80264 research outputs found
Sort by
High-Performance RF/mmWave Transceiver Building Blocks, Beamformers, and Their Design Automation
No full textThe mmWave frequency spectrum inherently provides larger bandwidth (BW) and higher data rates compared to lower frequency bands. This increased BW also enhances sensing resolution, making it particularly valuable for radar and imaging applications. Consequently, mmWave integrated circuits (ICs) have broad applications across diverse fields, including high-speed wireless communications, automotive systems, and human vital sign monitoring. However, mmWave circuit and system design presents several key challenges. (1) As the operating frequency approaches the upper limits of the technology, achieving high gain and low noise in amplifiers becomes increasingly difficult. This necessitates the exploration of new transistor technologies and innovative circuit topologies. (2) Although phased arrays have been extensively used in mmWave bands to mitigate the high path loss, existing phased arrays are mostly built based on analog beamforming, which can support only a single beam at any given time. This is insufficient to enable multi-user multiple-input multiple-output (MU-MIMO) technique, which is prevalent in sub-6GHz communication bands. (3) The design of RF/mmWave circuits has been a manual process, which is labor-intensive and time-consuming. This has limited the design productivity and caused extensive development overhead for the time to market.
This thesis focuses on addressing these challenges through three research thrusts, aiming to (1) enhance mmWave building block performance, (2) enable MU-MIMO for mmWave phased arrays, and (3) enhance RF/mmWave IC design productivity through AI-assisted design automation.
Chapter 2 and Chapter 3 present two LNA design examples that demonstrate how we enhance the mmWave building block performance through transistor-level innovations. Specifically, I proposed a systematic yet intuitive design approach, which can transform a conventional narrow-band LNA into a broadband implementation. It’s worth noting that this new design approach only updates LNA component values without changing the prevalent inductive degeneration LNA topology, thereby ensuring minimal design overhead. Utilizing this approach, for the first time, I achieved an octave LNA BW from 25 to 50 GHz, with state-of-the-art NF performance. I also developed a new output network synthesis methodology, which miniaturizes a wideband filter-based output network into a single on-chip transformer footprint. Our network outperforms existing transformer-based networks in terms of BW and in-band gain ripple, allowing for significantly improved control on its frequency response. I also proposed a new linearity enhancement scheme to improve the LNA IP3 performance.
Beyond component-level innovations demonstrated in Chapter 2 and Chapter 3, I also delved into the transceiver system design. Chapter 4 presents a novel notch steering scheme with its silicon implementation. This scheme integrates an auxiliary-path vector modulator (VM) into each antenna element to generate an interference-canceling beam. By spatially combining the array factors of the main beam and the interference-canceling beam, a deep spatial notch is created with minimal main-beam power degradation. Unlike the conventional ZF method that requires matrix inversion in digital for spatial notch creation, our scheme enables the computation of antenna weights in analog, thereby significantly reducing the computational cost and latency. The proposed notch steering scheme is also scalable. More auxiliary-path VMs can be added to achieve multiple spatial notches. And the design overhead remains minimum as the auxiliary-path VMs are transistor-only implementation. We demonstrated this idea using a 28-GHz four-element fully connected HBF TX array. It is capable of simultaneously transmitting two independent, wideband (400-MHz) DSs in the same polarization toward two directions, while achieving >35dB spatial interference rejection.
As RF/mmWave systems continue to push the limits of performance, the need for efficient, scalable design methodologies becomes critical. Traditional manual design process for RF and mmWave ICs is both time-consuming and labor-intensive, limiting innovation and slowing development cycles. Recent advancement in AI has demonstrated its powerful capability of using learning-based approaches to solve complex tasks. Inspired by these breakthroughs, Chapter 5 and Chapter 6 presents how the machine learning techniques can be applied in the RF/mmWave IC design automation. Ultimately, the goal is to enhance design productivity, while ensuring that key circuit performance metrics are on par with or even exceed those of state-of-the-art manual designs. Our approach adopts ”AI-human collaboration” to ensure explainable, trustworthy designs while leveraging AI’s full potential, which frees designers to focus on circuit insights and exploring new ideas. Chapter 5 demonstrates this flow in the GlobalFoundries 22-nm CMOS SOI process using two DCO designs: one covering 7.1–8.6 GHz and another at 3.8–4.6 GHz. Both achieved >192.4 dBc/Hz FoM and <1.5 kHz frequency resolution in a single inductor footprint. And a 28-GHz mmWave PA is also shown in Chapter 6, which achieves an OP1dB of 18.6 dBm with a power-added efficiency (PAE) of 41.6%.
Finally, Chapter 7 concludes this thesis and discusses the future research directions
High-throughput robotic strategy to implant ultraflexible neural devices
No full textNeural interfaces have shown remarkable potential in restoring sensory and motor function,
treating neurological disorders, developing technologies, and studying cognitive mechanisms.
Among neural recording and stimulation techniques, Micro-Electrode Arrays (MEAs) offer
the most significant temporal and spatial span, and recent advances in ultraflexible MEAs
have allowed for minimal foreign body response, thus improving the longevity and quality of the recorded signals. However, ultraflexible MEAs surgeries represent a considerable
challenge, even after comprehensive training, and carry the potential risk of inducing tissue damage or infection, ultimately contributing negatively to the performances of neural
probes. To overcome surgical hurdles, I present a high-throughput robotic strategy automating the insertion of up to 32 ultraflexible neural probes simultaneously. The system permits both speeds in the µm/s range for scarless insertion and high acceleration for
successful retraction of the shuttle wire. Tailored tungsten wire designs fabricated using
laser-micromachining enable tissue damage to be minimized. A user-friendly interface, with
real-time feedback and pre-coded procedures, enhances surgical precision and time efficiency
compared to non-assisted implantation strategies. Coupling functional imaging, laser scans,
and brain vasculature images, this innovative apparatus, achieving micrometer precision,
can target discrete layers and brain regions across the entire neocortex of rodents while
avoiding vascular structures. The successful development and deployment of this robotic
approach represent a first in inserting ultraflexible neural probes and holding promises of
streamlining a pivotal process in brain-computer interfaces
Pedogenic Carbonate Formation Mechanisms and their Trace Element Signatures
Get PDFSoil is the largest terrestrial reservoir of carbon. Pedogenic carbonates are of particular importance due to their stability, long residence times, and strength as an archive of soil water chemistry. Geochemical and textural differences among pedogenic carbonates indicate multiple formation mechanisms that reflect different environmental conditions and sources of carbon. Here, we measure trace elements in pedogenic carbonates from Dance Bayou, TX. The carbonates in this region have previously been shown to precipitate under both oxic and anoxic conditions, with distinct carbon isotope compositions reflecting the incorporation of atmospheric versus soil-respired CO2. We develop a cleaning method using sequential acetic acid leaches, quantify trace element concentrations using inductively coupled plasma mass spectrometry, pair the results with carbon and oxygen isotopic data, and develop a framework to interpret our measurements in the context of pedogenic carbonate formation mechanisms
Ultrastrong Light-Matter Coupling in Magnetic Materials
Get PDFHybrid light-matter coupled states, or polaritons, in magnetic materials have attracted significant attention due to their potential for enabling novel applications in spintronics and quantum information processing. While much attention has been paid to strong coupling of ferromagnetic structures to cavities, coupling between light and antiferromagnetic (AFM) or paramagnetic structures begs further study. Here, we investigate AFM magnon polaritons in NiO, demonstrating strong coupling at frequencies exceeding 1 THz, offering promise for advancements in terahertz magnonics.
Additionally, we examine ultrastrong coupling in the paramagnetic insulator Gd3Ga5O12, where the electron paramagnetic resonance (EPR) of Gd3+ ions behaves similarly to a two-level atom ensemble, described by the Dicke model. Observing temperature dependent behavior in Zeeman polaritons demonstrates that a spin--boson system is more compatible with the Dicke model and has advantages over boson--boson systems for pursuing experimental realizations of phenomena predicted for ultrastrongly coupled light--matter hybrids.
These results provide a groundwork for cavity-enabled devices, spintronics, and exploration of Dicke physics in condensed matter systems
Development of an Automatic Speed Modulation Control for Next Generation LVAD Based on Real-Time Cardiac Activity Monitoring
Get PDFCardiac disease is the leading cause of death and places significant strain on the
limited supply of donor hearts. Left Ventricular Assist Devices (LVADs) extend pa-
tient survival as a bridge to transplant and, increasingly, as destination therapy;
however, current devices are challenged by driveline infections and inadequate re-
sponsiveness to dynamic cardiovascular demands. Wireless charging offers promise
for mitigating driveline infections, but its success depends on reducing overall pump
energy consumption.
To address these issues, we’re developing an intelligent hybrid magnetic levitation
LVAD system. In this study, we introduce a novel, automatic speed modulation
algorithm that leverages the LVAD’s average magnetic levitation power consumption
as a surrogate indicator of patient activity state. The algorithm, implemented via
a finite state machine with threshold detection, dynamically adjusts pump speed
to optimize hemodynamic performance while minimizing energy usage. The system
was evaluated using a previously validated numerical mock circulatory loop (nMCL)
coupled with a detailed LVAD model, with simulations conducted across activity
states including sleep, rest, and light exercise. Comparative analysis demonstrated
that the speed modulation algorithm improved circulatory support during exercise
while reducing energy consumption during rest. This framework promises enhanced
energy efficiency and improved physiological compatibility, paving the way for future
LVAD designs and clinical applications
Open, But at What Cost? Examining Platform Openness and Its Market Implications
No full textPlatform openness has emerged as a critical governance strategy to facilitate platform growth and long-term success. My dissertation examines how platform openness affects complementors, users and through what mechanisms. In Chapter 1, I take a demand heterogeneity perspective and argue that increased platform openness attracts more herd adopters who have lower purchase intentions and higher exit rates. Chapter 2 further examines how openness influences complementors, particularly new entrants. By disentangling selection and competition mechanisms, I reveal that greater openness negatively affects new entrants' performance by flooding the market with low-quality complements and intensifying competitive pressures. Chapter 3 investigates the overlooked role of information environments, showing that platform openness amplifies users' reliance on informational cues, such as user reviews, thus significantly impacting complementor outcomes. Empirical analyses using a natural experiment from Steam's governance policy change provide robust evidence for all three studies. Together, my dissertation extends extant studies by adding more insights into the trade-offs of platform openness, making novel theoretical contributions to the literature of platform governance strategy, complementor competition dynamics, and effective information system design
The Gerber Method: Using Multipliers for Daily Box Office Prediction
No full textThe movie industry is important to the United States both culturally and financially. A core part of the movie industry is exhibition at the domestic box office. This paper proposes and implements a model to predict the daily box office gross of a film over the entire course of its time in theaters. Using a novel approach based on daily, weekly, and seasonal multipliers, the model creates a robust time series which it updates as new data becomes available. To do this, comparison movies are found using a K-nearest neighbors model; then the individual characteristics of these similar movies are combined with a set of predicted multipliers. The model is trained on a dataset of over 3,000 movies and their box office grosses from 2015 to 2025. This approach is not only novel, but the daily time series modeling is something no other paper has attempted. Overall, we find very strong results with a median weighted mean absolute percentage difference (WMAPD) of 0.2
Acoustically targeted measurement of transgene expression in the brain
No full textGene expression is a critical component of brain physiology and activity. Brain development, function, and plasticity relies on a regulated process of converting genetic information into functional products. However, monitoring gene expression in the living brain has been a significant challenge. The confined structure of the brain, protected by the cranium and shielded by the blood-brain barrier, has posed difficulty in non-invasive and sensitive measurement of gene expression with specificity. The aim of this thesis is to develop a new paradigm of technology capable of measuring gene expression in the brain non-invasively with cell-type, spatial, and temporal specificity. To achieve this, we combined focused ultrasound liquid biopsy and recovery of engineered protein markers that are designed to be expressed in neurons and exit into the brain’s interstitium. When ultrasound is applied to targeted brain regions, it temporarily opens the blood-brain barrier and releases the interstitial markers into the bloodstream. Once in blood, the markers can be readily detected from blood collection followed by compatible biochemical techniques. We call this Recovery of Markers through InSonation (REMIS). We demonstrated improved recovery of engineered Gaussia luciferase marker, under constitutive promoter, from the brain into the blood in every tested animal. Further, we implemented the markers to measure endogenous neuronal signaling activity by controlling the expression of the marker under a genetic circuit that responds to c-Fos when activated by enhanced neuronal activity. Lastly, we measured enhanced serum level of overexpressed human alpha-synuclein in the engineered Parkinson’s disease model mThy1-aSyn (Line61) mouse strain with REMIS. Overall, our work demonstrates the feasibility of combining engineered gene expression reporters and focused ultrasound liquid biopsy to noninvasively and specifically measure gene expression in the intact brain
The Industrial Organization of Financial Markets
Get PDFThe first chapter, ’Cannibalization and Scope Economies within Fund Families: The Impact of Passive Investing on Fund Fees,’ examines how the rise of passive funds, such as index funds and ETFs, affects price competition in the mutual fund industry through structural demand and supply estimation. By modeling mutual fund companies as multi-product producers offering both active and passive funds, the analysis incorporates multi-product pricing and economies of scale and scope. The results indicate that multi-product fund families increase fees by 1 to 1.5 basis points due to competition internalization, while economies of scope from manag- ing both types of funds substantially reduce fees. A counterfactual merger analysis reveals that price increases driven by market power outweigh the cost-efficiency benefits.
The second chapter, ’Two-sided Matching in IPO Underwriting Market,’ studies the factors that drive a firm to hire one investment bank as a lead manager over another, using a revealed preference approach. I estimate a two-sided matching model where an IPO-issuing firm and an investment bank (lead underwriter) select the best
counterpart from a set of available options at the time. The focus is on whether firms with prestigious underwriters are matched to firms with greater uncertainty to mitigate their uncertainty or to firms with less uncertainty to complement their lower risk. The results indicate that firms with greater risk at the time of the IPO, measured by earnings and firm age, are more likely to hire larger and prestigious underwriters. This suggests that underwriter scale and prestige could substitute for the issuing firm’s risk.
The third chapter, ’Weighting for Performance,’ is a coauthored work with Alan Crane and Kevin Crotty. We use mutual fund holdings to identify the portfolio weighting scheme that best characterizes each fund’s asset allocation choice. A ma- jority of funds are best characterized as equally weighted. Only a small minority of funds are classified as market-weighted, despite the pervasiveness of market-weighted benchmarks. The revealed weighting scheme varies systematically with fund char- acteristics, such as portfolio size, fund age, and expense ratios. Holding security selection fixed, we find that the performance of funds on average would improve using simple risk-parity based portfolio weights
Lightweight Physical-Layer Security Primitives for 5G-and-Beyond Wireless Communications
No full textThe development of 5G-and-beyond wireless communication represents a major transition toward faster, more intelligent, and more flexible connectivity. Compared with previous generations, 5G-and-beyond systems are designed not only for higher data rates and larger capacity, but also for lower latency, enhanced intelligence, and broader applicability. These capabilities enable a wide range of mission-critical applications, such as immersive AR/VR, intelligent transportation, remote robotic surgery, and drone-assisted communication, where communication quality is tightly coupled with safety, efficiency, or privacy. However, the open nature of wireless propagation also introduces significant security concerns. In particular, as wireless transceivers become more mobile, autonomous, and distributed, it becomes increasingly difficult to verify their identity and prevent eavesdropping. These concerns raise new requirements for the transmitter (TX), which must be able to identify itself as a legitimate source and prevent sensitive information from leaking to unintended receivers.
Traditionally, wireless security is achieved through digital cryptography. Although effective in many scenarios, cryptographic methods face four limitations when applied to future systems: (1) the added power and latency overhead becomes problematic for real-time bit-wise encryption, (2) key management becomes complex and power-hungry, and (3) physical signal leakage itself is not protected by encryption. To address these issues, physical-layer security (PLS) has gained increasing attention. By embedding security directly into the physical behavior of the TX, such as frequency, phase, amplitude, or time, PLS enables protection without relying on high-level cryptographic protocols. These techniques can be implemented with minimal latency, power, and area overhead, making them suitable for 5G-and-beyond systems where both performance and security are critical.
This thesis focuses on low-overhead, TX-based physical-layer security techniques that address two major security requirements: (1) identification of TX to the receiver, and (2) prevention of wireless eavesdropping. Three system-level designs are proposed, each implemented with custom application-specific integrated circuit (ASIC) TXs and modules, and demonstrated through measurement.
The first TX design addresses the identification problem. We propose a physical-layer identification TX that incorporates a digital physically unclonable function (PUF) to control its spectral regrowth. This creates a unique RF fingerprint (RFF) for each TX, beyond what is achievable with intrinsic process variation alone. A 2.4-GHz prototype is implemented in GlobalFoundries 45-nm CMOS SOI process with 4.7 dBm output power and 36% efficiency. Measurement results show significant improvement in RFF stability, uniqueness, and dynamic range compared to prior work.
On top of it, we further enhance the identification performance with feature extraction and identification model. We develop a lightweight neural network that extracts PSD features from TX signals and performs device identification. The model is optimized for low-power implementation and works seamlessly with the proposed hardware. In measurement, 240 devices are identified with over 99% accuracy, and 40 devices at unseen distance achieve over 95%, demonstrating strong generalization and robustness under various conditions.
The second TX design focuses on preventing sidelobe eavesdropping. We present a mm-Wave antenna subset modulation (ASM) TX operating at 28 GHz. By randomly selecting antenna subsets at the symbol rate, the transmitted I/Q symbols are scrambled outside the main direction, preventing eavesdropping without degrading performance in the desired direction. The ASIC is integrated with on-board antennas and includes a high-speed on-chip true random number generator (TRNG) for secure and unpredictable antenna selection. The system supports 1.2-Gb/s 64-QAM communication with ±2° information beamwidth and maintains high EVM performance. This work highlights the great potential and practicality of integrating ASM technology into future radios.
In summary, this thesis proposes and demonstrates three low-overhead PLS techniques at the transmitter level, targeting future communication systems with tight constraints on power, latency, and security. These methods provide a practical and efficient way to enhance wireless security without relying on complicated cryptographic operations, and can serve as a complementary layer of protection in 5G-and-beyond wireless networks