated modulation classification, signal detection, and specific emitter ID. Research in these areas has primarily focused on "one size fits all" networks that assume a fixed signal length in both training and inference. However, since some signals are more complex than others, due to channel conditions, transmitter/receiver effects, etc., being able to dynamically utilize just enough of the received symbols to make a reliable decision allows for more efficient decision making in applications such as electronic warfare and dynamic spectrum sharing. Additionally, the operator may want to get to the quickest possible decision. Recurrent neural networks have been shown to outperform other architectures when processing temporally correlated data, such as from wireless communication signals. However, compared to other architectures, such as CNNs, RNNs can suffer from drastically longer training and evaluation times due to their inherent sample-by-sample data processing. While traditional usage of both of these architectures typically assumes a fixed observation interval during both training and testing, the sample-by-sample processing capabilities of recurrent neural networks opens the door for "decoupling" these intervals. This is invaluable in real-world applications due to the relaxation of the typical requirement of a fixed time duration of the signals of interest. This work illustrates the benefits and considerations needed when "decoupling" these observation intervals for spectrum sensing applications. In particular, this work shows that, intuitively, recurrent neural networks can be leveraged to process less data (i.e. shorter observation intervals) for simpler inputs (less complicated signal types or channel conditions). Less intuitively, this works shows that the "decoupling" is dependent on appropriate training to avoid bias and insure generalization. / Master of Science / With the increase in spectrum congestion, intelligent spectrum sensing systems have become more important than ever before. In the field of Radio Frequency Machine Learning (RFML), techniques like deep neural networks and reinforcement learning have been used to develop more complex spectrum sensing systems that are not reliant on expert features. Architectures like Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) have shown great promise for applications like automated modulation classification, signal detection, and specific emitter ID. Research in these areas has primarily focused on "one size fits all" networks that assume a fixed signal length in both training and inference. However, since some signals are more complex than others, due to channel conditions, transmitter/receiver effects, etc., being able to dynamically utilize just enough of the received symbols to make a reliable decision allows for more efficient decision making in applications such as electronic warfare and dynamic spectrum sharing. Additionally, the operator may want to get to the quickest possible decision. Recurrent neural networks have been shown to outperform other architectures when processing temporally correlated data, such as from wireless communication signals. However, compared to other architectures, such as CNNs, RNNs can suffer from drastically longer training and evaluation times due to their inherent sample-by-sample data processing. While traditional usage of both of these architectures typically assumes a fixed observation interval during both training and testing, the sample-by-sample processing capabilities of recurrent neural networks opens the door for "decoupling" these intervals. This is invaluable in real-world applications due to the relaxation of the typical requirement of a fixed time duration of the signals of interest. This work illustrates the benefits and considerations needed when "decoupling" these observation intervals for spectrum sensing applications. In particular, this work shows that, intuitively, recurrent neural networks can be leveraged to process less data (i.e. shorter observation intervals) for simpler inputs (less complicated signal types or channel conditions). Less intuitively, this works shows that the "decoupling" is dependent on appropriate training to avoid bias and insure generalization.

Portable communication systems require, in part, high-efficiency radio frequency power amplifiers (RF PA) if battery lifetime is to be conserved. Conventional amplifier classifications and definitions are presented in a unified and concise format. The Bipolar Junction Transistor (BJT) and Metal-Semiconductor Field Effect Transistor (MESFET) are evaluated as active devices in high-efficiency RF PA designs. Two amplifier classes (class CE and class F) meet the system requirements of an 850 MHz operating frequency, a power output of 3 W, a battery supply voltage of 9 Vdc, and a sinusoidal-type signal to be amplified. Both classes are evaluated through recent research literature and simulated using the PSpice® computer simulation program. Class CE and class F are found to provide efficiencies exceeding 80 percent under the given system constraints.</p. / Master of Science

Yes / The potentially broad application area in engineering design using Genetic Algorithm (GA) has been widely adopted by many researchers due to its high consistency and accuracy. Presented here is the initial design of a wideband non-dispersive wire bow-tie antenna using GA for breast cancer detection applications. The ultimate goal of this design is to achieve minimal late-time ringing but at higher frequencies such as that located from 4 to 8 GHz, in which is desire to penetrate human tissue for near field imaging. Resistively loading method to reduce minimal ringing caused by the antenna internal reflections is implemented and discussed when the antenna is located in free space and surrounded by lossy medium. Results with optimised antenna geometry and different number of resistive loads are presented and compared with and without existence of scatterers.

Deep learning has become an ubiquitous part of research in all fields, including wireless communications. Researchers have shown the ability to leverage deep neural networks (DNNs) that operate on raw in-phase and quadrature samples, termed Radio Frequency Machine Learning (RFML), to synthesize new waveforms, control radio resources, as well as detect and classify signals. While there are numerous advantages to RFML, this thesis answers the question "is it secure?" DNNs have been shown, in other applications such as Computer Vision (CV), to be vulnerable to what are known as adversarial evasion attacks, which consist of corrupting an underlying example with a small, intelligently crafted, perturbation that causes a DNN to misclassify the example. This thesis develops the first threat model that encompasses the unique adversarial goals and capabilities that are present in RFML. Attacks that occur with direct digital access to the RFML classifier are differentiated from physical attacks that must propagate over-the-air (OTA) and are thus subject to impairments due to the wireless channel or inaccuracies in the signal detection stage. This thesis first finds that RFML systems are vulnerable to current adversarial evasion attacks using the well known Fast Gradient Sign Method originally developed for CV applications. However, these current adversarial evasion attacks do not account for the underlying communications and therefore the adversarial advantage is limited because the signal quickly becomes unintelligible. In order to envision new threats, this thesis goes on to develop a new adversarial evasion attack that takes into account the underlying communications and wireless channel models in order to create adversarial evasion attacks with more intelligible underlying communications that generalize to OTA attacks. / Master of Science / Deep learning is beginning to permeate many commercial products and is being included in prototypes for next generation wireless communications devices. This technology can provide huge breakthroughs in autonomy; however, it is not sufficient to study the effectiveness of deep learning in an idealized laboratory environment, the real world is often harsh and/or adversarial. Therefore, it is important to know how, and when, these deep learning enabled devices will fail in the presence of bad actors before they are deployed in high risk environments, such as battlefields or connected autonomous vehicle communications. This thesis studies a small subset of the security vulnerabilities of deep learning enabled wireless communications devices by attempting to evade deep learning enabled signal classification by an eavesdropper while maintaining effective wireless communications with a cooperative receiver. The primary goal of this thesis is to define the threats to, and identify the current vulnerabilities of, deep learning enabled signal classification systems, because a system can only be secured once its vulnerabilities are known.