Andrew D. Brown

Design and fabrication of a new vibration-based electromechanical power generator

A device is described for generating electrical power from mechanical energy in a vibrating environment. The design utilises an electromagnetic transducer and its operating principle is based on the relative movement of a magnet pole with respect to a coil. The approach is suitable for embedded remote microsystems structures with no physical links to the outside world. Simulation, modelling and test results following fabrication of a first prototype have demonstrated that generation of practical amounts of power within a reasonable space is possible. Power generation of more than 1 mW within a volume of 240 mm3 at a vibration frequency of 320 Hz has been obtained.

Overview of the SpiNNaker System Architecture

SpiNNaker (a contraction of Spiking Neural Network Architecture) is a million-core computing engine whose flagship goal is to be able to simulate the behavior of aggregates of up to a billion neurons in real time. It consists of an array of ARM9 cores, communicating via packets carried by a custom interconnect fabric. The packets are small (40 or 72 bits), and their transmission is brokered entirely by hardware, giving the overall engine an extremely high bisection bandwidth of over 5 billion packets/s. Three of the principal axioms of parallel machine design (memory coherence, synchronicity, and determinism) have been discarded in the design without, surprisingly, compromising the ability to perform meaningful computations. A further attribute of the system is the acknowledgment, from the initial design stages, that the sheer size of the implementation will make component failures an inevitable aspect of day-to-day operation, and fault detection and recovery mechanisms have been built into the system at many levels of abstraction. This paper describes the architecture of the machine and outlines the underlying design philosophy; software and applications are to be described in detail elsewhere, and only introduced in passing here as necessary to illuminate the description.

SpiNNaker: A 1-W 18-Core System-on-Chip for Massively-Parallel Neural Network Simulation

The modelling of large systems of spiking neurons is computationally very demanding in terms of processing power and communication. SpiNNaker - Spiking Neural Network architecture - is a massively parallel computer system designed to provide a cost-effective and flexible simulator for neuroscience experiments. It can model up to a billion neurons and a trillion synapses in biological real time. The basic building block is the SpiNNaker Chip Multiprocessor (CMP), which is a custom-designed globally asynchronous locally synchronous (GALS) system with 18 ARM968 processor nodes residing in synchronous islands, surrounded by a lightweight, packet-switched asynchronous communications infrastructure. In this paper, we review the design requirements for its very demanding target application, the SpiNNaker micro-architecture and its implementation issues. We also evaluate the SpiNNaker CMP, which contains 100 million transistors in a 102-mm2 die, provides a peak performance of 3.96 GIPS, and has a peak power consumption of 1 W when all processor cores operate at the nominal frequency of 180 MHz. SpiNNaker chips are fully operational and meet their power and performance requirements.

Optimizing the Jiles-Atherton model of hysteresis by a genetic algorithm

Modeling magnetic components for simulation in electric circuits requires an accurate model of the hysteresis loop of the core material used. It is important that the parameters extracted for the hysteresis model be optimized across the range of operating conditions that may occur in circuit simulation. This paper shows how to extract optimal parameters for the Jiles-Atherton model of hysteresis by the genetic algorithm approach. It compares performance with the well-known simulated annealing method and demonstrates that improved results may be obtained with the genetic algorithm. It also shows that a combination of the genetic algorithm and the simulated annealing method can provide an even more accurate solution than either method on its own. A statistical analysis shows that the optimization obtained by the genetic algorithm is better on average, not just on a one-off test basis. The paper introduces and applies the concept of simultaneous optimization for major and minor hysteresis loops to ensure accurate model optimization over a wide variety of operating conditions. It proposes a modification to the Jiles-Atherton model to allow improved accuracy in the modeling of the major loop.

Simulation of magnetic component models in electric circuits including dynamic thermal effects

It is essential in the simulation of power electronics applications to model magnetic components accurately. In addition to modeling the nonlinear hysteresis behavior, eddy currents and winding losses must be included to provide a realistic model. In practice the losses in magnetic components give rise to significant temperature increases which can lead to major changes in the component behavior. In this paper a model of magnetic components is presented which integrates a nonlinear model of hysteresis, electro-magnetic windings and thermal behavior in a single model for use in circuit simulation of power electronics systems. Measurements and simulations are presented which demonstrate the accuracy of the approach for the electrical, magnetic and thermal domains across a variety of operating conditions, including static thermal conditions and dynamic self heating.

Biologically-Inspired Massively-Parallel Architectures - Computing Beyond a Million Processors

The SpiNNaker project aims to develop parallel computer systems with more than a million embedded processors. The goal of the project is to support large-scale simulations of systems of spiking neurons in biological real time, an application that is highly parallel but also places very high loads on the communication infrastructure due to the very high connectivity of biological neurons. The scale of the machine requires fault-tolerance and power-efficiency to influence the design throughout, and the develop-ment has resulted in innovation at every level of design, including a self-timed inter-chip communic-ation system that is resistant to glitch-induced deadlock and ‘emergency’ hardware packet re-routing around failed inter-chip links, through to run-time support for functional migration and real-time fault mitigation.

On-chip and inter-chip networks for modeling large-scale neural systems

The real-time modeling of large systems of spiking neurons is computationally very demanding in terms of processing power, synaptic weight memory requirements and communication throughput. We propose to build a high-performance computer for this purpose with a multicast communications infrastructure inspired by neurobiology. The core component is a chip multiprocessor incorporating some tens of small embedded processors, interconnected by a NoC that carries spike events between processors on the same or different chips. The design emphasizes modeling flexibility, power-efficiency, and fault-tolerance, and is intended to yield a general-purpose platform for the real-time simulation of large-scale spiking neural systems

Modeling frequency-dependent losses in ferrite cores

We suggest a practical approach for modeling frequency-dependent losses in ferrite cores for circuit simulation. Previous work has concentrated on the effect of eddy-current losses on the shape of the B--H loop, but in this paper we look at the problem from the perspective of energy loss and propose a different network for accurately modeling power loss in ferrite cores. In power applications, the energy loss across the frequency range can have a profound effect on the efficiency of the system, and a simple ladder network in the magnetic domain is not always adequate for this task. Simulations and measurements demonstrate the difference in this approach from the RL ladder network models both in the small-signal and large-signal contexts.

Radio frequency (RF) time-of-flight ranging for wireless sensor networks

Position information of nodes within Wireless Sensor Networks (WSNs) is often a requirement in order to make use of the data recorded by the sensors themselves. On deployment the nodes normally have no prior knowledge of their position and thus a locationing mechanism is required to determine their positions. In this paper, we describe a method to determine the point-to-point range between sensor nodes as part of the locationing process. A two-way Time-of-Flight (TOF) ranging scheme is presented using narrow-band RF. The frequency difference between the transceivers involved with the point-to-point measurement is used to obtain a sub-clock TOF phase offset measurement in order to achieve high resolution TOF measurements. The ranging algorithm has been developed and prototyped on a TI CC2430 development kit with no additional hardware being required. Performance results have been obtained for the line-of-sight (LOS), non-line-of-sight (NLOS) and indoor condition. Accuracy is typically better than 7.0m RMS for the LOS condition over 250.0m and 15.8m RMS for the NLOS condition over 120.0m using a one-hundred sample average. Indoor accuracy is measured to 1.7m RMS using a 1000 sample average over 8.0m. Ranging error is linear and does not increase with increased transmitter-receiver distance. Our TOA ranging scheme demonstartes a novel system where resolution and accuracy are time dependent in comparison to alternative frequency dependent methods using narrowband RF.

Magnetic material model characterization and optimization software

The accurate characterization and modeling of magnetic materials are critical in simulating the performance analysis of electrical circuits incorporating magnetic components. Software has, therefore, been developed, including genetic algorithm-optimization techniques and metric-based goal functions to enable appropriate accuracy in the final model. Multiple loop optimization has been developed to allow a wide range of operating conditions to be used in the goal function, with appropriate weighting for the ultimate application. Sensitivity and Monte Carlo analyses ensure the models are stable and tolerant of parameter variations. Comparisons of simulated B-H curves with measured results demonstrate the capability of the software.