This thesis presents a theoretical basis and the practical implications for using small-scale (up to 50 kW) low-cost power generating systems comprising an ungoverned turbine, a three phase self-excited induction generator, a variable VAr source for voltage regulation, and an electronic load governor. The thrust of this work is towards the reduction of cost by replacing the more conventional synchronous generator by an induction machine and, by eliminating the site-specific fully-governed turbine. An electronic load governor, utilising zero-voltage switching to minimise waveform distortion, is used to regulate the system frequency by controlling the generator loading. The performance of the load governor is analysed in conjunction with both synchronous and induction generators, and the conditions under which the governed system is stable are determined. The induction machine is shown to have better operating characteristics than the synchronous machine and has the further advantages of low maintenance and ready availability making it particularly suitable for isolated applications. A theoretical explanation of the process by which an induction machine self-excites when capacitance is connected across its terminals is developed, and methods of guaranteeing excitation are proposed. It is shown that the remanent magnetism, present in the rotor, plays a significant role in the machine excitation and the conditions leading to loss of remanence are investigated. Once excited, the operation of the induction generator is analysed for both balanced and unbalanced loadings. The practical implications of these results are taken into account in proposals for single phase supplies, these being more desirable in smaller systems. A range of static VAr sources, including a switched capacitor VAr source and an inductively loaded ac-to-dc converter, are considered for controlling the machine’s excitation. The VAr sources are compared on the basis of harmonic distortion and system stability. The thesis considers the application of these devices to micro-hydro and wind generating systems. Preliminary investigations have been carried out into the use of two-speed induction machines to improve the efficiency of CSCF Constant Speed Constant Frequency wind generating schemes and into the use of centrifugal pumps as ungoverned turbines. Practical results obtained from laboratory experiments and tests on existing micro-hydro installations are included. It is shown that significant reductions in cost are obtainable by employing the methods discussed in this thesis. The resultant systems may incur a reduction in efficiency but there is little reduction in performance and in many respects the quality of supply is improved.

Digitised pictures have both cosmetic and scientific image aspects. This thesis presents an image coding system that compresses monochromatic digital image information in such a way that the cosmetic-or viewed-image aspects of a reconstructed picture are indistinguishable from the original to a (human) observer. The coding system is concerned with viewed aspects of digitised images and is based on a simplified 2-channel model of the human eye, with the image information being separated into slowly varying texture information and rapidly changing edge information corresponding to those two channels. Each of the two types of image data is coded using techniques suited to its individual characteristics. Texture information is coded using predictive waveform coding where the predictive filter coefficients are generated by linear predictive coding (LPC) techniques. Since the waveform prediction is not perfect, this thesis evaluates several different methods of differential pulse code modulation (DPCM) coding the residual prediction error. This signal is then used to improve the texture coding quality. Image edges are isolated from the digital image data by an asymptotically optimal edge detector known as the Laplacian of Gaussian (LOG) filter. Using two image edge models, the LOG isolated edges produce contours which are used to define edge position, shape, and "sharpness". The edge position data is then compressed using a chain-coding system. The thesis also considers the effect the edge characteristics have on the coding system and develops two methods for reducing the number of coded edges. Reconstruction of the coded image is in two parts corresponding to the two channels of the coding system. The texture information is reconstructed by predicting texture pixels and adding the decoded DPCM error signal. Edge information not contained in the texture coding channel is reconstructed using only edge information corresponding to the high spatial frequency channels of the human eye. These two channels are added together producing the final coded image reconstruction. Also considered are the effects of varying the original image sampling density. The 2-channel coding system is applicable to a wide variety of images, producing coded images with a high visual quality and coding bitrates as low as 0.6 bits/pixel (256 x 256 pixel images), while maintaining a relatively low computational overhead both at the encoder and the decoder.

Networked control systems (NCSs) are a type of distributed control systems where sensors, actuators, and controllers are interconnected through a communication network. This system setup has the advantage of low cost, °exibility, and less wiring, but it also inevitably invites some delays and data loss into the design procedure. The focus of this thesis is to address the problem of analysis and design of networked control systems when the communication delays are varying in a random fashion. This random feature of the time delays is typical for commercially used networks, such as a DeviceNet (which is a controller area network (CAN)) and Ethernet network. Models for communication network delays are ¯rst developed, in which Markov processes are used to model these random network-induced delays. Based on such models, we establish novel methodologies for stability analysis, control with disturbance attenuation, and fault estimation for a class of uncertain linear/nonlinear uncertain NCSs with random communication network-induced delays in both sensor-to-controller and controller-to- actuator channels. Data packet dropouts in the communication channels also have been taken into consideration in the modelling and design procedure. The main technique used in this thesis is based on the Lyapunov-Razumikhin method, which results in delay-dependent controllers. We ¯rst consider the design prob- lems for uncertain linear NCSs. In this case, state feedback controllers and dynamic output feedback controllers are designed to satisfy both stability and disturbance at- tenuation requirements for this class of NCSs. Moreover, a robust fault estimator that ensures the fault estimation error is less than a prescribed performance level is designed. We further go on to address the control problems for uncertain nonlinear NCSs. The nonlinear plant is ¯rst described by the T-S fuzzy model. Based on this model, stability analysis, disturbance attenuation, and fault estimation problems are studied for uncer- tain nonlinear NCSs. It should be noted that system uncertainties, disturbances and noises are addressed in both cases. The existence of such controllers and fault estimators are given in terms of the solvability of bilinear matrix inequalities. Iterative algorithms are proposed to change this non-convex problem into quasi-convex optimization problems, which can be solved e®ectively by available mathematical tools. Finally, to demonstrate the e®ectiveness and advantages of the proposed design methodologies in this thesis, numerical examples are given in each designed control systems. The simulation results show that the proposed design methodologies can achieve the prescribed performance requirements.

The aim of this thesis is the development of an understanding of some of the Limitations involved in the design of active filters. This introduces the idea of imposing an upper bound on the amplitude of the input voltage to an active filter such that the distortion, introduced, in the input stage of the operational amplifier, is Less then a value specified in the design, and investigating theoretically the errors in the frequency response of the actual circuit, caused by the operational amplifier which is considered linear but nonideal in that the analysis includes the finite input impedance and output conductance as well as the single-pole rolloff characteristic. The results of this theoretical research have been incorporated into a package of highly interactive computer programs, collectively called the Active Filter synthesis program, for the design of minimum-phase lowpass, highpass, bandpass, and bandstop active filters, the group-delay equalization of these filters and also of both unloaded, and loaded telephone Lines, and the displaying of their ideal frequency responses. The above-mentioned research arose from the design and construction of an experimental 600/1200 Baud, Modem, the filters in which provide a practical application of these developments. / Whole document restricted, but available by request, use the feedback form to request access.

Whole document restricted, see Access Instructions file below for details of how to access the print copy. / Correct coagulant dosage is necessary for the efficient operation of conventional drinking water treatment plants, yet no accurate or automated way of determining this exists. Streaming current (SC) is a measurement of charge on particles in water and is useful in feedback control of coagulant dosage. Analysis of the movement of change within a SC sensor can provide some explanation of its slow response, while signal processing utilising Fourier analysis improves the instrument's bandwidth. Presently inaccurate manual jar tests are the only way determine the SC required for best coagulation. An online automated jar tester is presented to improve on this. It uses an automatic sampling system that takes a sample from the process stream. An optimisation algorithm makes repeated step adjustments to the SC set point and gradually moves it in the direction of improving jar test results. The system was evaluated on both a small-scale model and a full-scale plant. Noise in the test measurements means the optimal set point cannot be located accurately enough, but the results indicate that this is possible. Greater accuracy would allow optimisation of turbidity and costs for multiple chemicals. A representative neural network model can be made of the dynamic relationship between coagulant dosage and streaming current in a scale model, with an alkali dosed to simulate a disturbance. In a rapid mixer, the measured response is significantly slower than the true response. Several common types of linear controller are designed and their performance at set point tracking and disturbance rejection is compared on this system. Model predictive control with a Kalman filter performs best in these tests, while the self-tuning regulator has benefits when the rate of set point change is slower. A non-linear feed-forward radial basis function network that adapts to the system's steady-state inverse can effectively augment a linear controller for this system. Adaptation rules based on vector eligibility are derived from dynamic back-propagation and extended to the general dynamic non-linear case. This can result in a useful and efficient feed-forward neural controller for dosing systems that can be represented by a Wiener model.

This thesis describes research undertaken as an initial step in the development of a nationwide land mobile coverage area prediction model for New Zealand. The most suitable approach is shown to be a computer-based method which includes corrections for both environmental and terrain features. Extensive field trials performed in Auckland at 76 and 465MHz to develop a propagation data base are described. Measurements have been made in a variety of environments over unobstructed, obstructed and mixed land-sea paths. An analysis of these measurements indicates that, subject to the availability of suitable topographic and environmental data bases and with the exception of two special cases, a plane earth based prediction method is suitable for incorporation in a nationwide land mobile coverage area prediction model for New Zealand. Recommendations are made for the implementation and further development of this model.

A new method for the synthesis of broadband impedance transformers having predictable passband frequency response is presented. The technique is based on the use of non-commensurate (i.e. unequal element length) transmission line networks. Through the use of approximations and computer optimization studies non-commensurate circuits are shown to be advantageous in distributed circuit design. The new method derives a non-commensurate circuit from a conventional commensurate prototype in such a way that the transmission matrix of pairs of elements in both circuits is made equal at one frequency. The transformation used ensures that the frequency response of the derived circuit closely matches that of the prototype circuit. Limitations on section characteristic impedances imposed either by constructional constraints or other practical realizability considerations are more easily met using the new design technique than when using conventional commensurate networks alone, Moreover, wide harmonically related passbands are largely avoided. The method finds greatest application where there are circuit length (or size) restrictions which must be met. Where such restrictions do not apply, conventional commensurate techniques are usually sufficiently flexible. The non-commensurate technique can however still provide benefits of even greater flexibility or better stop-band attenuation. The method presented is only an approximate equivalence and so an analysis of the technique is presented. The analysis establishes the degree of approximation. The use and application of the non-commensurate design technique are supported by an experimental investigation.

Three phase AC drives controlling cage induction motors have become widely accepted in industry, but one extant problem with this technology is that of increased acoustic noise emitted from the driven motor. This Thesis addresses the problem of the acoustic noise emitted from motors driven from voltage sourced PWM inverters and proposes a technique - Spread Spectrum Switching - for minimizing its effects. In the course of the work many other issues associated with real-time microprocessor-based PWM have also been advanced: • efficient microprocessor based PWM waveform generation, • harmonic analysis of generalized PWM waveforms, • compensation for the effects of power switch timing delays, and • compensation for the finite resolution of timers. The Thesis uses a variety of computational and analytical methods, backed by experimental observations, to quantify the improvement gained in each of these areas. Spread spectrum switching is a technique for eliminating the characteristically tonal structure of the acoustic noise emitted from a PWM inverter driven motor. Similar to the concept of spread spectrum communications, spread spectrum switching involves pseudo-randomly varying the instantaneous PWM switching frequency so that the energy of any PWM switching harmonics is dispersed over a wide bandwidth. This energy dispersion effectively eliminates any tonal components from the resultant motor acoustic noise while leaving the overall sound level largely unchanged; spread spectrum switching provides a significant qualitative yet minimal quantitative noise reduction. The PWM generation paradigm used in this Thesis is the recently reported Space Vector Modulation. A novel algorithm for microprocessor based space vector PWM generation is proposed, providing a basis for fast, efficient generation, even when overmodulating - a situation where many algorithms operate significantly more slowly. Furthermore, it is shown that the space vector method inherently generates a near optimum - in terms of motor harmonic loss - PWM waveform. However, when physically realized on a practical inverter such ideal PWM waveform s are corrupted by timing errors associated with both the inverter's power switches, predominantly the lockout time, and the finite resolution of hardware timers. Resolution corrected modulation is proposed for overcoming the problem of finite timer resolution and involves the use of integral feedback to account for any errors between ideal and physically realizable PWM switching times. This technique effectively provides 4 to 5 bits of added resolution to a given timer, allowing accurate waveform generation at low sinewave amplitudes and high switching frequencies using readily available, often microprocessor based, timers. Lockout times cause inverter output voltage errors, with consequent current zero crossing distortion, and a strategy for alleviating this problem is proposed and implemented in both a triangulation and space vector modulator. Two harmonic analysis techniques are proposed for analyzing PWM waveforms. The first technique is suitable for the analysis of regularly sampled PWM waveforms and has been used here to obtain closed form expressions for the harmonics of both space vector and asymmetrical triangulation PWM. These expressions show that PWM harmonics occur as a series of "combs" centered on multiples of the switching frequency. A second technique - the Directional Rotational Transform - is proposed for numerical analysis of general PWM waveforms. This technique uses an equivalent space vector representation of the PWM waveform, yielding the magnitude, phase and sequence (positive or negative) of the harmonics, and is useful in situations where each of the three phase waveforms is different, as in these cases Fourier Transform analysis of a single phase or line voltage only approximates the harmonics actually seen by the motor. The spectra generated using both these techniques compare favourably with those measured experimentally and, for synchronous PWM, those evaluated from Fourier Transforms. The culmination of modulation techniques presented in this Thesis yields a microprocessor based AC inverter drive featuring low acoustic noise emission at but a few kiloHertz switching rates and accurate PWM waveform generation using a single chip, low cost, micro-controller.

Studies of the families of variable length self-synchronising T-Codes are presented. The studies use Monte Carlo computer simulations to analyse the synchronising properties of the T-Codes and the relationships between these and their coding efficiency. Software and hardware implementations of the algorithms which are used to support these simulations are documented, and one of the encoding/decoding techniques presented is shown to be capable of operating at very high speed while being simple to implement. This technique is also shown to be applicable to the encoding and decoding of any variable length, exhaustive, instantaneous code. From the results of the simulations it is shown that some T-Codes consistently outperform others in the family with respect to their synchronising properties, and that the best of these codes have an average synchronisation delay of approximately 1.5 codewords. More importantly, it is shown that the T-Codes with the best synchronising performance are also the most efficient when encoding symbols from a specified information source, and that the coding efficiency of these codes is only nominally less than that of an optimal Huffman code generated for the source. A theoretical analysis of the synchronising properties is presented, and this is used to explain the relationships which exist between the average synchronisation delay and coding efficiency. These results are used to demonstrate that it is possible to choose a T-Code which has both the desirable properties of unrivalled synchronising performance and optimal coding efficiency, and guide-lines are given for choosing such a T-Code for a specified information source. This combination of a fast and simple encoding/decoding technique with codes which have unrivalled synchronising performance without sacrificing coding efficiency is offered as a solution to many of the problems normally associated with variable length coding techniques.

A new method for the synthesis of broadband impedance transformers having predictable passband frequency response is presented. The technique is based on the use of non-commensurate (i.e. unequal element length) transmission line networks. Through the use of approximations and computer optimization studies non-commensurate circuits are shown to be advantageous in distributed circuit design. The new method derives a non-commensurate circuit from a conventional commensurate prototype in such a way that the transmission matrix of pairs of elements in both circuits is made equal at one frequency. The transformation used ensures that the frequency response of the derived circuit closely matches that of the prototype circuit. Limitations on section characteristic impedances imposed either by constructional constraints or other practical realizability considerations are more easily met using the new design technique than when using conventional commensurate networks alone, Moreover, wide harmonically related passbands are largely avoided. The method finds greatest application where there are circuit length (or size) restrictions which must be met. Where such restrictions do not apply, conventional commensurate techniques are usually sufficiently flexible. The non-commensurate technique can however still provide benefits of even greater flexibility or better stop-band attenuation. The method presented is only an approximate equivalence and so an analysis of the technique is presented. The analysis establishes the degree of approximation. The use and application of the non-commensurate design technique are supported by an experimental investigation.