Bioimpedance spectroscopy is a study of the variation of tissue’s electrical properties, both conductive and dielectric, through a frequency spectrum (100Hz to 1MHz). It involves the application of AC signals to the surface of the tissue via electrode pairs, which can either be a current signal or a voltage signal, and the resulting surface signals are recorded via a separate or the same pair. The recorded signals are converted to impedance measurements via a demodulation procedure. The study of tissue’s electrical properties can provide with useful information regarding both its physiology and pathology. The main aim of the work described in the thesis is the design and development of novel CMOS current drivers, for wideband tetrapolar bioimpedance measurements (100Hz- 1MHz). The work presented resulted in the fabrication and experimental validation of two current driver circuits which offer superior performance relative to existing designs. The first design was fabricated using a 0.6 µm CMOS technology and occupies a silicon area of 0.64 mm2. It can deliver a maximum output current of 5 mAp-p operating from a ±9V supply (15V output compliance). The output impedance is 665 kΩ at 100 kHz and 372 kΩ at 500 kHz. The second design was fabricated using a standard 0.35 µm CMOS technology and occupies a silicon area of 0.4 mm2. It can deliver a maximum output current of 1 mAp-p operating from a ±2.5V supply (4V output compliance). The output impedance is higher than 1 MΩ up to 500 kHz reducing to 360 kΩ at 1MHz, thus further improving the performance up to 1MHz. A custom-made bioimpedance measuring device is also presented which incorporates the second current driver that has been tested in both in-vitro and ex-vivo experiments (using post-operative human colon cancer specimens). Experimental results verify a stable and reliable operation of the circuit up to 1MHz. Finally a third current driver, using a standard 0.35 µm HV CMOS technology, is presented (its silicon implementation is a future plan) which aims to improve upon certain performance limitations of the previous designs. Simulated results demonstrate an input/output phase (~2.08º) at 1MHz (better than any reported CMOS current driver), and an output impedance of 1MΩ at 1MHz. The current driver employs a fully automated continuous time common mode feedback (CMFB) loop to accurately set the output DC levels which has not yet been implemented on any existing CMOS design.

In spinal cord injured (SCI) individuals the neural pathway between the brain and the extremities is damaged. However, there is still the capacity to elicit muscle activation despite the absence of any supraspinal input. Recent studies have proposed epidural spinal cord electrical stimulation (ESCS) as a means to facilitate locomotor recovery in SCI humans. In epidural stimulation a number of electrodes are placed in the spine, outside the dura, and stimulus current pulses are used to ‘tune’ the spinal circuitries. Some rat studies have supported this concept, but further testing is required to increase our understanding and optimise the stimulation parameters. Testing protocols are currently limited by the available technology. More specifically, the number of electrodes one can use seriously limits the paradigms that could be investigated. For this reason, electrode arrays, as opposed to the conventional pairs of electrodes, can be used to investigate the effect of ESCS at different sites. The development of epidural electrode arrays for chronic testing in rats is a challenging task due to their small size. The difficulties increase radically when a large number of electrodes need to be independently controlled. It has been well documented in the literature that a large number of connections (wires) is highly undesirable because it either makes the implantation procedure more challenging, or, if the device is successfully placed in the body, it could imperil perfusion, result in infections, tissue damage, or simply cause the device to fail. The development of a flexible epidural electrode array suitable for chronic implantation in rats was the main goal of this work. For the first generation of the system, flexible passive 12-electrode arrays, using silicone rubber and annealed platinum foil, were designed and fabricated—suitable for use with an external stimulator. In vivo evaluation of these devices showed that they failed quickly, 87.5% of the connections after a week inside the spine of a rat. The failure analysis performed highlighted the need to reduce the number of connections to avoid inflammation and improve the mechanical stability of the implants. To overcome the connections ‘bottleneck’ without compromising the number of electrodes (which was necessary for the planned paradigm), our approach was to develop application-specific integrated circuits (ASICs) to be embedded on the arrays, acting as electrode drivers. The ASICs reduce the number of connections to 3, feature a very small silicon footprint (less than 0.36 mm2 core area), consume very low power (up to 114 μW during a full stimulation cycle), and allow for the necessary versatility for the testing with a real-time control system, developed by our collaborators (in the FP7 NEUWalk project). A custom designed ‘hub’, designed by Dr. Clemens Eder, is used to electronically – rather than manually – manage the stimulus parameters and the operation of the ASICs. It can be programmed via a graphical user interface (GUI) or the real-time controller. Moving to the second generation of the system, active (with embedded ASICs) epidural arrays were designed, developed and evaluated. For this version, platinum iridium foil, was preferred, due to its superior mechanical strength. The next part of the work focused on the the different aspects of the fabrication and assembly processes. More specifically, size restrictions related to the implantation site dictated the need to use thinned ASICs. To post-process the already fabricated chips, a method for purely mechanical silicon thinning at individual die level was developed and characterised. For the integration of the ASICs on the arrays an evaluation study was conducted to examine the mechanical reliability of the bonds produced by electrical rivet bonding. Combining all the above, a new fabrication process was developed for the active arrays. Despite the fact that, so far, chronic in vivo testing has not been yet implemented, the produced prototypes were electrically and mechanically evaluated in vitro, and results are satisfactory, as no failed tracks were observed during the chronic tests in the lab. The current setup allows power and data to be transferred to the implant real-time through a connector fixed on the rat’s head, while the animal is on a treadmill or on a runway. This implies that there is no need for a wireless system at this stage. However, more complex experiments where the rats would be able to move freely and interact with other rats unrestricted, developing a behaviour that is closer to their natural, could provide significant new knowledge in the future. Although there are still many things to understand regarding epidural stimulation and its effect before planning an experiment like this, this was kept in mind throughout the whole design and development phases of this system. On this basis, the developed subcomponents are compatible with a system level design of a fully implantable platform to be used in freely moving rats, stimulated for 3 – 4 hours per day. This system comprises the active electrode array, which is the focus of this thesis, together with a miniaturised, battery-powered implantable version of the previously mentioned hub (which is on-going work, and is not presented here).

This thesis presents a study of material characterisation methods and techniques, which can determine the effects of surface and structural defects on the properties of single crystal rubrene. A method of cleaving is devised to produce surfaces free from previous exposure to the ambient. This method reproducibly provides large terraces intersected by straight, well orientated step edges. Atomic force microscopy studies show the cleaved surfaces undergo environmentally dependent molecular reorganization. This results in the formation of nanoscale ‘beads’ at molecular step edges and narrow ‘fingers’, one molecule high. The beads show an insulating behaviour with increased conduction at the surrounding surface. The methodology of applying time of flight secondary ion mass spectroscopy to rubrene crystals is developed to study the chemical composition both at and below the surface. This shows a uniform oxide (C42H28O) covering the surface with an increased concentration of peroxide (C42H28O2) located at crystallographic defects. To investigate the effect of surface and defects on exciton dynamics in rubrene a confocal photoluminescence (PL) arrangement is designed and built. An extended PL distribution is imaged providing evidence of exciton diffusion within the material. This diffusion is seen to increase within the bulk with a suppression of emission at 603nm. Defects are seen to affect PL with emission of the 650nm PL band having a greater contribution in the presence of defects. This emission is also spatially displaced from the maximum intensity of the other bands. These results imply the existence of a defect mediated recombination pathway. These studies show that environmental reactions readily occur at the location of crystallographic defects and step edges. This is of importance to the operation of rubrene electrical devices. This work provides a set of techniques and developed methodologies which enable the characterisation of technologically important processes on rubrene. These should extend to other organic single crystals such as pentacene and tetracene.

The emergence of co-located Multiple Input - Multiple Output (MIMO) radar in recent years has meant that snapshot imaging of a target scene can be achieved without the large number of antennas and circuits of a conventional phased array radar, for a given angular resolution. This thesis presents two applications of MIMO radar systems. Firstly, the cross sectional imaging of ice shelves in Antarctica. Two trials were conducted using a single channel SFCW radar to collect data, one from a 2D MIMO array and another from a linear MIMO antenna layout. The results from the signal processing of the data are then provided to highlight the effectiveness of the MIMO antenna layout for ice shelf imaging. The second application is the identification of obstructions at railway level crossings, which require real-time detection of large targets within the boundaries of a level crossing. For real-time imaging, however, the use of multiple transmitters and receivers has now placed constraints on other parts of the radar system for fast data capture and signal processing. To alleviate this problem, the application of sparse sampling in angular space for MIMO radars is investigated to reduce hardware costs and the amount of data required for a snapshot image. Results from a practical demonstration of the sparse array using an FMCW radar as well as from further simulations using a combination of random number PDFs are presented. For ice shelf imaging, a stationary MIMO radar is capable of producing the depth profile without the need for a moving SAR system. A sparsely sampled MIMO antenna array can aid real-time detection of obstructions in level crossings with minimum hardware costs. The results presented in this thesis provide opportunities for further developments in MIMO radars for both geophysical and terrestrial applications.

Indium arsenide (InAs) nanowires are attracting a growing interest in the semiconductor industry as their remarkable properties make them ideal candidates for future applications in a wide variety of electronic, photonic and sensing devices. In the present work, InAs nanowires are grown via solid-source molecular beam epitaxy on Si (111) substrates without the use of heterocatalytic nanoparticle seeds. The native oxide layer forming easily on InAs nanowires must be removed prior to metallisation to achieve highly transparent contacts. We present a systematic comparative study of the contact resistance between InAs nanowires and metals following the use of (a) a wet etching in an ammonium polysulfide solution or (b) an argon milling process. Nanowires treated by the argon milling process with in situ deposition of metallic contacts exhibit a contact resistance which is more than one order of magnitude lower than that of nanowires treated with ammonium polysulfide. From fourpoint measurements, an upper bound of 1.4×10−7 .cm2 is extracted for the contact resistivity of metallic contacts on nanowires treated by the argon milling process. While the growth of semiconductor nanowires on silicon allows their direct integration with the established CMOS technology, the absence of a heterocatalyst usually results in a pronounced polytypism in the nanowires, proved to be detrimental to their optical and electrical properties. To solve this issue, InAs1−xSbx nanowires (0 _ x _ 0.15) were grown on silicon substrate via a catalyst-free MBE process. We observed a sharp decrease of stacking fault density in the InAs1−xSbx nanowire crystal structure with increasing antimony content. InAs0.85Sb0.15 nanowires exhibit a mobility three times larger than InAs nanowires. Finally both magnetic-field-dependent and gate-voltage-dependent measurements of universal conductance fluctuations and of localisation effects were performed on InAs and InAs1−xSbx nanowires at low temperature. From the analysis of the fluctuation amplitude and the correlation field, a phase-coherence length in the hundred nanometre range is observed for all nanowires below 10 K.

One-dimensional nanostructures such as semiconductor nanowires are very attractive for application in next generation electronics. This work presents an experimental study of InAs-based and ZnO-based nanowires grown by molecular beam epitaxy for electronics applications. InAs, InAsP and InAsSb nanowires were grown self-catalytically on silicon. Phosphorus incorporation was studied by means of HRTEM, XRD, EDX and PL. The phosphorus incorporation rate was shown to be 10 times smaller than that of arsenic. InAs and InAsP nanowires exhibit the wurtzite structure with a high density of stacking faults and phase boundaries. Conversely, InAsSb nanowires exhibit the zincblende structure with the density of stacking faults decreasing as the antimony content increases. Antimony incorporation and reduction of the stacking fault density improves the nanowire mobility. ZnO and ZnMgO nanowires and ZnO/ZnMgO core-shell nanowire heterostructures were grown by plasma-assisted molecular beam epitaxy on various substrates with gold particles as a growth catalyst. Nanowire growth was shown to occur only at temperatures between 700 and 850 C and Zn pressures between 1 and 3 10 7 Torr. A two-step growth procedure on silicon was implemented to increase the yield of nanowire growth. Mg incorporation was shown to be 4 times smaller than that of Zn. At Mg content higher than 20 %, MgZnO rocksalt phase segregation is observed in the as-grown samples. Core-shell nanowires were fabricated by growing the shell at a lower temperature of 500 C. ZnO nanowire field effect transistors were fabricated and optimised. High- and low-temperature transport measurements allowed determination of the bulk nanowire and contact properties. Nanowires grown on sapphire and silicon were compared. Nanowires grown on sapphire exhibit an extra donor that determines their low temperature conductivity and give a wider photoluminescence band-edge emission peak. A novel technique to measure the spectrum of deep traps in nanowire field effect transistors was implemented to study ZnO nanowires.

High performance digital technologies are rapidly requiring increasing signal bandwidths and higher resolutions, and are being used in ever increasing areas. To be able to achieve the increasing digital requirements, higher speed analog-to-digital conversion (ADC), with greater resolution, is required. At the highest speeds and signal frequencies, applications include, optical communications, radar, high speed test and measurement, and electronic warfare. Current electronic ADC technology is stagnating. These limitations can be attributed to poor timing jitter of electrical systems. Optical sources have been demonstrated with timing jitter orders of magnitude lower than the best demonstrated for electrical systems. This work presents an ADC with a photoconductive switch sampling system using an In- GaAs photoconductive switch with a telecommunication wavelength cavity-less optical source. The photoconductive switch is able to operate over a wide frequency bandwidth (>20GHz), and the optical source used is able to produce a low timing jitter pulse train at gigahertz repetition rates. The optical source developed here is based on a cavity-less design, with a single electro optic modulator. A description of the design and implementation of the optical source is presented. The key parameters of the optical source, for optical sampling, are tested alongside two mode locked lasers for comparison. The lasers are a semiconductor laser which is fundamentally mode locked, and a bre ring laser which is harmonically mode locked. The cavity less source has been tested to have the lowest timing jitter of the three optical sources. Typically low temperature grown (LT) GaAs is used as the photoconductive material for optical sampling, however due to the large band gap energy, optical sources with a wavelength of 800nm or less is required. The choice of lasers at this wavelength are limited compared to optical sources at longer wavelengths. To be able to work with a laser operating at telecommunication wavelengths a photoconductive switch with a band gap energy of around 0.8eV is required. InGaAs has an appropriate band gap energy to work with telecommunication wavelength lasers, however InGaAs ultrafast materials are not as well established as LT-GaAs. Nitrogen implanted InGaAs has been used as the ultrafast material for photoconductive switches here. This material has been tested alongside LT-GaAs switches for comparison. It has been measured that InGaAs switches have a high dark current compared to LT-GaAs switches, which is expected with the lower band gap energy. The InGaAs device has been measured to have lower insertion losses (for the same optical power at 800nm) and better linearity than the LT-GaAs device measured here. Finally, a photonic sampling ADC system using an InGaAs sampling switch and a telecommunication wavelength laser has been demonstrated for the rst time. This system demonstrated a sampling rate of 7.2GHz and an ENOB of 3.9 for signals up to 8GHz, with a bandwidth limited by the electronic ADC to 900MHz. A similar system using a LT-GaAs and a Ti:sapphire laser has previously been published, which demonstrated an ENOB of 5.9 for a bandwidth of 750MHz and sampling rate of 1.5GHz. The di erence in ENOB is due to the lower sampling rate used in the LT-GaAs system. However, the optical source used for the InGaAs system is cheaper, smaller and more robust than the Ti:Sapphire laser used with the LT-GaAs system.

The trend in the construction of the latest array antennas is such, that electronic beam scanning can be conducted for a significantly large angular sector. Indeed, as far as circular array antennas are concerned, the quality of the radiation pattern will not be impaired when scanned in the full azimuthal axis. Compared to planar and several different shapes of conformal antenna arrays, spherical array antennas have the unique property of scanning their beam with no deterioration in full spherical coverage (i.e. azimuth and elevation axes). Furthermore, multi-beam functioning and direction finding techniques can be applied to spherical array antennas. In contrast to antennas that use mechanical motors for steering, they exhibit higher speed and stability. Low Earth Orbit (LEO) satellite communications can utilize the aforementioned properties of spherical array antennas. Spherical harmonic (phase mode) theory constitutes the base on which the signal processing of the spherical array antenna is developed. In this analysis, spherical harmonics are employed, which take advantage of the sphere’s symmetry and regenerate themselves in the far-field. The spherical phase mode approach is able to provide considerable improvements and computational simplification in several array processing levels. Specifically, research efforts are concentrated toward the choice of an optimal uniform spherical distribution and the minimum number of array elements that are necessary in the process. Various antenna element distributions are investigated and compared to each other placing special emphasis on the design and fabrication efficiency. Additionally, the direction finding potential of a spherical array antenna is explored by applying one of the main DOA estimation algorithms used with spherical arrays, the spherical ESPRIT algorithm. A spherical array, called LISA, has been designed and a prototype demonstrator has been manufactured. The spherical array has a radius of 40 cm, and operates at 3 GHz, employing circular polarization. A multi-planar approach, with 240 triangular planar tiles approximating the spherical surface, is adopted. Special attention is given to the RF front end of the array, and especially to the development of the microstrip antenna design that is accommodated on each tile (three elements) and a bespoke feed network. Valuable simulation and measurement results are also provided, demonstrating satisfactory performance of the tested tiles and offering essential conclusions for discussion and further investigation.

The increases in power dissipation and transient fault rates experienced in applications based on state-of-the-art integrated processor technologies now lead to bottlenecks in energy consumption, processing-throughput scaling, and system reliability. Therefore, application and system designers are now beginning to accept the notion of error tolerance across the system stack, i.e., new methods that detect and tolerate (or mitigate) faults in exchange for resource efficiency at the application, runtime, compiler, architecture and hardware layers are now under intensive investigation. Within this context, this thesis proposes and develops novel approaches to data packing and numerical entanglement for accelerated, error-tolerant, signal processing applications. In particular, different data packing strategies suitable for integer and floating point inputs are studied, and they are subsequently used to create the new concept of numerical entanglement, which is proposed for highly-reliable numerical processing of integer data streams. The results of this thesis demonstrate that up to seven-fold decrease of processing time or energy consumption can be obtained if the signal processing application can operate under increased margins in the mean-squared error or signal-to-noise ratio of linear or sesquilinear processing kernels. In addition, for methods that scale processing throughput or energy consumption at the expense of reliability, our proposal for highly-reliable numerical processing of integer data streams is shown to detect all possible system-induced errors in one out of M input streams (M=3) with minimal overhead. Therefore, the thesis proposals can be used in software or hardware systems for resource scaling of error-tolerant signal processing applications.

An overview of current spectrum management methods was carried out and it was decided that to be able to accommodate the predicted explosive growth in wireless data services changes would have to be made to avoid spectrum depletion becoming a very serious future problem. In addition, a review of numerous spectrum exploitation studies and measurement campaigns showed that nationally spectrum was being significantly underexploited. To address this apparent anomaly, it was decided that finding a means for improving spectrum exploitation nationally may provide a potential solution for the spectrum depletion problem. This is referred to as spectrum sharing, the focus for the DEng research. A new spectrum management model named Opportunistic Spectrum Management (OSM) is proposed on the basis that all spectrum licenses are awarded on a non-exclusive usage basis and spectrum use is based on a multi-level priority structure, because certain types of services require performance and spectrum availability guarantees. The proposed model retains backward compatibility with the traditional spectrum management model and its recent enhancements. To fully exploit the OSM model, the thesis proposes an automated spectrum management system name Opportunistic Spectrum Sharing System (OSSS) capable of dynamically allocating spectrum and enforcing spectrum usage priority rules both nationally and in real-time. OSSS overcomes limitations of traditional sensing based cognitive radio technology primarily its inability to distinguish between different spectrum usage classes and the hidden node problem. A key issue associated with computing acceptable power levels in an automated spectrum sharing system produced over-optimistic results. The thesis shows how this can be extended to improve its accuracy without compromising its closed form nature. Finally, the thesis concludes with a discussion of the implications and commercial benefits of opportunistic spectrum sharing from the perspectives of various wireless stakeholders, which is supported by extracts from the author’s various publications.