SEARCH · Engineering Papers

Results for “input”

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 37 records · Page 2

All sources NASA NTRS DOE OSTI

NASA NTRS2011

High-Voltage-Input Level Translator Using Standard CMOS

proposed integrated circuit would translate (1) a pair of input signals having a low differential potential and a possibly high common-mode potential into (2) a pair of output signals having the same low differential potential and a low common-mode potential. As used here, "low" and "high" refer to potentials that are, respectively, below or above the nominal supply potential (3.3 V) at which standard complementary metal oxide/semiconductor (CMOS) integrated circuits are designed to operate. The input common-mode potential could lie between 0 and 10 V; the output common-mode potential would be 2 V. This translation would make it possible to process the pair of signals by use of standard 3.3-V CMOS analog and/or mixed-signal (analog and digital) circuitry on the same integrated-circuit chip. A schematic of the circuit is shown in the figure. Standard 3.3-V CMOS circuitry cannot withstand input potentials greater than about 4 V. However, there are many applications that involve low-differential-potential, high-common-mode-potential input signal pairs and in which standard 3.3-V CMOS circuitry, which is relatively inexpensive, would be the most appropriate circuitry for performing other functions on the integrated-circuit chip that handles the high-potential input signals. Thus, there is a need to combine high-voltage input circuitry with standard low-voltage CMOS circuitry on the same integrated-circuit chip. The proposed circuit would satisfy this need. In the proposed circuit, the input signals would be coupled into both a level-shifting pair and a common-mode-sensing pair of CMOS transistors. The output of the level-shifting pair would be fed as input to a differential pair of transistors. The resulting differential current output would pass through six standoff transistors to be mirrored into an output branch by four heterojunction bipolar transistors. The mirrored differential current would be converted back to potential by a pair of diode-connected transistors, which, by virtue of being identical to the input transistors, would reproduce the input differential potential at the output

Yager, Jeremy A.↗

NASA NTRS2013

SDR Input Power Estimation Algorithms

The General Dynamics (GD) S-Band software defined radio (SDR) in the Space Communications and Navigation (SCAN) Testbed on the International Space Station (ISS) provides experimenters an opportunity to develop and demonstrate experimental waveforms in space. The SDR has an analog and a digital automatic gain control (AGC) and the response of the AGCs to changes in SDR input power and temperature was characterized prior to the launch and installation of the SCAN Testbed on the ISS. The AGCs were used to estimate the SDR input power and SNR of the received signal and the characterization results showed a nonlinear response to SDR input power and temperature. In order to estimate the SDR input from the AGCs, three algorithms were developed and implemented on the ground software of the SCAN Testbed. The algorithms include a linear straight line estimator, which used the digital AGC and the temperature to estimate the SDR input power over a narrower section of the SDR input power range. There is a linear adaptive filter algorithm that uses both AGCs and the temperature to estimate the SDR input power over a wide input power range. Finally, an algorithm that uses neural networks was designed to estimate the input power over a wide range. This paper describes the algorithms in detail and their associated performance in estimating the SDR input power.

Nappier, Jennifer M.↗

NASA NTRS2021

Importance of Trade Dependencies for Agricultural Inputs: A Case Study of Finland

Approximately 80% of the world's population lives in countries that are dependent on food imports to sustain an adequate food supply. Besides these food imports, further dependencies also arise due to the requirements for agricultural inputs, including energy, fertilisers, feed and agricultural machinery. While the trade of agricultural inputs is relatively well represented in economic analysis, the quantification is often in terms of monetary values, and thus the assessments of the actual input quantities are very limited. In this paper, we develop a framework for analysis of the traded input quantities at the country-level and demonstrate its utility through an in-depth analysis of Finland's dependency on agricultural inputs. Further, we assess the importance of these inputs for domestic food production from a resilience perspective. We find that Finland, a country with relatively high food self-sufficiency but also a great dependency on imported agricultural inputs, experienced an increase in its dependency over the period 1996-2016. In case of trading partners, the story is more mixed: while trading partners for soybeans increased, those decreased in case of electricity while no significant change was observed in other commodities. In Finland, the rapeseed dependency on imports (ratio of net imports and consumption) grew from around 0% to almost 50% between the years 1987-2013. Changes for electricity remained substantially smaller, with dependency only slightly growing during the study period. Crude oil and soybean supplies are totally dependent on imports as neither is produced in Finland. Interestingly, the highest dependencies for agricultural inputs were on different countries than that for foodstuff imports. Thus, when identifying and assessing approaches to increase resilience to trade-related shocks, it is essential to understand both foodstuff and agricultural input dependencies within the global food system.

trade dependencies↗

DOE OSTI2024

Impact of Heat Input and Step Over Distance on Part Geometry for Multi-Bead WAAM Components

Abstract This study investigates the impact of heat input and overlap ratio or stepover distance on part geometry in wire arc additive manufacturing (WAAM). In this work, an experimental approach was taken to quantify the variation in height and aspect ratio of walls at different combinations of overlap and heat input. Multi-bead walls were deposited with the overlap ratio varying from ½, ⅔, and 1 at a low, medium, and high-power set point on the welding power supply. A relationship between heat input, overlap ratio, and the resultant part geometry is established. Trials with an overlap ratio of one and low to medium heat input exhibited greater lack of fusion defects than the high heat input trials. At higher heat input levels, the as-deposited geometry was found to have less variation in height and an increase in aspect ratio compared to lower heat input trials. The aspect ratio of deposited walls ranged from approximately 1, for the low-power trial with an overlap ratio of ½, to approximately 7, for the high-power trial with an overlap ratio of 1. Additionally, the low and medium power trials with an overlap ratio of ⅔ exhibited the least height variation in the top surface for those heat input levels with an average range of 2.9 mm and 1.8 mm respectively. For the high heat input trials, an overlap ratio of 1 was observed to have the least variation in height with a standard deviation of 0.42 mm and an average range of 1.7 mm; however, lack of fusion defects were present in the top surface and the part cross section.

Hussein, Zaky↗

NASA NTRS1975

Input design for identification of aircraft stability and control derivatives

An approach for designing inputs to identify stability and control derivatives from flight test data is presented. This approach is based on finding inputs which provide the maximum possible accuracy of derivative estimates. Two techniques of input specification are implemented for this objective - a time domain technique and a frequency domain technique. The time domain technique gives the control input time history and can be used for any allowable duration of test maneuver, including those where data lengths can only be of short duration. The frequency domain technique specifies the input frequency spectrum, and is best applied for tests where extended data lengths, much longer than the time constants of the modes of interest, are possible. These technqiues are used to design inputs to identify parameters in longitudinal and lateral linear models of conventional aircraft. The constraints of aircraft response limits, such as on structural loads, are realized indirectly through a total energy constraint on the input. Tests with simulated data and theoretical predictions show that the new approaches give input signals which can provide more accurate parameter estimates than can conventional inputs of the same total energy. Results obtained indicate that the approach has been brought to the point where it should be used on flight tests for further evaluation.

Gupta, N. K.↗

NASA NTRSDate not supplied

Practical Aspects of Multiple-Input Design for Aircraft System Identification Flight Tests

Practical aspects of multiple-input design methods for aircraft system identification flight tests are discussed. The orthogonal optimized multisine multiple-input design technique is explained and demonstrated, along with useful variations. Practical implementation and application issues associated with applying orthogonal optimized multisine inputs in flight are explained, and solutions are demonstrated using flight data from full-scale and subscale aircraft, as well as simulated flight data from an F-16 nonlinear simulation. Topics include truncated and extended excitation inputs, variations for control effectors with limited movement capabilities, aircraft dynamic response to excitation inputs, pilot implementation, statistical uncertainty quantification, multiple-input efficiency and effectiveness, flight testing with active feedback control, frequency-response estimation, control actuator modeling, and control interaction effects. Examples demonstrate that the orthogonal optimized multisine input design method is a practical, versatile, and effective approach for designing multiple excitation inputs for aircraft system identification flight tests.

Multiple-Input Design↗

DOE OSTI2023

Model Inputs, Outputs, and Scripts associated with: “Combined effects of stream hydrology and land use on basin-scale hyporheic zone denitrification in the Columbia River Basin”

This data package is associated with the publication “Combined effects of stream hydrology and land use on basin‐scale hyporheic zone denitrification in the Columbia River Basin”, published in Water Resource Research (Son et al.2022) available at https://doi.org/10.1029/2021WR031131. This data package includes the key model inputs/outputs of the river corridor model for the Columbia River Basin (CRB) and the model source codes used in the manuscript. The model is a carbon-nitrogen-coupled river corridor model (RCM), and the model is used to quantify hyporheic zone (HZ) denitrification at the NHDPLUS stream reach scales. The RCM used in this study combines empirical substrate models derived from observations and three microbially driven reactions, including two-step denitrification and aerobic respiration, are considered within the HZ. The key input data of the model are exchange flux, residence time, and stream solute (dissolved organic carbon (DOC), dissolved oxygen (DO), and nitrate concentrations). These inputs are constant over time and represent long-term averaged values. This study uses the RCM to explore the spatial patterns of HZ denitrification across reaches with different sizes and land use in the CRB. Our main objective is to use the RCM as a virtual reality model, and the machine-learning models as surrogates that encapsulate the complexities of the physics-based model while identifying the importance of different variables that are not evident in the model conceptualization. We do not include a direct comparison of the modeled HZ denitrification and measurements; however, the RCM can capture the overall spatial patterns of the HZ denitrification because the model inputs and its reaction networks are based on well-established theory and a physical-based model. The combination of the model-based predictions and a machine-learning approach (e.g., random forest) is used to improve our understanding of what variables of the model are associated with spatial patterns of the modeled denitrification across reaches with different sizes and land uses, and to develop a proxy model using measurable variables to reproduce the simulated patterns.This dataset contains five folders: (1) model_inputs, (2) model_outputs, (3) Rscripts, (4) figures, and (5) model_codes. It also contains a readme, file level metadata (FLMD), and data dictionary (dd). Please see the FLMD for a list of all the files contained in this data package and descriptions for each. The model_inputs folder contains the model inputs used to drive the model simulations. The model_outputs folder contains key model output files from the river corridor model. The Rscripts folder contains the Rscripts for pre- and post- processing model results. The figures folder contains the raw figures associated with the manuscript. The model_codes folder includes key model source codes/input files. All files are .jpg, .jpeg, .out, .e, .od, .dat, .sub, .F90, .0, .R, .sbx, .cpg, .sbn, .shx, .shp, .dbf, .prj, .tfw, .tif, .xml, .pdf, or .csv.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Model Inputs, Outputs, and Scripts associated with: “Spatial microbial respiration variations in the hyporheic zones within the Columbia River Basin”

This data package is associated with the publication “Spatial microbial respiration variations in the hyporheic zones within the Columbia River Basin” published in the Journal of Geophysical Research: Biogeosciences (Son et al. 2022) available at doi: 10.1029/2021JG006654. This data package includes the key model inputs/outputs of the river corridor model for the Columbia River Basin (CRB) and the model source codes, which were used in the manuscript. The model is a carbon-nitrogen-coupled river corridor model (RCM), and the model is used to quantify hyporheic zone (HZ) aerobic and anaerobic respiration at the NHDPLUS stream reach scales. The RCM used in this study combines empirical substrate models derived from observations and three microbially driven reactions to compute respiration of the HZ for each National Hydrography Dataset (NHD) reach within the CRB. The reactions in HZs of each NHD reach include anaerobic respiration and two-step anaerobic respiration via denitrification. Our HZ respiration estimates are limited to the lotic (or flowing) stream/river systems, and do not account for the respiration process in water column. Note that the RCM only simulates the HZ’s contribution to the dissolved carbon dioxide (CO2) concentrations in the streams, and the CO2 emissions to the atmosphere are not modelled. The model computes at hourly timesteps because of the fast reaction rates. The key input data of the model are exchange flux, residence time, and stream solute (dissolved organic carbon (DOC), dissolved oxygen (DO), and nitrate concentrations). These inputs are constant over time and represent long-term averaged values.This modeling framework successfully quantified HZ respiration components over multiple scales. It revealed key mechanisms driving the spatial variation of HZ aerobic and anaerobic respiration in reaches with varying hydrologic and substrate conditions. Thus, this modeling study offers a testing hypothesis in different river system (e.g., climate and biomes) for the HZ respiration processes, and can be used as a sampling design tool for large-scale HZ experimental studies.This dataset contains five folders: (1) model_inputs, (2) model_outputs, (3) Rscripts, (4) figures, and (5) model_codes. It also contains a readme, file level metadata (FLMD), and data dictionary (dd). Please see the FLMD for a list of all the files contained in this data package and descriptions for each. The model_inputs folder contains the model inputs used to drive the model simulations. The model_outputs folder contains key model output files from the river corridor model. The Rscripts folder contains the Rscripts for pre- and post- processing model results. The figures folder contains the raw figures associated with the manuscript. The model_codes folder includes key model source codes/input files. All files are .jpg, .jpeg, .out, .e, .od, .dat, .sub, .F90, .0, .R, .sbx, .cpg, .sbn, .shx, .shp, .dbf, .prj, .tfw, .tif, .xml, .pdf, or .csv.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2026

Input specific neural networks

Neural networks have emerged as powerful tools for mapping between inputs and outputs. However, their black-box nature limits the ability to encode or impose specific structural relationships between inputs and outputs. Many scientific and engineering problems, such as constitutive modeling in solid mechanics, require networks that can enforce convexity, monotonicity, or other structural constraints to ensure physical consistency. Here, we introduce the Input Specific Neural Network (ISNN), a new architecture that enables multiple, distinct constraints to be imposed on different input subsets for scalar-valued outputs. This framework unifies convex, monotone–convex, monotone, and arbitrary mappings within a single network for the first time. Two ISNN architectures with analytical first- and second-order derivatives are developed. We demonstrate the performance on synthetic toy problems, inverse problems in isotropic hyperelasticity, and finite element simulations. ISNNs achieve improved extrapolation behavior, require fewer invariant inputs than standard input convex networks for polyconvex potentials, and enable significant computational savings via manual differentiation. We also show how ISNNs can be used to learn structural relationships between inputs and outputs via a binary gating mechanism. Particularly, ISNNs are employed to model a homogenized anisotropic free energy potential in a decoupled multiscale setting. The network learns whether or not the potential should be modeled as polyconvex and retains only the relevant layers while using the minimum number of inputs. ISNNs provide a flexible foundation for embedding structural priors into neural networks, enhancing both interpretability and stability. They are broadly applicable across computational mechanics and other scientific domains requiring constrained functional relationships.

Jadoon, Asghar A. [Univ. of Texas, Austin, TX (Uni↗

NASA NTRS1977

Effects of control inputs on the estimation of stability and control parameters of a light airplane

The maximum likelihood parameter estimation technique was used to determine the values of stability and control derivatives from flight test data for a low-wing, single-engine, light airplane. Several input forms were used during the tests to investigate the consistency of parameter estimates as it relates to inputs. These consistencies were compared by using the ensemble variance and estimated Cramer-Rao lower bound. In addition, the relationship between inputs and parameter correlations was investigated. Results from the stabilator inputs are inconclusive but the sequence of rudder input followed by aileron input or aileron followed by rudder gave more consistent estimates than did rudder or ailerons individually. Also, square-wave inputs appeared to provide slightly improved consistency in the parameter estimates when compared to sine-wave inputs.

Cannaday, R. L.↗

NASA NTRS1984

Proposed input for determining longitudinal aerodynamic parameters for the space shuttle

A control input form to improve the identification of longitudinal aerodynamic parameters for the Space Shuttle, especially at low Mach numbers, has been proposed. This input combines features of several inputs currently used with the Shuttle to improve the response in angle-of-attack without exceeding pitch rate limits. The responses of the proposed input were generated using a simple three-degree-of-freedom simulation. An examination of the power spectral densities of these responses showed them to have more power near the vehicle natural frequency than doublet inputs previously used with the Shuttle. When the responses to the proposed input were corrupted with noise and processed using a maximum likelihood parameter extraction program, the identifiability of the parameters was improved over the identifiability of the same parameters using actual Shuttle responses from doublet input of a similar magnitude. This preliminary study indicates that the proposed input form should result in improved identifiability of longitudinal static and control parameters for the Shuttle vehicle.

Suit, W. T.↗

NASA NTRS1995

High input impedance amplifier

High input impedance amplifiers are provided which reduce the input impedance solely to a capacitive reactance, or, in a somewhat more complex design, provide an extremely high essentially infinite, capacitive reactance. In one embodiment, where the input impedance is reduced in essence, to solely a capacitive reactance, an operational amplifier in a follower configuration is driven at its non-inverting input and a resistor with a predetermined magnitude is connected between the inverting and non-inverting inputs. A second embodiment eliminates the capacitance from the input by adding a second stage to the first embodiment. The second stage is a second operational amplifier in a non-inverting gain-stage configuration where the output of the first follower stage drives the non-inverting input of the second stage and the output of the second stage is fed back to the non-inverting input of the first stage through a capacitor of a predetermined magnitude. These amplifiers, while generally useful, are very useful as sensor buffer amplifiers that may eliminate significant sources of error.

Kleinberg, Leonard L.↗

NASA NTRS2002

Convergence of limb, visceral, and vertical semicircular canal or otolith inputs onto vestibular nucleus neurons

The major goal of this study was to determine the patterns of convergence of non-labyrinthine inputs from the limbs and viscera onto vestibular nucleus neurons receiving signals from vertical semicircular canals or otolith organs. A secondary aim was to ascertain whether the effects of non-labyrinthine inputs on the activity of vestibular nucleus neurons is affected by bilateral peripheral vestibular lesions. The majority (72%) of vestibular nucleus neurons in labyrinth-intact animals whose firing was modulated by vertical rotations responded to electrical stimulation of limb and/or visceral nerves. The activity of even more vestibular nucleus neurons (93%) was affected by limb or visceral nerve stimulation in chronically labyrinthectomized preparations. Some neurons received non-labyrinthine inputs from a variety of peripheral sources, including antagonist muscles acting at the same joint, whereas others received inputs from more limited sources. There was no apparent relationship between the spatial and dynamic properties of a neuron's responses to tilts in vertical planes and the non-labyrinthine inputs that it received. These data suggest that non-labyrinthine inputs elicited during movement will modulate the processing of information by the central vestibular system, and may contribute to the recovery of spontaneous activity of vestibular nucleus neurons following peripheral vestibular lesions. Furthermore, some vestibular nucleus neurons with non-labyrinthine inputs may be activated only during particular behaviors that elicit a specific combination of limb and visceral inputs.

NASA Discipline Developmental Biology↗

NASA NTRS2015

Can Simulation Credibility Be Improved Using Sensitivity Analysis to Understand Input Data Effects on Model Outcome?

Model and simulation (MS) credibility is defined as, the quality to elicit belief or trust in MS results. NASA-STD-7009 [1] delineates eight components (Verification, Validation, Input Pedigree, Results Uncertainty, Results Robustness, Use History, MS Management, People Qualifications) that address quantifying model credibility, and provides guidance to the model developers, analysts, and end users for assessing the MS credibility. Of the eight characteristics, input pedigree, or the quality of the data used to develop model input parameters, governing functions, or initial conditions, can vary significantly. These data quality differences have varying consequences across the range of MS application. NASA-STD-7009 requires that the lowest input data quality be used to represent the entire set of input data when scoring the input pedigree credibility of the model. This requirement provides a conservative assessment of model inputs, and maximizes the communication of the potential level of risk of using model outputs. Unfortunately, in practice, this may result in overly pessimistic communication of the MS output, undermining the credibility of simulation predictions to decision makers. This presentation proposes an alternative assessment mechanism, utilizing results parameter robustness, also known as model input sensitivity, to improve the credibility scoring process for specific simulations.

Credibility↗

DOE OSTI2025

Automated shaker placement and regularized input estimation for MIMO testing.

Multi-input, multi-output (MIMO) testing is used in component qualification to reproduce operational responses in the laboratory. It is often preferred to single-input and base-shake testing because of the potential for equivalent or better tests using smaller actuators and shorter test suites. Given a target response, two key steps in MIMO test design are selecting actuator locations and solving for input loads. Actuator locations are often manually selected using expert judgment. If an automatic method is used, locations are usually determined by simulating the vibration control problem and minimizing a combination of the input energy and control residuals. To select a configuration, the relative importance of input energy and residuals must be specified. Specifying relative weights is, in general, a manual and subjective process. This paper develops an objective function that compares actuator configurations based on control accuracy and required input energy without any manual parameter tuning. The objective function uses an optimally selected tradeoff parameter for each candidate configuration. To choose actuator locations using the new objective function, a pivoting algorithm for integer programming problems is developed. Starting with an initial configuration (such as the one generated by a greedy algorithm), the pivoting algorithm guarantees an objective function decrease in each iteration until convergence is reached. In a simulation featuring a structure excited by a diffuse acoustic field, electrodynamic shaker locations and regularized inputs are solved for without any analyst-specified parameters. Simulations are performed in MIMO configurations where the number of target responses is less than, equal to, and greater than the number of actuators.

Multi-input multi-output↗

DOE OSTI2024

Reductive Analysis with Compiler-Guided Large Language Models for Input-Centric Code Optimizations

Input-centric program optimization aims to optimize code by considering the relations between program inputs and program behaviors. Despite its promise, a long-standing barrier for its adoption is the difficulty of automatically identifying critical features of complex inputs. This paper introduces a novel technique, reductive analysis through compiler-guided Large Language Models (LLMs), to solve the problem through a synergy between compilers and LLMs. It uses a reductive approach to overcome the scalability and other limitations of LLMs in program code analysis. The solution, for the first time, automates the identification of critical input features without heavy instrumentation or profiling, cutting the time needed for input identification by 44× (or 450× for local LLMs), reduced from 9.6 hours to 13 minutes (with remote LLMs) or 77 seconds (with local LLMs) on average, making input characterization possible to be integrated into the workflow of program compilations. Optimizations on those identified input features show similar or even better results than those identified by previous profiling-based methods, leading to optimizations that yield 92.6% accuracy in selecting the appropriate adaptive OpenMP parallelization decisions, and 20-30% performance improvement of serverless computing while reducing resource usage by 50-60%.

Input-Centric Optimization↗

DOE OSTI2024

Diversity of visual inputs to Kenyon cells of the Drosophila mushroom body

The arthropod mushroom body is well-studied as an expansion layer representing olfactory stimuli and linking them to contingent events. However, 8% of mushroom body Kenyon cells in Drosophila melanogaster receive predominantly visual input, and their function remains unclear. Here, we identify inputs to visual Kenyon cells using the FlyWire adult whole-brain connectome. Input repertoires are similar across hemispheres and connectomes with certain inputs highly overrepresented. Many visual neurons presynaptic to Kenyon cells have large receptive fields, while interneuron inputs receive spatially restricted signals that may be tuned to specific visual features. Individual visual Kenyon cells randomly sample sparse inputs from combinations of visual channels, including multiple optic lobe neuropils. These connectivity patterns suggest that visual coding in the mushroom body, like olfactory coding, is sparse, distributed, and combinatorial. However, the specific input repertoire to the smaller population of visual Kenyon cells suggests a constrained encoding of visual stimuli.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2024

RAP: Resource-aware Automated GPU Sharing for Multi-GPU Recommendation Model Training and Input Preprocessing

Ensuring high-quality recommendations for newly onboarded users requires the continuous retraining of Deep Learning Recommendation Models (DLRMs) with freshly generated data. To serve the online DLRM retraining, existing solutions use hundreds of CPU computing nodes designated for input preprocessing, causing significant power consumption that surpasses even the power usage of GPU trainers. To this end, we propose RAP, an end-to-end DLRM training framework that supports Resource-aware Automated GPU sharing for DLRM input Preprocessing and Training. The core idea of RAP is to accurately capture the remaining GPU computing resources during DLRM training for input preprocessing, achieving superior training efficiency without requiring additional resources. Specifically, RAP utilizes a co-running cost model to efficiently assess the costs of various input preprocessing operations, and it implements a resource-aware horizontal fusion technique that adaptively merges smaller kernels according to GPU availability, circumventing any interference with DLRM training. In addition, RAP leverages a heuristic searching algorithm that jointly optimizes both the input preprocessing graph mapping and the co-running schedule to maximize the end-to-end DLRM training throughput. The comprehensive evaluation shows that RAP achieves 78.3× speedup on average over CPU-based DLRM input preprocessing frameworks. In addition, the end-to-end training throughput of RAP is only 2.04% lower than the ideal case, which has no input preprocessing overhead.

Wang, Zheng↗