Accelerator Module Research Articles

A high-performance wearable thermoelectric generator with comprehensive optimization of thermal resistance and voltage boosting conversion

This paper develops a superhigh-performance wearable thermoelectric generator (WTEG) for harvesting body heat, achieving an output power density of 15.8 μW/cm2 in windless and moveless conditions, and 97.6 μW/cm2 for walking of 0.8 m/s. A novel WTEG configuration integrated with the porous sandwich substrate and direct-soldering Cu-foam heat sink is designed to significantly improve its flexibility and considerably reduce the thermal resistance at the cold/hot sides. A new compact low-voltage boosting converter is optimized to obtain a high conversion efficiency (such as >50%@100 mV) and allow low self-startup input voltage (20 mV) and stable output voltage. A semi-automatization manufacturing process of the self-powered wearable sensor system is also designed to integrate the WTEG module, energy management module, multi-sensor (including acceleration, temperature, humidity, heart rate, and blood oxygen) module, and Bluetooth module onto the flexible substrate. The experimental results indicate that the sandwich substrate increases WTEG performance by 25% and achieves about 40 mW/cm2 for the constant temperatures at the cold/hot sides with the temperature difference of 60 °C. When the cold side of WTEG is exposed to the ambient air, the Cu-foam heat sink enables a performance increase of 73.6% compared to that without a heat sink for the ambient temperature of 18 °C, which even reaches 302% under the wind speed of 2 m/s (obtaining 457.97 μW/cm2). It is interesting to find that the bending WTEG could significantly increase its performance by 45.6% (for the curvature radius of 20 mm) compared to the plane case due to the thermal condition improvement. For harvesting body heat, the performance of WTEG fixed on the forehead is better than that on the arm or shank. It achieves a superhigh average output power of 3.12 mW for walking at the ambient temperature of 18 °C, which could fully power the wearable multi-sensor health monitoring system continuously.

Grain boundary segregation and carbide precipitation in heat treated niobium superconducting radio frequency cavities

A fundamental understanding of superconducting radio frequency Nb cavity processing is necessary to achieve the desired improvement in their performance, which is needed for further upgrades of modern particle accelerators. To recognize the physical processes behind the losses in the accelerator modules, it is required to address not only the observed improvements but also the degradation occurring after different surface treatments. Here, we report on microscopic and spectroscopic studies of several cutouts from an extremely well performing cavity, which showed a systematic degradation after modified surface treatments and annealing conditions. Our results suggest that an abundance of low-angle grain boundaries surrounding the small sized grains can be related to the local superconductivity breakdown at high accelerating field gradients. Losses due to grain boundary segregated carbides are discussed to being most dominant and to leading to an anomalous Q-degradation of the whole cavity starting at low fields.

Design and Implementation of Embedded Real-Time English Speech Recognition System Based on Big Data Analysis

This article uses Field Programmable Gate Array (FPGA) as a carrier and uses IP core to form a System on Programmable Chip (SOPC) English speech recognition system. The SOPC system uses a modular hardware system design method. Except for the independent development of the hardware acceleration module and its control module, the other modules are implemented by software or IP provided by Xilinx development tools. Hardware acceleration IP adopts a top-down design method, provides parallel operation of multiple operation components, and uses pipeline technology, which speeds up data operation, so that only one operation cycle is required to obtain an operation result. In terms of recognition algorithm, a more effective training algorithm is proposed, Genetic Continuous Hidden Markov Model (GA_CHMM), which uses genetic algorithm to directly train CHMM model. It is to find the optimal model by encoding the parameter values of the CHMM and performing operations such as selection, crossover, and mutation according to the fitness function. The optimal parameter value after decoding corresponds to the CHMM model, and then the English speech recognition is performed through the CHMM algorithm. This algorithm can save a lot of training time, thereby improving the recognition rate and speed. This paper studies the optimization of embedded system software. By studying the fixed-point software algorithm and the optimization of system storage space, the real-time response speed of the system has been reduced from about 10 seconds to an average of 220 milliseconds. Through the optimization of the CHMM algorithm, the real-time performance of the system is improved again, and the average time to complete the recognition is significantly shortened. At the same time, the system can achieve a recognition rate of over 90% when the English speech vocabulary is less than 200.

Field Trial on 5G New Radio Over Satellite

5G New Radio (NR) is the 3rd Generation Partnership Project (3GPP) radio access technology for the next generation mobile communications network. A major evolution of 5G constitutes the integration of non-terrestrial networks including geostationary and low Earth orbit satellites. The seamless integration of satellites in the terrestrial mobile network requires significant adaptations within the radio access network and the development of new features in the core network to cope with the specific satellite channel characteristics. To date, the 5G control and data plane has been standardized to handle only continuous backhaul communication between the network components. However, a mobile satellite enabled next generation Node B (gNB) located in a vehicle or in a moving aerial platform needs to be able to handle frequent backhaul outages of various duration as well as longer signal delays as opposed to short terrestrial connections via fiber. In this paper, we report the results of an over-the-air (OTA) field trial comprising a mobile edge node connected to the 5G standalone core network components over a geostationary satellite. We analyze Transmission Control Protocol (TCP) acceleration and GPRS Tunneling Protocol (GTP)/TCP/Internet Protocol (IP) header compression features through the GTP. Moreover, the influence of short and long interruptions in the communication between the edge node and the central components on the entire system performance is investigated. The header compression and TCP acceleration modules were implemented on the satellite modems and are now part of the protocol stack of these devices. The results show up to 12% higher data rates for the 5G user equipment (UE), on a 1.5 MHz single carrier return link compared to deactivated TCP acceleration and header compression. We increased the data rate by 20% on the 4.5 MHz DVB-S2X forward link between the UE and 5G core. Moreover, our measurements reveal that even satellite-enabled gNB mobility is possible with the current Release 15 standard. After a short outage of the satellite connection due to shadowing, the UE can successfully re-establish the user plane connection to the core network. Our results will facilitate the full integration of satellite components in 5G through open and standard solutions.

Demonstration of a compact plasma accelerator powered by laser-accelerated electron beams

Plasma wakefield accelerators are capable of sustaining gigavolt-per-centimeter accelerating fields, surpassing the electric breakdown threshold in state-of-the-art accelerator modules by 3-4 orders of magnitude. Beam-driven wakefields offer particularly attractive conditions for the generation and acceleration of high-quality beams. However, this scheme relies on kilometer-scale accelerators. Here, we report on the demonstration of a millimeter-scale plasma accelerator powered by laser-accelerated electron beams. We showcase the acceleration of electron beams to 128 MeV, consistent with simulations exhibiting accelerating gradients exceeding 100 GV m−1. This miniaturized accelerator is further explored by employing a controlled pair of drive and witness electron bunches, where a fraction of the driver energy is transferred to the accelerated witness through the plasma. Such a hybrid approach allows fundamental studies of beam-driven plasma accelerator concepts at widely accessible high-power laser facilities. It is anticipated to provide compact sources of energetic high-brightness electron beams for quality-demanding applications such as free-electron lasers.

Performance Analysis of the χMD Matrix Solver Package for MODFLOW-USG.

The χMD matrix solver package is incorporated into USGS groundwater modeling software, such as MODFLOW-NWT, MODFLOW-USG, and MT3D. The solver is used to solve matrices assembled through numerical discretization of the groundwater flow equation, and solute transport equations. χMD has demonstrated its higher robustness, faster execution speed, and more efficient memory usage compared to the existing solvers for many types of groundwater flow problems. χMD uses preconditioned iterative Krylov-subspace methods and consists of preconditioning and acceleration modules. Because the solver package uses a variety of preconditioning features including level-based incomplete lower-upper (ILU) factorization method with a drop tolerance scheme, users must choose optimal preconditioning parameters to improve execution speed and robustness. In order to examine how the preconditioning parameters, ILU factorization level, and drop tolerance values affect the overall performance of the matrix solver, we evaluated five different groundwater model applications using MODFLOW-USG that include different numerical complexities. For those five cases, the number of discretization nodes varied from 10,000 cells to 730,300 cells. From the analysis, we found that the preconditioning parameters greatly affect execution times and memory usage of the preconditioning and acceleration procedures. In addition, a combination of the ILU level between five to seven and the drop tolerance value between 10-2 and 10-3 usually resulted in shorter overall execution time. Our study suggests that the users can elicit higher performance and robustness of the χMD matrix solver using this combination of the parameters and enhance computational efficiency of solving groundwater and solute transport problems.

Electrodynamic System of the Linear Induction Accelerator Module

This article presents the results of theoretical and experimental studies on the electrodynamic system of an accelerator module, which is a part of the linear induction accelerator (LIA) (BINP and RFNC-VNIITF). In the course of this research, 3-D models of the LIA accelerator module with the geometry close to the real one have been created. These models allowed us to calculate the main parameters of the electromagnetic eigenmodes for these modules. To verify the calculation results, an experimental stand permitting to register the spectral properties of the dipole modes has been built. The main results of the work are the electromagnetic field distributions of the modes in the accelerator module obtained in modeling, as well as the mode spectra calculated and registered in “cold” experiments (without beam). Based on the obtained data, the main features of the oscillation’s spectral composition and their $Q$ -factors have been established in the frequency range $f\approx 300{(\!{-}\!)}1100$ MHz, and the effective methods for suppressing these oscillations have been proposed. The development of such methods is absolutely necessary in order to reduce the increment of the beam breakup (BBU) instability, which is the main obstacle in the increase of the beam current in LIA.

Application of Face Recognition Method Under Deep Learning Algorithm in Embedded Systems

The Convolutional Neural Network (CNN) based on deep learning is introduced to propose two deep face detection algorithms and design an embedded face recognition system, in an effort to apply the deep learning algorithm to face detection and explore the embedded face recognition system. First, an optimized Multi-task Cascaded Convolutional Network (MTCNN) algorithm is proposed for the simulation transformation and crop preprocessing of the face image, denoted as OMTCNN. Second, a lightweight face recognition algorithm based on CNN is proposed to reduce the computational complexity of face recognition in the embedded system, denoted as LCNN. Finally, OMTCNN and LCNN are combined to construct the multi-core embedded face recognition system. Results demonstrate that OMTCNN shows good performance in determining face identity, and its training accuracy can reach 95.78%, significantly better than the unimproved MTCNN algorithm. On the Labeled Faces in the Wild (LFW) dataset, LCNN provides a correct rate of 98.13%, and the reduction in the complexity of the model itself increases the computational speed by as much as 6.4 times. The test results suggest that the designed face recognition system has good applicability on the embedded platform. The increase in the accuracy of the face recognition module and the introduction of the calculation acceleration module have significantly improved the face detection and recognition performance of the embedded system. The embedded face recognition system based on deep learning has practical application value.

The Case for FPGA-Based Edge Computing

Edge Computing has emerged as a new computing paradigm dedicated for mobile performance enhancement and energy efficiency purposes. Specifically, it benefits today’s interactive applications on power-constrained devices by offloading compute-intensive tasks to the edge nodes in close proximity. Meanwhile, FPGA is well known for its excellence in accelerating (domain-specific) compute-intensive tasks such as deep learning algorithms in a high performance and energy-efficient manner due to its hardware-customizable nature. In this paper, we make the first attempt to leverage and combine the advantages of these two, and proposed a new network-assisted computing model, namely FPGA-based edge computing. As a case study, we choose three computer vision (CV)-based mobile interactive applications, and implement their back-end computation engines on FPGA. By deploying such application-customized accelerator modules for computation offloading at the network edge, we experimentally demonstrate that this approach can effectively reduce response time for the applications and energy consumption for the entire system in comparison with traditional CPU-based edge/cloud offloading approach.

On the Heterogeneity of Existing Repositories of Movements Intended for the Evaluation of Fall Detection Systems.

Due to the serious impact of falls on the autonomy and health of older people, the investigation of wearable alerting systems for the automatic detection of falls has gained considerable scientific interest in the field of body telemonitoring with wireless sensors. Because of the difficulties of systematically validating these systems in a real application scenario, Fall Detection Systems (FDSs) are typically evaluated by studying their response to datasets containing inertial sensor measurements captured during the execution of labelled nonfall and fall movements. In this context, during the last decade, numerous publicly accessible databases have been released aiming at offering a common benchmarking tool for the validation of the new proposals on FDSs. This work offers a comparative and updated analysis of these existing repositories. For this purpose, the samples contained in the datasets are characterized by different statistics that model diverse aspects of the mobility of the human body in the time interval where the greatest change in the acceleration module is identified. By using one-way analysis of variance (ANOVA) on the series of these features, the comparison shows the significant differences detected between the datasets, even when comparing activities that require a similar degree of physical effort. This heterogeneity, which may result from the great variability of the sensors, experimental users, and testbeds employed to generate the datasets, is relevant because it casts doubt on the validity of the conclusions of many studies on FDSs, since most of the proposals in the literature are only evaluated using a single database.

Bayesian Multi-objective Hyperparameter Optimization for Accurate, Fast, and Efficient Neural Network Accelerator Design.

In resource-constrained environments, such as low-power edge devices and smart sensors, deploying a fast, compact, and accurate intelligent system with minimum energy is indispensable. Embedding intelligence can be achieved using neural networks on neuromorphic hardware. Designing such networks would require determining several inherent hyperparameters. A key challenge is to find the optimum set of hyperparameters that might belong to the input/output encoding modules, the neural network itself, the application, or the underlying hardware. In this work, we present a hierarchical pseudo agent-based multi-objective Bayesian hyperparameter optimization framework (both software and hardware) that not only maximizes the performance of the network, but also minimizes the energy and area requirements of the corresponding neuromorphic hardware. We validate performance of our approach (in terms of accuracy and computation speed) on several control and classification applications on digital and mixed-signal (memristor-based) neural accelerators. We show that the optimum set of hyperparameters might drastically improve the performance of one application (i.e., 52–71% for Pole-Balance), while having minimum effect on another (i.e., 50–53% for RoboNav). In addition, we demonstrate resiliency of different input/output encoding, training neural network, or the underlying accelerator modules in a neuromorphic system to the changes of the hyperparameters.

FPGA‐Enhanced Data Processing System Using PCM Technology

Due to unique features, Storage class memory (SCM) technologies such as Phase change memory (PCM) open up new opportunities for architect engineers. In such a scenario, we present a true PCM storage system with an FPGA storage controller to explore ISP benefits. Our contributions are summarized as follows. 1) Propose a heterogeneous ISP architecture which uses FPGA working as storage controller and accelerator; 2) Offer a new research direction for designing storage devices which inherently support ISP; 3) Present a novel storage system which eliminate data transfer; 4) The first evaluation of ISP on a real SCM device; 5) Demonstrate significant performance gains by using efficient data flow and consuming extremely small amounts of host resources. Besides, the proposed system can be extended to handle other kinds of data applications by adding corresponding accelerator and data conversion module in FPGA. Multinode system can be realized for more aggressive results.

Compact and tunable active-plasma lens system for witness extraction and driver removal

Plasma based technology will allow an unprecedented reduction of the size of accelerating machines. Both fundamental research and applied science and technology will take profit of this feature. The same compactness is required downstream the accelerator module, where the plasma-accelerated beams usually experience a large angular divergences growth. Therefore compact, strong and tunable focusing devices are needed. Active-plasma lenses have been demonstrated to be a compact and affordable tool to generate radially symmetric magnetic fields. We present a new scheme using active-plasma lenses and a metallic collimator to catch and transport the witness bunch while removing the driver. The considered case study is in the context of the EuPRAXIA project.

Actigraphic measurement of the upper limbs movements in acute stroke patients

BackgroundStroke units provide patients with a multiparametric monitoring of vital functions, while no instruments are actually available for a continuous monitoring of patients motor performance. Our aim was to develop an actigraphic index able both to identify the paretic limb and continuously monitor the motor performance of stroke patients in the stroke unit environment.MethodsTwenty consecutive acute stroke patients (mean age 69.2 years SD 10.1, 8 males and 12 females) and 17 bed-restrained patients (mean age 70.5 years SD 7.3, 7 males and 10 females) hospitalized for orthopedic diseases of the lower limbs, but not experiencing neurological symptoms, were enrolled. This last group represented our control group. The motor activity of arms was recorded for 24 h using two programmable actigraphic systems showing off as wrist-worn watches. The firmware segmented the acquisition in epochs of 1 minute and for each epoch calculates two motor activity indices: MAe1 (Epoch-related Motor Activity index) and MAe2 (Epoch-related Motor Activity index 2). MAe1 is defined as the standard deviation of the acceleration module and MAe2 as the module of the standard deviation of acceleration components. To describe the 24 h motor performance of each limb, we calculated the mean value of MAe1 and MAe2 (respectively MA1_24h and MA2_24h). Then we obtained two Asymmetry Rate Indices: AR1_24h and AR2_24h to show the motor activity prevalence. AR1_24h refers to the asymmetry index between the values of MAe1 of both arms and AR2_24h to MAe2 values.The stroke patients were clinically evaluated by NIHSS at the beginning (NIHSST0) and at the end (NIHSST1) of the 24 h actigraphic recordings.ResultsBoth MA1_24h and MA2_24h indices were smaller in the paretic than in the unaffected arm (respectively p = 0.004 and p = 0.004). AR2_24h showed a better capability (95% of paretic arms correctly identified, Phi Coefficient: 0.903) to discriminate the laterality of the clinical deficit than AR1_24h (85% of paretic arms correctly identified, Phi Coefficient: 0,698). We also found that AR1_24h did not differ between the two groups of patients while AR2_24h was greater in stroke patients than in controls and positively correlated with NIHSS total scores (r: 0.714, p < 0.001 for NIHSS, IC95%: 0.42–0.90) and with the sub-score relative to the paretic upper limb (r: 0.812, p < 0.001, IC95%: 0.62–0.96).ConclusionsOur data show that actigraphic monitoring of upper limbs can detect the laterality of the motor deficit and measure the clinical severity. These findings suggest that the above described actigraphic system could implement the existing multiparametric monitoring in stroke units.

A 47.14-$\mu\text{W}$ 200-MHz MOS/MTJ-Hybrid Nonvolatile Microcontroller Unit Embedding STT-MRAM and FPGA for IoT Applications

The demand for energy-efficient, high-performance microcontroller units (MCUs) for the use in power-supply-critical Internet-of-Things (IoT) sensor-node applications has witnessed a substantial increase. In response, research concerning the development of several low-power-consuming MCUs has been actively pursued. The performance level of such MCUs, however, has not been sufficient, thereby rendering them non-feasible for the use in IoT sensor-node applications that process a large number of received signals immediately followed by extraction of valuable information from them to limit data transferred to a data center. To realize next-generation IoT systems based on intelligent sensor-node application, ultra-low-power high-performance MCUs need to be developed. This paper presents an ultra-low-power-consuming and high-performance MCU configuration based on the spintronics device technology, using which all modules are non-volatilized, and any wasteful power consumption is eliminated by controlling the power supplied independently to each module. By incorporating a reconfigurable accelerator module, for performing various signal-processing procedures in sensor-node applications, and a memory controller, which can speed up the entire system by relaxing the data-transfer bottleneck of logic and memory, the proposed MCU configuration achieves ultra-low power consumption and high-speed operation. As confirmed by the results obtained via measurements performed on a fabricated chip, the proposed MCU design, on average, consumed 47.14 $\mu \text{W}$ power at an operating frequency of 200 MHz. This corresponds to the world’s highest signal-processing performance and energy efficiency of highly functional IoT sensor nodes powered by harvested energy

European XFEL Superconducting Cryomodules Characterization Toward Modules Acceptance and Future LLRF Operation

This paper describes some achievements of the low-level radio frequency (LLRF) tests performed to evaluate the superconducting cryomodules in preparation of their installation in the European X-ray Free Electron Laser (XFEL) accelerator. The software developed to characterize the cryomodules tested at the Accelerator Module Test Facility (AMTF) will be presented. The purpose of these tests is to evaluate some of the key parameters of the eight Tera Electronvolt Superconducting Linear Accelerator cavities, housed in every cryomodule. These parameters include: cavity quench gradients, cavity fundamental resonant modes, performance of the slow and fast frequency tuners, and characteristics of the input power coupler. These results impact the final decision on the cryomodule acceptance and its installation inside the linear accelerator (linac). In this approach, the MicroTCA.4-based LLRF system is used to control and assess the cryomodule performance. Middle layer servers are developed to verify the tests initial conditions, carry the test, and log the results in a dedicated database. The tests results provide a basis for the cryomodule validation, but also serve as a reference useful during the commissioning and operating phase of the accelerator.

From SPARC_LAB to EuPRAXIA@SPARC_LAB

Following the promising results obtained at the SPARC_LAB test-facility in Frascati (Italy), we have recently submitted a proposal to develop a new facility driven by a plasma accelerator module for extended and user-oriented applications. The new multi-disciplinary user-facility will be equipped with a soft X-ray Free Electron Laser (FEL) operating with energies larger than 1 GeV. This design study is performed to be fully compatible with the EuPRAXIA design study. Here, the latest layout and beam parameters are presented.

Use of a Compact Neutral Particle Analyzer for Studying Thermal and Suprathermal Ions in Plasma Discharges with Neutral Beam Injection

A compact neutral particle analyzer that has been modified for studying thermal and suprathermal ions in neutral-beam heated plasma is described. Results of the analyzer calibration performed using an atomic beam are presented. The possibility of adjusting the analyzer channel energies by varying the voltage applied to the acceleration module of the analyzer is tested. The measured ion temperatures and suprathermal-ion spectra are presented for discharges with the neutral beam injection at the TUMAN-3M tokamak.

Design of Compact Accelerator Module of the Induction Synchrotron

Induction synchrotron (IS) has become the focus of improving the intensity of high energy ion beams in recent years. But because of the pulse voltage of the induction accelerating cell cannot adjust quickly, causing the mismatch of the magnetic deflection field and accelerate electric field, there is a serious beam loss problem in IS. The simulation of beams longitude dynamics proves that increasing the number of accelerating cells can effectively reduce the loss. In this paper, a novel compact induction accelerator module is introduced. Compared with the existing induction accelerator module of KEK, the new design can increase the number of accelerating cells to more than five times in the same installation space and improves the voltage top drop of the acceleration pulse significantly.

Compact ultracold electron source based on a grating magneto-optical trap

The ultrafast and ultracold electron source, based on near-threshold photoionisation of a laser-cooled and trapped atomic gas, offers a unique combination of low transverse beam emittance and high bunch charge. Its use is however still limited because of the required cold-atom laser-cooling techniques. Here we present a compact ultracold electron source based on a grating magneto-optical trap (GMOT), which only requires one trapping laser beam that passes through a transparent accelerator module. This makes the technique more widely accessible and increases its applicability. We show the GMOT can be operated with a hole in the center of the grating and with large electric fields applied across the trapping region, which is required for extracting electron bunches. The calculated values of the applied electric field were found to agree well with measured Stark shifts of the laser cooling transition. The electron beams extracted from the GMOT have been characterised. Beam energies up to 10 keV were measured using a time-of-flight method. The normalised root-mean-squared transverse beam emittance was determined using a waist scan method, resulting in $\epsilon = 1.9 \rm{nm}$. The root-mean-squared transverse size of the ionisation volume is $30 \mu\rm{m}$ or larger, implying an electron source temperature in the few-10K range, $2-3$ orders of magnitude lower than conventional electron sources, based on photoemission or thermionic emission from solid state surfaces.