Integration of alternative materials and devices with CMOS will expand functionality and improve performance of established applications in the era of diminishing returns from Moore's Law scaling. In particular, integration of thin-film magnetic materials will enable improvements in energy efficiency of digital computing applications by enabling integrated power conversion and management with on-chip power inductors. Integrated voltage reg- ulators will also enable fine-grained power management, by providing dynamic scaling of the supply voltage in concert with the clock frequency of synchronous logic to throttle power consumption at periods of low computational demand. Implementation of integrated power conversion requires high capacity energy storage devices. This is best achieved with integration of thin-film magnetic materials for high quality on-chip power inductors. This thesis describes a body of work conducted to develop integrated switch-mode voltage regulators with thin-film magnetic power inductors. Soft-magnetic materials and inductor topologies are selected and optimized, with intent to maximize efficiency and current density of the integrated regulators. Custom integrated circuits are designed and fabricated in 45nm-SOI to provide the control system and power-train necessary to drive the power inductors. A silicon interposer is designed and fabricated in collaboration with IBM Research to integrate custom power inductors by 2.5D chip stacking, enabling power conversion with current density greater than 10A/mm2. The concepts and designs developed from this work will enable significant improvements in performance-per-watt of future microprocessors.

The design of the next-generation Internet infrastructure is driven by the need to sustain the massive growth in bandwidth demands. Novel, energy-efficient, optical networking technologies and architectures are required to effectively meet the stringent performance requirements with low cost and ultrahigh energy efficiencies. In this thesis, a cross-layer communications platform is proposed to enable greater intelligence and functionality on the physical layer. Providing the optical layer with advanced networking capabilities will facilitate the dynamic management and optimization of optical switching based on performance monitoring measurements and higher-layer attributes. The cross-layer platform aims to create a new framework for networks to incorporate packet-scale measurement subsystems and techniques for monitoring the health of the optical channel. This will allow for quality-of-service- and energy-aware routing schemes, as well as an enhanced awareness of the optical data signals. This thesis first presents the design and development of an optical packet switching fabric. Leveraging a networking test-bed environment to validate networking hypotheses, advanced switching functionalities are demonstrated, including the support for quality-of-service based routing and packet multicasting. The investigated cross-layering is based on emerging optical technologies, enabling packet protection techniques and packet-rate switching fabric reconfiguration. Coupled with fast performance monitoring, the platform will achieve significant performance gains within the endeavor of all-optical switching. Allowing for a more intelligent, programmable optical layer aims to support greater flexibility with respect to bandwidth allocation and potentially a significant reduction in the network's energy consumption. The ultimate deliverable of this work is a high-performance, cross-layer enabled optical network node. The experimental demonstration of an initial prototype creates a dynamic network element with distributed control plane management, featuring fast packet-rate optical switching capabilities and embedded physical-layer performance monitoring modules. The cross-layer box enables an intelligent traffic delivery system that can dynamically manipulate optical switching on a packet-granular scale. With the goal of achieving advanced multi-layer routing and control algorithms, the network node requires an intelligent co-optimization across all the layers. The proposed cross-layer design should drive optical technologies and architectures in an innovative way, in order to fulfill the void between the design of basic photonic devices and the networking protocols that use them. The performance of the entire network -- from the optical components, to the routing algorithms and user applications -- should be optimized in concert. This contribution to the area of cross-layer network design creates an adaptable optical pipe that is extremely flexible and intelligent aware of both the physical optical signals and higher-layer requirements. The impact of this work will be seen in the realization of dynamic, energy-efficient optical communication links in future networking infrastructures.

It is an ongoing effort to improve field-effect transistor (FET) performance. With silicon transistors approaching their physical limitations, alternative materials that can outperform silicon are required. Graphene, has been suggested as such an alternative mainly due to its two-dimensional (2D) structure and high carrier velocities. The band structure limits achievable bandgaps, preventing digital electronic applications. This, however, does not rule out analog electronic applications at high frequencies, where the full potential of improved carrier speeds in graphene can be exploited. In this thesis, the high-bias characteristics of graphene FETs are investigated. Current saturation as well as the effect of ambipolar conduction on the current-voltage characteristics are studied. A field-effect model is developed that can capture the effects of the unique band structure, such as a density-dependent saturation velocity. The effect of channel length scaling in these devices is studied down to 100-nm channel length with the aid of pulsed-measurement techniques. Transistors RF performance and bias dependence of high frequency behavior is explored. Novel fabrications methods are developed to improve FET performance. A technique is developed to grow metal-oxides on graphene surface for efficient gate coupling. An alternative approach to making high quality devices is realized by incorporating hexagonal-boron nitride as a gate dielectric. These transistors exhibit the potential of graphene electronics for high-performance analog electronic applications.

Energy Harvesting Active Networked Tags (EnHANTs) are a new type of wireless device in the domain between RFIDs and sensor networks. Future EnHANTs will be small, flexible, and self-powered devices that can be attached to everyday objects that are traditionally not networked to enable "Internet of Things" applications. This work describes the design and development of the EnHANT prototypes and testbed. The current prototypes use thin-film photovoltaics optimized for indoor light harvesting, form multihop networks using ultra-low-power Ultra-Wideband Impulse Radio (UWB-IR) transceivers, and implement energy harvesting adaptive networking protocols. The current testbed enables the evaluation of different algorithms by exposing individual prototypes to repeatable light conditions based on real-world irradiance data. New approaches to characterizing the energy available to energy harvesting devices were explored. A mobile data-logger was used to record the intensity of ambient light, determine the light source, and record the acceleration from motion during different real world activities. These traces were used to model the behavior of photovoltaic and inertial energy harvesters in real world deployments and can be replayed in the EnHANTs testbed. In addition, new techniques to evaluate the efficiency of different photovoltaic technologies under indoor illumination were developed. A proof-of-concept system was built to characterize photovoltaics under a standardized set of conditions in which the radiant intensity and spectral composition of the light source were systematically varied. Techniques to structure student research projects within the EnHANTs project were developed. Project-based learning approaches were implemented to engage students using real-world system development constraints. A survey of the students showed that this approach is an effective method for developing technical, professional, and soft skills. Open source hardware was also applied to EnHANTs project and extended into other domains. A laboratory-based class in flat panel display technology was developed. The course introduces fundamental concepts of display systems and reinforces these concepts through the fabrication of three display devices. A lab kit platform was developed to enable remote students to use low-cost, course specific hardware to complete the lab exercises remotely. This platform was also applied to external projects targeted at non-university students. A workshop was developed to teach artists, designers, and hobbyists how to design and build custom user interfaces using thin-film electronics and rapid prototyping tools. Surveys of the students and workshop participants showed that this platform is an effective teaching tool and can be easily adapted and expanded.

Merging chemical and biomolecular sensors with silicon integrated circuits has the potential to push complex electronics into a low-cost, portable platform, greatly simplifying system- level instrumentation and extending the reach and functionality of point of use technologies. One such class of sensor, the thin-film bulk acoustic resonator (FBAR), has a micron-scale size and low gigahertz frequency range that is ideally matched with modern complementary metal-oxide-semiconductor (CMOS) technologies. An FBAR sensor can enable label-free detection of analytes in real time, and CMOS integration can overcome the measurement complexity and equipment cost normally required for detection with acoustic resonators. This thesis describes a body of work conducted to integrate an array of FBAR sensors with an active CMOS substrate. A monolithic fabrication method is developed, which allows for FBAR devices to be built directly on the top surface of the CMOS chip through post-processing. A custom substrate is designed and fabricated in 0.18 µm CMOS to support oscillation and frequency measurement for each sensor site in a 6×4 array. The fabrication of 0.8-1.5 GHz FBAR devices is validated for both off-chip and on-chip devices, and the integrated system is characterized for sensitivity and limit of detection. On-chip, parallel measurement of multiple sensors in real time is demonstrated for a quantitative vapor sensing application, and the limit of detection is below 50 ppm. This sensor platform could be used for a broad scope of label-free detection applications in chemistry, biology, and medicine, and it demonstrates potential for enabling a low-cost, point of use instrument.

This thesis investigates some of the practical issues related to the implementation of level-crossing ADCs in nanometer CMOS. A level-crossing ADC targeting minimum power is designed and measured. Three techniques to circumvent performance limitations due to the zero-crossing detector at the heart of the ADC are proposed and demonstrated: an adaptive resolution algorithm, an adaptive bias current algorithm, and automatic offset cancelation. The ADC, fabricated in 130 nm CMOS, is designed to operate over a 20 kHz bandwidth while consuming a maximum of 8.5 uW. A peak SNDR of 54 dB for this 8-bit ADC demonstrates a key advantage of level-crossing sampling, namely SNDR higher than the classic Nyquist limit.

Motivated by the ever increasing demand of wireless communication for larger capacity and higher quality, wireless communication system grows from a single-pair point-to-point communication system to a multiple-transceiver pair communication network. Various new communication techniques, for example, cooperative communication, interference management, multi-carrier communication, are employed to enhance the system capacity and improve the communication quality. Even for some single-pair communication scenarios, due to the different quality demands for different types of information messages, more advanced coding schemes should be designed to provide more protection for more important information messages, for example, the system emergency message. This thesis proposes several coding schemes to address the above questions. More specifically, the proposed coding schemes are summarized as follows. Message-wise error protection is a new unequal error protection scheme where in a codebook some special messages are more protected than other ordinary messages. We propose the first practical coding scheme for message-wise error protection based on LDPC codes, where codeword flipping is employed to separate the special message codewords from the ordinary message codewords. We consider a half-duplex 4-node joint relay system with two sources, one relay, and one destination, where the relay combines the information from both sources and transmits it to the destination together with both sources. We propose joint network and channel coding schemes based on the superposition coding (SC) and the Raptor coding (RC), and design practical Raptor codes for the proposed coding schemes. We propose novel coding and decoding methods for a fully connected K-user Gaussian interference channel. Each transmitter encodes its information into multiple layers and transmits the superposition of those layers. Each receiver performs a twofold task by first identifying which interferers it should decode and then determining which layers of them should be decoded. We propose practical coding schemes that employ the quadrature amplitude modulations (QAM) and Raptor codes. We propose group decoding and the associated rate allocation schemes for the multi-relay assisted interference channels, where both the relays and the destinations employ constrained group decoding. We consider two types of relay systems, the hopping relay system with no direct source-destination links, and the inband relay system with direct source-destination links. For each relay type, our objective is to design the relay assignment and group decoding strategies at the relays and destinations, to maximize the minimum information rate among all source-destination pairs. We consider a distributed storage system employing some existing regenerate codes where the storage nodes are scattered in a wireless network. The existing full-downloading approach, where the data collector downloads all symbols from a subset of the storage nodes for data reconstruction, becomes less efficient in wireless networks. This is because that, due to fading, the wireless channels may not offer sufficient bandwidths for full downloading. We propose a partial downloading scheme that allows downloading a portion of the symbols from any storage node, and formulate a cross-layer wireless resource allocation problem for data reconstruction employing such partial downloading. We derive necessary and sufficient conditions for the data reconstructability for partial downloading, in terms of the numbers of downloaded symbols from the storage nodes. We also propose channel and power allocation schemes for partial downloading in wireless distributed storage systems.

Silicon photonics has grown rapidly since the first Si electro-optic switch was demonstrated in 1987, and the field has never grown more quickly than it has over the past decade, fueled by milestone achievements in semiconductor processing technologies for low loss waveguides, high-speed Si modulators, Si lasers, Si detectors, and an enormous toolbox of passive and active integrated devices. Silicon photonics is now on the verge of major commercialization breakthroughs, and optical communication links remain the force driving integrated and Si photonics towards the first commercial telecom and datacom transceivers; however other potential and future applications are becoming uncovered and refined as researchers reveal the benefits of manipulating photons on the nanoscale. This thesis documents an exploration into the unique guided-wave and nonlinear properties of deeply-scaled high-index-contrast sub-wavelength Si waveguides. It is found that the tight confinement inherent to single-mode channel waveguides on the silicon-on-insulator platform lead to a rich physics, which can be leveraged for new devices extending well beyond simple passive interconnects and electro-optic devices. The following chapters will concentrate, in detail, on a number of unique physical features of Si waveguides and extend these attributes towards new and interesting devices. Linear optical properties and nonlinear optical properties are investigated, both of which are strongly affected by tight optical confinement of the guided waveguide modes. As will be shown, tight optical confinement directly results in strongly vectoral modal components, where the electric and magnetic fields of the guided modes extend into all spatial dimensions, even along the axis of propagation. In fact, the longitudinal electric and magnetic field components can be just as strong as the transverse fields, directly affecting the modal group velocity and energy transport properties since the longitudinal fields are shown to contribute no time-averaged momentum. Furthermore, the vectoral modal components, in conjunction with the tensoral nature of the third-order susceptibility of Si, lead to nonlinear properties which are dependent on waveguide orientation with respect to the Si parent crystal and the construction of the modal electric field components. This consideration is used to maximize effective nonlinearity and realize nonlinear Kerr gratings along specific waveguide trajectories. Tight optical confinement leads to a natural enhancement of the intrinsically large effective nonlinearty of Si waveguides, and in fact, the effective nonlinearty can be made to be almost 10^6 times greater in Si waveguides than that of standard single-mode fiber. Such a large nonlinearity motivates chip-scale all-optical signal processing techniques. Wavelength conversion by both four-wave-mixing (FWM) and cross-phase-modulation (XPM) will be discussed, including a technique that allows for enhanced broadband discrete FWM over arbitrary spectral spans by modulating both the linear and nonlinear waveguide properties through periodic changes in waveguide geometry. This quasi-phase-matching approach has very real applications towards connecting mature telecom sources detectors and components to other spectral regimes, including the mid-IR. Other signal processing techniques such as all-optical modulation format conversion via XPM will also be discussed. This thesis will conclude by looking at ways to extend the bandwidth capacity of Si waveguide interconnects on chip. As the number of processing cores continues to scale as a means for computational performance gains, on-chip link capacity will become an increasingly important issue. Metallic traces have severe limitations and are envisioned to eventually bow to integrated photonic links. The aggregate bandwidth supported by a single waveguide link will therefore become a crucial consideration as integrated photonics approaches the CPU. One way to increase aggregate bandwidth is to utilize different eigen-modes of a multimode waveguide, and integrated waveguide mode-muxes and demuxes for achieving simultaneous mode-division-multiplexing and wavelength-division-multiplexing will be demonstrated.

Analog-to-digital converters (ADCs) are analog pre-processing systems that convert the real life analog signals, the input of sensors or antenna, to digital bits that are processed by the system digital back-end. Due to the various issues associated with CMOS technology scaling such as reduced signal swings and lower transistor gains, the design of ADCs has seen a number of challenges in medium to high resolution and wideband digitization applications. The various chapters of this thesis focus on efficient design techniques for ADCs that aim to address the challenges associated with design in scaled CMOS technologies. This thesis discusses the design of three analog and mixed-signal prototypes: the first prototype introduces current pre-charging (CRP) techniques to generate the reference in Multiplying Digital-to-Analog Converters (MDACs) of pipeline ADCs. CRP techniques are specifically applied to Zero-Crossing Based (ZCB) Pipeline-SAR ADCs in this work. The proposed reference pre-charge technique relaxes power and area requirements for reference voltage generation and distribution in ZCB Pipeline ADCs, by eliminating power hungry low impedance reference voltage buffers. The next prototype describes the design of a radiation-hard dual-channel 12-bit 40MS/s pipeline ADC with extended dynamic range, for use in the readout electronics upgrade for the ATLAS Liquid Argon Calorimeters at the CERN Large Hadron Collider. The design consists of two pipeline A/D channels with four MDACs with nominal 12-bit resolution each, that are verified to be radiation-hard beyond the required specifications. The final prototype proposes Switched-Mode Signal Processing, a new design paradigm that achieves rail-to-rail signal swings with high linearity at ultra-low supply voltages. Switched-Mode Signal Processing represents analog information in terms of pulse widths and replaces the output stage of OTAs with power-efficient rail-to-rail Class-D stages, thus producing Switched-Mode Operational Amplifiers (SMOAs). The SMOAs are used to implement a Programmable Gain Amplifier (PGA) that has a programmable gain from 0-12dB.

Detection and estimation, two classical statistical signal processing problems with wellestablished theories, are traditionally studied under the fixed-sample-size and centralized setups, e.g., Neyman-Pearson target detection, and Bayesian parameter estimation. Recently, they appear in more challenging setups with stringent constraints on critical resources, e.g., time, energy, and bandwidth, in emerging technologies, such as wireless sensor networks, cognitive radio, smart grid, cyber-physical systems (CPS), internet of things (IoT), and networked control systems. These emerging systems have applications in a wide range of areas, such as communications, energy, the military, transportation, health care, and infrastructure. Sequential (i.e., online) methods suit much better to the ever-increasing demand on time-efficiency, and latency constraints than the conventional fixed-sample-size (i.e., offline) methods. Furthermore, as a result of decreasing device sizes and tendency to connect more and more devices, there are stringent energy and bandwidth constraints on devices (i.e., nodes) in a distributed system (i.e., network), requiring decentralized operation with low transmission rates. Hence, for statistical inference (e.g., detection and/or estimation) problems in distributed systems, today's challenge is achieving high performance (e.g., time efficiency) while satisfying resource (e.g., energy and bandwidth) constraints. In this thesis, we address this challenge by (i) first finding optimum (centralized) sequential schemes for detection, estimation, and joint detection and estimation if not available in the literature, (ii) and then developing their asymptotically optimal decentralized versions through an adaptive non-uniform sampling technique called level-triggered sampling. We propose and rigorously analyze decentralized detection, estimation, and joint detection and estimation schemes based on level-triggered sampling, resulting in a systematic theory of event-based statistical signal processing. We also show both analytically and numerically that the proposed schemes significantly outperform their counterparts based on conventional uniform sampling in terms of time efficiency. Moreover, they are compatible with the existing hardware as they work with discrete-time observations produced by conventional A/D converters. We apply the developed schemes to several problems, namely spectrum sensing and dynamic spectrum access in cognitive radio, state estimation and outage detection in smart grid, and target detection in multi-input multi-output (MIMO) wireless sensor networks.