42 research outputs found

    Adaptive memory power management techniques for HPC workloads

    No full text
    The memory subsystem is responsible for a large fraction of the energy consumed by compute nodes in High Performance Computing (HPC) systems. The rapid increase in the number of cores has been accompanied by a proportional increase in the DRAM capacity and bandwidth. Thus, the memory system consumes a significant amount of the power budget available to a compute node. There is a broad research effort focused on power management techniques using DRAM low-power modes. However, memory power management still presents many challenges towards Exascale. In this thesis, the potential of Dynamic Voltage and Frequency memory Scaling (DVFS) is studied considering the ability to select different frequencies for different memory channels. The approach adopted is based on tuning voltage and frequency dynamically to maximize the energy savings while maintaining performance degradation within tolerable limits. It was observed that HPC workloads rarely require maximum bandwidth, and the bandwidth demand placed by applications is spread over different channels. Also, HPC applications do not use all the bandwidth in a sustained manner, and they have phases where this bandwidth demand is not at its peak. Hence applications can tolerate reduction in bandwidth to save energy. Channel access patterns of applications are studied to determine the potential additional energy savings by controlling channels independently. Evaluation of proposed DVFS algorithm is conducted through a novel hybrid evaluation methodology that includes simulation and executions on real hardware. Results show the large potential of adaptive memory power management techniques based on DVFS for HPC workloads.M.S.Includes bibliographical referencesby Karthik Elangova

    Impedance-Source DC-to-AC/DC Converter

    No full text
    This article presents a novel impedance-source-based direct current (DC)-to-alternating current (AC)/DC converter (Z-Source DAD Converter). The Z-Source DAD converter converts the input DC voltage into AC or DC with buck or boost in the load voltage. This Z-Source DAD conversion circuit is a single-stage power conversion system. This converter circuit converts the input DC voltage into variable-magnitude output DC voltage or converts the DC voltage into a variable-magnitude output AC voltage. The higher voltage magnitude in boost mode can be controlled by controlling the shoot-through (ST) state timing of the converter. MATLAB-Simulink simulation and microcontroller-based hardware circuit results are presented to demonstrate power conversion with the buck and boost features of the Z-Source DAD converter for both types of output voltages. The simulation and experimental results show that the Z-Source DAD converter converts the given DC supply into AC or DC with buck or boost in the output load voltage

    Real Time Hardware-in-Loop Implementation of LLC Resonant Converter at Worst Operating Point Based on Time Domain Analysis

    No full text
    The inductor inductor capacitor (LLC) resonant topology has become more popular for deployment in high power density and high-efficiency power converter applications due to its ability to maintain zero voltage switching (ZVS) over a wider input voltage range. Due to their ease of operation and acceptable accuracy, frequency domain-related analytical methods using fundamental harmonic approximation (FHA) have been frequently utilized for resonant converters. However, when the switching frequency is far from the resonant frequency, the circuit currents contain a large number of harmonics, which cannot be ignored. Therefore, the FHA is incapable of guiding the design when the LLC converter is used to operate in a wide input voltage range applications due to its inaccuracy. As a result, a precise LLC converter model is needed. Time domain analysis is a precise analytical approach for obtaining converter attributes, which supports in the optimal sizing of LLC converters. This work strives to give a precise and an approximation-free time domain analysis for the exact modeling of high-frequency resonant converters. A complete mathematical analysis for an LLC resonant converter operating in discontinuous conduction mode (DCM)—i.e., the boost mode of operation below resonance—is presented in this paper. The proposed technique can confirm that the converter operates in PO mode throughout its working range; in addition, for primary MOSFET switches, it guarantees the ZVS and zero current switching (ZCS) for the secondary rectifier. As a function of frequency, load, and other circuit parameters, closed-form solutions are developed for the converter’s tank root mean square (RMS) current, peak stress, tank capacitor voltage, voltage gain, and zero voltage switching angle. Finally, an 8 KW LLC resonant converter is built in the hardware-in-loop (HIL) testing method on RT-LAB OP-5700 to endorse the theoretical study

    A Brief Review of Hydrogen Production Methods and Their Challenges

    No full text
    Hydrogen is emerging as a new energy vector outside of its traditional role and gaining more recognition internationally as a viable fuel route. This review paper offers a crisp analysis of the most recent developments in hydrogen production techniques using conventional and renewable energy sources, in addition to key challenges in the production of Hydrogen. Among the most potential renewable energy sources for hydrogen production are solar and wind. The production of H2 from renewable sources derived from agricultural or other waste streams increases the flexibility and improves the economics of distributed and semi-centralized reforming with little or no net greenhouse gas emissions. Water electrolysis equipment driven by off-grid solar or wind energy can also be employed in remote areas that are away from the grid. Each H2 manufacturing technique has technological challenges. These challenges include feedstock type, conversion efficiency, and the need for the safe integration of H2 production systems with H2 purification and storage technologies

    System Architecture, Design, and Optimization of a Flexible Wireless Charger for Renewable Energy-Powered Electric Bicycles

    No full text
    Wireless power transmission (WPT) is one of the breakthroughs in effortless electric vehicle (EV) charging technology. Different types of wireless charger topologies were proposed and implemented to meet various constraints like power transfer efficiency, wireless transfer distance, and misalignment tolerance. Yet the coupling separation and the transfer efficiency are still underdeveloped for contactless charging of medium- and low-power EVs like e-cycles and e-scooters. For achieving the high-distance WPT in the vehicles which are prone to misalignment issues, series-series (SS) compensated WPT is used. The conventional SS-compensated WPT uses a voltage-fed converter for the power conversion. But the combination of these topologies allows reverse current flow in the system, which will affect the transfer efficiency and life span of the source. To prevent this, a reverse blocking diode or a current-fed converter can be used. Though the reverse current problem can be solved, these approaches seem to reduce the power transfer efficiency further. This article tries to optimize the current-fed converter-based SS-WPT to achieve higher coupling separation, higher power transfer efficiency, and higher misalignment tolerance than the conventional designs. To achieve this, the input inductor of the current-fed converter and the primary coil of the SS-WPT are tuned without affecting the magnetic resonance condition. The transfer efficiency was found to be 94% at a coupling separation of 200 mm, which is 20% more than the conventional voltage source inverter-based, renewable energy-powered SS-WPT charging efficiency. After proving the concept in prototype design, the results are validated by testing the same in a real-time electric cycle. </p

    Finite Element Analysis of Human Bone Structures on the Cell Broadband Engine

    No full text
    With the current advances in bone imaging and progress in numerical techniques, the micro structural Finite Element analysis (FEM) of human bone for stiffness and strength assessment for individual fracture risk prediction, with a massive potential for parallelism as become a signi?cant candidate for investigation in the current multicore processors. This master thesis work focuses on investigating the credibility of Finite Element analysis of the human bone structure on the IBM Cell processor.Microelectronics & Computer EngineeringElectrical Engineering, Mathematics and Computer Scienc
    corecore