SEARCH · Search NASA

Results for “bandwidth optimization”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 649 records · Page 36

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Strain-concentration for fast, compact photonic modulation and non-volatile memory

A critical figure of merit (FoM) for electro-optic (EO) modulators is the transmission change per voltage, d T / d V . Conventional approaches in wave-guided modulators maximize d T / d V via a high EO coefficient or longer light-material interaction lengths but are ultimately limited by material losses and nonlinearities. Optical and RF resonances improve d T / d V at the cost of spectral non-uniformity, especially for high- Q optical cavity resonances. Here, we introduce an EO modulator based on piezo-strain-concentration of a photonic crystal cavity to address both trade-offs: (i) it eliminates the trade-off between d T / d V and waveguide loss—i.e., enhancement of the resonance tuning efficiency d v c / d V for the fixed EO coefficient, waveguide length, and cavity Q —and (ii) at high DC strains it exhibits a non-volatile (NV) cavity tuning Δ v c ,NV for passive memory and programming of multiple devices into resonance despite fabrication variations. The device is fabricated on a scalable silicon nitride-on-aluminum nitride platform. We measure d v c / d V =177±1MHz/V, corresponding to Δ v c =40±0.32GHz for a voltage spanning ±120V with an energy consumption of δ U /Δ v c =0.17nW/GHz. The modulation bandwidth is flat up to ω BW,3dB /2 π =3.2±0.07MHz for broadband DC-AC and 142±17MHz for resonant operation near a 2.8 GHz mechanical resonance. Optical extinction up to 25 dB is obtained via Fano-type interference. Strain-induced beam-buckling modes are programmable under a “read-write” protocol with a continuous, repeatable tuning range of 5±0.25GHz, allowing for storage and retrieval, which we quantify with mutual information of 2.4 bits and a maximum non-volatile excursion of 8 GHz. Using a full piezo-optical finite-element-model (FEM) we identify key design principles for optimizing strain-based modulators and chart a path towards achieving performance comparable to lithium niobate-based modulators and the study of high strain physics on-chip.

Wen, Y. Henry (ORCID:0009000685423628)↗

NASA NTRS2016

Phase-Retrieval Uncertainty Estimation and Algorithm Comparison for the JWST-ISIM Test Campaign

Phase retrieval, the process of determining the exitpupil wavefront of an optical instrument from image-plane intensity measurements, is the baseline methodology for characterizing the wavefront for the suite of science instruments (SIs) in the Integrated Science Instrument Module (ISIM) for the James Webb Space Telescope (JWST). JWST is a large, infrared space telescope with a 6.5-meter diameter primary mirror. JWST is currently NASA's flagship mission and will be the premier space observatory of the next decade. ISIM contains four optical benches with nine unique instruments, including redundancies. ISIM was characterized at the Goddard Space Flight Center (GSFC) in Greenbelt, MD in a series of cryogenic vacuum tests using a telescope simulator. During these tests, phase-retrieval algorithms were used to characterize the instruments. The objective of this paper is to describe the Monte-Carlo simulations that were used to establish uncertainties (i.e., error bars) for the wavefronts of the various instruments in ISIM. Multiple retrieval algorithms were used in the analysis of ISIM phase-retrieval focus-sweep data, including an iterativetransform algorithm and a nonlinear optimization algorithm. These algorithms emphasize the recovery of numerous optical parameters, including low-order wavefront composition described by Zernike polynomial terms and high-order wavefront described by a point-by-point map, location of instrument best focus, focal ratio, exit-pupil amplitude, the morphology of any extended object, and optical jitter. The secondary objective of this paper is to report on the relative accuracies of these algorithms for the ISIM instrument tests, and a comparison of their computational complexity and their performance on central and graphical processing unit clusters. From a phase-retrieval perspective, the ISIM test campaign includes a variety of source illumination bandwidths, various image-plane sampling criteria above and below the Nyquist- Shannon critical sampling value, various extended object sizes, and several other impactful effects.

Design Analysis↗

NASA NTRS1984

Two-dimensional computer simulation of EMVJ and grating solar cells under AMO illumination

A computer program, SCAP2D (Solar Cell Analysis Program in 2-Dimensions), is used to evaluate the Etched Multiple Vertical Junction (EMVJ) and grating solar cells. The aim is to demonstrate how SCAP2D can be used to evaluate cell designs. The cell designs studied are by no means optimal designs. The SCAP2D program solves the three coupled, nonlinear partial differential equations, Poisson's Equation and the hole and electron continuity equations, simultaneously in two-dimensions using finite differences to discretize the equations and Newton's Method to linearize them. The variables solved for are the electrostatic potential and the hole and electron concentrations. Each linear system of equations is solved directly by Gaussian Elimination. Convergence of the Newton Iteration is assumed when the largest correction to the electrostatic potential or hole or electron quasi-potential is less than some predetermined error. A typical problem involves 2000 nodes with a Jacobi matrix of order 6000 and a bandwidth of 243.

Gray, J. L.↗

DOE OSTI2016

SU-F-I-14: 3D Breast Digital Phantom for XACT Imaging

Purpose: The X-ray induced acoustic computed tomography (XACT) is a new imaging modality which combines X-ray contrast and high ultrasonic resolution in a single modality. Using XACT in breast imaging, a 3D breast volume can be imaged by only one pulsed X-ray radiation, which could dramatically reduce the imaging dose for patients undergoing breast cancer screening and diagnosis. A 3D digital phantom that contains both X-ray properties and acoustic properties of different tissue types is indeed needed for developing and optimizing the XACT system. The purpose of this study is to offer a realistic breast digital phantom as a valuable tool for improving breast XACT imaging techniques and potentially leading to better diagnostic outcomes. Methods: A series of breast CT images along the coronal plane from a patient who has breast calcifications are used as the source images. A HU value based segmentation algorithm is employed to identify breast tissues in five categories, namely the skin tissue, fat tissue, glandular tissue, chest bone and calcifications. For each pixel, the dose related parameters, such as material components and density, and acoustic related parameters, such as frequency-dependent acoustic attenuation coefficient and bandwidth, are assigned based on tissue types. Meanwhile, other parameters which are used in sound propagation, including the sound speed, thermal expansion coefficient, and heat capacity are also assigned to each tissue. Results: A series of 2D tissue type image is acquired first and the 3D digital breast phantom is obtained by using commercial 3D reconstruction software. When giving specific settings including dose depositions and ultrasound center frequency, the X-ray induced initial pressure rise can be calculated accordingly. Conclusion: The proposed 3D breast digital phantom represents a realistic breast anatomic structure and provides a valuable tool for developing and evaluating the system performance for XACT.

60 APPLIED LIFE SCIENCES↗

NASA NTRS2011

Laser Truss Sensor for Segmented Telescope Phasing

A paper describes the laser truss sensor (LTS) for detecting piston motion between two adjacent telescope segment edges. LTS is formed by two point-to-point laser metrology gauges in a crossed geometry. A high-resolution (<30 nm) LTS can be implemented with existing laser metrology gauges. The distance change between the reference plane and the target plane is measured as a function of the phase change between the reference and target beams. To ease the bandwidth requirements for phase detection electronics (or phase meter), homodyne or heterodyne detection techniques have been used. The phase of the target beam also changes with the refractive index of air, which changes with the air pressure, temperature, and humidity. This error can be minimized by enclosing the metrology beams in baffles. For longer-term (weeks) tracking at the micron level accuracy, the same gauge can be operated in the absolute metrology mode with an accuracy of microns; to implement absolute metrology, two laser frequencies will be used on the same gauge. Absolute metrology using heterodyne laser gauges is a demonstrated technology. Complexity of laser source fiber distribution can be optimized using the range-gated metrology (RGM) approach.

Liu, Duncan T.↗

DOE OSTI2015

Coherence Conversion for Optimized Resolution in Optical Measurements - Example of Femtosecond Time Resolution Using the Transverse Coherence of 100-Picosecond X-Rays

A way is proposed to obtain a femtosecond time resolution over a picosecond range in x-ray spectroscopic measurements where the light source and the detector are much slower than that. It is based on the invariance of the modulus of the Fourier transform to object translations. The method geometrically correlates time in the sample with x-ray amplitudes over a spatial coordinate, and then takes the optical Fourier transform through far-field diffraction. Thus, explicitly time-invariant intensities that encode the time evolution of the sample can be measured with a slow detector. This corresponds to a phase-space transformation that converts the transverse coherence to become effective in the longitudinal direction. Because synchrotron-radiation sources have highly anisotropic coherence properties with about $10^5$ longitudinal electromagnetic-field modes at 1 eV bandwidth, but only tens to hundreds transverse modes, coherence conversion can drastically improve the time resolution. Reconstruction of the femtosecond time evolution in the sample from the Fourier intensities is subject to a phase ambiguity that is well-known in crystallography. However, a way is presented to resolve it that is not available in that discipline. Finally, data from a demonstration experiment are presented. The same concept can be used to obtain attosecond time resolution with an x-ray free-electron laser.

Adams, Bernhard W.↗

NASA NTRS1980

Analytical design of multispectral sensors

An optimal design based on the criterion of minimum mean square representation error using the Karhunen-Loeve expansion was developed to represent the spectral response functions from a stratum based upon a stochastic process scene model. From the overall pattern recognition system perspective, the effect of the representation accuracy on a typical performance criterion (the probability of correct classification) is investigated. The optimum sensor design provides a standard against which practical (suboptimum) operational sensors can be compared. An example design is provided and its performance is illustrated. Although developed primarily for the purpose of sensor design, the procedure has potential for making important contributions to scene understanding. Spectral channels which have narrow bandwidths relative to current sensor systems may be necessary to provide adequate spectral representation and improved classification performance.

Wiersma, D. J.↗

DOE OSTI2024

Optimal Spectral Resolution for Infrared Studies of Solids and Liquids

Due to a legacy originating in the limited capability of early computers, the spectroscopic resolution used in Fourier transform infrared spectroscopy and other systems has largely been implemented using only powers of two for more than 50 years. In this study, we investigate debunking the spectroscopic lore of, e.g., using only 2, 4, 8, or 16 cm −1 resolution and determine the optimal resolution in terms of both (i) a desired signal-to-noise ratio and (ii) efficient use of acquisition time. The study is facilitated by the availability of solids and liquids reference spectral data recorded at 2.0 cm −1 resolution and is based on an examination in the 4000–400 cm −1 range of 61 liquids and 70 solids spectra, with a total analysis of 4237 peaks, each of which was also examined for being singlet/multiplet in nature. Of the 1765 liquid bands examined, only 27 had widths <5 cm −1 . Of the 2472 solid bands examined, only 39 peaks have widths <5 cm −1 . For both the liquid and solid bands, a skewed distribution of peak widths was observed: For liquids, the mean peak width was 24.7 cm −1 but the median peak width was 13.7 cm −1 , and, similarly, for solids, the mean peak width was 22.2 cm −1 but the median peak width was 11.2 cm −1 . While recognizing other studies may differ in scope and limiting the analysis to only room temperature data, we have found that a resolution to resolve 95% of all bands is 5.7 cm −1 for liquids and 5.3 cm −1 for solids; such a resolution would capture the native linewidth (not accounting for instrumental broadening) for 95% of all the solids and liquid bands, respectively. After decades of measuring liquids and solids at 4, 8, or 16 cm −1 resolution, we suggest that, when accounting only for intrinsic linewidths, an optimized resolution of 6.0 cm −1 will capture 91% of all condensed-phase bands, i.e., broadening of only 9% of the narrowest of bands, but yielding a large gain in signal-to-noise with minimal loss of specificity.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

NASA NTRS2008

JPEG 2000 Encoding with Perceptual Distortion Control

An alternative approach has been devised for encoding image data in compliance with JPEG 2000, the most recent still-image data-compression standard of the Joint Photographic Experts Group. Heretofore, JPEG 2000 encoding has been implemented by several related schemes classified as rate-based distortion-minimization encoding. In each of these schemes, the end user specifies a desired bit rate and the encoding algorithm strives to attain that rate while minimizing a mean squared error (MSE). While rate-based distortion minimization is appropriate for transmitting data over a limited-bandwidth channel, it is not the best approach for applications in which the perceptual quality of reconstructed images is a major consideration. A better approach for such applications is the present alternative one, denoted perceptual distortion control, in which the encoding algorithm strives to compress data to the lowest bit rate that yields at least a specified level of perceptual image quality. Some additional background information on JPEG 2000 is prerequisite to a meaningful summary of JPEG encoding with perceptual distortion control. The JPEG 2000 encoding process includes two subprocesses known as tier-1 and tier-2 coding. In order to minimize the MSE for the desired bit rate, a rate-distortion- optimization subprocess is introduced between the tier-1 and tier-2 subprocesses. In tier-1 coding, each coding block is independently bit-plane coded from the most-significant-bit (MSB) plane to the least-significant-bit (LSB) plane, using three coding passes (except for the MSB plane, which is coded using only one "clean up" coding pass). For M bit planes, this subprocess involves a total number of (3M - 2) coding passes. An embedded bit stream is then generated for each coding block. Information on the reduction in distortion and the increase in the bit rate associated with each coding pass is collected. This information is then used in a rate-control procedure to determine the contribution of each coding block to the output compressed bit stream.

Watson, Andrew B.↗

DOE OSTI2019

Outage probability and average BER estimation of FSO system employing wavelength diversity

Free space optical (FSO) communication has gained significant attention among researchers for high bandwidth and secured data transfer applications. However, its performance is limited by the local atmospheric conditions. This paper focusses on an FSO system employing a diversity method to compensate signal loss occurred due to atmospheric turbulence conditions varying from moderate to strong. We derive a closed-form expression for two important parameters, viz.: average bit error rate (BER) and outage probability over gamma–gamma distribution channel. The numerical results for different combining schemes are presented in this paper. Enhancement in reliability and performance of FSO communication system with wavelength diversity is investigated. Effect of different wavelengths 850 nm, 1310 nm and 1550 nm on average BER is analysed and minimum BER is achieved for optimal combining at 1550 nm at diversity order R=3.

42 ENGINEERING↗

DOE OSTI2016

A Business Case for Nuclear Plant Control Room Modernization

This paper presents a generic business case for implementation of technology that supports Control Room Modernization (CRM). The analysis presented in two forms; 1) a standalone technology upgrade, and 2) a technology upgrade that is built upon and incremental to a prior business case created for Mobile Work Packages (MWP). The business case contends that advanced communication and networking and analytical technologies will allow NPP to conduct control room operations with improved focus by reducing human factors and redundant manpower, and therefore operate with fewer errors. While some labor savings can be harvested in terms of overtime, the majority of savings are demonstrated as reduced time to take the plant off line and bring back on line in support of outages. The benefits are quantified to a rough order of magnitude that provides directional guidance to NPPs that are interested in developing a similar business case. This business case focuses on modernization of the operator control room and does not consider a complete overhaul and modernization of a plants instrument and control systems. While operators may be considering such an investment at their plants, the sizable capital investment required is not likely supported by a cost/benefit analysis alone. More likely, it is driven by obsolescence and reliability issues, and requires consideration of mechanical condition of plant systems, capital depreciation, financing, relicensing and overall viability of the plant asset over a 20-year horizon in a competitive market. Prior studies [REF] have indicated that such a modernization of plant I&C systems, alone or as part of a larger modernization effort, can yield very significant reductions in O&M costs. However, the depth of research and analysis required to develop a meaningful business case for a plant modernization effort is well beyond the scope of this study. While CRM as considered in this study can be easily integrated as part of grander plant modernization effort, it can also be considered as a stand-alone project, implemented as a supervisorial layer over and above existing systems. CRM is enabled by a suite of technologies, which are described in further detail in the body of this report. They include: • Modernized control room with interactive displays • High-bandwidth wireless networks • Mobile devices • Component identification technology • Mobile wireless video cameras • Smart or task based operator displays • Computer-Based Procedures (CBP)/Automated Mobile Work Packages (MWP) • Intelligent plant configuration • Advanced data analytics An analysis was conducted to determine how these technologies might impact outage operations at an NPP. The analysis concluded that outage management capabilities would be improved with application of these technologies, and result in annual benefits for the plant. These improved capabilities were identified and described in further detail in this report. • Reduction in need redundant manpower (verification, etc) • Improved critical path shutdown and restart duration during outages • Reduction in human factors and errors • Remote watches • Paperless outage coordination • Bulk work optimization

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

NASA NTRS2019

Tunable mid-wave infrared spectral filters based on GexSbyTez for multispectral imaging

The mid-wave infrared (MWIR) spectrum contains a wealth of invaluable information including the spectral ‘fingerprint’ of many chemical species and has applications in remote sensing and astronomical imaging. Traditionally, filtering for the MWIR is achieved by means of passive multilayer interference (dichroic) filters, Fabry-Perot-based micro-electro-mechanical system (MEMS) filters, liquid crystal tunable filters, and focal plane array (FPA) filters. For accurate multispectral imaging applications, these approaches suffer from various limitations such as: having moving parts; exhibiting slow response times, and; having limited spectral bandwidth / resolution. Recently, there has been significant interest toward ‘active’ spectral imaging technologies, whereby the ability to provide electrically tunable narrowband filtering—spanning the entire MWIR—is highly desirable. In this work we introduce a new spectral imaging technology, namely actively tunable optical transmission filters using the phase-change material (PCM) GexSbyTez (GST). The GST exhibits a large, reversible change in its refractive index across the MWIR upon a phase transition (from amorphous to crystalline). This refractive index modulation governs the filter’s optical characteristics. Through an optical stimulus or applied voltage—which changes the GST state from amorphous to crystalline—one can actively tune the filter with MHz speed. We incorporate GST into optimized guided mode resonance (GMR) and plasmonic nanohole array (PNA) device architectures, enabling <10nm spectral resolution and tunable operation across 3~5µm. Our proposed PCM-MWIR filter will be able to extract the maximum amount of ‘useful’ information within the atmosphere for remote Earth sensing measurements at operational speeds orders of magnitude faster than current airborne-based sensors. Moreover, it may enable affordable SmallSat-based MWIR instrumentation which is complimentary to other observation systems. The MISSE (Materials International Space Station Experiment-Flight Facility) experiment has been proposed as a testbed for a PCM-based tunable MWIR filter module to allow exposure of the module to the low Earth orbit space environment. This will provide valuable data regarding the robustness of the filter to withstand the radiation and atomic oxygen environment and allow assessment of the technology for use in space applications.

Kim, Hyun Jung↗

DOE OSTI2026

Paradigm for universal quantum information processing with integrated acousto-optic frequency beamsplitters

Frequency-bin encoding offers tremendous potential in quantum photonic information processing, in which a single waveguide can support hundreds of lightpaths in a naturally phase-stable fashion. This stability, however, comes at a cost: arbitrary unitary operations can be realized by cascaded electro-optic phase modulators and pulse shapers, but require nontrivial numerical optimization for design and have thus far been limited to discrete tabletop components. In this article, we propose, formalize, and computationally evaluate a new paradigm for universal frequency-bin quantum information processing using acousto-optic scattering processes between distinct transverse modes. We show that controllable phase matching in intermodal processes enables 2 × 2 frequency beamsplitters and transverse-mode-dependent phase shifters, which together comprise cascadable FRequency-transverse-mODe Operations (FRODOs) that can synthesize any unitary via analytical decomposition procedures. Modeling the performance of both random gates and discrete Fourier transforms, we demonstrate the feasibility of high-fidelity quantum operations with existing integrated photonics technology, highlighting prospects of parallelizable operations achieving 100% bandwidth utilization. Our approach is realizable with CMOS technology, opening the door to scalable on-chip quantum information processing in the frequency domain.

Lukens, Joseph M. [Purdue Univ., West Lafayette, I↗

NASA NTRS2004

Traveling Wave Tube (TVT) RF Power Combining Demonstration for use in the Jupiter Icy Moons Orbiter (JIMO)

The Jupiter Icy Moons Orbiter (JIMO) is set to launch between the years 2012 and 2015. It will possibly utilize a nuclear reactor power source and ion engines as it travels to the moons of Jupiter. The nuclear reactor will produce hundreds of kilowatts of power for propulsion, communication and various scientific instruments. Hence, the RF amplification devices aboard will be able to operate at a higher power level and data rate. The initial plan for the communications system is for an output of 1000 watts of RF power, a data rate of at least 10 megabits a second, and a frequency of 32 GHz. A higher data rate would be ideal to fully utilize the instruments aboard JIMO. At NASA Glenn, one of our roles in the JIMO project is to demonstrate RF power combining using multiple traveling wave tubes (TWT). In order for the power of separate TWT s to be combined, the RF output waves from each must be in-phase and have the same amplitude. Since different tubes act differently, we had to characterize each tube using a Network Analyzer. We took frequency sweeps and power sweeps to characterize each tube to ensure that they will behave similarly under the same conditions. The 200 watt Dornier tubes had been optimized to run at a lower power level (120 watts) for their extensive use in the ACTS program, so we also had to experiment with adjusting the voltage settings on several internal components (helix, anode, collector) of the tubes to reach the full 200 watt potential. from the ACTS program. Phase shifters and power attenuators were placed in the waveguide circuit at the inputs to the tubes so that adjustments could be made individually to match them exactly. A magic tee was used to route and combine the amplified electromagnetic RF waves on the tube output side. The demonstration of 200 watts of combined power was successful with efficiencies greater than 90% over a 500 MHz bandwidth. The next step will be to demonstrate the use of three amplifiers using two magic tees by adding a 200 watt Dornier tube to the Varian and Logimetrics combined setup for a total of 400 watts. After that we will use two 200 watt Dorniers for 400 watts and eventually four 200 watt Dornier tubes to demonstrate 800 watts. After demonstrating the success of power combining, we will need to verify the integrity of a modulated signal sent through the combined tubes. The purpose will be to see what effects separating and recombining will have on the modulated signal and also what effect it will have on combining efficiency. A Bit Error Rate (BER) will be determined by a Bit Error Rate Tester (BERT) by comparing the random information it transmits to what it receives back. The process began with two 100 watt tubes, a Varian and a Logimetrics, salvaged

Downey, Joseph A.↗

NASA NTRS1998

Architectural Optimization of Digital Libraries

This work investigates performance and scaling issues relevant to large scale distributed digital libraries. Presently, performance and scaling studies focus on specific implementations of production or prototype digital libraries. Although useful information is gained to aid these designers and other researchers with insights to performance and scaling issues, the broader issues relevant to very large scale distributed libraries are not addressed. Specifically, no current studies look at the extreme or worst case possibilities in digital library implementations. A survey of digital library research issues is presented. Scaling and performance issues are mentioned frequently in the digital library literature but are generally not the focus of much of the current research. In this thesis a model for a Generic Distributed Digital Library (GDDL) and nine cases of typical user activities are defined. This model is used to facilitate some basic analysis of scaling issues. Specifically, the calculation of Internet traffic generated for different configurations of the study parameters and an estimate of the future bandwidth needed for a large scale distributed digital library implementation. This analysis demonstrates the potential impact a future distributed digital library implementation would have on the Internet traffic load and raises questions concerning the architecture decisions being made for future distributed digital library designs.

Biser, Aileen O.↗

DOE OSTI2019

The Institute for Sustained Performance, Energy, and Resilience

The Green Queue framework is a framework designed to automate the development and deployment of customized architecture- and application-aware power savings recipes for large-scale HPC applications achieving up to 21% and 32% energy savings on HPC production applications run at scale. Additional work focused on the memory sub-system involved our methodology that uses application and machine characterization information to build predictive machine learning models that can accurately quantify phase-level sensitivity to the reduced per core memory bandwidth resulting from changes in the memory bus frequency to reduce the power. We evaluated the predictive capability of the model on real applications and validated them at a fine grain level by looking at 43 individual computational phases or application hotspots as well as the whole application. For more than 91% of the application hotspots, the prediction error is less than 10% (15) . Building from these validated performance and power models collaborations among SUPER team members developed an automated end-to-end system to reduce the complexity of developing and deploying machine learning models for performance, power, and energy. The new framework Automatic Multi-objective Modeling with Machine Learning (AutoMOMML) enabled multi-objective optimizations (power and performance) for HPC workloads.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2019

Analyzing the suitability of contemporary 3D-stacked PIM architectures for HPC scientific applications

Scaling off-chip bandwidth is challenging due to fundamental limitations, such as a fixed pin count and plateauing signaling rates. Recently, vendors have turned to 2.5D and 3D stacking to closely integrate system components. Interestingly, these technologies can integrate a logic layer under multiple memory dies, enabling computing capability inside a memory stack. This trend in stacking is making PIM architectures commercially viable. In this work, we investigate the suitability of offloading kernels in scientific applications onto 3D stacked PIM architectures. We evaluate several hardware constraints resulted from the stacked structure. We perform extensive simulation experiments and in-depth analysis to quantify the impact of application locality in TLBs, data caches, and memory stacks. Our results also identify design optimization areas in software and hardware for HPC scientific applications.

Peng, Ivy Bo↗

DOE OSTI2023

Digital Modeling on Large Kernel Metamaterial Neural Network

Deep neural networks (DNNs) utilized recently are physically deployed with computational units (e.g., CPUs and GPUs). Such a design might lead to a heavy computational burden, significant latency, and intensive power consumption, which are critical limitations in applications such as Internet of Things (IoT), edge computing, and usage of drones. Recent advances in optical computational units (e.g., metamaterial) have shed light on energy-free and light-speed neural networks. However, the digital design of the metamaterial neural network (MNN) is fundamentally limited by its physical limitations, such as precision, noise, and bandwidth during fabrication. Moreover, the unique advantages of MNN’s (e.g., light-speed computation) are not fully explored via standard 3×3 convolution kernels. In this paper, we propose a novel large kernel metamaterial neural network (LMNN) that maximizes the digital capacity of the state-of-the-art (SOTA) MNN with model re-parametrization and network compression, while also considering the optical limitation explicitly. The new digital learning scheme can maximize the learning capacity of MNN while modeling the physical restrictions of meta-optics. With the proposed LMNN, the computation cost of the convolutional front-end can be offloaded to fabricated optical hardware. The experimental results on two publicly available datasets demonstrate that the optimized hybrid design improved classification accuracy while reducing computational latency. In conclusion, the development of the proposed LMNN is a promising step towards the ultimate goal of energy-free and light-speed AI.

97 MATHEMATICS AND COMPUTING↗