Mar
2023Passive mask-based lensless cameras encode depth information in their measurements for a certain depth range. Early works have shown that this encoded depth can be used to perform 3D reconstruction of close-range scenes. However, these approaches for 3D reconstructions are typically optimization based and require strong hand-crafted priors and hundreds of iterations to reconstruct. In this work, we propose FlatNet3D—a feed-forward deep network that can estimate both depth and intensity from a single lensless capture. FlatNet3D is an end-to-end trainable deep network that directly reconstructs depth and intensity from a single lensless measurement using an efficient physics-based 3D mapping stage and a fully convolutional network. Our algorithm is fast and produces high-quality results, which we validate using both simulated and real scenes captured using PhlatCam.
Another interesting aspect of mask-based lensless cameras is their ability to perform computation in the optical domain with operations defined by the mask patterns and configurations. In this work, we exploit this particular ability of lensless cameras to design privacy-preserving cameras using optical encryption. Existing lensless camera designs fail to preserve the privacy of the scene in which they are deployed, due to inherent flaws in their design. To overcome this, we propose OpEnCam - novel lensless camera design for optical encryption. OpEnCam is a lensless camera that encrypts the incoming light before capturing it using optical masks’ modulating ability. Recovery of the original scene from an OpEnCam measurement is possible only if one has access to the camera’s encryption key, defined by the unique optical elements of each camera. Our OpEnCam design suggests two major improvements over existing lensless camera designs - (a) the use of two co-axially located optical masks, one stuck to the sensor and the other a few millimeters above the sensor, and (b) the design of mask patterns derived heuristically from signal processing ideas. We show, through experiments, that OpEnCam is robust against point-source and blind ciphertext-only attacks while still maintaining the imaging capabilities of existing lensless cameras when the key is known. We validate the efficacy of OpEnCam using simulated and real data.
Finally, we explore the possibility of designing the masks of lensless cameras together with inference algorithms. In this work, we propose a learning-based framework for designing the mask for thin lensless cameras. To highlight the effectiveness of the learned lensless systems, we learn a phase mask for multiple computer vision tasks using physics-based neural networks. Specifically, we learn the optimal mask for the following tasks- 2D scene reconstructions, optical flow estimation, and face detection. We show that the mask learned through this framework is better than heuristically designed masks, especially for small sensor sizes that allow lower bandwidth and faster readout. We verify the performance of our learned phase mask on real data.
Mar
2023Gallium nitride (GaN) is becoming a mainstream semiconductor for power and radio-frequency applications. Due to the superior material properties of GaN and the intrinsic advantages of the high electron mobility transistor (HEMT) structure, there has been an increased demand for GaN based HEMTs. Conventional GaN HEMTs are depletion mode or normally-ON devices. On the other hand, high performance normally-OFF GaN HEMTs eliminate the requirement of negative supply voltage and makes the circuit configurations simple. A gate electrode surrounding the channel can provide enhanced control over the electrons in the two-dimensional electron gas (2-DEG) resulting in a shift of threshold voltage towards the positive direction. Recently, fin-shaped HEMT (Fin-HEMT) structures have been introduced to use this advantage. They can realize superior gate controllability, and enhancement-mode operation with a higher current ON/OFF ratio as compared with the planar counterparts. To facilitate the use of these transistors for circuit design, robust compact models that accurately capture the behaviour of the device are needed.
In this talk, a charge based compact model for drain current in Fin-HEMTs will be presented. The model considers the effects of both the top and side gates on the charge densities inside the device through modified gate capacitance formulations. First, a closed form expression for 2-DEG density in Fin-HEMTs will be discussed. The total drain current which is modeled as the sum of 2-DEG and double-gate field effect transistor (FET) currents will be elaborated subsequently. A special feature of the model is that it can be used for different fin configurations by changing the corresponding equivalent circuits. Finally, the model validation with experimental data will be presented.
Mar
2023As the demand for high-speed and reliable wireless communication continues to grow, energy efficiency has become a critical factor in the design of wireless networks. One emerging technology that has greatly interested the research community is the cell-free massive MIMO (multiple-input-multiple-output) architecture. Cell-free massive MIMO has been demonstrated to improve the performance over the small-cell schemes and thus is a promising technology capable of achieving the demands of 6G communication standards. However, while the system's capabilities are impressive, its high energy consumption presents a significant challenge in achieving energy-efficient wireless communication.
In this talk, we will focus on energy-efficient user-centric cell-free massive MIMO systems, which aim to maximize the network's sum rate while meeting the energy constraint. We will discuss the challenges in designing such systems, including the optimization of power coefficients of users and antenna selection coefficients at access points (APs). We will begin with a brief introduction to cell-free (CF) and user-centric cell-free massive MIMO (UC CF-mMIMO) and discuss our derived lower bound on the achievable uplink rate of users in a UC CF-mMIMO system in the presence of a mixed analog-to-digital converter (ADC) resolution profile at the APs. We will then describe our optimization problem mathematically. We will present an alternating optimization solution approach based on successive convex approximation (SCA) and binary particle swarm optimization (BPSO), which has shown to be an effective method. Additionally, we will discuss a complete meta-heuristic-based solution approach, which can be used as an alternative solution for applications where latency is the critical metric.
Please attend if you are interested to know more about user-centric cell-free massive MIMO.
Mar
2023In wireless communication systems, high frequency acoustic resonators are used as filters and frequency reference oscillators. Commonly used resonators are surface acoustic wave resonator, FBAR and quartz crystal resonator. Crystal resonators are used as clock generator due to high quality factor and excellent temperature frequency stability. The drawback of these resonators is their bulky size and incompatible fabrication process, which hamper the integration of the high performance oscillators with CMOS circuits. One possible candidate to replace the bulky oscillators is high Q-factor micromachined MEMS resonator, which provides an avenue to bring low power multi modal filters, oscillators and sensors on-chip. Thin film piezoelectric on Silicon (TPoS) technology enables high frequency resonators with low motional impedance, relatively high Q-factor and high linearity. The TPoS resonator comprises of a piezoelectric layer sandwiched between two metal electrodes on Silicon, which is low acoustic loss material. In this work, we had studied the effect of physical dimensions such as length, width and thickness of the TPoS MEMS resonator operating at 1 GHz on its performance. The effects of geometrical dimensions, the order of excitation, and number of anchors on Q-factor were modeled considering different loss mechanisms and excellent agreement with experimental data is achieved. For a resonator of 225 μm width excited in its 23rd mode of resonance at 990 MHz, the measured motional resistance, unloaded quality factor in vaccum and linear thermal coefficient of frequency were 107 Ω, 9556 and -28 ppm respectively. The measured unloaded quality factor is the highest reported in literature for TPoS resonators in this frequency range.
Mar
2023Controlled manipulation of microparticles with optical tweezers is a sensitive tool for observing, detecting, and studying the properties of local environments in fields such as biology and soft matter. The translational and rotational degrees of freedom of a probe particle each store unique information about the local environment. Scientists use several techniques, such as optical tweezers, magnetic tweezers, and microscope imaging tools, to control and track the position of particles. However, the out-of-plane rotations of particles, in particular, have not been extensively studied. In this work, I present techniques to generate and detect out-of-plane rotations induced by single and dual beam traps. Additionally, I demonstrate the use of out-of-plane rotations to probe the mechanical properties of air-water interfaces and generate accelerated self-assembly at the interface using thermal flows. This work provides valuable insights and information about various systems in the fields of biology, soft matter, and others.
Mar
2023This talk will be a narrative on how biosensor activities started in our group at IIT Bombay. The perspective that I'll offer in this talk will be that of an embedded systems developer attempting to make biosensors. I'll talk about our journey of the last few years, highlighting choices we made and breakthroughs that worked in our favor, as well as challenges and failures that have made us work harder. I'll highlight our work on an optical bio-sensing platform, which originated as a solution for sensing cardiac disease biomarkers, and has evolved to fit many other applications. I'll also talk about our work on electrochemical sensors using PCBs as electrodes for DNA and biomarker sensing.
Bio:Siddharth Tallur is an Associate Professor in Electrical Engineering department at IIT Bombay. Prior to joining IIT Bombay in 2016, he worked on developing MEMS inertial sensor products as MEMS product and applications engineer at Analog Devices Inc. in Wilmington MA, USA. Siddharth graduated with Ph.D. in Electrical and Computer Engineering at Cornell University, and received the best thesis award for his work on opto-mechanical oscillators. He is an alumnus of IIT Bombay, having graduated with B.Tech. in Electrical Engineering in 2008. His research interests include developing low-cost and low-power smart sensor systems for structural health monitoring and biosensing.
This talk is being organised under the auspices of the Indian Nanoelectronics User’s Programme – Idea to Innovation (INUP-i2i), supported by the Ministry of Electronics and Information Technology (MeitY). More details of this programme are available at https://inup.iitm.ac.in/
Mar
2023This thesis explores the design and fabrication of miniaturized tunable optical gratings. Two approaches are explored, the first being a period-tunable optical grating in the visible range. The grating device is designed to work in the transmission mode, and is fabricated using stretchable soft material. The device consists of a prestretched elastomer membrane with printed electrodes on either side of the membrane and sinusoidal grating profile on one side. The central region which is transparent is used as the optical element. The design and fabrication protocol of the device is discussed. The electrical characterization of the fabricated device gives a grating period tunability of 34.4% on DC electrical actuation of 5.5 kV.
Feb
2023The boundaries of existing cities are expanding rapidly due to the exponential growth in urban population. Therefore, the existing water distribution network (WDN) system needs to be extended to the newly developed areas to meet the additional demand. The optimal design of a sub-network planned for network expansion requires multiple simulations under various constraints. Simulating the additional sub-network along with the existing network takes a lot of CPU time. The size of the problem becomes larger when the stochastic nature of domestic demands, optimal design and layouts, control, and operation of various hydraulic components are considered. In this study, network reduction methodologies are developed for single and multi-port connections using the Thevenin theorem. The number of network elements in the equivalent network obtained using the Thevenin theorem is significantly less than the ones obtained by the existing WDN reduction methods. It is possible to reorganize and expand a large existing network from a prior knowledge of the most sensitive portion of the large network. The accuracy and robustness of the proposed reduction methodologies are investigated on realistic WDNs by comparing the results with the established hydraulic simulator, EPANET. The applications of a single-port Thevenin equivalent network for optimal subnetwork design and for finding the economic diameter for maximum power transfer are evaluated. Therefore, the proposed network reduction methods can greatly benefit hydraulic engineers. Finally, we shall discuss an equivalent network construction methodology based on multiple measurements at different nodal points of a large network for which a complete layout information is unavailable.
Feb
2023Thin noncontact angle sensors that provide reliable, high-resolution output even in a dusty and humid environment are very valuable in industrial applications. In this talk, a novel design to realize a thin eddy-current-based angle sensor will be presented. The proposed sensor is realized by converting the shaft, whose angle of rotation is to be measured, into the sensing element. The modification to the shaft is minimal; a small surface groove is introduced without affecting the mechanical strength. The stationary part of the sensor consists of two layers of flexible square-planar coils. Depending on the angular position of the shaft, the inductances of the planar coils get modified. These are measured using a specially designed circuitry, optimized for this sensor. The output for the entire circle range (360o) is derived from the inductance values of each coil using a successive approximation algorithm developed for this purpose. Finite-element analysis was used to design the sensor and analyze the axial/radial misalignment of the rotor. A sensor prototype was built and tested. The output showed a resolution of 0.1o and a worst-case linearity error of 0.9%. The prototype sensor dimensions are selected such that it fits a standard steering column. The proposed sensor is thin, easy to manufacture at low cost, tolerant to axial vibration by design, and has a 360o sensing range. The details will be presented in the talk.
Feb
2023This talk presents two methods proposed for estimating the state of n-qubit quantum systems as part of my M.S. work (More focus on the second method).
In order to motivate the talk, a simple quantum system namely "qubit" will be discussed. This will be done by explaining the polarization of light (Electromagnetic radiation) which realizes a qubit just as an on-off switch realizes a classical bit. The state of polarization of light and the measurements of it will be discussed which will then lead to formally stating the problem considered in this work.
At the end of any quantum operation carried on a quantum computer, there is a need to estimate the state of quantum systems. Although these computations are "quantum", the estimation of the quantum states (represented by a complex matrix) can be performed using classical computers from their measurements. To this end, we propose two methods. The first method incorporates prior knowledge about the state and estimates the quantum state by inverting the linear relation between the state and the measurements obtained. This method provides a closed form expression for the estimate because of which the estimation is performed much faster than any iterative methods. Two caveats of this method are as follows. 1) It requires an informationally complete set of measurements which is exponential in the size of the system (n). 2) The estimate may not be positive semidefinite which is a property of the quantum state.
To overcome the above said caveats, in the second work, a neural network for quantum estimation is designed based on a popular iterative algorithm named Singular Value Thresholding (SVT). The proposed network estimates low-rank quantum states from fewer measurements than that is required for an informationally complete set. Since the network is tailored to quantum state tomography, the estimate is always a valid quantum state. The proposed network is computationally more efficient than SVT in the sense that a 3-layered network (equivalent of 3 SVT iterations) outperforms SVT (in terms of fidelity) which converges in 1700 iterations.
Jan
2023A smart grid is an electricity network based on digital technology that is used to supply electricity to consumers via two-way digital communication, which allows for monitoring, analysis, control and communication to help improve efficiency, reduce energy consumption and cost, and maximize the transparency and reliability of the energy supply chain. The smart grid was introduced with the aim of overcoming the weaknesses of conventional electrical grids by using smart net meters.
Speaker Bio:
Dr. Ramu Ramanathan received the B.Tech from I.I.T., Madras, in 1977, the M.Sc.E. from University of New Brunswick, in 1979, and the Ph.D. from Virginia Tech in 1982. His experience spans a wide variety of organizations including the private and government sector including a regulated and unregulated arena for utilities. He had senior executive positions at leading companies. He has started a new business line within the major corporation and successfully grown the business by more than 500%. He has turned around companies and also started new businesses by getting funding from investors. He is currently the president of Maxisys Inc. He was a member of the board of directors for TIE (Global International Entrepreneurs Organization). He has published more than 50 technical papers, and reports and offered IEEE tutorials on optimization techniques.
Jan
2023In this talk, I will focus on two research thrusts in energy and latency-efficient edge intelligence. The first thrust is inspired by the fact that spiking neural networks (SNNs) are a promising low-power alternative to compute and memory-expensive DNNs, due to their high activation sparsity and use of accumulates (AC) instead of expensive multiply-and-accumulates (MAC). However, most existing SNNs require multiple time steps for acceptable inference accuracy, hindering real-time deployment and increasing energy consumption. This motivates the need for novel training frameworks to optimize source DL models, their conversion to target SNNs, and subsequent SNN fine-tuning, which I will discuss in this talk. The second thrust is inspired by the fact that modern image sensors generate huge amounts of high-resolution visual data that typically must be transferred to downstream CPUs/accelerators for computer vision (CV) processing. This data transfer requires significant bandwidth and can dominate the total energy consumption, particularly when the CV processing is heavily optimized using SNNs. I will discuss a novel in-sensor computing paradigm based on CNNs, coupled with intelligent algorithm-hardware co-design that can mitigate this concern. This paradigm is the first to demonstrate the feasibility of enabling all the computational aspects of modern CNN layers inside image sensors. Coupled with the optimized SNNs, this paradigm can reduce the total system energy-delay product (EDP) of several on-device CV workloads by two orders of magnitude compared to existing approaches without significantly affecting the performance.
Biography
Gourav Datta (Graduate Student Member, IEEE) received the B.Tech. degree in instrumentation engineering with a minor in electronics and electrical communication engineering from the Indian Institute of Technology, Kharagpur in 2018. He is currently pursuing a Ph.D. degree at the University of Southern California, USA. His research interests include energy and latency-efficient algorithm-hardware co-design for machine learning at the edge. He is a finalist of the Qualcomm Innovation fellowship 2022 (North America), a recipient of the IEEE Graduate Fellowship on Applied Superconductivity, and a Ming Hsieh Ph.D. scholar. During his Ph.D. tenure, he has published 20 peer-reviewed papers in top-tier venues such as Scientific Reports, Frontiers in Neuroscience, TCAS-I, DATE, and WACV among others.
Jan
2023Recent advancements in fields like Internet of Things (IoT), augmented reality and robotics have led to an unprecedented demand for miniature cameras with low cost that can be integrated anywhere and can be used for distributed monitoring. Lensless imaging has emerged as a potential solution for realizing ultra-miniature cameras by eschewing the bulky lens in a traditional camera. However, the reduction in the size and cost of these imagers comes at the expense of their image quality due to the high degree of multiplexing inherent in their design. Without a focusing lens, lensless cameras rely on computational algorithms to recover the scenes from multiplexed measurements.
Current iterative optimization-based lensless scene reconstruction algorithms produce noisier and perceptually poorer images. In this work, we propose a non-iterative deep learning-based reconstruction approach that results in a significant improvement in image quality for lensless scene reconstructions. Our approach, called FlatNet-separable, integrates the lensless imaging model with a convolutional neural network to provide high-quality reconstructions. It does so by having two stages—(a) an inversion stage that uses the physics of the separable forward model to learn a mapping from the measurement to intermediate image space, and (b) an enhancement stage that uses a fully convolutional network to enhance the intermediate image. We use a separable FlatCam prototype to perform our experiments. We demonstrate the efficacy of FlatNet-separable on a large-scale real paired dataset collected using a monitor display capture setup and unpaired measurements of challenging real scenes in front of the camera.
Although FlatNet-separable provides photorealistic reconstructions for a separable lensless prototype, it cannot provide any meaningful reconstruction when the separability of the model is broken. Moreover, recent trends in lensless imaging have shown that non-separable models have superior properties compared to separable ones. To overcome this shortcoming of FlatNet-separable, we propose a novel architecture called FlatNet-gen. FlatNet-gen, like FlatNet-separable, uses two stages—(a) an inversion stage and (b) an enhancement stage. However, to account for the non-separability of the model, we learn an inverse of the forward process in the Fourier domain. We use a non-separable PhlatCam prototype to perform our experiments. We demonstrate the efficacy of FlatNet-gen on both display and direct captured data that we collected. To demonstrate the ability of FlatNet-gen to reconstruct in uncontrolled illuminations, we collect a paired dataset using a PhlatCam-Webcam. Finally, with the help of this dataset, we show that FlatNet-gen can provide high-quality scene reconstructions in uncontrolled illuminations.
Jan
2023Modern electrical power systems are migrating towards renewable energy, which is fraught with difficulties. The grid's diminishing inertia is one of the key challenges to the grid’s transition towards renewable energy sources.Grid Inertia (GI) is a measure of total kinetic energy available in the grid that balances the supply mismatch immediately after a disturbance. It represents the grid's inherent ability to withstand power imbalances. A significant amount of inertia was provided by traditional power plants, which are being replaced by renewable sources with low or no inertia. With reduced inertia, the stability of the power system is affected, increasing the probability of grid outages. Therefore, it is crucial for the grid operators to monitor grid inertia for a secured, stable and economic operation of the grid.
In this seminar talk, a novel method to estimate regional inertia of a power system from switching events such as connection or disconnection of generators or loads is presented. The method uses Synchronous Phasor Measurement Units (PMUs) installed at the generators or tie-lines of the region. In the event of detection of frequency events, like i.e., power imbalance, a system of linear equations is formulated and solved to obtain the inertia. The performance of the method is demonstrated using Kundur Two-Area and IEEE-39 bus test systems. In addition, the effect of loads’ response that resembles inertia due to voltage and frequency dependence is examined. An analysis is carried out to study how the inertia-like response of loads varies with respect to various system parameters.
Jan
2023This research is about the optimal covariance steering (CS) problem for discrete time linear stochastic systems modelled using moment-based ambiguity sets. To hedge against the uncertainty in the state distributions while performing covariance steering, distributionally robust risk constraints are employed during the optimal allocation of the risk. Specifically, a distributionally robust iterative risk allocation (DR-IRA) formalism is used to solve the optimal risk allocation problem for the CS problem using a two-stage approach. The upper-stage of DR-IRA is a convex problem that optimizes the risk, while the lower-stage optimizes the controller with the new distributionally robust risk constraints. The proposed framework results in solutions that are robust against arbitrary distributions in the considered ambiguity set. Finally, I will demonstrate the proposed approach using numerical simulations.
Brief Bio:
Currently, I am a postdoctoral research fellow working with Dr. Anders Rantzer at the department of automatic control in Lund University, Sweden. I was born in Nagercovil, Tamilnadu. I finished my undergraduate studies in Electrical & Electronics Engineering from the Government College of Technology, Coimbatore. I completed my Masters in Electrical Engineering from Arizona State University, USA where I worked with Dr. Armando Rodriguez on my thesis on Missile Target Engagement for bank-to-turn missiles. Subsequently, I moved to The University of Texas at Dallas for my PhD under Dr. Tyler Summers where I worked on security of cyberphysical systems and risk bounded motion planning using the distributionally robust optimization techniques. My current research interests include learning based and adaptive control, risk bounded motion planning, anomaly detection in cyber physical systems.
Jan
2023Analog/mixed-signal (AMS) circuits are a crucial component of modern-day integrated circuits whose layouts continue to be handcrafted, unlike their digital counterparts, which have been predominantly automated. This manual effort is one of the leading causes of low design productivity, long design/layout cycles, high development costs, and respins due to failures. Traditionally, AMS layouts generated automatically have been unable to match the performance of handcrafted ones. Advanced process nodes restrict layout geometries and have lesser freedom of choice. Although the number of AMS blocks in designs continues to increase, a significant portion of these requires modest performance. Even for high-performance designs, automated layouts can help reduce design/layout iterations. Recent advances in machine learning help solve the AMS layout generation problem in a manner that was not previously possible. These advances make it amenable for AMS layout automation. This talk will present (a) an open-source automated AMS layout synthesis tool developed from the ground up with these considerations and (b) a specific problem of automatic well-island generation and well-tap insertion.
Biography:
Ramprasath received B. Tech in Electronics and Communication Engineering from the NIT Trichy in 2009 and PhD in Electrical Engineering from IIT Madras in 2016. He is currently a postdoctoral associate at the University of Minnesota, working on open-source AMS layout synthesis and hardware architectures for machine learning inference. Before this, he was at IBM, Synopsys, and Qualcomm, developing tools and methodologies for automation of back-end-of-line design.
Jan
2023Magnetic Resonance Imaging (MRI) is a powerful diagnostic tool for a variety of disorders, but its utility is often limited by its slower speed and higher cost compared to modalities like CT or X-Rays. Reducing the time required for a scan would decrease the cost of MR imaging and allow for acquiring scans in situations where a patient cannot stay still for the scan duration. To accelerate the imaging time, MR scan data is sub-sampled, which brings aliasing artifacts in the image domain. Recently, artificial neural networks (ANN) have received attention in medical imaging particularly in producing artifact free images from sub-sampled scans. However, deep learning methods have certain limitations.
To begin with, the existing networks haven’t simultaneously utilized the information available in image and frequency domain. Second, though a large number of models have been proposed for multi-channel MRI reconstruction, these models need to be compared on a common platform and studied for generalizability when these models are tested on datasets acquired with different coils. Third, fully sampled data needed for supervised training of models is difficult to acquire. Algorithms need to be developed to learn useful features from easily available under-sampled data. Finally, there is a need to verify how under-sampled data can be used to enhance the generalizability of deep learning models. In this work, we develop networks to address the limitations mentioned earlier.
First, we propose Dual-Encoder-Unet which takes in both zero filled k-space and under sampled image as input and simultaneously optimizes both the domain data for reconstruction. Second, we participated in a global Multi-channel MR Image Reconstruction challenge and proposed a deep learning based cascaded architecture and evaluated its performance across different models. Third, we propose a self-supervision-based pretext learning algorithm, inspired from autoencoder for efficient feature extraction from under-sampled data. Finally, we propose three different pretext learning algorithms for multi-channel MR Image reconstruction and evaluated how pre-training can enhance the generalizability of the models. We have extensively validated all the proposed methods with suitable experiments.
Jan
2023Magnetic Resonance Imaging (MRI) is a valuable clinical diagnostic modality for spine pathologies with excellent characterization for infection, tumor, degenerations, fractures and herniations. However in surgery, image-guided procedures continue to rely on CT and fluoroscopy, as MRI slice resolutions are typically insufficient. Building upon state-of-the-art single image super-resolution, we propose a reference-based, unpaired texture-transfer strategy for deep learning based MRI super-resolution (SR). In this work, we implement reference based super-resolution both as an attachable module without a trainable feature extractor and as a complete end-to-end trainable SR network.
First, we use the scattering transform based module to relate the texture features of image patches to unpaired reference image patches, and we include a loss term for multi-contrast texture consistency. Secondly, we propose a fully trainable transformer network for the reference based super resolution (RefSR) with lifting scheme based texture extractor. The commonly used VGG feature extractors in RefSR contain redundant and irrelevant information and neglect the high frequency details of reference image. The usage of data-adaptive, lifting scheme based wavelet transform represents the texture information in multiple scales and frequency sub-bands. The texture matching is carried out in the low frequency sub-band and the multi-scale swapped features are incorporated into the model at different stages. We compare our models with state-of- the-art networks using three publicly available datasets and observe improvement in PSNR and SSIM metrics for 4x super-resolution.
Jan
2023Knee arthroscopy is a complex procedure, mostly due to the 2D limited field of view provided by the arthroscope, the strong hand-eye coordination required of the operator, and poor ergonomics. In many cases, it may lead to unintended injuries to the patient and/or postoperative complications. Employing a real-time guidance imaging, such as ultrasound (US), could bring a significant improvement in the outcome. Registration of partial view intra-operative US to pre-operative MRI is an essential step for such image-guided minimally invasive surgeries. In this work, we present an automatic, landmark-free 3D multimodal registration of pre-operative MRI to 4D US (high-refresh-rate 3D-US) for enabling guidance in knee arthroscopy. We focus on the problem of initializing registration in the case of partial views.
The proposed method utilizes a pre-initialization step of using the automatically segmented structures from both modalities to achieve a global geometric initialization. From the US volumes, the required structures are segmented using a pre-trained Mask R-CNN deep network. For MRI, the segmentations are achieved using deformable registration of an open-source SPL knee atlas with the acquired MRI volumes using ANTs (Advanced Normalization Tools). This is followed by computing distance maps of the procured segmentations for registration in the distance space. Following that, the final local refinement between the MRI-US volumes is achieved using the LC2 (Linear correlation of linear combination) metric.
The method is evaluated on 11 cases spanning six subjects, with four levels of knee flexion. The errors obtained through the developed registration algorithm and inter-observer difference values are found to be comparable. We have shown that the proposed algorithm is simple, robust and allows for the automatic global registration of 3D US and MRI that can enable US-based image guidance in minimally invasive procedures.
Jan
2023Packaging is undergoing a major paradigm shift and promises to take up the lag caused by the slowing down
of CMOS scaling. In this paper, we examine these shifts that have been driven by the scaling of key
packaging metrics such as bump pitch, trace pitch, inter-die spacing and alignment. The goal of advanced
packaging is to enable the same benefits that Moore/Dennard scaling has accomplished for CMOS viz.
density, performance, power, and cost and can make packaged chip assemblies comparable to monolithic
SoCs using these metrics with the additional advantage of heterogeneity. The vehicles that advanced
packaging employs are somewhat different: dielets/chiplets, advanced assembly techniques, simplified inter-
chip communication protocols and cost optimization via the use of optimized heterogeneous technologies.
Another important aspect of advanced packaging is the adoption and adaptation of silicon technology
methods to packaging. In this talk we will discuss the technologies and some instantiation examples that we
have developed at UCLA.
Biography
Subramanian S. Iyer (Subu) is Distinguished Professor and holds the Charles P. Reames
Endowed Chair in the Electrical Engineering Department and a joint appointment in the Materials
Science and Engineering Department at the University of California at Los Angeles. He is Director
of the Center for Heterogeneous Integration and Performance Scaling (UCLA CHIPS). Prior to that
he was an IBM Fellow. His key technical contributions have been the development of the world’s
first SiGe base HBT, Salicide, electrical fuses, embedded DRAM and 45nm technology node used
to make the first generation of truly low power portable devices as well as the first commercial
interposer and 3D integrated products. He has been exploring new packaging paradigms and
device innovations that may enable wafer-scale architectures, in-memory analog compute and
medical engineering applications. He is a fellow of IEEE, APS, iMAPS and NAI as well as a
Distinguished Lecturer of IEEE EDS and EPS. He is on the Board of Governors of IEEE EPS. He
is a Distinguished Alumnus of IIT Bombay and received the IEEE Daniel Noble Medal for emerging
technologies in 2012 and the 2020 iMAPS Daniel C. Hughes Jr Memorial award and the iMAPS
distinguished educator award in 2021. Prof. Iyer is currently Prof. Ramakrishna
Rao Visiting Chair Professor at CeNSE, IISc, Bengaluru.
Jan
2023The power system becomes an attractive choice for cyber attacks because of many interconnections. Thus, it is crucial to understand attack vulnerabilities and their impact on the system. This research presents the Vulnerability Assessment (VA) of grid control systems to data injection cyber-attacks (CA). The analysis, carried out from a power system engineer's perspective, illustrates a successful stealth attack that can be implemented on the system with minimal system knowledge by injecting false data into sensor and actuator measurements. A combination of an eavesdropping attack (EDA) and a False Data Injection Attack (FDIA) is used in the attack model. Further, a singular spectrum analysis-based method extracts system's dynamics under regular operation. A projection-based distance tracking method is proposed for detecting grid control system attacks. Two variations of the algorithm, viz-a-viz, single-variate and multi-variate algorithms, are proposed for detection at different levels of the power system. The proposed methodology is robust, adaptive, and computationally efficient, especially considering system and measurement noises. The three main features of the proposed method are: (i) uses standard SCADA data and does not require attack data, (ii) can be integrated with the existing grid control system with minimal hardware, and (iii) is independent of system configuration and upgrades. Formal hypothesis testing is also proposed to determine the detection threshold and also make the method adaptive. The attack detection algorithm can successfully detect different attacks, including stealth attacks on multiple sensors. The algorithm is tested on an IEEE 39-bus New England test system, 300 bus test system, and 1888 bus RTE system. IoT-based hardware was used for establishing the robustness of the algorithm. The proposed method was found to be reliable, fast, robust, and scalable under noisy measurements compared to existing methods.
Jan
2023The character of emerging distribution systems has dramatically changed in the past two decades due to the infusion of DER, new digital protective devices, intelligent sensors, modern communication, computing and control systems and others at both the primary and secondary distribution levels. These are becoming looped and may be networked in the future to improve system performance. In fact, the entire distribution system should be considered as one comprehensive system.
The IEEE-SA sponsored guide (P2030.12) is the subject of the seminar. It covers the design and selection of protective devices and the coordination between them for various modes of operation of the microgrid. They include grid-connected, islanded modes and transitions between modes. This guide will facilitate the deployment of protection systems, given the challenge of protecting equipment and assets in the different modes of operation of the microgrid. The guide proposes different approaches: centralized and decentralized, passive and active, detection and taking proper actions to protect the microgrid dependably and securely and its equipment to assure resiliency, reliability, safety, and other performance measures.
The seminar presentation will cover the following topics: definition, types, and structures of microgrids, objectives and challenges, design configurations and considerations, system studies, control system coordination, requirements for Advanced Distribution Management System (ADMS) including Microgrid Energy Management System (MEMS), communication structures and system testing. This is an on-going effort, and the status of current progress will be discussed.
Bio:
Dr. S. S. Venkata was with the University of Washington (UW), Seattle, Washington where he taught since 1979 and he continues with his affiliation. He held positions at West Virginia University and the University of Massachusetts, Lowell for eight years. From 1996 to 2002, he was Professor and Chairman of the Department at ISU.In 2003, he was the Palmer Chair Professor of Electrical and Computer Engineering Department at Iowa State University (ISU), Ames, Iowa. He was the founding Dean and Distinguished Professor of Wallace H. Coulter School of Engineering at Clarkson University, Potsdam, New York during 2004-2005. He was Vice President with KEMA Inc. for six years during 2005-2010.
He is also the President of Venkata Consulting Solutions LLC (VCS). He provides R&D support and services for the electric power industry and training for future engineering personnel. He was with GE Power/Alstom Grid Inc. from January 2011 to September 2017 as Principal Scientist.
Dr. Venkata is a Life Fellow of the IEEE. At the IEEE level, he served as a member of the IEEE Fellows Committee for six years during 2010 to 2015 and 2021. He also served on the Power and Energy Society (PES) Board as Vice-President, Publications during 2004-07. In addition, he served the PES at various levels for the past 55 years. In 2016 he received the Robert M. Janowiak Outstanding Leadership and Service Award from ECEDHA. He also received the IEEE PES Douglas M. Staszesky Distribution Automation Award in 2015. In 2000 he received the Third Millennium Award from the IEEE. In 1996 he received the Outstanding Power Engineering Educator Award from the IEEE Power Engineering Society.
His latest book published in 2022 is on "Electric Power and Energy Distribution Systems: Models, Methods, and Applications" (IEEE Press).
Jan
2023This work deals with enhancement in extreme imaging conditions, specifically in low light and in underwater scenario. Supervised networks address the task of low-light enhancement using paired images. However, collecting a wide variety of low-light/clean paired images is tedious as the scene needs to remain static during imaging. In this paper, we propose an unsupervised low-light enhancement network using context-guided illumination-adaptive norm (CIN). Inspired by coarse-to-fine methods, we propose to address this task in two stages. In stage-1, a pixel amplifier module (PAM) is used to generate a coarse estimate with an overall improvement in visibility and aesthetic quality. Stage-2 further enhances the saturated dark pixels and scene properties of the image using CIN. Different ablation studies show the importance of PAM and CIN in improving the visible quality of the image. Next, we propose a region-adaptive single input multiple output (SIMO) model that can generate multiple enhanced images from a single low-light image. The objective of SIMO is to let users choose the image of their liking from a pool of enhanced images. Human subjective analysis of SIMO results shows that the distribution of preferred images varies, endorsing the importance of SIMO-type models. Lastly, we propose a low-light road scene (LLRS) dataset having an unpaired collection of low-light and clean scenes. Unlike existing datasets, the clean and low-light scenes in LLRS are real and captured using fixed camera settings. Exhaustive comparisons on publicly available datasets, and the proposed dataset reveal that the results of our model outperform prior art quantitatively and qualitatively.
For the underwater scenario, several supervised networks exist that attempt to remove haze information from underwater images using paired datasets and pixel-wise loss functions. However, training these networks requires large amounts of paired data which is cumbersome, complex and time-consuming. Also, directly using adversarial and cycle consistency loss functions for unsupervised learning is inaccurate as the underlying mapping from clean to underwater images is one-to-many, resulting in an inaccurate constraint on the cycle consistency loss. To address these issues, we propose a new method to remove haze from underwater images using unpaired data. Our model disentangles haze and content information from underwater images using a Haze Disentanglement Network (HDN). The disentangled content is used by a restoration network to generate a clean image using adversarial losses. The disentangled haze is then used as a guide for underwater image regeneration resulting in a strong constraint on cycle consistency loss and improved performance gains. Different ablation studies show that the haze and content from underwater images are effectively separated. Exhaustive experiments reveal that accurate cycle consistency constraint and the proposed network architecture play an important role in yielding enhanced results. Experiments on UFO-120, UWNet, UWScenes, and UIEB underwater datasets indicate that the results of our method outperform prior art both visually and quantitatively.
Jan
2023 There have been several recent widespread oscillation events in the eastern and western American power interconnections, and in Europe, which were caused by forced oscillations resonating with system modes. For instance, large MW oscillations were seen throughout the eastern interconnection on January 11, 2019, from a 0.25 Hz forced oscillation at a power plant in Florida. These events are related to the phenomenon of interarea resonance when a forced oscillation interacts with system interarea modes with close-by frequencies. The resonance effect can lead to long distance propagation of a forced oscillation, even when the associated system mode is well-damped, depending on the sensitivity of the source location of the forced oscillation. This talk will summarize the mathematical theory of interarea resonance based on linear system analysis and will present several signal processing algorithms for recognizing and analyzing such events online using synchrophasor measurements. The methodology will be illustrated on recent system oscillation events in the North American interconnections.
About the Speaker
Mani V. Venkatasubramanian is a Boeing Distinguished Professor in Electrical Engineering at Washington State University (WSU), Pullman, WA. He also serves as the Director of the Energy Systems Innovation Center (ESIC) at WSU. He received his M.S. and D.Sc. in Systems Science and Mathematics from Washington University, St.Louis, MO, and B.E. (Hons). In Electrical and Electronics Engineering from Birla Institute of Technology and Science, Pilani, India. He was an invited member of the working groups that studied the 1996 Western interconnection blackouts and the 2003 Northeastern blackout. He serves as the Chair of the IEEE PES Working Group on Power System Dynamic Measurements. He is a Fellow of IEEE.
Jan
2023Normal form analysis is used to assess the nonlinear behaviour; it can also be used for siting controllers. Normal form analysis is straight-forward for systems governed by explicit ordinary differential equations. The presence of Static VAr Compensator (SVC) or Static Synchronous Compensator (STATCOM) results in the governing equations that are not in this form; we show that even these cases are amenable to normal form analysis. Normal form analysis can be used to find the locations of Power System Stabilizer (PSS), SVC and STATCOM that result in the best damping of rotor swings. In the methods available in the literature, analysis employing normal forms for selection of the best location of PSS depends on initial conditions of state variables. We employ indices that are independent of initial conditions for finding the locations of PSS, SVC and STATCOM. Case studies are conducted to analyze the effect of siting PSS, SVC and STATCOM on the damping of rotor swings; comparison of the performance of the method of normal forms with that of the linearization method to determine the location of damping controllers is presented.
Dec
2022Software Defined Networking (SDN) is driving transformations in
Research and Education (R&E) networks, enabling innovations in network
research, enhancing network performance, and providing security
through a policy-driven network management framework. The Holland
Computing Center (HCC) at the University of Nebraska-Lincoln (UNL)
supports scientists studying large datasets, and has identified a need
for flexibility in network management and security, particularly with
respect to identifying data flows. This problem is addressed through
the deployment of a production SDN with a focus on integrating network
resource management for large-scale GridFTP data transfers. We will
discuss SNAG (SDN-managed Network Architecture for GridFTP transfers),
an architecture that enables the SDN-based network management of
GridFTP file transfers for large-scale science datasets. SNAG proposes
a novel architecture that combines application- and network-layer
coordination with SDN-enabled network management to manage, monitor
and account for network resources used by GridFTP transfers. We will
also discuss ScienceSDS, a novel software defined security framework
for securely monitoring large-scale science datasets over a software
defined networking and network functions virtualization (SDN/NFV)
infrastructure. Finally, we will discuss other recent and ongoing
projects in the areas of Optical networks, Internet of Things and
Future Internet.
Brief Biography
Byrav Ramamurthy is currently a Professor at the School of Computing
at the University of Nebraska-Lincoln (UNL), where he has been on the
faculty since August 1998. He is the author of the book "Design of
Optical WDM Networks - LAN, MAN and WAN Architectures" and a co-author
of the book "Secure Group Communications over Data Networks" published
by Kluwer Academic Publishers/Springer in 2000 and 2004
respectively. He serves on the Editorial Board of IEEE/ACM
Transactions on Networking (ToN). He serves as an Editor-in-Chief for
the Springer Nature Photonic Network Communications (PNET) journal. He
serves as the Steering Committee Chair for IEEE ANTS conference
(2023-25).
He received his BTech degree from IIT, Madras, India and his MS and
PhD degrees from the University of California, Davis, USA. His
research areas include optical networks, cybersecurity and Internet of
Things (IoT). His research is supported by the U.S. National Science
Foundation, U.S. Department of Energy, U.S. Department of Agriculture,
AT&T Corporation, Agilent Tech., and OPNET Inc.
Dec
2022Modem Architecture Evolution from 4G to 5G, including evolution to support new 5G Redcap and e-Redcap standards.
Dec
2022Designing of proper grounding system is essential for ensuring the reliable and safe operation of power system equipment. The system's earth impedance ought to be as low as possible to safely protect the power system apparatus from the negative impacts of high currents due to faults and lightning. The behavior of grounding electrodes under lightning impulse voltages/currents is an important factor needs to be considered while designing lightning protection systems for power systems as well as buildings. In this talk, Impact of water, acid rain and bentonite on ionization characteristics of soil under lightning impulse voltage will be discussed besides the Impact of Lime Stabilization on Expansive Soil for Grounding.
Dec
2022Grid-tie inverters (GTIs) have a wide range of useful applications. Few examples of GTIs are active front end converter (AFEC), static synchronous compensator (STATCOM), shunt active filter (SAF) and GTI for distributed generation applications (DGAs). The research work spans different aspects of analysis, design and control related to GTIs. An improved phase locked loop (PLL), analysis of converter side inductor ripple current, a new approach to LCL filter design and current controller design are the broad research aspects of this work. In this seminar, the proposed improvements in PLL are discussed.
One of the most important aspects in the control of GTIs is proper synchronization of the converter with the utility grid. The synchronization is usually achieved by phase-locked loops (PLLs). The commonly used PLL structure is the synchronous reference frame based PLL (SRF-PLL) which is implemented in rotating direct-quadrature axis (dq axis). Under non-ideal grid voltage conditions, the fundamental positive-sequence voltage of the utility appears as a dc component in the dq frame while the abnormalities like harmonics, unbalance and dc-offset appear as ac components. The fundamental positive-sequence voltage is the information of interest for synchronization and controls and is to be separated.
Dec
2022In an islanded microgrid (IMG), droop control effectively shares real and reactive power demands among distributed generators (DGs), thereby regulating the frequency and voltage in high X/R ratio networks. However, there is a strong coupling between the real power voltage and reactive power frequency in medium and low-voltage microgrids due to the low X/R ratio. For effective droop control, the coupling between the real power voltage and reactive power frequency should be eliminated through decoupling factors. Existing methods in the literature on decoupled droop control do not consider the effect of changing network conditions. This study proposes a decoupling method for improved power sharing among parallel-operated inverter-based DGs. Each DG injects a second-harmonic voltage, based on which a second harmonic self-impedance is calculated at the DG terminal. With certain assumptions, the fundamental self-impedance is computed and used to determine the decoupling factors. The proposed method was implemented on two test cases and compared with existing droop methods. Simulation results clearly demonstrate that the proposed method improves power sharing accuracy, with varying load and network topological conditions, as compared to existing methods.
Dec
2022CMOS ICs for generating 10Vppd sinusoids at 1kHz and 10kHz with THD less than -140dBc to test high-resolution ADCs are demonstrated. Capacitor nonlinearity cancellation and opamp output conductance nonlinearity suppression are incorporated into an active-RC band-pass filter and oscillator.
Dec
2022Speech is one of the most widely used forms of communication. A text-to-speech (TTS) synthesiser is an important speech technology which generates speech corresponding to a given text. Traditional approaches to training a TTS system, such as unit selection synthesis (USS) and hidden Markov model (HMM) based synthesis, rely on language-specific modules and accurate segmented boundaries at the sub-word level. The hand-crafting of these modules makes system building quite difficult and time-consuming. With the advent of neural network-based end-to-end (E2E) approaches, training TTS systems has become easier when a large amount of data is available for a language. Systems can be trained quickly using accurate
In this talk, I will present two main contributions of our work: (1) A multi-language character map (MLCM) to handle the issue of different scripts across languages. (2) A language-family based perspective to system building. The objective is to exploit the phonotactic properties of language families, where small amounts of accurately transcribed data across languages can be pooled together to train TTS systems. Experiments in low-resource and zero-shot scenarios highlight the efficacy of the proposed approaches.
Dec
2022To satisfy the ever-increasing need for high data rate over wired channels, multi gigabits per second full-duplex serial links are in huge demand in the industrial and automotive sectors. Whereas 1 Gb/s full-duplex serial links for automotive application are already available in the market, research is going on currently to develop 5 or 10 Gb/s serial links to satisfy customer needs. A major bottleneck in achieving higher data rates is the ADC in the receiver path. Medium resolution ADCs with hundreds of MHz to a few GHz sampling rate is essential for such applications. In this work, our goal is to fulfil the ADC speed requirements while solving the area and power penalties associated with commonly used solutions like flash or pipelined architecture.
The flash topology suffers from the drawback of exponential growth in power and area with the resolution. For a resolution of 6-8 bit, this becomes very prominent. Pipelined ADCs, on the other hand, needs a linear residue amplifier which is power consuming and difficult to design in lower technology nodes. Asynchronous SAR ADCs are very power efficient and can achieve few hundreds of MHz sampling speed for moderate resolution ADCs. We present two high-speed power and area efficient SAR ADC cores
which can be used in time-interleaved architectures to achieve sampling speeds in the range of a few GHz. A 6-bit 500-MS/s ADC core is designed for a multi Gb/s ADC
application that consumes only 3.08 mW power. Whereas a 200-MS/s 8-bit SAR ADC is designed to be employed in an existing 1-Gb/s serial link application which consumes
2.8 mW power. Both these designs are fabricated in a low-cost 65 nm technology.
Dec
2022Optimization is ubiquitous in science and many areas of engineering, especially robotics, power distribution networks, signal processing, guidance and navigation of aerial and autonomous vehicles and machine learning. The majority of these applications are subjected to real-time constraints, depending on certain characteristics of the system and typically vary with respect to time. For example, the guidance of a robot in a dynamic environment, traffic engineering in computer networks, economic dispatch problems in power generating units, and neural network learning are some examples where time-varying (TV) optimization is inevitable.
Several methods exist to solve time-invariant convex optimization problems, including Newton’s method, subgradient methods, interior point method and primal-dual dynamics. In this work, a projected primal-dual dynamical system is proposed to solve an inequality constrained nonsmooth convex optimization problem. Then the proposed projected dynamical system is employed to solve the extended Fermat-Torricelli Problem (eFTP) of finding a point in R n , from which the sum-of-distances to a finite number of nonempty, closed and convex sets is minimum. Along with the point that minimizes the sum-of-distances, the eFTP also solves for the points in each convex set at which the minimum sum-of-distances is achieved.
In most real-time applications, either the objective function or the constraints of an optimization problem may change with respect to time, and it leads to an optimal point/set of optimal points at each time instant. Consequently, the optimal points of the problem at each time instant form an optimizer trajectory and hence tracking the optimizer trajectory calls for the need to solve the TV optimization problem. A second-order continuous-time gradient-flow approach is proposed in this work to track the optimizer trajectory of unconstrained and constrained TV convex optimization problems with a strongly convex objective function. First, to track the optimizer trajectory of unconstrained and equality constrained TV optimization problems, finite-time and fixed-time prediction-correction based dynamical system approaches are presented in this work that guarantees the tracking of the optimizer trajectory within a finite time. Later, a projected primal-dual dynamical system based on the prediction-correction technique is proposed to track the optimizer trajectory of an inequality constrained TV convex optimization problem. The trajectories of the proposed projected dynamical system track the optimizer trajectory of the inequality constrained TV optimization problem asymptotically. The uniform asymptotic stability of the proposed dynamical system to the optimizer trajectory is shown via Lyapunov-based analysis. Finally, the TV version of the eFTP is considered to illustrate the applicability of the prediction-correction based projected primal-dual dynamical system proposed in the thesis.
Dec
2022A considerable level of even ordered harmonics are observed in the grid voltage measured at different industrial sites. These even ordered harmonics create triple the grid frequency component in the DC side of Adjustable Speed Drives (ASDs). Due to this, a 150 Hz current component is created in the electrolytic capacitors of ASDs and considerably impacts the heating in the capacitor. Also, Total Harmonic Distortion (THDi) of the current drawn by the ASD is adversely affected due to the even ordered harmonics. To address this problem, a cost-effective approach based on suitable triggering of the SCRs in the bridge of ASD is proposed. This approach may suppress the effect of even ordered harmonics on capacitor ripple current and improve the source current THDi. The proposed strategy is analyzed by simulations under different grid impedance and harmonic combination cases, and validated by a hardware test setup.
Dec
20225G systems have been successfully deployed by hundreds of operators worldwide with many more in the pipeline. Even as 5G systems are getting deployed, 5G standards continue to evolve through 5G Advanced in 3GPP and eventually leading into 6G. Several exciting technologies are on the cusp of adoption in 5G Advanced which will act as the building blocks for defining 6G. In this presentation, we will present a snapshot of how 5G features evolved across 3GPP releases, the expected timeline of 6G standardization and discuss some key technologies that are on the horizon for 6G. Some example technologies that will be discussed include Giga MIMO for enabling new coverage spectrum in upper mid-band, AI/ML for wireless, full duplex radio for macro deployments, intelligent surfaces to help adapt the RF environment to optimize the communication link, and joint communication/sensing. We will discuss results from a combination of real world data from OTA test networks and comprehensive system level simulations. While the technologies discussed show excellent promise towards defining a new generation of cellular communication, there are interesting problems that need the attention of both the academia and the industry to bring these technologies to fruit
Bio:
Dr. Kiran Mukkavilli is a Sr. Director of Engineering in Qualcomm Wireless Research where he currently leads the systems research team for sub-24GHz technologies for 5G Advanced and 6G. He joined Qualcomm in 2003, holds more than 150 granted U.S. patents, and has been involved in system design, standardization, implementation, and commercialization of a variety of technologies at Qualcomm. He was one of the principal architects of MediaFLO, a mobile broadcast solution from Qualcomm, and played a key role in the product development, standardization, and commercialization efforts. He was also the systems design lead responsible for commercialization of the UMTS modem in Qualcomm Snapdragon 800/801 products. In his current role, he is responsible for R&D efforts for PHY/MAC system design, standardization, and prototyping for Sub 24GHz aspects of 5G NR. He successfully led Qualcomm’s effort to set up industry’s first end to end 3GPP Rel 15 spec compliant 5G NR call with the leading infra vendors using prototype UE implementation. Dr. Mukkavilli received his M.S. and Ph.D. in electrical engineering from Rice University and holds a Bachelor of Technology from Indian Institute of Technology, Madras.
Dec
2022Integrated photonic devices in general work on the basis of precise phase control of guided mode(s). Especially, waveguides fabricated in silicon photonics platform suffer from fabrication imperfection related phase errors as well as due to thermal phase fluctuations (thermo-optic coefficient ) Such phase error causes serious performance degradation in resonator devices. On the other hand, devices like microring resonator (MRR), MicroDisc resonator (MDR), DBR based Fabry-Perot resonator (FPR) are very much popular in designing versatile functional photonic integrated circuits (PICs) such as high-speed optical transceiver, quantum/neuromorphic photonic processor chip, microwave photonic processor chip etc. These resonator devices are actively reconfigured via thermo-optic and/or electro-optic effects. It is also observed that typical thermo-optic drift (due to ambient temperature fluctuation and thermal cross-talk) of resonance wavelengths is in the order of 70 which is a big value for high-Q resonators ). Therefore, in a large scale integrated photonic circuit it is really a great challenge to stabilize all the resonator devices independently; global temperature stabilization of entire photonic chip becomes detrimental for their individual reconfigurability (thermo-optic and/or electro-optic).
In this talk, we will discuss first the state-of-the-art techniques for resonance stabilization used in silicon photonic integrated circuits. Afterward, we will present some of our preliminary investigation and scopes for future research in this direction.
Dec
2022Broadcast systems typically use a network of synchronized towers that transmit the same content. Such single-frequency networks (SFNs) are ideal when the same content is to be delivered to the entire subscriber base. Newer standards like ATSC 3.0 and 5G NR support the delivery of regional content, such as targeted news and weather, through local service areas (LSAs). Using orthogonal frequency bands to broadcast this content leads to poor spectral efficiency. However, reusing the same frequency band to deliver distinct local content can lead to co-channel interference (CCI) at the boundaries between the LSAs. This interference can disrupt the broadcast signal and degrade the viewer experience. In this talk, we propose two approaches that can be used to manage the CCI in such reuse-1 deployments. The first approach scales the local content powers near the LSA boundary, while the second introduces a restricted orthogonality of local content at the boundary region. These schemes are compared in terms of SINR and spectral efficiency. Simulation results show that the proposed schemes improve the local content coverage area without compromising the spectral efficiency and the trade-offs involved in their selection.
Dec
2022With the extensive use of online social networks that enable large-scale opinion exchanges, the study of opinion dynamics has gained more importance. In classical opinion dynamics, an agent updates its opinion based on its neighbors on a social graph or the proximity of its opinion with others. Linear dynamics capture graph-based interactions, and bounded confidence dynamics capture pairwise opinion-dependent interactions. Stochastic bounded confidence (SBC) dynamics is a recent framework that generalizes classical dynamics by modeling the inherent stochasticity and noise (errors) in real-world opinion exchanges. The asymptotic behavior of these dynamics has already been studied in the literature. However, the findings do not shed light on their behavior in finite time, which is often of interest in practice.
In this work, we characterize the evolution of opinions over a finite time period. We focus primarily on characterizing the finite time behavior of SBC dynamics of two agents and on a bistar graph. Such dynamics closely model bipartisan democratic societies with two leaders/ parties with large followings. In particular, we derive high-probability bounds for the opinion difference between two agents and demonstrate a strong concentration of opinion difference around zero under the conditions that lead to asymptotic stability of the dynamics. This work is crucial for analyzing general multi-agent dynamics.
Dec
2022Carnatic music (CM) employs a profusion of continuous pitch variation called gamakas in addition to the usual 12 musical notes per octave. However, CM notation is in terms of svaras with little or no gamaka information and, therefore, cannot be synthesized. Previous work on extracting gamaka information in a descriptive transcription for synthesis is fairly recent, and treats the pitch curve as a whole. In this research, we aim to automatically extract a descriptive transcription for CM by separating the pitch curve into its components. Towards this end, we define a constant pitch note (CPN) as a segment of the pitch curve whose pitch is within empirical limits. The pitch-curve is then viewed as consisting of CPNs and transients, which are the segments of the pitch curve outside CPNs. We further define stationary points, or STAs, as the maxima and minima of transients.
A histogram of pitch-values folded to one octave has significant values between the musical notes due to gamakas. By contrast, histograms of only CPN pitch-values show sharp peaks at notes in the raga. We further propose a novel view of a CPN in CM as an upward anchor and/or a downward anchor depending on the direction of adjacent pitch movements. The peaks in the histograms of upward and downward anchors in a raga are detected as anchor-targets. Next, we separate STAs into maxima (max-STAs) and minima (min-STAs). We detect max-STA targets and min-STA targets from the peaks of the respective histograms. The anchor-targets and STA-targets explain not only the notes and gamakas in a CM-raga, but also serve as a reference for component-wise precision measurement. The difference in measured precision, ~20 cents for CPNs and ~60 cents for STAs, suggests that transcription should also be done component-wise.
We further propose the use of anchor-specific STA-targets and obtain them from max-STAs and min-STAs adjacent to each anchor in the raga. We found that treating STAs as being in the state of an anchor or of a transient, is beneficial in quantizing to the anchor-specific targets. We propose state based transcription (SBT) using maximum likelihood sequence estimation. The pitch-value of each CPN or STA is quantized to the target corresponding to its state in the estimated sequence. These quantized pitch-values in semitones, and the timing information of CPNs and STAs, constitute the descriptive transcription. We use cosine interpolation to generate a pitch curve from the descriptive transcription and feed this pitch curve to a five-harmonic synthesizer. In a subjective listening test, expert-ratings of synthesized samples show that the descriptive transcription captures gamakas.
Synthesized audio samples will be played during the talk to support the technical points above. A sample of a re-synthesized audio track mixed with the original source is available at:
on.soundcloud.com/RSx3c
For discernibility, the two sources are occasionally played individually.
Dec
2022Electric Market Energy Trading is gaining more importance now-a-days due to the transformation from vertically integrated to horizontally restructured power systems. This feature enables the participation of renewables to participate in Electric Markets. A ‘Bidding Strategy and Trading Platform’ for renewables will be presented in this talk.
Strategic Bidding is an important task which is much needed for Wind Power Plants(WPP) so as to reduce imbalances between accepted bids and actual output. However, these imbalances do not allow the WPP to participate in Electric Markets due to the penalty for bid deviation. A joint Double Auction Bidding Strategy is proposed and developed for the WPP to coordinate the Pumped Storage Hydro Plant (PSHP) and Demand Response (DR) to reduce the WPP actual output imbalances. A convex Strategic Bidding Model is formulated for WPP, PSP, and DR in both day-ahead energy and ancillary service markets by considering Upward Spinning Reserve (UPSR) and Downward Spinning reserve (DWSR) respectively. In addition, fixed, shift-able, curtailable, and incremental loads are considered.
Recent advancements in Digitization, Automation, and Integration of renewables at the distribution level are the main impetuses that cultivate prosumers' interest in participating in the Peer-To-Peer (P2P) Electricity Trading Platform (ETP). The Distributed System Operator (DSO) at the Electricity Trading Platform (ETP) is usually not involved in validating energy or monetary transactions. In this context, a novel Electricity Trading Framework (ETF) is proposed and implemented using block-chain distributed ledger system. This trading framework eliminates the need for a Central Aggregator (CA) or DSO in the P2P trading process. Smart contracts have been used for collecting bids/offers and monetary settlements. The final work broadly involves integration of optimization, smart contracts, and monetary settlement layers.
Dec
2022The microstructure of thin films of organic semiconductors can significantly affect the whole range of optoelectronic processes, from light absorption to the generation of free charge carriers, their transport as well as recombination. Thus, understanding and being able to tune this microstructure along with being able to measure its role in the energetic disorder is a critical step towards the commercialisation in organic optoelectronic devices such as thin-film transistors and organic solar cells (OSC). In the first part of this presentation, we report research on the development of the thin film microstructure in-situ during vacuum deposition of a-6T and ZnPc, two prototypical materials for OSC, using grazing incidence wide and small angle scattering (GIWAXS/GISAXS).
Dec
2022As a result of densification, the performance of the wireless networks has become highly interference-limited and energy inefficient. To overcome this problem, interference mitigation techniques such as Successive Interference Cancellation (SIC) can be used to decode multiple packets simultaneously at the receiver. In this context, we analyze a SIC-based Slotted Aloha (SIC-SA) Medium Access Control (MAC) protocol for wireless networks. We derive expressions for packets decoding probability and optimal transmission probability of the nodes of the SIC-SA MAC protocol. Our derivation is based on the order statistics of Independent and Identical / non-Identical exponentially distributed received-signal-powers from the nodes under the Rayleigh channel condition. Throughput, delay, and energy efficiency of the SIC-SA MAC protocol have been derived and validated against simulation. The effect of path loss exponent, SINR threshold, and the number of nodes on the performance of SIC-SA have been studied. The performance of SIC-SA in a network of nodes distributed randomly according to the Poisson Point Process has been analyzed. Extension of our analysis to Power Domain Non-Orthogonal Multiple Access (NOMA) has been demonstrated. We also analyzed the impact of imperfect estimation of channel state information and imperfect SIC at the receiver. Results show an improvement in performance metrics of SIC-SA over the traditional Slotted Aloha.
Dec
2022Quantum key distribution (QKD) promises unconditional security based on quantum mechanics principles. Fiber-based and free space-based QKD systems have been implemented, but the chip-based implementation of QKD systems is essential for large-scale deployment of QKD systems. Silicon pho-tonics technology platform is a potential candidate for on-chip QKD because of its advantages, such as being a low-cost CMOS-compatible fabrication process, robust, and compactness. Proof of concept on-chip QKD has already been implemented using polarization encoding and time-bin encoding in the silicon photonics platform. Time-bin encoding is preferred because polarization encoding suffers from birefringence in fiber optics communication systems. One of the essential components in the time-bin encoding QKD system is the delay line interferometer (DLI). The desired long waveguide delay line can be implemented in a compact spiral architecture. However, fabrication imperfection leads to unpredictable delay outcomes. Thermal optic tuning of long waveguide delay line is found to be an inefficient solution for this remedy. Therefore, a plenty of scope is there for the design and demonstration of tunable delay lines by integrating thermo-optically/electro-optically actuated resonant/non-resonant silicon photonic components.In this talk, we will discuss various configurations of on-chip delay line devices and their merits/demerits. We will also present the design, fabrication, and characterization of passive spiral waveguides and their time delay measurements.
Dec
2022Digital watermarking technology is used for document security,
ownership and copyright enforcement, product tracing and counterfeit
deterrence, retail checkout and recycling. Usually a secret data
carrying signal is embedded in a host, such as an image, video or
audio such that it is imperceptible to a human, but can be recovered
easily with a computer. Signal Rich Art is a recent advancement in the
field, in which we embed the signal in visible artistic/texture
patterns which are cognitively imperceptible.
This will be an interactive talk with a lot of pretty pictures. It is
recommended that attendees download the free Digimarc Discover app,
available for both Apple and Android devices from the respective app
stores. The app will be able to decode the information from the images
presented on the slides using the smartphone camera. Digimarc, a
company based in Portland, Oregon, USA, is a world leader in
applications and intellectual property of digital watermarking with
over a thousand patents.
About the speaker:
Ajith Kamath is a graduate of IITM (BTech, EE, 2000) and NCSU (PhD,
EE, 2005). He has worked in R&D at ArrayComm and is currently at
Digimarc
Dec
2022Decentralized cellular networks where infrastructure ownership is distributed between several independent entities have emerged to increase overall network accessibility. In this setting, users can connect to any nearby, untrusted base station with no prior subscription to the operator. This talk will introduce Proof of Service as a method to address the primary challenges of ensuring reliable and trustworthy performance in this setting. Proof of Service redesigns the OCS to implement its escrow and charging functionalities separately using smart contracts and state channels respectively. An integral part of our design is the verification of delivered service to ensure that the user only pays commensurately. To do this, we introduce the notion of two-sided measurements where both the users and the providers independently assess the cellular service delivered. A key challenge is ensuring that these measurements reliably result in agreement. Our preliminary findings show that reconciling measurements from different layers of the cellular stack to arrive at a diverse set of matching observations is challenging but not impossible. Finally, this will be followed by a discussion of our ongoing implementation using Magma. This work is being done in collaboration with Witness Chain.
Bio :
Milind Kumar Vaddiraju is a second year PhD student in the ECE department at UIUC advised by Prof. Pramod Viswanath. He received his B.Tech in Electrical Engineering from IIT Madras in 2020. His current research focuses on decentralizing wireless technology and involves designing and evaluating new architectures that enable the organic growth of a network by individual entities in a trust-free setting. Before pursuing this direction of work, he spent two years at the Indigenous 5G Testbed at IIT Madras. He is a recipient of the ECE Distinguished Research Fellowship, James M. Henderson Fellowship and the Dilip and Sandhya Sarwate Fellowship at UIUC. More details can be found at www.milindkumarv.com
Dec
2022Future Wireless Local Area Networks (WLANs) are expected to cater to dense WLAN use case scenarios such as airports, smart homes, outdoor stadiums, etc., having numerous stations (STAs) deployed with diverse traffic requirements. The IEEE 802.11ax is the current WLAN standard proposed to meet these requirements. In this seminar, we discuss the Static and Dynamic resource allocation methodologies of the Multi-User Orthogonal Frequency Division Multiple Access (MU-OFDMA) feature of the IEEE 802.11ax network. In contrast to Static MU-OFDMA, in the case of Dynamic MU-OFDMA system, resource units (RUs) are allocated dynamically between the schedule and random access (RA) RUs based on the heterogeneous traffic requirements of the stations (STAs). We have modeled the Static MU-OFDMA system using the Fluid limit approach and the Dynamic MU-OFDMA system using the Discrete-Time Markov Chain (DTMC) Methodology. The tradeoff between throughput and RA capability (rate of new users accessing the system) of a static & dynamic system are studied and validated against simulation.
Further, we discuss the impact of the non-saturated traffic models, such as the Poisson and Pareto traffic models, on the Throughput of a WLAN system and have also studied the impact of the Multi-TID-based packet aggregation feature on the Throughput of the IEEE 802.11ax WLAN systems.
Dec
2022
Whispering gallery modes (WGMs) are specific resonances of a wave field that are strongly confined inside a micro-resonator, potentially providing a quality factor of 106 - 109. Such WGM could be used for sensing applications by trapping the particles on to the surface of the micro-resonators. Recently, microbottle resonators (MBR) have attracted much attention since it promises a customizable mode structure while maintaining a favorable Q/V ratio and good confinement. However, the excitation of WGMs in such MBRs is a challenging task and conventional approaches involve a rather cumbersome exercise of exciting these WGMs using a tapered fiber, limiting its use for only laboratory demonstrations.
We propose a novel approach to excite the WGMs in a microbottle resonator using an optical beam possessing orbital angular momentum (OAM). We observe that the helical propagation of the OAM, its azimuthally varying phase, as well as its transverse electric field profile are consistent with that of the WGM in a microbottle resonator. Such an observation is also confirmed by the similarity in the analytical solutions of the field profiles for both the high charge OAM and the WGM. We have also performed FEM/FDTD simulations to simulate the WGM in a microbottle resonator and find that the simulation results are consistent with the analytical results. We have also investigated the purity of the OAM modes that could be generated using a phase only spatial light modulator (SLM). Such results, along with a roadmap for future experiments will be discussed during this talk.
Dec
2022
Multilevel inversion can be obtained using dual-inverter fed open end winding induction motor (OEWIM) drive. Field oriented control (FOC) and predictive current control (PCC) methods for the OEWIM drive with common-mode voltage (CMV) elimination are proposed. Both the proposed methods effectively control the speed of the OEWIM while eliminating the CMV in the dual inverter. Also, an improvised PCC method is proposed to completely obliterate the effect of the dead time in the OEWIM drive. The proposed PCC is formulated using a voltage based objective function instead of a conventional current based one. The candidate switching combinations for the evaluation of the objective function are thoughtfully chosen based on the position of the reference voltage vector to reduce the computational burden drastically, thereby improving the drive’s performance. Also, two improved PCC methods for the operation of the OEWIM drive with a single dc source and a floating capacitor, wherein a sector-based selection of the candidate switching combinations for the evaluation of the objective function are proposed that drastically reduced the computations, thereby facilitating a reduction in sampling time. Both the PCC methods effectively regulate the capacitor voltage, thus generating a symmetrical three-level voltage at the OEWIM terminals.
Later, a weighting factor less cascaded PCC (CPCC) for the OEWIM drive with a single dc source and a floating capacitor bank is presented. Sequential evaluation of the objectives is introduced in this work, where the choice of the sequence of the control objectives is very critical. The second control objective is evaluated only for the two best cases obtained from the evaluation of the first objective. Unlike the generalized sequential predictive control, it is demonstrated that the system will completely collapse if the sequence of the control objectives is reversed. The proposed CPCC also uses a voltage based objective function with a sector-based selection of the candidate switching combinations to reduce the computational burden drastically.
All the methods proposed in the present work and analysed in detail, simulated and are experimentally verified.
Dec
2022Small signal model of PWM converters are derived from averaging method. In the case of resonant converters, the averaging method fails because tank currents and voltages have no dc component. Moreover, the phenomenon of "beat frequency dynamics" is not addressed in the averaging method. Although numerous modeling techniques are presented in literature to model resonant converter; they are either involved math or complex numerical computations. The state trajectories of resonant converters exhibit rotating behaviour and are analysed using Fundamental Harmonic Approximation (FHA).
Nov
2022
CMOS process compatible silicon photonics technology finds a variety of applications in the area of communication, sensing, biomedical, automotive etc. Optical interconnect for data centers has been one of the initial successes of the technology in the field of communications. Silicon photonics transceiver plays a vital role for low cost and low power optical interconnect solutions for catering to ever-increasing data traffic. The key component of the transceiver is a modulator, which encodes electrical data onto an optical carrier. High speed modulators have been implemented popularly by Mach-Zehnder Interferometer (MZI) or Microring resonator (MRR) configurations. Though MRR based modulators are wavelength selective (operable only around resonant wavelengths), they offer smaller footprint and lower energy consumption per bit. Microdisc resonator and ditributed Bragg reflector (DBR) based Fabry-Perot cavity can be good alternatives to MRR. However, they have been relatively less investigated. All these modulators are designed on the basis of thermo-optic and plasma dispersion effects; the former effect is typically used for phase correction/detuning whereas the latter effect is used for high speed modulation. Our literature survey reveals that plenty of scope is there to improve the figures of merit, especially in case of resonator based modulators since they are thermally unstable and exhibit bistability at higher optical power levels.
In this talk, we will discuss various methods to improve the performance of plasma dispersion modulator and the impact of high input optical power in resonator based modulators. We will also present the design, fabrication and characterization of a passive microdisc resonator, along with its thermo-optic behavior and optical bistability at high optical power levels.
Nov
2022
Early detection and diagnosis of certain diseases like cancer is limited by the ability to obtain high resolution images in deeper body cavities and hollow organs. The present imaging techniques for cancer screening such as ultrasound, MRI and CT have limited ability to detect pre-cancerous lesions. Techniques such as PET are expensive and has issues of low availability. Therefore, it is imperative to develop micro-endoscopes with minimal invasiveness to image cavities inaccessible to standard endoscopes using relatively inexpensive but robust optical components. Free space optical systems can yield the necessary resolution to retrieve structural information for early detection. But they cannot be integrated inside a minimally invasive endoscope penetrating through the body to sufficient depth. Therefore, fiber optics needs to be integrated with miniaturized optical system for endoscopy.
Fluorescence imaging endoscopy is widely explored by researchers to visualize the molecular changes of the tissue. Single photon fluorescence imaging currently employed in the commercial endoscopes for high resolution imaging is limited by tissue penetration depth and phototoxicity. Research trends suggest the scope of two photon fluorescence technique for deeper tissue imaging with high resolution. The conventional two photon imaging is limited to a free space laboratory setup and non-medical applications. The scope of double clad fiber coupler for fiber optic fluorescence imaging is being explored in our study.
In this seminar, we discuss the optical design of an endoscope for two photon fluorescence imaging. The design focus on the getting a homogenous resolution in the full field of view by correcting the optical aberrations and thereby improving the image quality. We also discuss the characterization and demonstration of imaging quality with gradient refractive index lenses as the objective lens for endoscopy. As two photon fluorescence imaging involves two wavelengths, the correction of chromatic aberration is a subject of investigation. We present the incorporation of a phase correction element to address the issue of chromatic focal shift.
Nov
2022
Speaker Verification uses speech as a biometric to verify the identity claimed by a speaker. Using speech for identity verification is very convenient and can be used along with other verification techniques to improve the verification performance. Speaker Verification with speaker embeddings obtained from trained Neural Networks is very popular. X-vectors and R-vectors are well known speaker embeddings. They are obtained from Time-Delay Neural Networks and Residual Neural Networks. These networks suffer from slow building of context with layers. So self-attention based networks which can capture global context in all of its layers can be explored for getting better speaker embeddings. Once the embeddings are obtained, Probabilistic Linear Discriminant Analysis (PLDA) is done conventionally to score the embeddings of a trial pair. PLDA is not a Neural Network based technique. So there is a lot of scope to explore Neural Network based architectures for scoring a trial pair.
In this talk, we will discuss the performance of self-attention based architectures, which take global context into consideration in all of its layers for generating speaker embeddings and trial pair scoring.
Nov
2022Low-light image enhancement has been an actively researched area for decades and has produced excellent night-time single-image restoration methods. Years of research has produced state-of-the-art algorithms exploiting techniques ranging from histogram equalization to retinex theory and more recently, convolutional neural networks for low-light enhancement. Despite these advances, the existing literature on low-light enhancement has two major limitations: a) current methods by design are limited to single image enhancement, even though the plight of low-light conditions is equally shared by all optical systems and b) no existing method can restore extremely dark night-time images captured in near-zero lux conditions with a reasonable computational budget. Addressing these problems can endow night-time capabilities to several devices/applications such as smartphones and self-driving cars.In this thesis, we present deep learning architectures for restoring extreme low-light images captured in different modalities, namely -- single/monocular image restoration, stereo image restoration and Light Field restoration. A practical low-light solution must also respect constraints such as limited GPU memory and processing power and should strike a balance between network latency, memory utilization, model parameters, and reconstruction quality. Existing methods, however, only target restoration quality and compromise on speed and memory requirements, raising concerns about their real-world usability.Our models are exceptionally lightweight, remarkably fast, and produce a restoration that is perceptually at par with state-of-the-art computationally intense models.For monocular image restoration, we do most of the processing in the higher scale-spaces, skipping the intermediate-scales wherever possible. Also unique to our model is the potential to process all the scale-spaces concurrently, offering an additional 30% speedup without compromising the restoration quality. Pre-amplification of the dark raw-image is an important step in extreme low-light image enhancement. Most of the existing state of the art methods need GT exposure value to estimate the pre-amplification factor, which is not practically feasible. Thus, we propose an amplifier module that estimates the amplification factor using only the input raw image and can be used off-the-shelf with pre-trained models without any fine-tuning. We show that our model can restore an ultra-high-definition 4K resolution image in just 1sec. on a CPU and at 32fps on a GPU and yet maintain a competitive restoration quality. We also show that our proposed model, without any fine-tuning, generalizes well to subsequent tasks such as object detection. We also propose a light-weight and fast hybrid U-net architecture for low-light stereo image enhancement. In the initial few scale-spaces, we process the left and right features individually because the two features do not align well due to large disparity. At coarser scale-spaces, the disparity between left and right features decreases and the networks receptive field increases. We use this fact to reduce computations by simultaneously processing the left and right features, which also benefits epipole preservation. As our architecture does not use any 3D convolution for fast inference, we use an Epipole-Aware loss module to train our network. This module computes quick and coarse depth estimates to better enforce the epipolar constraints. Extensive benchmarking in terms of visual enhancement and downstream depth estimation shows that our architecture not only performs significantly better but also offers 4-60 x speed-up with 15-100 x lower floating point operations, suitable for real-world applications.To facilitate learning-based techniques for low-light LF imaging, we collected a comprehensive LF dataset of various scenes. For each scene, we captured four LFs, one with near-optimal exposure and ISO settings and the others at different levels of low-light conditions varying from low to extreme low-light settings. We also propose the L3F-wild dataset that contains LF captured late at night with almost zero lux values. Existing single-frame low-light enhancement techniques do not harness the geometric cues present in different LF views and so lead to either blurry or too noisy restorations. Hence we propose deep neural network architectures for LFs. Our networks not only perform visual enhancement of each LF view but also preserve the epipolar geometry across views. We achieve this by extracting both global and view-specific features and later appropriately fusing them using our RNN-inspired feedforward network. Our LF networks can also be used for low-light enhancement of single-frame images, despite they being engineered for LF data. We do so by proposing a transformation to convert any single-frame DSLR image into a pseudo-LF. This allows the same architecture to be used for both LF and single-image low-light enhancement. With all the above advantages intact, our latest LF based neural network offers considerable speed-up with a significantly lower memory footprint.
Nov
2022Angle sensors that are thin and insensitive to dust, oil, and moisture, are highly in demand in industry, particularly for automotive, robotic, and aerospace applications. Although numerous angle sensors based on various measurement mechanisms have been reported, most of them have disadvantages such as bulkiness, needs complex or specialized manufacturing processes, sensitivity to misalignments, and noticeable power consumption. In this talk, a new, thin non-contact angle sensor based on the variable reluctance technique will be presented. The proposed sensor has a disc-shaped eccentric rotor and a stator with two printed circuit boards with circular coils. These parts are easy to manufacture using established but less expensive processes. The rotor and stator are positioned such that the inductances of the coils change as the angle changes. The inductance values are processed using a simple algorithm to obtain the sensing angle. For the stator, a dual coil structure is employed which ensures negligible sensitivity to the axial movement of the rotor. The sensor was designed with the help of finite element analysis. A prototype of the sensor has been built and tested. Static and dynamic performance tests were performed on the prototype. The resolution and worst-case linearity error of the prototype were found to be 0.06 o and 0.7% respectively. The features and performance factors of the sensor are apt for industrial and automotive applications. The details will be presented in the talk.
Nov
2022Cyber-Physical Systems (CPS) for Modern Digital Power Networks are important from the point of view of false data attacks or manipulations on digital devices like smart meters, phasor measurement units, intelligent electronic devices, etc. The study presents cyber-physical modeling of a power system from an attack detection and mitigation point of view. The research deals with generating attack vectors to spoof the bad data detection strategies and detect them using machine learning techniques. In this work, we present three problems relevant to the smart grids, which are the i) attack vector generation problem (data attack), and ii) attack detection (protection) iii) Placement of the detection algorithm on the network. This first and second part provides a novel technique to generate and detect data integrity attacks in smart grids. It also gives an optimization algorithm for generating FDIA against state estimation algorithms present at the control center. The formulation for generating AC state estimation attack with full information and limited information along with DC state estimation attacks is given. It further proposes a combining technique for the voting based ensemble learning technique (MVCC) to detect FDIA in smartgrids. The model is then tested on an IEEE 24 bus system and 39 bus new England system by generating false data injection attacks and detecting them. The detection strategy is compared with most of the existing state of the art Machine learning algorithms, ensemble algorithms and conventional weighted least square algorithm and is found to have a better performance. Finally in the third part an optimal position of a centralized controller is found, for a smart grid network to place the algorithm. The model takes into consideration the delay associated with each communication line along with data congestion and node degree. This model is tested with IEEE 24 bus system, and a comparison is made based on different locations of the controllers. We identified node availability, and their respective power flows during communication line contingencies.
Nov
2022Cyber-Physical Systems (CPS) for Modern Digital Power Networks are important from the point of view of false data attacks or manipulations on digital devices like smart meters, phasor measurement units, intelligent electronic devices, etc. The study presents cyber-physical modeling of a power system from an attack detection and mitigation point of view. The research deals with generating attack vectors to spoof the bad data detection strategies and detect them using machine learning techniques. In this work, we present three problems relevant to the smart grids, which are the i) attack vector generation problem (data attack), and ii) attack detection (protection) iii) Placement of the detection algorithm on the network. This first and second part provides a novel technique to generate and detect data integrity attacks in smart grids. It also gives an optimization algorithm for generating FDIA against state estimation algorithms present at the control center. The formulation for generating AC state estimation attack with full information and limited information along with DC state estimation attacks is given. It further proposes a combining technique for the voting based ensemble learning technique (MVCC) to detect FDIA in smartgrids. The model is then tested on an IEEE 24 bus system and 39 bus new England system by generating false data injection attacks and detecting them. The detection strategy is compared with most of the existing state of the art Machine learning algorithms, ensemble algorithms and conventional weighted least square algorithm and is found to have a better performance. Finally in the third part an optimal position of a centralized controller is found, for a smart grid network to place the algorithm. The model takes into consideration the delay associated with each communication line along with data congestion and node degree. This model is tested with IEEE 24 bus system, and a comparison is made based on different locations of the controllers. We identified node availability, and their respective power flows during communication line contingencies.
Nov
2022
The thrust to integrate more features on a single die has made system-on-chips (SoCs) increasingly complex. These SoCs routinely deploy several data converters together with digital backend to realize single chip solutions for both established and upcoming applications e.g. Automotive Radar, LiDAR, and 5G. Essentially, SoCs are the new-era PCBs integrating analog/mixed-signal IPs with complex digital signal processing engines.
Omni Design Technologies (Omni), Inc. is a developer of disruptive analog/mixed-signal IP cores with the mission to provide a wide range of low-power, high-performance embedded circuits for system-on-chip (SoC) development. In this presentation, I will first introduce Omni followed by a brief overview of our IP portfolio and targeted application space e.g. 5G, LiDAR among others. I will then walk you through our LiDAR solution followed by salient details and measurements results of our 10-b, 2.5GS/s ADC, and 12-b, 7GS/s DAC, both designed in advanced FinFET processes. I will conclude with brief descriptions of our propriety SWIFT™ technology and notch-less time interleaving background calibration.
Speaker Bio:
Dr. Vaibhav Tripathi received PhD and MS degrees in Electrical Engineering from Stanford University and B. Tech in Electrical Engineering from Indian Institute of Technology (IIT), Kanpur. He has been with Omni Design since 2018, where he is now the Vice President of Technology. In 2010, he was with Texas Instruments, Dallas, working on high-speed comparators. From 2014-2016, he worked in the high-speed data converters group at Maxim Integrated, where he developed a low-power 14-b, 250 MS/s pipelined-SAR ADC. From 2016-2018, he was with Samsung Semiconductor, where he co-led the development of 10-b, 2GS/s time-interleaved SAR ADC. In addition, he was also involved in several RF projects such as low-power BLE receiver for IoTs and MM-Wave phase shifters for 5G. Dr. Tripathi is a recipient of the Stanford Graduate Fellowship and Analog Devices Outstanding Student Designer Award 2013, and his research interests include high-speed data converters and RF/MM-Wave frontends.
About:
Omni Design Technologies, Inc. is a developer of disruptive, ultra-low power semiconductor embedded circuits (IP cores), including ultra-low power analog circuits, highly-efficient interface circuits and connected sensors. Our patented and proprietary technology offers solutions that use dramatically lower power and provide superior performance, architected from the ground up to take advantage of deep sub-micron CMOS processes. Our IP offerings target SoCs that address a wide range of application areas including IoT, test and measurement, high speed interfaces, communications, medical imaging, and sensor hubs. Omni Design Technologies, Inc. is a privately held company with offices in Milpitas, CA, Fort Collins, CO, Boston, MA and Bangalore, India. Omni Design was founded in 2015 by a team of semiconductor industry veterans, technologists, and experienced entrepreneurs with a successful track record of delivering high performance analog solutions that advance the state of the art.
Nov
2022Single photon sources are crucial for modern quantum key distribution systems. This work investigates single-photon generation from semiconductor quantum dots using a semi-classical approach. Eigenmode analysis of typical pyramidal quantum dots is performed using the Finite Element Method (FEM). FEM discretization of the simulation domain, development of a weak form of time-independent Schrodinger wave equation, derivation of element and global matrices, Eigenmode solution, and mode visualization are described in detail. The effect of strain, wetting layer, and physical dimensions on the emission wavelength are studied, and empirical relations are obtained. The quantum dot simulation can further be integrated with photonic microcavity simulation and the second-order coherence function of a single photon source with a semiconductor quantum dot in photonic microcavity may be estimated. The talk concludes with a discussion of the current research status and the future directions to take.
Nov
2022
The flux density variation pattern (FDVP) in various sections of a trapezoidal back-EMF, brushless DC (BLDC) machine stator core is non-sinusoidal. In fact, it is trapezoidal in nature. Due to this reason, core loss estimation becomes a challenging step in the process of designing these machines. Standard finite element analysis (FEA) based machine design tools only take material data sheet values as input and use standard frequency domain (FD) models to estimate loss. The underlying problem in this process is that the loss data provided in material data sheets are for pure sinusoidal FDVP. Available methods that address this problem are computationally intensive and consume a fairly long time. This work proposes a simple theoretical core loss estimation method for these machines.
An FD loss expression for trapezoidal FDVP, has been derived from more general time domain models, in this work. The FDVPs in various sections of a BLDC machine stator have been studied and found to be related to the machine design parameters. These relationships have been verified by 2D time stepped FEA. Based on these relationships and the derived FD model, core loss expression for no-load has been proposed.
Various designs have been studied to understand the effects of loading on the FDVP and core loss. Then suitable modifications to loss expressions have been proposed to incorporate these loading effects. Modified core loss expressions have been verified by comparing their loss estimates with numerically calculated losses from time stepped FEA generated FDVPs.
Experimental studies have been conducted on a toroidal M45 steel core sample to verify the FD loss calculation model that has been used to derive the loss expressions. The final core loss expressions have been validated by experimentally measuring core losses of a 5 kW, 10000 rpm BLDC motor.
Nov
2022It is anticipated that a hundredfold increase in data rate is required to satisfy the demands of next-gen mobile networks. One method that will aid in increasing the capacity is increasing the bandwidth. Frequencies between 30 - 300 GHz, known as the Millimeter-wave frequencies, offer several hundreds of MHz bandwidths. However, the migration to mm-wave frequencies affects the power at the receiver and thereby limits the coverage area of the base stations. Thus, the streamlined and effective power amplification is the key to achieving cost-effective and definitive beamforming.
Nov
2022In this talk, we consider the problem of secure packet routing at the maximum achievable rate in multi-hop Quantum Key Distribution (QKD) networks. We consider a practical setup where a QKD protocol randomly generates symmetric private key pairs for secure communication over each link in a network. The quantum key generation process is modeled using a stochastic counting process. Packets are first encrypted with the available quantum keys and then transmitted on a point-to-point basis over the links. A fundamental problem in this setting is the design of a secure and capacity-achieving routing policy that takes into account the time-varying availability of the encryption keys and finite link capacities. To address this problem, we propose a new secure throughput-optimal policy called Tandem Queue Decomposition (TQD). The TQD policy integrates the QKD process with the state-of-the-art Universal Max Weight routing policy. We show that the TQD policy solves the problem of secure and efficient packet routing for a broad class of traffic, including unicast, broadcast, multicast, and anycast. In brief, the TQD policy works by reducing the problem to the generalized network flow problem without the key availability constraints over a transformed network. The proposed policy is then proven stable by using a Lyapunov drift argument on the virtual queues.We note that the previous setting makes a rather strong assumption that the nodes can be fully trusted and the adversary can eavesdrop on the links only. To address this, in our ongoing work, we investigate the problem of secure throughput-optimal routing when the nodes can only be partially trusted with a pre-specified set of beliefs. Finally, we demonstrate the competitiveness of the TQD policy over other existing algorithms by numerically comparing them on a simulator built on top of the state-of-the-art OMNeT++ network simulator platform.
Nov
2022The optical phase profile of light is used in many diverse fields like optical metrology, optical topography, imaging applications, etc. While the transverse phase profile of a beam contains important information, it cannot be detected directly using devices like a camera. Interference is used to convert the hidden phase information of a beam to visible intensity information. The most widely used interference method involves the interference of the unknown object beam with a known tilted reference beam. There are several algorithms that retrieve phase information from such a single off-axis interference pattern. The choice of an algorithm for Optical Phase Retrieval (OPR) depends on the type of sample and noise present, which may be different for various applications. In order to choose the right algorithm, it, therefore, becomes important to understand how each algorithm works in different circumstances. In seminar 1, three main algorithms were introduced in detail, which are the Fourier Transform (FT) method, the continuous wavelet transform (CWT) method, and the Hilbert Transform (HT) method. The mathematical background of these algorithms were introduced, and a few initial experiment results were presented. In this talk, the different aspects of choosing one method are explained. Factors such as the tilt angle, visibility of fringes, noise, etc. are considered. The main parameters that will change the performance of the methods are the spatial frequency content of the object beam, the fringe frequency introduced by the tilt angle between the beams and noise. An integral part of the OPR method is the phase unwrapping step, which also plays a vital role in the quality of the retrieved phase profile. The performance of the combination of the OPR method and standard phase unwrapping methods is evaluated using average RMS error and structural similarity index measure (SSIM). Experimental results of quantitative phase imaging (QPI) of human blood samples to study the cell structure will be presented, and the performance of the methods will be discussed. We have noticed that the methods can give good-quality retrieved phase profiles for lower spatial frequency applications. However, when the spatial frequency is higher, the CWT method has some disadvantages over the other two methods. The HT method is much more prone to noise, but works with higher spatial frequency components compared to the CWT method. The FT method is easy to implement and gives good results, provided the interference patterns spectral requirements are satisfied. The talk will end with a detailed summary of how to choose the best algorithm, as well as discussions on future work.
Nov
2022Though graphene exhibits high charge carrier mobility , practically synthesized large-area graphene films suffer from poor conductivity. This is due to the poly-crystalline nature and the presence of other defects, and low intrinsic carrier density when operated near the charge neutrality point (CNP). The application of graphene as the transparent conducting electrode (TCE) in optoelectronic devices requires that it possess low sheet resistance comparable to that of the existing transparent conducting oxides (TCO) such as fluorine/indium doped tin oxides. This can be addressed by using multi-layer films or by doping graphene films. Here we investigate the surface doping of graphene films by organic small molecule interactions and its dynamics by exploring the inhomogeneity due to varying interactions at the adsorption site. Experimental evidence of defect-specific doping effects of organic dopant on graphene with artificially induced controlled defects was recorded. Density functional theory (DFT) based calculations were performed to validate these observations qualitatively. We show that the degree of doping is governed by the nature of graphene film at the localized dopant-interaction site. The relative change in resistance scales proportionally to the initial resistance (or defect density) of the graphene film. This inhomogeneity in graphene doping is adopted to describe the doping-induced discrepancies in the transfer characteristics of graphene channel transistors when operated around CNP. A two-dimensional Weierstrass transform of an ideal Id-Vg characteristics - based on the assumption that the defect distribution to be perfectly random - is hypothesized to capture the deviations from ideal behavior when operating around CNP. This is verified with the reported experimental data along with DFT calculations.Unconventional doping based on polymer ferroelectrics was explored using advanced electrostatic force probe microscopic (EFM) techniques. Ferroelectric doping was realised with polyvinylidene fluoride-trifluoro ethylene (PVDF-TrFE) nano-thin films (about 40 nm - enabling the low voltage poling desired for this study). The ferroelectric characterization of PVDF-TrFE using dynamic contact EFM and the induced doping effect in graphene using Kelvin probe force microscope (KPFM) provides a clear insight into the ferroelectric-field dependent doping and parasitics triboelectric surface charge induced doping of graphene. Electrostatic doping-induced Fermi-level tuning is used to map the nature of electronic states in reduced graphene oxide thin films. This was enabled by implementing a modified KPFM setup with in-situ gating-based time-resolved measurements. A hybrid model based on DFT and the experimental analysis was employed to probe the nature of states around CNP in these materials. The reflection of these electrostatic effects as parasitic in the electrical performance at the device level was explored. The experimental devices were fabricated with graphene/poly(3-hexyl thiophene): phenyl [6,6] C61 butyric acid methyl ester bulk heterojunction blend/aluminum diode structure and compared with standard reference devices made with indium tin oxide/poly (3,4-ethylene dioxythiophene) polystyrene sulfonate in place of graphene. On comparison, the graphene devices showed larger reverse bias currents. This was attributed to the intra-device level self-gating effects of the graphene electrodes by the counter cathode under normal operations. An analytical model was developed to capture this effect and very verified along with numerical TCAD simulations. In the last part of the thesis, an in-situ measurement setup to study the dynamics of small molecular doping of graphene was developed. The choice of Si-graphene heterojunction solar cell for this study enabled the investigation of dynamic doping of these devices due to the availability of exposed graphene surfaces in this device architecture. The measurements of the fabricated heterojunction solar cells with this setup enabled a facile route to optimize the thickness of dopant (4.1 ± 0.2 nm) to achieve peak-efficiency operations.
Nov
2022Hypothesis testing under controlled sampling has received significant interest recently. One such setting with controlled sampling is the multi-armed bandit setting. In this work, we consider sequential multi-hypothesis testing in a multi-armed bandit setting involving K arms. We consider the case where each arms signal follows a distribution from a vector exponential family. The actual parameters of the arms are unknown to the decision maker. The decision maker incurs a delay cost for delay until a decision and a switching cost whenever the decision maker switches from one arm to another. Thus we have a sequential decision making problem where the decision maker gets only a limited view of the true state of nature at each stage, but can control his view by choosing the arm to observe at each stage.The goal of this work is to develop sequential policies to identify the true hypothesis as quickly as possible while ensuring that probability of false detection is below a constraint. We focus on problems where there is definite structure associated with the parameter set of the arms. This work is divided into two parts: (a) an odd arm identification problem where, as the name suggests, one among the set of K arms has a distribution different from that of the rest of the arms, and (b) a general multi-hypothesis testing setting of which a wide range of problems including the odd arm identification, best arm identification, multiple anomaly detection and partition identification problems are special cases.In each of these cases, an information-theoretic lower bound on the total cost (expected time for a reliable decision plus total switching cost) is first identified, and a variation on a sequential policy based on the generalised likelihood ratio statistic is then studied. Due to the vector exponential family assumption, the signal processing at each stage is simple; the associated conjugate prior distribution on the unknown model parameters enables easy updates of the posterior distribution. The proposed policies, with a suitable threshold for stopping, are shown to satisfy the given constraint on the probability of false detection. The policies are also shown to be asymptotically optimal in terms of the total cost among all policies that satisfy the constraint on the probability of false detection.
Oct
2022Recently, there has been an increasing focus on control techniques for distributed energy resources (DERs) that emulate the behavior of a synchronous generator (SG). This type of control is known as synchronverter (SV) or virtual synchronous generator (VSG), which has the advantage of better stability. This research involves different improvements to the SV control of DERs under various conditions and configurations, which are briefly given as follows and the same will be presented in the talk. Initially, a simple gradient descent-based pre-synchronization control for SV scheme is proposed, which does not disturb local load connection to DER during the synchronization. The parameters of the proposed method are designed using steady state and transient response analysis, as compared to trial and error parameters tuning in previous virtual current-based methods. Experimental validations are presented for the proposed pre-synchronization control under all initial conditions, transient response analysis, local load changes, and grid integration.
Oct
2022Silicon on Insulator(SOI) has been popular for high-performance designs and is preferred over bulk CMOS technology in certain applications. The presence of buried oxide(BOX) in SOI allows for lower junction capacitance and suppresses the short channel effects making it an attractive option in scaled technologies. However, the lower thermal conductivity of the BOX inhibits heat dissipation to the substrate. This results in heat being confined to the channel region leading to a substantial rise in device temperature, in turn degrading the performance of the device. BSIMSOI models this self-heating using a single-T-node model in SPICE. However, the model assumes that the heat is confined to the same device while in reality, most of the heat flows to the neighbouring devices via metal interconnect. Therefore, it is essential to model the heat exchange between devices at the circuit level.
Oct
2022High demand for reliability, low latency, seamless connectivity lead to the 5G era. 5G NR has gained importance in academic and industrial research communities in recent times as its millimetre wave (mmWave) band operations offer a promising alternative for next-generation wireless communications. For the initial connection establishment of User Equipments (UEs) with the base station (gNB), 5G cellular networks employ a random access procedure where UEs transmit a randomly choosen preamble for acheiving uplink synchronization and subsequent uplink grant. It is of utmost importance that we evaluate the factors that aid and those that limit the random access procedure in achieving its envisaged goal. In our work, we develop an analytical model for random access procedure for connection establishment in 5G when beamforming is employed. The random access medium is modelled as a multi-channel slotted Aloha system. We use an equilibrium point analysis framework to derive the performance metrics namely, the rate at which UEs can lock on to a gNB and the average and variance of the time taken by a new UE to establish connection with the base station. We analyse the impact of the retransmission limit and the number of available preambles for random access on the performance of connection establishment of a UE. Our analysis brings out the importance of the choice of base station configurations with respect to the user demographics of a 5G NR cell.
However, the susceptibility of mmWave signals to severe path loss and shadowing requires the use of highly directional antennas. Since the narrow beams are vulnerable to blockages, the interference behaviour becomes the key factor in building the network. Hence, we propose a blockage model by considering the distribution of buildings in the cell as a Poisson point process. We analytically derive the blockage probability using the queuing theory. The effective success probability of the random access procedure has been presented through extensive simulations.
Oct
20225G-NR offers a full range of mobile broadband services, including massive connectivity from human-held intelligent devices to sensors and machines. More importantly, it can support critical machine-based communications with instant action and ultra-high reliability. Several physical and virtual network interfaces, such as FAPI, ORAN, and PCI-E, play a prominent role in achieving high throughput for 5G base stations. In our work, we concentrate primarily on the architectural implementation of one such prime interface, FAPI, abbreviated as the Functional Application Programming Interface. Broadly, FAPI is an interface that helps communicate between the MAC layer of the Layer 2 protocol stack and the PHY Layer in wireless systems. In contrast to the FAPI in 4G LTE, FAPI in 5G NR offers advanced error indication messaging, reliability and secured communication for messages and structure exchange procedures between L2 and L1 at higher throughput. In this work, we propose an FPGA-based implementation methodology for FAPI that manages the physical layer and allows synchronised data flow between MAC and PHY. We look at the architectural implementation of FAPI in the 5G testbed.
Oct
2022Electric vehicles (EVs) are an excellent alternative to gasoline-powered vehicles when it comes to sustainability and the reduction of environmental pollution. Inductive power transfer (IPT) based charging has been gaining popularity over plug-in EVs as it helps to have reduced battery capacity as well as improved convenience and safety while charging. IPT in EVs could be either static or dynamic. In the research work conducted and presented in the thesis, some of the important sensing challenges faced in the design of the IPT system for the charging of EVs are addressed. The efficiency of the IPT reduces drastically as the misalignment between the primary and secondary pads increases. To help to improve the misalignment and hence the efficiency of the wireless charging, a magnetoresistive (MR) sensor-based system is designed and developed. In the design of the IPT highway, one of the challenges faced is the detection of EV, to enable charging, as it travels along the highway. A sensor system was developed to detect the EV approaching the primary (ground) pad laid on the highway, so as to enable charging when it detects sufficient coupling between the primary and secondary (vehicle) pads. Another challenge faced in commercial IPT systems is that the primary pad should be able to interoperate with different types of secondary pads. For this, an MR sensor-based sensing technique was developed to identify the type of secondary pad so that the primary is energized in the correct mode. Following this a suitable sensing technique was developed for vehicle-to-grid (V2G) applications, to reliably detect the ground pad and sense its configuration before the actual V2G power transfer is initiated. The sensing techniques designed to address these challenges were optimized for the number of sensors and the sensor positions, using numerical analysis and finite element analysis. This was followed by detailed experimental studies to validate the functionality of the sensor systems. The proposed sensor systems need very less additional hardware which in turn reduces the overall system cost and its complexity. It can be easily integrated into new as well as existing primary/secondary pads. As the primary coil isn’t fully powered during the sensing phase, it saves power as well as limits human exposure to unwanted magnetic fields. As these systems are based on magnetic sensing, the performance is less likely to be affected due to dust, snow, oil, etc. Details of the work done and conclusions made will be presented in the talk.
Oct
2022Recent research has established that the local Lipschitz constant of a neural network directly influences its adversarial robustness. We exploit this relationship to construct an ensemble of neural networks which not only improves the accuracy, but also provides increased adversarial robustness. The local Lipschitz constants for two different ensemble methods - bagging and stacking - are derived and the architectures best suited for ensuring adversarial robustness are deduced. The proposed ensemble architectures are tested on MNIST and CIFAR-10 datasets in the presence of white-box attacks, FGSM and PGD. The proposed architecture is found to be more robust than a) a single network and b) traditional ensemble methods.
Oct
2022In this dissertation, we study two problems in the field of nonparametric hypothesis testing – anomaly detection and clustering. A finite collection of S data streams is observed. Each data stream is an i.i.d sequence drawn from arbitrary, unknown probability distribution. In the anomaly detection problem, a small subset of the streams are drawn from a distribution q while the remaining streams are drawn from a distribution and our task is to identify the anomalous streams (q) using as few samples as possible while ensuring a certain accuracy. In the more general clustering problem, the distributions that generate the data streams themselves form clusters based on their proximity to each other based on some underlying distance metrics. Here, our objective is to design sequential tests to segregate the data streams into clusters using as few samples as possible while ensuring a certain accuracy. In our work, we propose universal sequential tests for both anomaly detection and clustering. The tests are universally exponentially consistent and stop in finite time almost surely. Moreover, we also characterize the rate of growth of the stopping time in the limit of vanishing error probability. Computer simulations show that the proposed tests outperform the existing fixed sample size tests for these problems in all cases.
Oct
2022Information flow or transfer in a dynamical network refers to the flow of information from a source entity to a receiver. It quantifies the contribution of an event towards the occurrence of another event, thereby enabling us to determine which event will have greater accuracy in predicting the outcome of a different event. The theory of information transfer has various applications in the field of biological science such as studying the flow of information from one neuron to another, understanding the effect of protein concentrations in gene
Oct
2022The switched reluctance motor (SRM) is being actively explored as a viable alternative for permanent magnet-based motors for the use in automotive applications, by virtue of its favourable characteristics such as permanent magnet-free construction, high fault tolerance, ruggedness, low manufacturing cost, etc. However, the very principle of operation of SRM introduces two serious drawbacks in its performance, which are torque ripple and acoustic noise. Torque Sharing Function (TSF) based control is a well-established method for the reduction of torque ripple in Switched Reluctance Motor (SRM) drives. There have not been much investigations carried out in the literature on the noise benefits of TSF based control and on the impact of the commutation angles of TSF definition on the acoustic noise in SRM. In this talk, the advantages of TSF based control in noise reduction of SRM are presented, in detail. A new TSF profile, which is characterized by a slower variation of the radial force and hence reduced vibration on the stator, is proposed. Further, the effects of variation of turn-on and overlap angles of TSF definition on the vibration and torque ripple in a closed loop control are analysed, using
Oct
2022Computers revolutionized our society in ways that could not even be imagined at their advent in the mid-20th century. Quantum computers hold similar promises. One of the key challenges is the development of single-photon emitters (SPE) that can be useful for a variety of quantum-enabled technologies. An extensive search is ongoing for such single-photon emitter sources, and none have yet shown all the desired properties for technological applications. We propose to investigate SPE that is needed for the emerging quantum-enabled technological world. Advances in the field of nanotechnology, lithography, and stamping techniques have enabled the fabrication of various devices with 2D Transition Metal Dichalcogenides (TMDCs) materials with tunable physical properties. The ability to stack monolayers of dissimilar TMDCs provides great flexibility in creating designer structures for a given application. There are recent reports establishing a strong, anti-bunched, stable and narrow linewidth emission from these materials. We aim to study the prospects of SPE in these materials.
Oct
2022
The rapid increase in demand for electric power has led to phenomenal growth in power generation and transmission.This has warranted the development of compact, reliable, and cost-effective insulation systems. Developing new insulating materials and structures, evaluating their long-term performance, and suitable diagnostic testing methods become the responsibilities of high voltage engineering. Appraising the teachers from engineering colleges, researchers and industry personnel on these recent developments has become essential. Further, to boost the required research activity, generate a pool of high-voltage engineering researchers and meet the demands of the industries and academia, IIT Madras High voltage group is scheduling a two-day outreach program from 8-9 October 2022.
Contact:
Prof. R. Sarathi
Email: sarathi@ee.iitm.ac.in
Tel: 044 2257 4436/5424
Download Brochure
Oct
2022Synthetic ester-based fluids are emerging as potential candidate for insulation in power transformers. The uniform dispersion of silica nanofiller in to the ester based fluid could further enhance its electrical properties. Identification of Optimized quantity of nanofiller for dispersion and Impact ofsurfactant for enhanced electrical properties are discussed in the presentation. In addition, the rheological properties and fluorescent properties of the ester-based nanofluid are elaborated in the presentation.
Sep
2022Recent advances in 3D fabrication have allowed the development of 3D memory over the logic die. The 3D memory presents itself as a viable solution to the memory wall problem. The 3D memory has stacked DRAM layers connected with Through Silicon Vias (TSVs), responsible for the very high bandwidth. Much effort has been made recently to improve the performance and power of the NMP architectures. Memory Centric Networks (MCNs) are advanced memory architectures that use NMP architectures. MCNs are multiple stacks of 3D memory units equipped with requirement-based processing cores, allowing numerous threads to execute concurrently.
Sep
2022Satellite images are typically subject to multiple photometric distortions. Different factors affect the quality of satellite images, including changes in atmosphere, surface reflectance, sun illumination, viewing geometries etc., resulting in multiple photometric distortions in the satellite images. In supervised networks, the availability of paired datasets is a strong assumption. Consequently, many unsupervised algorithms have been proposed to address this problem. These methods synthetically generate a large dataset of degraded images using image formation models. A neural network is then trained with an adversarial loss to discriminate between images from distorted and clean domains. However, these methods yield suboptimal performance when tested on real images that do not necessarily conform to the generation mechanism. Also, they require a large amount of training data and are rendered unsuitable when only few images are available. To address these important issues, we propose a distortion disentanglement and knowledge distillation framework for satellite image restoration. Our algorithm requires only two images: the distorted satellite image to be restored and a reference image with similar semantics. Ablation studies show that our proposed mechanism successfully disentangles distortion. Exhaustive experiments on different timestamps of Google-Earth images and on publicly available datasets, LEVIR-CD and SZTAKI, show that our proposed mechanism can tackle a variety of distortions and outperforms existing state-of-the-art restoration methods visually as well as on quantitative metrics.
Sep
2022The advent of Erbium Doped Fiber Amplifiers (EDFAs) revolutionized optical communication links with its large bandwidth, relatively large gain at low noise figures and crosstalk, polarization insensitivity, high conversion efficiency, and fundamentally - ability to amplify in the optical domain. However, Amplified Spontaneous Emission (ASE) noise accumulation limits the link reach, necessitating the need for regeneration. Although cascaded optical all-repeater/all-regenerator are well researched, the analysis lacks demonstration of unconditional superiority of the latter over the former in terms of BER and the monetization of BER advantage into power saved or extra optical reach for the all-regenerator link over the all-repeater link. Motivated by the above, we propose a simple yet general framework for the performance limit estimation of repeated/regenerated links and demonstrate the same for typical fiber optical and free-space optical links. An abstract model is introduced first, which is then employed to analyze practical amplified(repeated)/regenerated cascaded optical links. Certain approximations to simplify the analysis and computations are also discussed, which should help reduce computational complexity and enable fast decision-making in dynamically reconfigurable optical networks. Further, a multihop Free Space Optical (FSO) link is analyzed using the same and compared against a commercial optical link simulation tool. The same can also be extended to analyzing other multihop links like underwater and satellite optical links.
Sep
2022
Rhythm is a fundamental dimension of music. Music Information Retrieval (MIR) on rhythm has been primarily focused on Western music (WM). The rhythm analysis of Indian art music Rhythm analysis in IAM has received attention in recent years. Unlike WM, the rhythm accompaniments in IAM are tonic based and also extempore or improvisational similar to that melody in IAM. IAM has an advanced rhythmic framework based on the tala (rhythm cycle), which is quite different from the western notions of rhythm. With a complex and sophisticated rhythmic structure and framework, IAM poses a significant challenge to state-of-the-art rhythm modeling tools MIR models developed for Western music cannot be directly applied to IA. While MIR-based analyses have received some attention for the analysis of mridangam (a percussion instrument) used in Carnatic music (CM), MIR-based analysis of tabla in Hindustan music (HM) has hardly received any attention. In this work, we aspire to primarily analyze the MIR rhythm aspects of tabla in HM and also compare the same with that of mridangam and other percussion instruments used in CM.
The proposed work consists of two parts. The first part addresses rhythm-related selective analysis using a large amount of music data combined with domain knowledge and expertise. The initial analysis deals with the rhythm-based structural segmentation and labeling of tabla solo audios. Later sections deal with the rhythm analyses by comparing and contrasting the percussion technique of tabla and mridangam in both accompaniment and solo performances. The second part analyzes various results from the first part from the human perception and cognition point of view. This part of the research intends to study the effect of various rhythmic and melodic variations of Indian art music on the human brain.
The focus of the talk will be first to review the research efforts on the analysis of percussion instruments in music and then focus our attention on efforts in Indian art music. Finally, we present our preliminary studies on tabla gharana recognition from solo tabla recordings. Teaching practices and performances of tabla are based on stylistic schools called gharana-s. Gharana-s are characterized by their unique playing technique, finger posture, improvisations, and compositional patterns (signature patterns). Recognizing the gharana information from a tabla performance is hence helpful in characterizing the performance. We present two different approaches to the task. The first approach consists of a transcription of tabla audio followed by a signature pattern search. The second approach consists of deep learning models that combine convolutional neural networks (CNN) and long short-term memory (LSTM) networks. The LSTM networks are trained to classify the gharana-s by processing the sequence of extracted features from CNNs.
Sep
2022
Permanent magnet synchronous motors (PMSMs) are widely used in electric vehicles (EVs) due to their superior performance than induction machines. However, the operating speed range of these machines is restricted due to the limitations in the maximum battery voltage available in the vehicle. In this work, a dual three-phase PMSM with zero degree winding displacement (DTP0-PMSM) with six-step operation is proposed to enhance the power rating and operating speed range compared to conventional 3-phase PMSMs. Dual three-phase PMSM also ensures zero circulating current even during six-step operation, reducing the copper loss compared to commonly used split-phase PMSMs. Specific applications like military vehicles, medium/heavy-duty trucks and off-road EVs demand high starting torque, high overload capability and wide constant power operating region. A dual 3-phase PMSM with unequal turns ratio (uneq0-PMSM) is also proposed to satisfy such special requirements with minimal overrating of the converter DC bus. An added advantage of uneq0-PMSM is that the shape of torque-speed characteristics can be varied during the design by changing the winding split ratio to perfectly suit the load requirement. The proposed concepts were experimentally validated on a custom-designed and fabricated 3kW IPMSM prototype.
Methods for improving PMSM starting while employing sensorless vector control are also proposed in this work. In the first part of the work, a quick and smooth changeover of PMSM control from open-loop (I-f) starting to closed-loop sensorless vector control is achieved using a novel pulse-off starting method. In the second part, a torque controller is proposed to vary the frequency slope dynamically and consequently prevent pole slipping instability while performing the open-loop start of PMSM. Both the proposed methods are highly suitable for various applications like compressors, pumps, fans and heating ventilation and air conditioning (HVAC) systems, and also for critical applications like electric ship propulsion, and emergency heat and smoke exhaust. The proposed methods pertaining to sensorless vector control were experimentally verified on a 25kW PMSM drive.
Sep
2022
The huge gamut of today’s internet-connected embedded devices has led to increasing concerns
regarding the security and confidentiality of data. To address these requirements, most embedded
devices employ cryptographic algorithms, which are computationally secure. Despite such mathematical
guarantees, as these algorithms are implemented on a physical platform, they leak critical information in
the form of power consumption, electromagnetic (EM) radiation, timing, cache hits and misses, and so on,
leading to side-channel analysis (SCA) attacks. My work focusses on developing advanced side-channel
attacks using machine learning (ML) and circuit-level low-overhead generic countermeasures. I will
present a cross-device deep learning-based profiling power side-channel attack (X-DeepSCA) which can
break the secret key of an AES-128 encryption engine running on an Atmel microcontroller using just a
single power trace. Thus, physical leakage coupled with ML techniques can increase the threat surface of
embedded devices significantly.
Despite all these advancements, most works till date, both attacks as well as countermeasures, treat the
crypto engine as a black box, and hence most protection techniques incur high power/area overheads. In
this talk, I will present the first white-box modeling of the EM leakage from a crypto hardware, leading to
the understanding of the genesis of the EM leakage. Combining the 2 key techniques – current-domain
signature attenuation (CDSA) and local lower metal routing shows >350x signature suppression in
measurements on our fabricated 65nm CMOS test chip, leading to SCA resiliency beyond 1B
encryptions. This key principle of killing the physical side-channel leakage at its source achieved a 100x
improvement in both EM and power SCA protection over the prior works with comparable overheads.
Next, considering the continuous growth of wearable and implantable devices around a human body, this
talk also focuses on analyzing the security of medical/personal devices, particularly for the internet-ofbody (IoB) and proposes electro-quasistatic human body communication (EQS-HBC) to form a covert
body area network. While the traditional wireless body area network (WBAN) signals can be intercepted
even at a distance of 5m, the EQS-HBC signals can be detected only up to 0.15m, which is practically in
physical contact with the person. Thus, this pioneering work proposing EQS-HBC promises >30x
improvement in private space compared to the traditional WBAN, enhancing physical security. In the long
run, EQS-HBC can potentially enable several applications in the domain of connected healthcare,
electroceuticals, augmented and virtual reality, and so on.
Finally, I will conclude with my vision towards developing secure, efficient, and ubiquitous IoT/IoB devices
and cyber-physical systems through the combination of physical fields and their interaction with the
network and system, which adds a new dimension to the analysis of the cyber-physical security.
Finally, I will conclude with my vision towards developing secure, efficient, and ubiquitous IoT/IoB devices
and cyber-physical systems through the combination of physical fields and their interaction with the
network and system, which adds a new dimension to the analysis of the cyber-physical security.
Bio: Debayan Das received his PhD and MS in Electrical and Computer Engineering from Purdue
University, USA in 2021 and the Bachelor of Electronics and Telecommunication Engineering degree
from Jadavpur University, India, in 2015. He is currently a Research Scientist at Intel Corporation, USA.
Prior to his Ph.D., he worked as an Analog Design Engineer at a startup based in India. He has interned
with the Security Research Lab, Intel Labs, USA, over the summers of 2018 and 2020. His research
interests include mixed-signal IC design and hardware security.
Dr. Das was a recipient of the IEEE HOST Best Student Paper Award in 2017 and 2019, the Third Best
Poster Award in the IEEE HOST 2018, and the 2nd Best Demo Award in HOST 2020. In 2019, one of his
papers was recognized as a Top Pick in Hardware and Embedded Security published over the span of
the last six years. He was recognized as the winner (third place) of the ACM ICCAD 2020 Student
Research Competition (SRC). During his Ph.D., he has been awarded the ECE Fellowship during 2016–
2018, the Bilsland Dissertation Fellowship in 2020–2021, the SSCS Pre-doctoral Achievement Award in
2021, and the Outstanding Graduate Student Research Award by the College of Engineering, Purdue
University, in 2021 for his outstanding overall achievements. He has authored/co-authored more than 45
peer-reviewed conferences and journals including 2 book chapters and 1 US patent. He has been serving
as a primary reviewer for multiple reputed journals and conferences including JSSC, TCAS-I, TVLSI,
TCAD, Design & Test, TODAES, JETCAS, TBME, IEEE Access, IoTJ, DAC, VLSI Design, HOST.
Sep
2022We study the problem of best arm identification in a federated learning multi-armed bandit setup with a central server and multiple clients. Each client is associated with a multi-armed bandit in which each arm yields {em i.i.d.} rewards following a Gaussian distribution with an unknown mean and known variance. The set of arms is assumed to be the same at all the clients. We define two notions of best arm---local and global. The local best arm at a client is the arm with the largest mean among the arms local to the client, whereas the global best arm is the arm with the largest average mean across all the clients. We assume that each client can only observe the rewards from its local arms and thereby estimate its local best arm. The clients communicate with a central server on uplinks that entail a cost of $Cge0$ units per usage per uplink. The global best arm is estimated at the server. The goal is to identify the local best arms and the global best arm with minimal total cost, defined as the sum of the total number of arm selections at all the clients and the total communication cost, subject to an upper bound on the error probability. We propose a novel algorithm {sc FedElim} that is based on successive elimination and communicates only in exponential time steps and obtain a high probability instance-dependent upper bound on its total cost. The key takeaway from our paper is that for any $Cgeq 0$ and error probabilities sufficiently small, the total number of arm selections (resp. the total cost) under {sc FedElim} is at most~$2$ (resp.~$3$) times the maximum total number of arm selections under its variant that communicates in every time step. Additionally, we show that the latter is optimal in expectation up to a constant factor, thereby demonstrating that communication is almost cost-free in {sc FedElim}.
Sep
2022Exact inference in Bayesian networks is intractable and has an exponential dependence on the size of the largest clique in the corresponding clique tree (CT), necessitating approximations. Factor based methods to bound clique sizes are more accurate than structure based methods, but are expensive since they involve inference of beliefs in a large number of candidate structure or region graphs. We propose an alternative approach for approximate inference based on an incremental build-infer-approximate (IBIA) paradigm, which converts the Bayesian network into a data structure containing a sequence of linked clique tree forests (SLCTF), with clique sizes bounded by a user-specified value. We show that our algorithm for incremental construction of clique trees always generates a valid CT and our approximation technique preserves the joint beliefs of the variables within a clique.In this seminar, I will present results obtained using the proposed algorithm for incremental construction of CTs and inference of prior beliefs using the IBIA framework. The framework was used to evaluate signal and transition probabilities for large digital circuits as well as several other Bayesian network benchmarks. The results show a significant reduction in error when compared to other approximate methods with competitive runtimes.
Sep
2022Quantum key distribution (QKD) provides a means of generating secret random bits, or keys, for cryptographic purposes, between two distant parties. QKD has drawn a lot of attention in last two decades because of its unconditional security guaranteed by quantum mechanics, such as no-cloning theorem. Experiments with 2-pulse Differential Phase-Shift Quantum Key Distribution (DPS-QKD) with quantum bit error rate (QBER) of 21% and quantum random number generator (QRNG) with two different entropies (arrival time of photon and path superposition) were presented in Seminar-I. Key generation efficiency, and security, in DPS-QKD improve with an increase in the number of optical delays or time-bin superpositions. We demonstrate the implementation of superposition states using time-bins, with two different approaches. In Type-A, we use an optical pulse and create superposition states with optical splitters and path delays. Similar superposition states are created, in Type-B, by applying direct phase modulation within a single weak coherent pulse. In this talk, we will discuss the equivalence between both the approaches, and implementation of higher-order superposition states of Type-B in DPS-QKD. We have established 4-state DPS-QKD, over 105 km of single mode optical fiber, with a QBER of less than 15% at a secure key rate of 2 kbps. To optimize the performance of QKD test-bed, gated single photon detector (SPD) was characterized with sub- picosecond weak coherent pulses. We have also shown that with temporal filtering, the QBER reduced to less than 10%, but with a 20 % reduction in key rate.
Sep
2022
Silicon germanium heterojunction bipolar transistors (SiGe HBTs) are rapidly evolving to cater the increased functionality and speed demands of the modern communication systems (4G, 5G & upcoming 6G). With the emerging mm-wave and THz market, the precise characterization and modeling of the devices are crucial to optimize the circuit-performance and minimize the number of overall design cycles. In this work, a very high frequency (up to 500 GHz) measurement is carried out and corresponding data is analyzed using two finite element tools, one solving the EM equations and the other solving the semiconductor equations. The presented methodology provides one with sufficient confidence in the adopted characterization techniques and results. Precisely, it allows one to differentiate between the accurate and erroneous characterizations. Further, the need for proper calibration and de-embedding techniques in high-frequency characterization is emphasized by investigating the s-parameters corresponding to a narrow-band amplifier at 170 GHz suitable for G-band radar applications.
On the other hand, intense research on the innovative design of BiCMOS compatible SiGe HBTs with increased speed and breakdown voltage is being pursued across the globe. In this direction of cutting-edge research, we present two unconventional SiGe HBT architectures with improved RF performance metrics. Our first design is based on a nanowire architecture that ensures reduced lateral parasitics leading to an fMAX above 900 GHz. The second architecture is an SOI-based lateral SiGe HBT that demonstrates an fMAX above 2.7 THz.
Sep
2022
It is predicted that, across the world, there would be over 7.5 billion users accessing the internet through wireless and wireline networks by 2030 [cybersecurityventures.com]. Fiber optic communication system being the backbone for mobile data and internet traffic; this proliferation of internet usage directly translates to a corresponding increased data-rate and bandwidth requirements in access, metro and long haul fiber optic links. The high data rate is accomplished either by employing wavelength division multiplexing (WDM) or by using advanced modulation formats which utilizes amplitude, phase and polarization of the electric field to encode digital data. The use of advanced modulation formats necessitates larger signal to noise ratios for error-free detection of data. Each optical amplifier in a fiber optic link, in addition to providing optical gain, also generates noise. Thus, there is a trade-off between the achievable spectral efficiency and the link length. Modulating outside the traditional communication bands towards an ultra-wide band option is another potential solution to improve the capacity of fiber links.
Traditional erbium doped fiber amplifiers restrict the wavelengths of operation to the C- (1530-1565 nm) and the L-band (1565-1625 nm), and have typical noise figure values of > 5 dB. An alternative to doped fiber amplifiers is the use of phase sensitive amplifiers (PSA). PSAs offer the possibility of adjusting the gain spectra and reducing the noise figure to even 0 dB. Optimal choice of a nonlinear medium to be used in PSA designs that can offer noise figure better than the commercial EDFAs, with least complexity/energy consumption is a subject of investigation. Nonlinear semiconductor optical amplifiers (SOA) have an ultra-wide band gain spectrum and have the potential to be used as nonlinear medium for both PSA and ultra-wide band amplification. In this seminar, we discuss the possibility of using SOA as an amplifier and as a nonlinear medium. We also discuss the characterisation details of a dispersion oscillating fiber for its potential use as a nonlinear medium for PSA.
Sep
2022
Images and videos captured in our day-to-day life are often degraded due to various inevitable elements such as blur, rain, haze, etc. The majority of such visual degradations are spatially varying. Existing restoration algorithms aim to recover the original clean information using input-agnostic and spatially-invariant processing in purely convolutional architectures. They fail to adequately address the inherent spatial and temporal variations observed in degraded inputs. Our work presents pixel-adaptive neural networks for various restoration tasks, where the network can adjust its behavior depending on the pixel content. Blur is a common phenomenon when capturing a dynamic scene with handheld devices. We design a pixel adaptive and feature attentive image deblurring network that can handle large blur variations across different spatial locations and different images. Our content-aware global-local filtering module considers not only the global dependencies but also the neighboring information dynamically for each pixel. We extend this framework to the spatio-temporal domain to handle blurry videos. We adaptively extract the additional information from the temporal axis to refine each frame using non-local attention. Our work is more efficient than existing designs while outperforming them by a large margin.
Subsequently, we address removing bad-weather-caused degradations such as rain streaks, haze, and raindrops from images. We delve deeper into the limitations of the pixel-adaptive modules and present a two-stage neural framework. We decompose the restoration task into two stages of degradation localization and degraded region-guided restoration. The additional knowledge of the degraded regions helps the network focus on restoring the most difficult regions. We demonstrate that this knowledge transfer technique can be equally beneficial for the highly ill-posed image inpainting task. We design a distillation-based approach, where we provide direct feature-level supervision while training. This additional guidance helps the network to reach a much better optima and produce superior results. We conduct extensive evaluations on multiple datasets to demonstrate the superiority of our method for each task separately.
Sep
2022The increase in greenhouse gas due to fossil fuel-powered generators, industries, and automobiles has led to severe climate change over the past decade. This led to an increase in renewable energy sources for power generation and Electric Vehicles (EVs) in the automobile industry. This work focuses on the effective utilization of renewable energy sources consisting of wind energy and solar PV systems for electric vehicle battery charging by interfacing through dc bus. The proposed dc bus voltage controller maintains a constant voltage at the dc bus under various environmental conditions like wind gusts, zero solar irradiation, and grid interruption. The design and stability aspects of the dc bus voltage controller are analyzed in detail using frequency domain analysis. The robustness of the proposed controller is tested with recorded field data of wind velocity and solar irradiation. The feasibility and efficacy of the proposed controller design are confirmed for the EV battery charging station using real-time simulation studies. The results are compared with the IEEE 519 standard, verifying the efficacy of the proposed controller design.
Sep
2022Team CoE-CPPICS cordially invites all interested individuals to register for the 10th webinar talk under the "PICS- Today Webinar Series"
Sep
2022Reusing the same spectrum in every cell is a common feature of 4G and 5G systems. Such systems are plagued by co-channel interference (CCI), especially for sub-1 GHz frequency band deployments, which have low path loss and hence high coverage. The carrier frequency generated by each transmitting tower will have slight variations primarily due to local oscillator deviations, leading to distinct carrier frequency offsets (CFOs) for the desired signal and the interfering signals. For an orthogonal frequency division multiplexing (OFDM) based system, CFO causes inter-carrier interference (ICI) to be generated during the demodulation process. Management of CCI and ICI are critical to accurately recovering the desired signal. In this work, we propose an advanced receiver consisting of a spectrally efficient joint channel estimator, a computationally simple time-domain ICI compensation algorithm, and a CFO-corrected joint detector to recover the desired information from the received signal. It is shown that the proposed methods enable accurate signal recovery even in severe CCI environments which are possible in reuse-1 UHF band cellular deployments.
Sep
2022In the development of electric vehicles (EVs) and plug-in hybrid electric vehicles (PHEVs), the battery charger plays an important part. The number of all-electric (EVs) and hybrid electric vehicles (HEVs) in India is expected to expand substantially in the next years. On-board batteries in EVs/HEVs require complete or partial recharging via a low voltage utility network that supplies home or light industrial customers. A battery charger is required for the utility connection, which is made up of a unidirectional or bidirectional power electronic converter that converts AC to DC or vice versa. Because of the smaller size of the high-frequency isolation transformer, a two-stage AC/DC and DC/DC power conversion is widely used in battery charging systems. This thesis is organized into seven chapters and focuses on the design and development of AC-DC and DC-DC converters for a 5 kW, 84V battery charger.
Sep
2022Faulty elements in a phased array antenna lead to the undesired radiation pattern and substandard system performance. Therefore, fault diagnosis in a phased array is an inevitable task to ensure the proper functioning of a communication system. A compressive sensing-based inverse problem is formulated to recover the sparse solutions and locate the faulty elements from the minimum number of far-field measurements. We propose a diagnosis method from fixed probe measurements and varying excitations. The non-convex lp norm (0 < p <1) minimization problem is solved using the IRL1-ADMM algorithm. Further reduction in the number of measurements is guaranteed by optimizing the excitations to minimize the mutual coherence of the system measurement matrix.
Sep
2022Optical coherence tomography (OCT) is a well-known imaging technology in the biomedical field and a popular commercial technology in ophthalmology. In terms of image resolution, it is superior to ultrasound imaging but has limited access to objects well below the surface. Given that it has a few micrometers of longitudinal and lateral resolution, it can also be used in non-biomedical applications such as sub-surfacing imaging of fabricated semiconductor chips, historical art conservation, agriculture sciences, etc. There are always certain trade-offs to be kept in mind when using a technology for a specific application, as the design should meet the imaging requirements.
Aug
2022
We consider a system with a large number of identical servers into which jobs arrive as a Poisson process. Job sizes are independent and exponentially distributed. Upon arrival, a job is assigned to d servers chosen uniformly at random. Servers split their effort equally amongst all jobs they are currently serving (processor sharing). When the total work done on a job by all servers to which it has been assigned equals its job size, the job departs the system.
The system can be analysed exactly only if d=1, i.e., there is no parallelism. We present a mean-field analysis if d is 2 or more, analogous to the cavity method in statistical physics. This involves analysing an index queue assuming all other queues are in their (unknown) equilibrium distribution, and yields a fixed-point equation for this unknown distribution, which can be solved numerically.
We provide evidence from simulations that the mean-field analysis is correct. We also present some very preliminary steps towards rigorously justifying the mean-field analysis. Finally, we describe a connection between the above model and a random graph/ hypergraph process.
Aug
2022Deep learning techniques have been applied to optimize a specific function such as coding, modulation, or equalization in wireless and optical communication systems. Such a modular implementation allows the individual system components to be optimized and analyzed separately and thus presents a convenient way of building the communication link.
Aug
2022
Landmark codes underpin reliable physical layer communication, e.g., Reed-Muller, BCH, Convolutional, Turbo, LDPC and Polar codes: each is a linear code and represents a mathematical breakthrough. The impact on humanity is huge: each of these codes has been used in global wireless communication standards (satellite, WiFi, cellular). Traditionally, the design of codes has been driven by human-ingenuity and hence the progress is sporadic. Can we automate and accelerate this process of discovering codes?
In this talk, I will talk about KO codes, a new family of computationally efficient deep-learning driven codes that outperform the state-of-the-art reliability performance on the standardized AWGN channel. KO codes beat state-of-the-art Reed-Muller and Polar codes, under the low-complexity successive cancellation decoding, in the challenging short-to-medium block length regime on the AWGN channel. We show that the gains of KO codes are primarily due to the nonlinear mapping of information bits directly to transmit real symbols (bypassing modulation) and yet possess an efficient, high performance decoder. The key technical innovation that renders this possible is the design of a novel family of neural architectures inspired by the computation tree of the Kronecker Operation (KO) central to Reed-Muller and Polar codes. These architectures pave the way for the discovery of a much richer class of hitherto unexplored nonlinear algebraic structures. And more interestingly, despite having a lot of encoding and decoding structure, KO codes exhibit striking similarity to random Gaussian codes!
Speakers bio:
Ashok is an incoming postdoctoral associate at EPFL with Prof. Michael Gastpar. He recently obtained his PhD in ECE from UIUC, advised by Prof. Pramod Viswanath. He also obtained his Masters in ECE (advised by Prof. Yihong Wu) from UIUC in 2017 and Bachelors in EE (advised by Prof. Vivek Borkar) with a minor in Mathematics from IIT Bombay in 2015. His current research interests are in theoretical and algorithmic aspects of machine learning and information theory. He is a recipient of Best Paper Award from ACM MobiHoc 2019. He is also a recipient of several graduate student awards and fellowships including Joan and Lalit Bahl Fellowship (twice), Sundaram Seshu International Student Fellowship, and is a finalist for the Qualcomm Innovation Fellowship 2018. Outside research, he likes to learn new languages, watch and read about international films, remembering movie trivia and cooking. For more details about him, please visit makkuva2.web.engr.illinois.edu
Aug
2022
Biography of the Speaker :
Pilsoon Choi (Senior Member, IEEE) received the Ph.D. degree in EECS from KAIST in 2004. While with KAIST, he developed the first IEEE 802.15.4 CMOS radio for low-power wireless sensor network which is now Internet-of-Things (IoTs). From 2004 to 2011, he was with Samsung Electronics Company Ltd., where he developed the first wirel ess codec SoC for real-time HD video streaming and was also involved in international standardization. Since 2012, he has been leading a communication circuit design team of the Low Energy Electronic Systems (LEES) research program. He is currently a Research Scientist with MIT. His research interests include RF signal-processing and energy-processing circuits design in III-V and CMOS technologies.
Affiliation of the Speaker :
Research Scientist, MIT, Cambridge, USA
Abstract :
III-V devices and circuits have been playing an important role in wireless communications while CMOS devices and circuits are required for digital and analog building blocks design. For future mobile applications, III-V will be crucial for the power efficiency of battery-powered devices, high linearity for carrier-aggregated signals, and small form factor of phased-array radios. This webinar starts with system level design issues for RF/mmWave circuits design, introduces LEES monolithic integration process that incorporates III-V HEMTs into standard CMOS integrated circuits in a wafer scale, and presents some III-V + CMOS design examples with outstanding performance that cannot be achieved with CMOS only. III-V HEMTs and their monolithic integration with CMOS circuits will tackle challenging problems in realizing future mobile devices.
Aug
2022Non-orthogonal multiple access (NOMA) is a major 5G technology that leverages resources in a non-orthogonal way to improve the spectral efficiency and number of user connections in a wireless network. One of the key issues with the multi-cell NOMA system is that cell-edge users may experience lower bit rates due to inter-cell interference (ICI) from nearby cells. Although joint transmission coordinated multi-point (JT-CoMP) transmission can help to mitigate this problem, cell-edge users performance is limited by inter-NOMA-user interference (INUI) from users in the same resource block (RB) as the cell-edge user.
Aug
2022
Solution-processed organic field-effect devices have been a low-cost alternative to conventional device technologies specifically for low-frequency flexible electronics, rigid electronics, and specific large area applications. Organic metal-insulator-semiconductor capacitors (OMISCAP) and organic thin-film transistors (OTFT) are fundamental building blocks in organic circuits. It is required to counter some fundamental limitations in OMISCAP and OTFT, i.e., frequency response, mobility, and contact effect. These fundamental limitations exist due to inferior charge injection from the metal to the semiconductor bulk.
In the first part of the presentation, the frequency response of an OMISCAP will be discussed. The cut-off frequency of dispersion (fT) for an OMISCAP typically decides the bandwidth of organic field-effect devices, which is a crucial parameter for circuits. The fT dependency on the dielectric constant of polyvinyl-4-phenol (PVP) polymer dielectric and on the injection barrier at the metal-semiconductor interface will be discussed. Impact of charge injection through the metal-semiconductor interface is equally important in deciding the performance of OTFT in terms of contact resistance, mobility etc. which will be discussed in the second part of the presentation. Typically, in solution-processed bottom gate bottom contact (BGBC) OTFT, the metal thickness elevated above the substrate leads to poor morphology and thus high contact resistance. Contact resistance usually leads to deterioration of performance such as high switch-on voltage and hysteresis, which are the fundamental bottleneck for circuit applications. The ways to mitigate the contact resistance by varying the electrode thickness and fabricating recessed electrodes on polymer dielectric will be discussed further. The planar OTFT substrate by fabricating the recessed source, drain and gate was found to reduce the contact resistance by two orders and improve the mobility by three-fold along with a significant reduction in hysteresis and switch-on voltage close to zero.
Aug
2022Power flows and voltage profiles in power transmission systems for power system planning are going to present in this seminar. Model formation and simulation of a southern transmission network for power flow control are presented in this work. Power flow correctional devices such as Variable Shunt Reactors (VSR) and On Load Tap Changers (OLTC) play an important role in the power flow control and voltage profile improvement. The suitability and application of VSR and OLTC to control the power flows and regulate the 220 kV and 400 kV voltage levels of the substations inside the grid are presented in detail. Power Loss improvement as a result of the application of VSR and OLTC study is discussed. The simulation results are validated in Dig-SILENT Power Factory for the year 2015 and 2020 data.
Aug
2022A three-coil sensor is designed and developed for sensing the level of a nonconductive liquid kept in a container. Though the sensor is a general displacement sensor, the structure of the sensor makes it best suited for liquid level measurement in a sealed container. The proposed sensor is made of two identical coils, say coil 1 and coil 2 of trapezoidal (wedge) shape, fixed on the outside but on opposite sides of the container of the liquid whose level is to be measured. They are placed with a 180-degree orientation with respect to each other. A third coil, namely coil 3, which is shorted at the ends with a capacitor is made to float on the surface of the liquid inside the container. The two trapezoidal coils are excited with a sinusoidal signal at the resonance frequency of the third coil-capacitor combination. The current drawn by the two trapezoidal coils are then processed to obtain an output that is linear to the position (and hence liquid level) of the floating coil. The proposed technique is evaluated analytically, and its efficacy demonstrated through simulation studies. Finite element analysis conducted using COMSOL also correlates with the results of the analytical studies. A prototype model developed and tested in the laboratory, not only established the practicality of the proposed sensor, but again validates the analytical and finite element simulation studies. Through the analysis, simulation, and prototype testing it is found that the proposed sensor suffers from end-effects and thus has a useful range of only 60% of the length of the sensor. It is shown here that by using a soft ferromagnetic core of high permeability only on the third coil, the endeffects can be removed. It nicely turns out that apart from removing the end-effects resulting in the entire 100 % of the sensor range to be useful, the introduction of the core also reduces the worst-case error from 3 % to 2%.
Aug
2022Wireless applications need power efficient wide-band analog-to-digital converters (ADCs) with high dynamic range. A continuous time delta sigma modulator (CTDSM) is attractive in such applications due to its resistive input impedance as well as implicit antialiasing features. Adopting a CTDSM ADC for wide bandwidths necessitates very high sampling rates in order to achieve desired performance targets. This offers challenges in terms of comparator meta-stability which directly affects loop timing closure as well as data dependent jitter. In addition, high frequency operation is accompanied by increased inter symbol interference (ISI) as well as increased power dissipation. Our work aims to analyse these challenges in detail and offer solutions for overcoming the same.
Aug
2022Learning with dependent data with streaming algorithms is very important in real world applications from time series forecasting to reinforcement learning and control systems. Here the data is often assumed to be Markovian. The limits of learning and the principles behind algorithm design in this context are poorly understood compared to the i.i.d. data setting. In many important cases we can utilize independence in the noise process to obtain algorithms and guarantees which perform near optimally with Markovian data, matching the rates obtained with i.i.d. data. In this talk, I will consider two questions with this theme:
Aug
2022Integrated photonics has gained a lot of interest in high-speed optical interconnects, high-capacity information processing systems and many other allied fields due to the evolution of cost-effective CMOS compatible silicon photonics technology. Transparency and low-loss optical waveguide design are the key for large scale functional photonic integrated circuit implementation. In general, silicon waveguide core is used for the realization of various passive and active components in the silicon-on-insulator (SOI) platform. However, a typical silicon waveguide possesses certain limitations owing to its lower band-gap and higher losses
Aug
2022Recently, significant progress has been made in developing autostereoscopic platforms for displaying real-world 3D scenes. Light fields are the best emerging choice for computational multi-view autostereoscopic displays since they provide an optimized solution to support direction-dependent outputs simultaneously without sacrificing the resolution. Such multi-view light field displays can accommodate numerous viewing directions, continuous motion parallax, greater depth-of-field and wider field-of-view. However, the primary challenge in handling light fields for any application are their enormous volume and requirements for storage and transmission. It is therefore critical to develop efficient representation and coding solutions for light fields suitable for display and streaming applications.
In this talk, we present two novel light field representations, coding and streaming schemes. The first is a hierarchical algorithm based on transmittance patterns of low-rank multiplicative layers and Fourier disparity layers for different light field scanning patterns. In another scheme, intrinsic redundancies in light field subsets of the same scanning patterns are eliminated through low-rank representation using Tucker decomposition with tensor sketching for various ranks and sketch dimension parameters. Both the proposed integrated formulations follow hierarchical scanning orders to operate on view subsets and analyze the approximated light field by sampling it in the depth dimension by decomposing the scene as a discrete sum of Fourier Disparity layers. This exploits additional intra-view, inter-view, and other redundancies among adjacent views in horizontal and vertical directions to further allow scalable light field coding.
Besides, the representation is appropriate for more general rendering by shifting approximated light field sub-aperture images in the depth dimension instead of two angular dimensions. We can reconstruct intermediate viewpoints from low-rank approximated views at different bitrates without any additional disparity maps. The proposed schemes are flexible to realize a range of multiple bitrates at the decoder within a single integrated system. Moreover, our hybrid low-rank approximation and encoding schemes can also be incorporated as a complement to other existing or future light field coding algorithms. Compression performance of the schemes analyzed on real light fields shows substantial bitrate savings compared to state-of-the-art codecs, while maintaining good reconstruction quality.
Aug
2022Recently, significant progress has been made in developing autostereoscopic platforms for displaying real-world 3D scenes. Light fields are the best emerging choice for computational multi-view autostereoscopic displays since they provide an optimized solution to support direction-dependent outputs simultaneously without sacrificing the resolution. Such multi-view light field displays can accommodate numerous viewing directions, continuous motion parallax, greater depth-of-field and wider field-of-view. However, the primary challenge in handling light fields for any application are their enormous volume and requirements for storage and transmission. It is therefore critical to develop efficient representation and coding solutions for light fields suitable for display and streaming applications.
In this talk, we present two novel light field representations, coding and streaming schemes. The first is a hierarchical algorithm based on transmittance patterns of low-rank multiplicative layers and Fourier disparity layers for different light field scanning patterns. In another scheme, intrinsic redundancies in light field subsets of the same scanning patterns are eliminated through low-rank representation using Tucker decomposition with tensor sketching for various ranks and sketch dimension parameters. Both the proposed integrated formulations follow hierarchical scanning orders to operate on view subsets and analyze the approximated light field by sampling it in the depth dimension by decomposing the scene as a discrete sum of Fourier Disparity layers. This exploits additional intra-view, inter-view, and other redundancies among adjacent views in horizontal and vertical directions to further allow scalable light field coding.
Besides, the representation is appropriate for more general rendering by shifting approximated light field sub-aperture images in the depth dimension instead of two angular dimensions. We can reconstruct intermediate viewpoints from low-rank approximated views at different bitrates without any additional disparity maps. The proposed schemes are flexible to realize a range of multiple bitrates at the decoder within a single integrated system. Moreover, our hybrid low-rank approximation and encoding schemes can also be incorporated as a complement to other existing or future light field coding algorithms. Compression performance of the schemes analyzed on real light fields shows substantial bitrate savings compared to state-of-the-art codecs, while maintaining good reconstruction quality.
Aug
2022Epoxy micro-nanocomposites are gaining importance to be used as potential insulation structures in power apparatus. Methodical experimental studies were carried out for investigating the electrical, thermal and mechanical properties of epoxy micro-composites as well as epoxy micro-nanocomposites. In real-time, these insulation structures are operated under harsh environmental conditions. Hence, the impact of various ageing conditions such as gamma irradiation, water ageing, ultraviolet irradiation and corona ageing on space charge as well as thermo-mechanical properties of epoxy micro-nanocomposites were studied in detail. The aged epoxy micro-nanocomposites were classified by employing principal component analysis and artificial neural network analysis to laser induced breakdown spectroscopy data.
Jul
2022Rayleigh Scattering based Distributed Acoustic Sensors are capable of both detecting and quantifying dynamic perturbations such as vibrations along the complete length of the sensing fiber. They are widely used in intrusion sensing, pipeline monitoring, structural health monitoring and in seismology fields. The basic working principle of a distributed sensing system is the same as that of Phase OTDR, which extracts the phase of the backscattered signal from an optical fiber. The phase of the signal varies with position of the fiber as well as with time due to refractive index variations induced by dynamic strains. Extracting the phase for consecutively obtained OTDR traces can give the relative phase change at any point along the fiber with time.
In this talk, we discuss the modeling of a DAS based on Rayleigh scattering, analysis of different phase extraction methods through simulation and experimental studies on phase extraction using coherent detection. Challenges and limitations in this field and performance enhancing techniques are also discussed.
Jul
2022With ever-increasing network traffic with an exponential growth rate of up to 60% per year, based on the application segment and geographical location, the next generation fiber-optic networks must focus on the spectrally efficient ways of providing high-capacity transport infrastructure for both long-haul and short-haul systems (<100 km). One of the methods to scale the capacity is by employing higher cardinality of modulation, thereby encoding more bits in a symbol, coupled with using a higher symbol rate of transmission. However, on scaling the modulation order, the allowed phase perturbation to be still within the error-free detection, a.k.a phase margin, reduces. Thus the performance becomes detrimental to small non-idealities or phase fluctuations in the system due to laser phase noise and transceiver IQ imbalance. Therefore, we need advanced digital signal processing (DSP)algorithms to correct these impairments.
In this talk, we present a novel DSP algorithm to correct for the laser phase noise, even in the presence of transmitter IQ imbalance, by intelligently adapting the decision boundaries. We show the improvement in performance by evaluating the algorithm for up to 32 GBd PM-16QAM (256 Gbps) modulation and show that it has less computational complexity than the conventional DSP algorithm. We also present a novel Geometric parameter extraction-based algorithm for receiver IQ imbalance correction. This pilot-free algorithm only uses less than 5% symbols from the received frame to estimate the statistical parameters, which are then used to correct the imbalanced data. We show the efficacy of this algorithm in simulations and through experiments for up to 80 GBd PM-16QAM (640 Gbps) signal.
Jul
2022Femtosecond (fs) laser micromachining is one of the most flexible manufacturing technologies to create micron sized features. Its ability to accurately and reproducibly create structures in a wide range of materials makes it an indispensable technology in a wide array of applications such as micro-cutting of cardiac stents, micro-scribing of silicon/thin film solar cells, micro-lithography of electronic chips, micro scribing/cutting of electrodes in flat panel displays, surface micro texturing of automotive engine parts, microfabrication of MEMS devices. The advantage of femtosecond pulses over the nanosecond (ns) or picosecond (ps) is their ability to deposit energy into a material in a very short time. It will remove or modify it before thermal processes originate. The ultrafast lasers vaporize matter without generating heat ("cold ablation"). The energy deposition process takes place quickly when compared to atomic relaxation procedures.
Jul
2022Software Defined Networking (SDN) provides a means to sunder the control and the data planes of traditional IP network devices. Further, it accommodates a centralized control logic that facilitates real-time configurable network devices, and addresses network resilience, traffic management, and access control. Therefore, many organisations are upgrading their networks with SDN switches. However, such a procedure incurs huge upgrading cost and is impractical to upgrade all IP routers to SDN switches at once. Thus, a feasible solution is to upgrade only a few of the IP routers with SDN switches. This kind of network consisting of both IP routers and SDN switches is referred to as a hybrid IP/SDN network. The IP routers which are replaced by SDN switches are called as candidate switches. The goal of our work is to propose a Candidate Selection Algorithm (CSA) that minimises the number of candidate switches while covering all Single Link Failures (SLFs) in the network.
Jul
2022Full-Duplex (FD) communication is a promising technique that can potentially double the system throughput compared to conventional Half Duplex (HD) transmission. Recent experiments have demonstrated near-perfect self-interference suppression techniques in the PHY layer to aid the possibility of FD communication. In this talk, we present a Synchronized Mode Full-Duplex (SM-FD) wireless LAN MAC protocol with appropriate modification of the IEEE 802.11 standard. The proposed SM-FD MAC protocol maintains synchronization between nodes and Access Point (AP) to increase efficiency. SM-FD MAC protocol supports WLAN having both FD and HD capable nodes by establishing both symmetrical links and asymmetrical links. To this end, we have utilized the packet-alignment-based capture effect in an efficient way. The proposed SM-FD MAC protocol has been analyzed using a Markov chain model. The analytical results for RTS and data collision probability, saturated throughput, delay, and energy spent per packet transmission have been derived and validated against simulation.
Jul
2022Artificial neural networks (Deep Learning) have shown excellent accuracy in image classification and recognition tasks. For performing these tasks accurately, efficient training of the network is essential. CMOS-based architectures have been commonly used as accelerators for training the networks. But in this case, the computing unit requires data from the off-chip memory, which leads to time delay and energy consumption. In-memory computing with non-volatile memory can significantly improve energy efficiency and latency due to the ability for one-step vector-matrix multiplication and local storage of network parameters. Resistive Random Access Memory (RRAM) is a viable candidate among emerging non-volatile memory technologies as it provides high device density, multi-bit storage, and low latency in write/read operations. The ability of conductance modulation in RRAM devices is analogous to that of a biological synapse. The first part of this talk discusses the need for in-memory computing. We then discuss different emerging non-volatile memory technologies that are at our disposal. We then discuss the operating principles of RRAM, the advantages of RRAM technology, and the technological challenges for practical applications.
Various high-k oxides have been previously used as an insulator for the RRAM technology. However, Silicon Oxide (SiOx) can still emerge as a great candidate due to the advantages of availability, low cost, and excellent CMOS compatibility. In the final part of this talk, we discuss the development of a SiOx RRAM technology. We use Inductively Coupled- Plasma Enhanced Chemical Vapor Deposition (ICP-PECVD) to develop SiOx as the active switching layer. We design and fabricate Au/SiOx/Ti/Au RRAM cross-point devices using photolithography and electron-beam evaporation. The devices exhibit stable, bipolar operation after an initial electroforming step. Detailed electrical characterizations are performed to evaluate the performance in terms of operating voltages, dynamic range, endurance, retention, and synaptic weight updates. We vary the thicknesses of the SiOx switching layer as well as the thickness of the Ti top electrode to optimize the RRAM performance. Detailed investigations have also been carried out to understand the resistance switching mechanism.
Jul
2022Organic Light Emitting Diodes (OLED) based displays in smartphones and wearables have seen a steady rise over the last few years. The next (3rd) generation Thermally Activated Delayed Fluorescence (TADF) Organic Light Emitting Diodes (OLEDs) avoid heavy metal atoms used in the current 2nd generation OLEDs. However, the major challenge in TADF OLEDs is the roll-off of external quantum efficiency (EQE) at high intensities. In our work, we used a TADF host 2,6-Bis(9,9-diphenylacridin-10(9H)-yl)pyrazine (PrDPhAc). The efficiency roll-off is presumed to be caused by Triplet-Triplet Annihilation (TTA) and charge imbalance. To minimize EQE roll-off, the recombination zone needs to be in the middle of the emissive layer. The recombination zone deviates from the center of the emissive layer at high voltages due to the difference in electron and hole mobility in the respective transport layers in OLED. A detailed study of electron and hole-only devices with different thicknesses of organic layers and different injection layers are fabricated and analysed. J-V characteristics show that electron injection is less than hole injection, which could be the reason for charge imbalance in OLED. 1,4,5,8,9,11-Hexaazatriphenylenehexacarbonitrile (HATCN) and Poly(3,4-ethylene dioxythiophene)-poly(styrene sulfonate) PEDOT: PSS were used as the hole injection layers.
Jul
2022Due to the presence of neutral-point voltage of four-leg voltage source converter (VSC) in the natural reference frame, the voltage dynamics of VSC are coupled through control inputs of all phases. This coupling leads to difficulty in assigning appropriate values to the control input variables using SMC with a conventional sliding surface. To eliminate this coupling, a new sliding variable for each converter leg is proposed. The proposed sliding variable for each leg is a function of the respective filter-capacitor voltage and the fictitious voltage which is a function of converter neutral-point voltage (NPV). Thus, the control of converter NPV is possible in addition to the control of compensator voltages using the proposed scheme. Further, the optimum value of sliding co-efficient, to maximize the existence region of sliding mode, is found for a four-leg VSC-based dynamic voltage restorer (DVR). The performance of a DVR with the proposed control scheme under various operating conditions is validated through detailed experimental studies carried out on a laboratory prototype of a four-leg DVR.
Jul
2022The idea of combining a large number of laser beams to generate a single high power laser beam opens up a wide range of high power laser applications in different areas including medical, defense and space exploration. Fiber lasers have unique advantages compared to other types of laser systems including efficient electrical to optical conversion, excellent beam quality due to single transverse mode of operation and the option for power scaling through master oscillator power amplifier (MOPA) schemes. Beyond kiloWatt power level, an attractive option for power scaling is beam combination of multiple fiber amplifiers. There are different approaches in beam combining such as spectral beam combining, coherent beam combining and polarization beam combining. Of these, coherent beam combining (CBC) offers much promise for scaling to more than 100 beams. CBC may be achieved using tiled aperture configuration or filled aperture configuration.
Jul
2022Control and Optimization Group, EE, and pCoE for Network Systems Learning, Control, and Evolution (CENS) would like to invite you to the workshop on "Linear Transfer Operators: From Data-driven Analytics to Prediction and Control in Dynamical Systems", July 19-22, 2022. The workshop will be conducted by Prof. Umesh Vaidya, Clemson University, USA.
Jul
2022Low-cost sensing is a relatively new paradigm for air quality monitoring at high spatial and temporal resolutions. However, data obtained by using this technique is less reliable due to various error sources such as atmospheric conditions and parameter drift. Calibration, transforming raw sensor measurements to the one generated by reference-grade instruments, is a promising approach to improving the accuracy of low-cost sensors (LCS) data. The calibration of LCS can be framed as a machine learning model to minimize the error between the reference instrument and LCS. This talk presents a quantitative analysis of regression and classification models in machine learning to calibrate LCS in air quality monitoring. Our approach is application-specific, where we divide the stationary applications of LCS into low, moderate and high concentration applications to determine a better calibration model for respective applications. Results are verified with real-time data obtained from SensurAir, an LCS device that we designed and deployed in two different locations in Chennai.
M V Narayana EE18D302
Jul
2022Department Of Electrical Engineering
Department Degree Distribution Programme
13 JULY 2022
Chief Guest
Prof. Rao R. Tummala
Distinguished Professor, Georgia Tech, USA
Program Notifications and Map(revised) to venue
Live Streaming
Jul
2022Prof. Santosh Devasia and Dr. Anuj Tiwari of the Univ. of Washington, Seattle will be giving a short course
"Inversion-based Feedforward Control for Precision Tracking,"
Jul
2022Dear All
The following webinar is organised by the INAE Chennai Chapter in their Webinar Series.
Prof.Bhaskar Ramamurthi, FNAE , Professor, Electrical Engineering and Former Director, IIT Madras will deliver a lecture " Towards an Atmanirbar Telecom Network" on Friday 8th July 2022 at 6.30 PM.
Jul
2022Workshop on "Grid-Connected Inverters: Operations and Control" on 6th, 7th and 8th July, 2022. Sponsored by C-DAC and NaMPET. Coordinators: Dr. N. Lakshminarasamma and Dr. Arun Karuppaswamy
Apr
2022IIT Madras , ARCI Chennai / Hyderabad along with INAE Chennai Chapter is planning to organise a two day National Conference on Energy Technologies (NCET) at IIT Madras, during 29-30th April 2022 The Conference aims to bring together leading academic scientists, researchers, research scholars and industry experts to exchange and share their experiences and research results on all aspects of energy storage systems and technologies including battery, fuel cell, super capacitors solar and wind energy.
Feb
2022PhD SEMINAR TALK – II Title: Distributed Energy Management using Consensus Algorithms Date: 14.02.2022 Time: 2:30 P.M Venue: Google meet Google Meet Link : meet.google.com/qrx-hsxn-qti Speaker: Ms. Naina P M (EE15D024) Guide: Dr. K.S Swarup Abstract The presence and penetration of distributed generators (DGs), transforms the power system grid into a decentralized network. The Energy Management Operations, which employ conventional centralized algorithms may not be suitable, effective and accurate in the presence of DGs. The distributed economic dispatch operation using the leader-follower consensus algorithm was presented in Seminar talk I. In Seminar talk II, we present distributed energy management of virtual power plants (VPP) using two different consensus-based algorithms. The ‘proportional consensus algorithm’ uses a consensus gain function which is a function of the number of iterations and the maximum delay applied. In contrast, proportional integral consensus uses an additional integral term to reduce the effect of noise and communication delay. The effectiveness of the proposed work is illustrated in three different systems i.) five bus microgrid, ii.) 15 node VPP, and iii.) IEEE 39 bus system. This approach helps to avoid the VPP coordinator, thereby preserving the privacy of the VPP participants.
Feb
2022Ph D Seminar - II Title: Dynamical system approaches to solve time-varying convex optimization problems Date: 14.02.2022 Time: : 03:00 P.M Venue: Google Meet Link: https://meet.google.com/pba-nswy-fsc Speaker: Ms. Rejitha Raveendran (EE17D016) Guide: Dr. Arun D. Mahindrakar Abstract A time-varying (TV) optimization problem arises in many real-time applications, where either the objective function or the constraints change continuously with time. Consequently, the optimal points of the problem at each time instant form an optimizer trajectory and hence tracking the optimizer trajectory calls for the need to solve the TV optimization problem. The work focuses on a dynamical system approach to solve TV convex optimization problems by analyzing convergence of the trajectories of a dynamical system to the optimizer trajectory of the underlying optimization problem. Prediction-correction algorithms are the frequently used approaches to track the optimizer trajectories of an unconstrained and equality constrained TV convex optimization problem exponentially. In the first part of the work we modify the existing prediction-correction dynamical system to achieve the convergence within a predefined fixed-time from all initial conditions. Later we propose a prediction-correction based projected primal-dual dynamical system to track the optimizer trajectory of a TV inequality constrained convex optimization problem with strongly convex objective function. Finally, we consider the TV extended Fermat -Torricelli problem (eFTP) of minimizing the sum-of-squared distances to a finite number of nonempty, closed and convex TV sets to illustrate the applicability of the proposed projected dynamical system.
Feb
2022PhD SEMINAR TALK – II Title: Distributed Energy Management using Consensus Algorithms Date: 14.02.2022 Time: 2:30 P.M Venue: Google meet Google Meet Link : meet.google.com/qrx-hsxn-qti Speaker: Ms. Naina P M (EE15D024) Guide: Dr. K.S Swarup Abstract The presence and penetration of distributed generators (DGs), transforms the power system grid into a decentralized network. The Energy Management Operations, which employ conventional centralized algorithms may not be suitable, effective and accurate in the presence of DGs. The distributed economic dispatch operation using the leader-follower consensus algorithm was presented in Seminar talk I. In Seminar talk II, we present distributed energy management of virtual power plants (VPP) using two different consensus-based algorithms. The ‘proportional consensus algorithm’ uses a consensus gain function which is a function of the number of iterations and the maximum delay applied. In contrast, proportional integral consensus uses an additional integral term to reduce the effect of noise and communication delay. The effectiveness of the proposed work is illustrated in three different systems i.) five bus microgrid, ii.) 15 node VPP, and iii.) IEEE 39 bus system. This approach helps to avoid the VPP coordinator, thereby preserving the privacy of the VPP participants.
Feb
2022Ph.D Seminar Talk - II Title : Conducting Polymer Based Capacitive Humidity Sensing- Exploring PANI-SSA as an Active Material Speaker : Debajyoti Biswas (EE14D302) Date : February 11, 2022 (Friday) Time : 5-00 PM Venue : Online (Webex) Meeting link : https://iitmadras.webex.com/iitmadras/j.php?MTID=m8aeb7733b87d6b275f2175cf31394573 Guide : Dr. Soumya Dutta Co-Guide : Dr. Susy Varughese (CH) ABSTRACT The emerging concerns about environment protection, food storage requirements, medical and chemicals storage etc. have led to widespread research in humidity sensors. Modern hygrometers exploit changes in inherent material properties such as resistance or capacitance of the sensing materials upon exposure to humidity and predict ambient humidity based on these variations. In this context, the possibility of a resistive humidity sensor based on conducting polymer polyaniline (PANI) doped with a novel organic dopant, sulfosuccinic acid (SSA), has been explored in our laboratory. The PANI-SSA-based resistive sensor exhibited considerable humidity sensing with sensitivity of 1.7%/%RH, limit of detection (LOD) of 8%RH, and linearity covering a wide range of relative humidity (RH) from 30% to 80%. This presentation will primarily emphasize on the origin of the sensing response in terms of activation energy of electron hopping. Later on, novel device structures incorporating capacitive humidity sensing will be discussed. The first part of this talk will cover the detailed analysis of resistive humidity sensor using temperature- and RH-dependent conductivity study. The following part of the talk will include two all-organic device structures for humidity sensing: one is based on metal-insulator-metal (MIM) capacitor and the other is based on metal-insulator-semiconductor (MIS) capacitor using PANI-SSA as a humidity-sensitive conducting layer in both the structures. A noticeable increase in capacitance upon exposure to humidity in MIM based device structure resulted in sensitivity of 1.9%/ %RH, LOD of 0.591% and linearity covering 0% to 90%RH. In MIS capacitor-based devices, the entire capacitance-voltage (C-V) characteristics was observed to shift upon exposure to humidity. In addition, a change in relative C-V characteristics with respect to humidity level in the moderate accumulation regime can offer another degree of freedom in humidity measurement. The variation in capacitance in step with humidity was attributed to conducting polymer-assisted dipole alignment leading to a modification in dielectric property of the insulator in the device stack.
Feb
2022PhD Seminar - I Title: Understanding the Li-Ion Battery pack degradation in the field using field-test and lab-test data Date: 11.02.2022 Time: 03:00 P.M Venue: Google Meet, Joining Link: https://meet.google.com/wga-odnk-neu Speaker: Mr. Sushant Mutagekar (EE15D206) Guide: Dr. Ashok Jhunjhunwala Co-Guide: Dr. Prabhjot Kaur (CEO, CBEEV) Abstract Li-Ion cell manufacturers do provide some information in relation to their cell performance at different but constant charge/discharge rates and at different and constant temperatures, but hardly any of these can be extended to field conditions, where charge/discharge rates and temperature are continuously varying. This paper attempts to take the primary cell testing data gathered in a lab environment and create a first-order model for battery behavior in real-life conditions. At the same time, real-life data for battery packs in 3-wheeler vehicles are obtained on a continuous basis, when driven or when the battery is being charged. The battery packs used are of 1.25kWh without any forced cooling, and the vehicles are used in the city of Chennai in India, where in 24 hours, the average temperature varies from 25°C to 48°C in summers and from 18°C to 32°C during winters. The State of Health estimate obtained from the model (and the lab data for cells) is then compared with actual field data; further variances between the two and possible reasons for such variance are discussed. The results show that with some care, the battery behavior in real-life can be reasonably predicted.
Jan
2022Speaker: Mr. Sandeep V Nair Guide: Dr. Kamalesh Hatua ABSTRACT Permanent magnet synchronous motors (PMSMs) are widely used in electric vehicles (EV) due to their superior performance than induction machines. However, the operating speed range of these machines is restricted due to the limitations in the maximum battery voltage available in the vehicle. In this work, a dual three-phase PMSM with zero degree winding displacement (DTP0-PMSM) with six-step operation is proposed to achieve double the power rating and enhanced operating speed range compared to conventional 3-phase PMSMs. Dual three-phase PMSM also ensures zero circulating current even during six-step operation, reducing the copper loss compared to commonly used split-phase PMSMs. Specific applications like military vehicles, medium/heavy-duty trucks and off-road EVs demand high starting torque, high overload capability and wide constant power operating region. A dual 3-phase PMSM with unequal turns ratio (uneq0-PMSM) is also proposed to satisfy such special requirements with minimal overrating of the converter DC bus. An added advantage of uneq0-PMSM is that the shape of torque-speed characteristics can be varied during the design by changing the winding split ratio such that it perfectly suits the load requirement. The proposed concepts were experimentally validated on a custom-designed and fabricated 3kW PMSM prototype.
Jan
2022Network topology refers to the structure of a network represented as a graph with edges connected between a set of nodes. This topology information is essential for monitoring, analysis, optimization, and control of networks. In practice, at times, network topology may be unknown, only partially known or incorrectly reported. To address these issues, this work aims to solve three related problems with respect to topology of a class of networks known as conserved networks, from flow data: (i) topology identification - identifying the complete topology, (ii) topology completion - inferring unknown edges when the topology is partially known, and (iii) topology verification - verifying if there are any errors in the reported topology. The results and methodologies to address these problems are developed using multivariate data analysis, graph theory, and control theory. The algorithms developed based on the proposed methodologies are shown to be of polynomial time complexity. The theoretical findings are corroborated through simulations.
Jan
2022Speaker: Ms. Amulya (EE17D003) Guide: Dr. K.S Swarup Abstract Considering the importance of Cyber security studies on grid control systems, a Cyber-Security Vulnerability Analysis of multi-area load frequency control (MA-LFC) to cyber-attacks and a single-variate spectral analysis-based real-time detection technique was presented in Seminar Talk-I. In this talk, we present a multi-level attack detection and a real-time mitigation/response methodology. The proposed detection technique is a combination of spectral analysis and hypothesis testing for the early detection of attacks. The proposed detection method is i) fast, ii) adaptive to system changes, and is iii) scalable. The proposed work is illustrated for three different systems namely, i) 39 bus New-England test system, ii) IEEE 300-bus system, and iii) 1888 France RTE system to illustrate its effectiveness and scalability. Once the attacks are detected, it is important to devise effective response strategies to mitigate the impact of the attacks for stable system operation. In this talk, we present the gaps in attack response studies and define a methodology for real-time attack mitigation or response.
Jan
2022Speaker : S Ragul (EE16D031) Date : January 24, 2022 (Monday) Time : 3:00 pm Venue : Google Meet (meet.google.com/ejj-vaxb-ega) Guides : Dr. Debdutta Ray and Dr. Soumya Dutta ABSTRACT : In the first part of the talk, studies on polymer ferroelectric doping of graphene and reduced graphene oxide (rGO) thin films investigated using advanced electrostatic force probe microscopic (EFM) techniques will be presented. The solution spin casting at elevated temperature of polyvinylidene fluoride-trifluoro ethylene (PVDF-TrFE) into Nano-thin films enabled the low voltage poling desired for this study. The ferroelectric characterization of PVDF-TrFE using dynamic contact EFM and the induced doping effect in graphene using Kelvin probe force microscope (KPFM) will be discussed. These studies can be used to design ferroelectric-based FETs that can achieve lower sub-threshold slopes. Further, electrostatic doping induced Fermi-level tuning is used to map the nature of the density of states around the charge neutrality point in rGO thin films on SiO2/Si substrates. This was enabled by implementing a modified KPFM setup with in-situ gating-based time-resolved measurements. The results are in agreement with the DFT and tight-binding model-based calculations. Following this, the reflection of these electrostatic effects as parasitic in the electrical performance at the device level was explored. The experimental devices were fabricated with graphene/poly (3-hexyl thiophene) : phenyl [6,6] C61 butyric acid methyl ester bulk heterojunction blend/aluminum diode structure and a standard reference device made of indium tin oxide/poly (3,4-ethylene dioxythiophene) polystyrene sulfonate in place of graphene. On comparison, the graphene devices showed larger reverse bias currents. This was mapped to the intra-device self-gating effects of the graphene electrodes by the counter cathode under normal operations. Details of the analytical equations adopted to capture this effect, along with the numerical TCAD results, will be presented. The developed model can also be adapted to other devices like light-emitting diodes, perovskite-based devices, thin-film a-Si devices and FETs with appropriate modifications. In the last part of the talk, the development of an in situ measurement setup to study the dynamics of small molecular doping of graphene will be presented. The choice of Si-graphene heterojunction solar cell enabled the dynamic doping studies at the device level due to the availability of exposed graphene surfaces in their device architecture. The response of the fabricated heterojunction solar cells measured with this setup will be discussed.
Jan
2022Automatic Speech Recognition (ASR) on low-resource devices (like mobile phones, wearables, etc) is in high demand. Cloud-based ASRs have inherent dependencies on the quality and security of the network. And in several localized applications like ATMs or information kiosks, depending on high-speed internet merely for the ASR can be overkill. Alternatively, the offline speech decoder does the Viterbi decoding locally. The large speech models are stored in the external memory (DRAM) and based on the speech input; the most probable word sequence is obtained. However, the energy cost of single DRAM access is almost two orders more than single on-chip memory (OCM) access. On the other hand, the SRAMs that are used for OCM, are expensive than DRAMs in terms of their design. This precludes us from using large OCM. Hence, we need to reduce the number of DRAM accesses and at the same time reduce the OCM storage, without adversely affecting the performance or accuracy. In this work, the tradeoff between external memory and OCM is addressed using a binary search tree – max heap (BST-MH) data structure. This, along with the cache, uses much less OCM than the existing works with very minimal degradation of accuracy. Consequently, the need for external memory accesses is also reduced, while decoding the speech in real-time. The results of its implementation on the Xilinx FPGA along with the comparison with other works are discussed in this seminar
Jan
The minimum specific on-resistance achievable for a target breakdown voltage for a conventional p-n junction is limited by the unipolar material limit. Super junctions (SJs) were introduced to reduce the specific on-resistance below this limit. We derive material-independent closed-form solutions for the design of both an ideal as well as a practical SJ and illustrate the calculations for materials such as silicon, silicon carbide, gallium nitride, and diamond. The commercialization of SJ in silicon carbide material has been problematic due to the p-pillar fabrication difficulty. We propose Charge Sheet Super-junction (CSSJ) as a viable alternative to SJ in silicon carbide material as it eliminates the need for fabricating p-pillars while achieving up to 45 % lower specific on-resistance than SJ. We also derive analytical equations for the design of CSSJ for breakdown voltage in the range of 100 V to 10 kV.