Thesis (M. Eng.)(Electronic)--University of Pretoria, 2005. / Summaries in English and Afrikaans. Includes bibliographical references.

Thesis (MScEng)--Stellenbosch University, 2004. / ENGLISH ABSTRACT: The purpose of this thesis is to implement the Frequency Shift Keying (FSK) modem modules for the Software Defined Radio (SDR), using different Linux sound Application Programming Interfaces (API's). The FSK modulation scheme, uses coherent detection with matched filters. The modules are implemented using the Linux operating system and the programming language used is C/C++. Simulation is performed first and then followed by implementation in real-time, using APIs. The APIs used are Open Sound System (OSS), Advanced Linux Sound Architecture (ALSA), Jack Audio Connection Kit (JACK) and PortAudio (PA). In real-time two computers are used, one acting as the modulator and the other as the demodulator. The two sound cards are connected by an audio cable. Results obtained using OSS, ALSA and JACK are satisfactory. Recommendations are subsequently made for the suitable API(s). Both ALSA and JACK are the best APIs for the implementation. / AFRIKAANSE OPSOMMING: Die doel van hierdie tesis is om modules vir 'n frekwensieskuif-sleuteling (FSK) modem vir 'n Sagteware-gedefinieerde Radio (SDR) te implementeer, deur gebruik te maak van verskillende Linux klankargitekture (API's). Die FSK modulasieskema gebruik koherente deteksie met aangepaste filters. Die modules is geïmplementeer met behulp van die Linux bedryfstelsel en CjC++ is gebruik as programmeringstaal. Simulasies is eers gedoen, gevolg deur 'n reële-tyd implementasie. Die klankargitekture wat gebruik is, is Open Sound System (OSS), Advanced Linux Sound Architecture (ALSA), Jack Audio Connection Kit (JACK) en PortAudio (PA). Twee rekenaars is gebruik vir die reële-tyd stelsel, waar een as die modulator optree en die ander een as die demodulator. Die twee klankkaarte is verbind deur 'n klankkabel. Bevredigende resultate is verkry met behulp van OSS, ALSA en JACK. Aanbevelings is gevolglik gemaak vir toepaslike API's. Beide ALSA en JACK is die geskikste API's vir die implementasie.

Thesis (MScEng)--Stellenbosch University, 2005. / ENGLIAH ABSTRACT: Data transmission plays a cardinal role in today's society. The key element of such a system is the antenna which is the interface between the air and the electronics. To operate optimally, many antennas require baluns as an interface between the electronics and the antenna. This thesis presents the problem definition, analysis and performance characterization of baluns. Examples of existing baluns are designed, computed and measured. A comparison is made between the analyzed baluns' results and recommendations are made. / AFRIKAANSE OPSOMMING: Data transmissie is van kardinale belang in vandag se samelewing. Antennas is die voegvlak tussen die lug en die elektronika en vorm dus die basis van die sisteme. Vir baie antennas word 'n balun, wat die elektronika aan die antenna koppel, benodig om optimaal te funktioneer. Die tesis omskryf die probleemstelling, analiese en 'n prestasie maatstaf vir baluns. Prakties word daar gekyk na huidige baluns se ontwerp, simulasie, en metings. Die resultate word krities vergelyk en aanbevelings word gemaak.

Thesis (MScEng)--Stellenbosch University, 2005. / ENGLISH ABSTRACT: Different types of RF and Microwave receiver systems architectures are investigated. The superheterodyne receiver characteristics are studied qualitatively. A three components system (LNA, Mixer and VeO) is selected and the components are first measured individually, and then measured as a single system. A mono filar axial mode helix antenna is designed, simulated using FEKO and constructed. A narrow band interdigital bandpass filter is designed using the capacitance network and its transformation, and it is simulated using two different simulation packages, eST Microwave Studio and AWR Microwave Office. Two filter prototypes are fabricated using cylindrical rods. An open circuited microstrip resonator oscillator is designed and implemented through the negative resistance method. Its performance is verified using the harmonic balance simulator in AWR MWO. A single balanced mixer is designed and implemented using 1800 hybrid junction (rat-race) and it is also fabricated using microstrip transmission line technology. All designed components performed well in comparison to their predicted performance. The LNA and three of the designed components (Filter, Oscillator and a Mixer) are connected together and characterized as a single four block system. The system performed well except for the IF signal which shift to a lower frequency due to oscillator load pulling. / AFRIKAANSE OPSOMMING: Die argitektuur van verskeie RF en Mikrogolf ontvanger stelsels word ondersoek. Die "superheterodyne" ontvanger eienskappe is kwalitatief bestudeer. 'n Stelsel wat uit drie komponente bestaan (LNA, Menger en VeO) is bestudeer. Die individuele komponente is eers gemeet en daarna die stelsel as 'n eenheid. 'n "Mono filar" aksiale modus heliks antenna is ontwerp, gesimuleer met FEKO en gebou. Verder is 'n nouband inter-digitale banddeurlaat filter ontwerp met behulp van die kapasitiewe network metode en sy transformasie. Hierdie ontwerp is gesimuleer deur van twee verskillende sagteware pakette gebruik te maak, naamlik eST Microwave Studio en AWR Microwave Office. Twee filter prototiepes is gebou met silindriese stafies. 'n Oopgeslote mikrostrook ossilator is ontwerp en geimplementeer deur negatiewe weerstand te gebruik. Die stelsel se prestasie is bevestig deur gebruik te maak van 'n harmoniese balans simulator in AWR MWO. Verder is 'n enkelbalans menger ontwerp en geimplimenteer via 'n 1800 hibriede verbinding ("rat-race"). Hierdie menger is gebou met mikrostrook transmisielyn. Al die ontwerpte komponente werk goed in vergelyking met die voorspelde resultate. Die LNA, filter, ossilator en menger word gekoppel en gekarakteriseer as 'n vier blok stelsel. Die stelsel werk goed behalwe vir 'n skuif in die IF-sein by laer frekwensies a.g.v. belasting.

Thesis (MScEng) -- Stellenbosch University, 2002. / Some digitised pages may appear illegible due to the condition of the original hard copy. / ENGLISH ABSTRACT: Harmonic mixers are capable of extended frequency operation by mixing with a harmonic of the LO (local oscillator) signal, eliminating the need for a high frequency, high power LO. Their output spectra also have certain characteristics that make them ideal for a variety of applications. The operation of the harmonic mixer is investigated, and the mixer is analyzed using an extension of the classic mixer theory. The synthesis of harmonic mixers is also investigated, and a design procedure is proposed for the design and realization of a variety of harmonic mixers. This design procedure is evaluated with the design and realization of two harmonic mixers, one in X-band and the other in S-band. Measurements suggest that the procedure is successful for the specific applications. / AFRIKAANSE OPSOMMING: Harmoniese mengers kan by hoer frekwensies gebruik word as gewone mengers deurdat hulle gebruik maak van ‘n harmoniek van die LO. ‘n Hoe-frekwensie, hoe-drywing LO word dus nie benodig nie. Die mengers se uittreespektra het ook ‘n aantal karakteristieke wat hulle goeie kandidate maak vir ‘n verskeidenheid van toepassings. Die werking van die harmoniese menger word ondersoek deur uit te brei op die klassieke menger-teorie. Die ontwerp van die harmoniese menger word vervolgens ondersoek, waama ‘n ontwerpsprosedure voorgestel word vir die ontwerp van ‘n verskeidenheid van harmoniese mengers. Hierdie prosedure word getoets met die ontwerp en realisering van twee harmoniese mengers, een in X-band en die ander in S-band. Vanuit die metings is dit duidelik dat die ontwerpsprosedure geslaagd is vir die spesifieke geval.

The Thesis can be broadly classified into 2 sections. Initially it attempts to solve an existing long-term problem in the spark plug manufacturing industry, namely reliability and criteria for spark plug production testing. The second section of the Thesis develops the preliminary work and explores the fundamental theory of ignition sparks in preparation for the main study of ignition system and ignition spark mathematical modelling. Section 1 of the Thesis begins by introducing spark plug faults and existing production faults detection equipment and methods. An evaluation of potential new fault detection and classification systems follows, in two phases: detection and analysis. A suitable electrical test and neural network based production test system is then derived from the preceding work. This first section of the Thesis culminates in a factory evaluation of the prototype system at the sponsoring company. Section 2 of the Thesis uses experience gained from Section 1 to effect an investigation into the fundamentals of ignition spark development, leading to the mathematical modelling of Kettering-type ignition systems and associated spark profiles. Finally, the practical benefits of the modelling are discussed with respect to potential real automotive applications. There are two main novel aspects of the work. (1) The use of a neural network to analyse and classify ignition spark waveforms is believed to be a novel idea which can be further extended in many areas beyond the spark plug testing application discussed here. (2) Ignition system and spark modelling is a relatively unexplored research area. This work has resulted in an increase in the level of complexity of model and therefore a potential increase in precision.

This thesis investigates areas of neural networks and their application to aspects of image processing. Three neural network models, namely the backpropagation network, the Hopfield network and the competitive network, are studied. First, the learning algorithms for backpropagation networks and competitive neural networks are studied and new algorithms are developed. The applications of Hopfield neural networks to image coding, and of backpropagation networks to image data compression, are then investigated, and new techniques are presented. The learning algorithms for feed-forward neural networks are studied. A modified backpropagation algorithm is developed for accelerating the learning of backpropagation networks. Simulation results for four benchmark tasks are presented which show that the new method provides improved acceleration of learning, and global convergence characteristics. The learning algorithms for competitive neural networks are investigated. By incorporating principles of statistical mechanics into competitive learning, a new simulated annealing procedure for training competitive neural networks, the Stochastic Competitive Learning Algorithm (SCLA), is developed. Simulation results are presented which show that the SCLA is insensitive to the initial values of weight vectors, and can achieve a lower cost function value than other established competitive learning algorithms, and is therefore a valuable tool for data clustering. The computational power of Hopfield neural networks is employed for image coding applications. A new Hopfield neural network based block truncation coding (HNNBTC) technique is developed. The new HNNBTC is shown to provide improved performance over established block truncation coding techniques both in terms of mean square error performance and visual quality of the coded images. Two variations of the HNNBTC technique are also presented which are shown to provide increased compression ratios without sacrificing much of the visual quality of the coded images. The application of backpropagation networks to image data compression are investigated, and two new techniques are developed. The ceniralised MLP (CMLP) image data compression scheme is developed, which is aimed at improving the networks' generalisation capabilities, thereby enabling them to effectively compress a wide range of novel images. The learning edge patterns using MLPs (LEPMLPs) scheme, for image data compression is developed, which is aimed at improving the reconstruction of edges, thus improving the visual quality of the reconstructed images. Tabulated experimental results and example reconstructed novel images, for the original method and the new techniques, are presented, which demonstrate the improved image compression perfonnance gained using these new techniques.

Synthetic aperture radar (SAP) is gaining increasing importance in the field of remote sensing. The main feature of image data transmitted from orbital satellite synthetic aperture radar is all-weather performance, which explains its importance in many applications. Unfortunately SAP image data is susceptible to "speckle noise", which means that considerable processing is required before full use can be made of the data. This thesis describes research performed into the enhancement of linear features, corresponding to ships' wakes, in noisy SAP satellite images. A summary of noise reduction techniques is described, and the suitability of each to the problem of enhancing linear features in SAR images is discussed. A survey of available hardware is also included, with an assessment of its suitability for the selected technique, namely the Radon transform. A transputer network is described in detail, together with OCCAM software to perform the Radon transform within a few seconds. The Radon transform, performed via the two-dimensional frequency domain, uses a technique based on the Fourier slice theorem, and in this thesis a previously reported ambiguity is confirmed and analysed. The development of a new Radon transform is described, which enables transformation into Radon space (for contrast enhancement), and back into image space, all via the frequency domain. This new transform was also implemented in OCCAM on the transputer network, and it is shown that the new transform has advantages over a previously reported line enhancement techniq'ie in that it eliminates the ambiguity found in the standard Radon transform, and reduces computation considerably when the inverse transform is also required.

Thesis (PhD (Electrical and Electronic engineering))--University of Stellenbosch, 2005. / Individuals can be identified by their handwriting. Signatures are, for example, currently used as a biometric identifier on documents such as cheques. Handwriting recognition is also applied to the recognition of characters and words on documents—it is, for example, useful to read words on envelopes automatically, in order to improve the efficiency of postal services. Handwriting is a dynamic process: the pen position, pressure and velocity (amongst others) are functions of time. However, when handwritten documents are scanned, no dynamic information is retained. Thus, there is more information inherent in systems that are based on dynamic handwriting, making them, in general, more accurate than their static counterparts. Due to the shortcomings of static handwriting systems, static signature verification systems, for example, are not completely automated yet. During this research, a technique was developed to extract dynamic information from static images. Experimental results were specifically generated with signatures. A few dynamic representatives of each individual’s signature were recorded using a single digitising tablet at the time of registration. A document containing a different signature of the same individual was then scanned and unravelled by the developed system. Thus, in order to estimate the pen trajectory of a static signature, the static signature must be compared to pre-recorded dynamic signatures of the same individual. Hidden Markov models enable the comparison of static and dynamic signatures so that the underlying dynamic information hidden in the static signatures can be revealed. Since the hidden Markov models are able to model pen pressure, a wide scope of signatures can be handled. This research fully exploits the modelling capabilities of hidden Markovmodels. The result is a robustness to typical variations inherent in a specific individual’s handwriting. Hence, despite these variations, our system performs well. Various characteristics of our developed system were investigated during this research. An evaluation protocol was also developed to determine the efficacy of our system. Results are promising, especially if our system is considered for static signature verification.

Thesis (PhD (Electrical and Electronic Engineering))--University of Stellenbosch, 2003. / The imaging of platiniferous reefs in the Bushveld Igneous Complex (BIC) is of great economical and sociological importance. Borehole radar technology has been identified as a viable mapping tool to be used in day-to-day mining operations, but a critical assessment has to be made on the feasibility of this postulation. The system analysis made of the borehole radar deployed in the BIC is presented in this dissertation. The analysis is done using a specific example - the GeoMole borehole radar system. A novel procedure, based on the basic theory of electromagnetic radiation and propagation, is proposed by which the entire physical radar system can be characterized. The power transmitted by an unconventional borehole-deployed transmitter is estimated by a sequence of free space measurements, numerical simulations and theoretic derivations and approximations. Antenna transfer functions (magnitude and phase) are numerically simulated for a variety of deployment configurations. The total system transfer function of the receiver analogue and digital chain is determined. This enables the calculation of the radar's performance figures necessary to determine the applicability of the radar in a specific geological setting. A radar system is only complete when considered in its environment. The BIC is a stratified system of numerous rock layers. An in-depth study is done on the propagation of radiowaves in stratified lossy media. Only the case for non-magnetic media is discussed in this dissertation. The developed theory is used to predict the system response to a typical transmitted radar pulse in the UG1 - UG2 stratigraphy of the BIC, determine the maximum detection range of reef horizons and estimate the reflectivity of the reefs. Resolution is one of the key parameters that determine the performance and accuracy of imaging. An algorithm is proposed, developed and tested by which the resolution of the system is increased and overlapping echoes become resolvable. Even though some of the techniques are developed with a specific system in mind, the applicability of the concepts and algorithms is universal.