An 8-bit asynchronous single-ended Successive Approximation Register (SAR) Analog-to-Digital Converter (ADC) for the realization of a perovskite photovoltaic (PV) sensing application is presented. Various techniques have been proposed to enhance speed and power efficiency. A bootstrap transmission gate-based sample and hold circuit is proposed to improve sampling linearity and reduce power consumption. The sample and hold circuit is configured to operate at a 90% duty cycle, hence, giving it enough time for the conversion. A two-stage latch dynamic comparator is implemented to ensure high-speed and low power consumption. The SAR ADC also implements an asynchronous SAR logic block, controlled by an internal clock, which eliminates the need for an external clock. The proposed SAR ADC was designed in a UMC 180nm CMOS technology. The SAR ADC is tested with a sinusoidal analog input of 500 mV at 10 kHz frequency and a 4 MHz input clock frequency. The SAR ADC consumes a power of 2.67 mW at a supply of 1.5 V. The proposed ADC is an important unit for application in the fast growing perovskite-based PV industry.

Given the widespread use of electric vehicles, they can be used in their idle state to provide services when connected to the grid or supply power in islanded operation using the bidirectional charger. Today, when connected to the grid, reversible chargers employ grid-following control, and when operating in islanded conditions, they use constant voltage constant frequency control. However, limitations exist for both control strategies. In this context, grid forming control can be used to operate in all cases. This paper showcases the potential of grid-forming control when applied to a reversible charger for different operation modes of electric vehicles.

In this paper, we discuss a general framework, namely Opportunistic Distributed Learning (ODL), which allows any node in the network to initiate a learning task while opportunistically leveraging local, unused distributed resources to collaboratively execute this task.

Progress in signal processing seems to converge towards perceptually inspired algorithms. This work proposes a novel signal processing framework, namely Cochlear Transform (CT), inspired by the active spiral cochlear mechanics. CT reflects how the spiral cochlea transforms the input stimuli to spatially supported and tonotopically organized basilar membrane responses, resulting in orthogonal cochlear modes with bio-inspired Time-Frequency Representation (TFR). We treat an input signal as a propagating wave that travels from the cochlear base to apex, along a path that takes a spiral shape from θ = 0◦ to θ = 990◦, respectively. More importantly, we showcase the generalizability of the CT to non-acoustic signals by modulating them to the highest sensitivity region of the cochlea.We try to achieve balance between bio-inspiration and adaptive signal processing. Accordingly, we demonstrate that CT attains important characteristics, such as filterbank with perceptual tuning when excited by White Gaussian Noise, noise robustness, and the cochlear decomposition orthogonality. Numerical examples of synthetic and real signals are given to demonstrate the robustness of CT and comparisons are made with Empirical Wavelet Transform (EWT) and the Variational Mode Decomposition (VMD). CT reveals new unseen and interpretable features not only of acoustic signals, but also real-life biological and seismic signals that stem from the active cochlear mechanics including acoustic distortion effects. This study is the first to demonstrate that mimicking both the active function and the structure of the human cochlear results inenhanced and interpretable TFR resolution, which allows the generalization of the approach to non-acoustic signals. CT resembles spectro-temporal receptive fields necessary for downstream perceptual auditory tasks with a range of application in data analysis, enhancing synergies between machine intelligence and bio-inspired perception. It also paves the way for bio-inspired feature engineering methods, including future CT extensions to applications for multivariate signals and 2D image processing.

Grid-forming control plays an essential role in the modernization of HV electrical transmission grids, particularly in mitigating challenges imposed by large voltage disturbances. In this context, the concept of virtual power has gained prominence as a potential approach that improves stability following such disturbances. This paper illustrates the use of virtual power method in enhancing the large disturbance stability given by the grid-forming control. An adaptive virtual impedance is also proposed to further improve system dynamics. The methods are evaluated through large disturbance stability analysis and time-domain simulations.

A document by Yatharth Ranjan . Click on the document to view its contents.

Here, we address the challenge of generalizing object detection and labeling in remote sensing data from one dataset to another. It is the variability in different altitudes, geographical variances, weather conditions, and object size across datasets that state-of-the-art DNN largely fails to generalize. Contrastive-based unsupervised domain adaptation attempts to bridge the gap by producing discriminating features for frames and instances across different datasets, and we have shown some progress on the domain alignment on local, global, and instance-aware features for remote sensing data. In this research, we propose using support-guided pseudo-labeling on the target domain instances to enable instance domain adaptation on the source and target datasets. Next, we introduce the contrastive loss function with multiple positive examples to make the model more generalized of the variable appearance of a particular class over images and domains. Also, we proposed debiased contrastive learning based on class probabilities to address the challenge of false negatives in the unsupervised framework. We show the advantages of the proposed model on satellite (DIOR and DOTA2.0) and drone (Visdrone and UAVDT) image datasets.

A gate sequence of single qubit transformations may be condensed into a single microwave pulse that maps a qubit from an initialized state directly into the desired state of the composite transformation. Here, machine learning is used to learn the parameterized values for a single driving pulse associated with a  transformation of three sequential gate operations on a qubit. This implies that future quantum circuits may contain roughly a third of the number of single qubit operations performed, greatly reducing the problems of noise and decoherence. There is a potential for even greater condensation and efficiency using the methods of quantum machine learning.

The integration of renewable energy resources is a prevailing trend in modern power systems. Wind and solar energy are key among these renewables, linked to the grid via power electronic converters. These converters require precise synchronization of their output with the grid, achieved through a phase-locked loop (PLL). The PLL’s ability to accurately track the grid voltage phase directly impacts synchronization stability. Moreover, stability relies on the converter’s controller, loads, and grid stiffness. The inclusion of these cross-linked dynamics culminates in a complex analytical model. However, commonly used reduced-order models often omit load dynamics, limiting the obtained insights. Addressing these limitations, this paper introduces a comprehensive method to model a 3- phase grid-tied inverter with an SRF-PLL and local loads, utilizing dynamic phasors. The approach evaluates the impact of system parameters, especially load configuration, on synchronization stability via eigenvalue analysis of the linearized state-space model. This model is apt for investigating weak grids and islanded distributed generation (DG) with diverse local loads. The model showcases stability boundaries, aiding the design of robust renewable energy systems.

This research paper investigates free space optic (FSO) system by integrating orbital angular momentum (OAM) and polarization division multiplexing (PDM) techniques resulting in the data rate enhancement. Simulation of OAM-PDM-FSO transmission is analyzed by using Optisystem. A successful transmission of 80 Gb/s overall capacity is investigated by means of individual OAM modes (LG0,0, LG0,13, LG0,40, and LG0,80), enhancing the spectral efficiency of the system under different climatic conditions. System performance is studied for various modulation schemes including return-to-zero (RZ), non-return-to zero (NRZ) and offset-quadrature-phase shift keying (OQPSK). The best performance has been shown by means of OQPSK modulation technique. 8x10 Gb/s transmission under the influence of various external weather conditions such as rain, fog and dust is reported. A maximum FSO link range of 6 km is realized under clear weather conditions whereas a minimum FSO range of 1.25 km is achieved under heavy dust conditions. Performance of proposed hybrid OAM-PDM system is evaluated at different link lengths in terms of Q-factor, log (BER), received power and eye diagrams.

In this paper, we investigate the effectiveness of a diffusion model to generate an augmented labeled dataset of synthetic images. The creation of synthetic images becomes indispensable in condition monitoring applications where capturing a diverse range of damages for training a supervised deep-learning model is impracticable or infeasible. To evaluate the potential of this technique, we explore the case of damage classification in synthetic fiber ropes. The approach involves training and fine- tuning a stable diffusion model to create synthetic photorealistic images, representing three distinct types of damages commonly found in fiber ropes. We assessed the efficacy of our method by training and testing the Resnet50 classifier on the synthetic dataset. Our results show that the classifier’s performance remains stable when up to 75% of the images used for training are synthetically generated. This outcome suggests that synthetically generated images produced by a finely-tuned diffusion model possess sufficient quality to serve as a viable addition to real images in data augmentation.

As computational demands continue to evolve in  the modern era, the choice of hardware architecture plays a  pivotal role in optimizing the performance of compute-intensive  applications. This research paper delves into the exploration  and comparison of three prominent hardware architectures:  x86, ARM, and RISC-V, within the context of compute-intensive  applications. The study begins with a comprehensive overview  of these architectures, highlighting their distinctive features,  and strengths. Subsequently, we investigate their suitability and  adaptability in diverse compute-intensive workloads. Our  analysis encompasses a wide spectrum of parameters, including  computational throughput, power efficiency, scalability, and  architectural flexibility. We scrutinize the architectural  intricacies that impact the execution of compute-intensive tasks,  shedding light on both the advantages and limitations of each  architecture. We used the gem5 simulator to compare these  Instruction Set Architectures (ISA). We run different  benchmarks on gem5 with different ISA and different  configurations and compare the result. Based on these results  we predict which architecture is better in which scenario. Gem5  is not a cycle accurate simulator but it’s a model accurate. In  conclusion, “Exploring Architectural Variations: x86, ARM,  and RISC-V in Compute-Intensive Applications” offers a  comprehensive insight into the nuances of hardware selection  for compute-intensive workloads. Our findings aid system  architects, researchers, and technology enthusiasts in making  informed decisions about the most suitable architectural choice  for their specific compute-intensive applications, ultimately  contributing to advancements in computational performance  and efficiency

Transcranial magnetic stimulation (TMS) is a non-invasive method for treating neurological and psychiatric disorders. It is being tested as an experimental treatment for patients with mild to moderate traumatic brain injuries (mTBI). Due to the complex, heterogeneous composition of the brain, it is difficult to determine if targeted brain regions receive the correct amount of electric field (E-field) induced by the TMS coil. E-field distributions can be calculated by running time-consuming finite element analysis (FEA) simulations of TMS on patient head models. Using machine learning, the E-field can be predicted in real-time. Our prior work used a Deep Convolutional Neural Network (DCNN) to predict the E-field in healthy patients. This study applies the same DCNN to mTBI patients and investigates how model depth and color space of E-field images affect model performance. Nine  DCNNs were created using combinations of 3, 4, or 5 encoder and decoder blocks with the color spaces RGB, LAB, and YCbCr. As depth increased, training and testing peak signal-to-noise ratios (PSNR) increased and mean squared errors (MSE) decreased. The depth 5 YCbCr model had the highest training and testing PSNRs of 34.77 dB and 29.08 dB and lowest training and testing MSEs of 3.335 * 10-4 and 1.237 * 10-3 respectively. Compared to the model in our prior work, models of depth 5 have higher testing PSNRs and lower MSEs and, except for RGB. Thus, DCNNs with depth 5 and alternative color spaces, despite losing information through color space conversions, resulted in higher PSNRs and lower MSEs.

This letter describes a lattice encoder for the NIST-recommended post-quantum encryption algorithm: Kyber. The key idea is to refine the analysis of Kyber decoding noise. We prove that Kyber decoding noise can be bounded by a sphere. This result shows the Kyber encoding problem is essentially a sphere packing in a hypercube. Lattice codes are then constructed to ensure denser packing and a lower decryption failure rate (DFR). Compared with the original Kyber encoder, the communication cost is reduced by up to 32.6%, the DFR is reduced by a factor of up to 2^{85}. The security arguments remain the same.

Remote sensing has enabled large-scale crop classification to understand agricultural ecosystems and estimate production yields. Since a few years, machine learning is increasingly used for automated crop classification. However, in most approaches the novel algorithms are applied to custom datasets containing information of few crop fields covering a small region and this often leads to models that lack generalization capability.  In this work, we propose a multi-modal contrastive self-supervised learning approach to obtain a pre-trained model for crop-classification without the use of labeled data. Such multi-modal self-supervised learning exploits the synergies of different data sources to obtain a richer representation of the data. We build our analysis by adapting the DENETHOR dataset developed for a part of Eastern Germany to our usecase.  We use the publicly available Sentinel2 and commercial Planetscope data. While Sentinel2 has higher spectral resolution, Planetscope has finer spatial resolution. For an end-user application, only one source is required. In this work, we analyze and compare the performance of our multi-modal self-supervised model against the uni-modal contrastive self-supervised model using the SCARF algorithm. In addition, we also compare our multi-modal self-supervised model with a supervised model. We find that our multi-modal pre-trained model surpasses the uni-modal and supervised models in almost all test cases.

RF photonic transversal signal processors, which combine reconfigurable electrical digital signal processing and high-bandwidth photonic processing, provide a powerful solution for achieving adaptive high-speed information processing. Recent progress in optical microcomb technology provides compelling multi-wavelength sources with compact footprint, yielding a variety of microcomb-based RF photonic transversal signal processors implemented by either discrete or integrated components. Although operating based on the same principle, processors in these two forms exhibit distinct performance. This letter presents a comparative investigation into their performance. First, we compare the performance of state-of-the-art processors, focusing on the processing accuracy. Next, we analyze various factors that contribute to the performance differences, including tap number and imperfect response of experimental components. Finally, we discuss the potential for future improvement. These results provide a comprehensive comparison of microcomb-based RF photonic transversal signal processors implemented using discrete and integrated components and provide insights for their future development.

This research delves into user preferences concerning structured (the subject follows an exact script to command the robot) and unstructured (the subject commands the robot in a conversational way) robot interaction through natural spoken language. Data was gathered from 30 adult participants who completed two distinct tasks involving both structured and unstructured commands. The study examines correlations between robot errors and user perceptions, as well as how past or present failures impact participants’ perception of robot utility. Three hypotheses are formulated, and the paper offers a comprehensive overview of the study’s aims, methodologies, and principal findings, which were ascertained using paired t-Test and Kendall-Tau correlations. The study indicates that participants showed a preference for the unstructured task in contrast to the structured one. Analysis of the data revealed interesting correlations between the user perception of the robot and the robot errors. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

A document by Prodipto Das . Click on the document to view its contents.