Search

Bibliography: leaves 149-155., This Thesis is a report of research done into the use of a GUNN Diode Millimetre Wave Source as both the Receiver and the Transmitter of a system. The Thesis is divided into three main sections. The first of these sections (Chapter 2) "sets the scene" so to speak. It presents certain preliminary fundamentals that are essential for placing the results obtained later into their proper context. Areas discussed are NOISE FIGURE, TRANSMISSION PATH LOSS and GUNN DIODE operation. The second section (Chapter 3) considers the operation of the device as a mixer. Its operation as a Microwave Power Source is a well-known concept and it will not be discussed here. This section covers such aspects as CONVERSION LOSS, NOISE FIGURE and BIAS CIRCUIT STABILITY. The third section considers those points arising from the second section which seem to be of interest. These include such aspects as Better Correlation of results and theory and ways of applying the device to modern Waveguide structures. The Thesis ends with a Conclusion, Appendix and a Bibliography.

Includes bibliographical references., Continuous Phase Modulation (CPM) possesses characteristics that make it very attractive for many applications. Efficient non-linear power amplifiers can be used in the transmitters of constant envelope CPM schemes. CPM also allows for the use of simple limiters in the demodulator rather than linear receivers with gain control. These characteristics not only increases the life of the power source, but it improves circuit reliability since less heat is generated. In some applications, such as satellite transmitters, where power and circuit failure is very expensive, CPM is the most attractive choice. Bandwidth efficiency, also, is very attractive, and improves as the order of the scheme increases (together with reduction in modulation index). Still further improvement is obtained through pulse shaping which normally result in partial response schemes as opposed to full-response (CPFSK) schemes. The inherent memory or coding gain of CPM increases the minimum distance, which is a figure of merit for a scheme's error performance. The length of the inherent memory is the constraint length of the scheme. Successful extraction of this inherent memory result in improved power efficiency. By periodic variation of the modulation index as in multi-h CPFSK, a sub class of CPM, coding gain or inherent memory can be significantly improved. CPM demodulation is also less sensitive to fading channels than some other comparable systems. Well-known schemes such as GSM digital mobile systems, DECT and Iridium all use some form of CPM to transport their information. These implementations are normally pulse-shaped FSK or MSK and are used for the reasons above, except that their receivers do not always exploit the inherent memory. Unfortunately, though, when one wants to exploit the inherent memory of higher level CPM schemes, all these attractive characteristics are offset by the complexity of the receiver structures which increases exponentially in complexity as the order or constraint length is increased. Optimum receivers for binary CPFSK were first described by Osborne and Luntz [19] in 1974 and their research was later extended by Schonhoff [26] to include M-ary CPFSK. These receivers evaluate likelihood functions after observing the received signal for a certain number of symbol intervals, say N, then calculate a set of likelihood parameters on which a likelihood ratio test regarding the first symbol is based. These receivers are complex and impractical but does provide valuable insight. This is called maximum likelihood sequence estimation (MLSE). Another way to do MLSE would be to correlate all possible transmitted sequences (reference signals at the demodulator) over a period of N symbol intervals with the received sequence. The first symbol of the reference sequence with which the received sequence has the largest correlation, is decoded as the most likely symbol. The number of reference sequences required at the receiver grow very fast as the observation period increases. Up to now, only the lowest order CPM schemes have feasible optimal receiver structures. The only practical solution thus far for the MLSE of higher order schemes is the use of software implementations of which the Viterbi algorithm is the most popular. Through recursive or sequential processing of data per interval, the number of matched filters required can be reduced. However, for schemes beyond a certain order and constraint length, the Viterbi algorithm's consumption of computational resources reduces its feasibility. Research into CPM is focused mainly on the quest for simpler demodulators and decoders or lower order schemes with better coding gain. In order to gain further insight into CPM, research is approached from different angles.

The report details the design of a dedicated Viterbi decoder chip set for an Ungerboek (3,2/3) Trellis Coded Modulation code. It was the specific intention of the thesis to design a system that could be implemented on standard Field Programmable Gate Arrays (FPGA) yet still be able to cope with high bit rates. The focus of the research was to both evaluate and modify the existing VLSI design techniques and to develop new techniques to make this possible. Trellis Coded Modulation refers to a specific sub-class of convolutional codes that ire an example of coded modulation. In coded modulation there is a direct link between the encoding and modulation processes aimed at improving the performance of the code by introducing redundancy in the signal set used to transmit the code. Ungerboek developed a technique for mapping the encoded words onto points in the signal set, called mapping by set partitioning, that maximises the Euclidian distance between adjacent codewords, and hence maximises the minimum distance between any two output sequences in the code. The Viterbi algorithm is a maximum likelihood decoder for convolutional codes such as TCM. The operation of the Viterbi algorithm is based on using soft decision decoding to produce an estimate of how well the received sequence corresponds with any of the allowed code sequences. The code sequences which most closely matches the received sequence is then decoded to form the output of the decoder. A central problem in implementing systems using TCM with Viterbi decoding is that although the encoder is a relatively simple device, the decoder is not. The complexity of the Viterbi decoder for any given TCM scheme will be the major drawback in implementing the scheme. As such techniques for reducing the complexity of Viterbi decoders are of interest to developers of communication systems. The algorithms describing the implementation and operation of the Viterbi algorithm can be categorised into three main layers. The top layer holds the theoretical algorithm itself, in the second layer are the set of algorithms that describe the broad techniques used to manipulate the theoretical algorithm into a form in which it can be implemented, and the third layer of algorithms describe the implementations themselves. The work contained in this thesis concentrates on the second two layers of algorithms.

Bibliography: page 79., This thesis sets out the procedures and results of an investigation into the performance capabilities of multi-h CPFSK digital modulation. Multi-h CPFSK is a coded modulation scheme. It exhibits high bandwidth efficiency and good envelope properties and is therefore well suited to use in satellite communications channels. However, the power efficiency of multi-h CPFSK signals is largely dependent on the particular multi-h code used, as well as the optimality of the decoder. Optimal decoding of multi-h CPFSK signals is achieved by means of the Viterbi Algorithm, applied to the multi-h phase trellis diagram. The Viterbi path metric is given by the squared euclidean distance between the particular path and the received signal.

Bibliography: p. 64., This dissertation investigates the synchronisation of Multi-h signals with a long constraint length. However, Multi-h codes exhibiting a long constraint length invariably also exhibit a large denominator q. Since traditional methods for synchronising Multi-h signals utilises a q-th power law device, where the frequency spectrum of the qth power of the signal renders all the necessary frequencies for synchronisation, we suspect that a large q could be detrimental. When simulated, it turns out that the qth power law device fails to deliver distinct (and useable) frequency components at a q size of about 8. Unfortunately, the most useful codes have denominators starting at a size of q equals 32. This called for a novel approach to synchronisation. One device that shows much potential is the new Massey-Hodgart coherent MSK demodulator. In a significant departure from standard quadrature structures, this MSK demodulator uses matched filter detection with a pair of reference signals at the two MSK signaling frequencies; an optimal maximum-likelihood bit decision is then formed over two bit intervals. The reference signals are recovered by a pair of decision-switched Costas loops, which are tightly integrated with the demodulator structure. The goal was to modify the Massey-Hodgart MSK demodulator into a Multi-h synchroniser that contained matched filter detection for all the frequencies in the Multi-h signal. The reference frequencies would still be decision switched and recovered by Costas loops.

Bibliography: leaves 82-91., Under certain atmospheric conditions, multipath propagation can occur. The interaction of radio waves arriving at a receiver, having travelled via paths of differing length, results in the phenomenon of frequency-selective fading. This phenomenon manifests as a notch in the received spectrum and causes a severe degradation in the performance of a digital radio system. As the total power in the received bandwidth may be unaffected, the Automatic Gain Control is not able to correct for this distortion, and so other methods are required. The dissertation commences with a summary of the phenomenon of multipath as this provides the context for the investigations which follow. The adaptive equalizer was developed to combat the distortion introduced by frequency-selective fading. It achieves this by applying an estimate of the inverse of the distorting channel's transfer function. The theory on adaptive equalizers has been well established, and a summary of this theory is presented in the form of Wiener Filter theory and the Wiener-Hopf equations. An adaptive equalizer located in a 34MBit/s QPSK digital radio is required to operate at very high speed, and its digital hardware implementation is not a trivial task. In order to reduce the cost and complexity, a compromise was proposed. If the tap weights of the equalizer could be represented by power-of-two binary numbers, the equalizer circuitry can be dramatically simplified. The aim of the dissertation was to investigate the performance of this simplified equalizer structure and to determine whether a power-of-two equalizer was a viable consideration.

The purpose of this report is to set out the results of an investigation into the artificial neural network (ANN) decoding of multi-h continuous phase modulation (CPM) schemes. Multi-h CPM schemes offer forward error correction (FEC) capabilities for continuous transmission, digital communication systems. Multi-h CPM is reported to be a bandwidth efficient alternative to other FEC techniques such as convolutional coding, while neural networks allow for high speed decoding. A neural network decoder was found in [12], where it had been used for the decoding of a convolutional code. This neural network structure by Xiao-an Wang and Stephen 'B. Wicker implements the Viterbi Algorithm (VA). All the necessary decoding information is contained in the interconnections of the ANN, and can be found by inspection of the state trellis diagram of the convolutional code. The decoder therefore requires no training. Since all the computation is done by analogue neurons and shift registers, the neural network reduces to a hybrid digital-analogue implementation of the VA. The use of analogue neurons allows the structure to be used for high data rate communications. Furthermore, the decoder is reported to be suitable for VLSI implementation.

The continuity properties of the CPFSK signal at it's symbol-period boundaries reveals an inherent memory contained in the transmitted signal. This is utilized as an error correction property. Furthermore, it was shown that the Multi-h CPFSK construction can be accomplished through the combination of a block constructing the memoryless component of the signal and either a block of digital logic circuitry or a continuous phase encoder constructing the memory component. The implementation of the first method was seen to function through simulations performed by using the TESLA simulation package. An extensive search for good Multi-h CPFSK h-sets was performed. The criteria for determining the performance of these h-sets was the Probability of Error gains over Minimum Shift Keying. The method of search was novel to this work. Specifically, a genetic search algorithm known as the Population Based Incremental Learning algorithm was utilized. The algorithm was implemented through the C++ programming language Faster error correction convolutional decoding algorithms were reviewed. Certain decoders exhibit lighter hardware demands, and in specific applications, are less susceptible to erasure problems. The Fano algorithm was selected as the best alternative to the Viterbi algorithm and was modified for the CPFSK implementation. The functionality of the implementation was tested using a C++ simulation. Various structures used to implement the synchronization and demodulation of Multi-hCPFSK were investigated. The most comprehensive structure that could be found was a scheme developed by Premji and Taylor using maximum likelihood techniques. This scheme was selected as it can be easily modified for the use with the large state, high speed implementation of non-symmetric Multi-h CPFSK investigated in this thesis. The PBIL algorithm was found to be an efficient method for finding good h-sets with large numbers of phase states. Theoretical gains over MSK using this method were found to be significant. It was concluded that the Fano decoder is highly applicable to the demodulator structure proposed in this thesis and is a preferred alternative to the Viterbi decoder under specific circumstances.

Spread Spectrum systems are found in many flavours, used in many applications and have existed since the early days of radio communications. The properties of spread spectrum do however place restrictions on the design, and often make the implementation expensive and complex. When using spread spectrum to provide a basic communications infrastructure, many factors need to be considered. These include supplying the appropriate technology at the right cost. To achieve this a trade-off against performance is often required. One of the more difficult aspects of Spread Spectrum design is the synchronisation of the spreading waveform. The primary characteristic of pseudonoise sequence synchronisation is the need for two levels of synchronisation namely acquisition (course synchronisation) and tracking (fine synchronisation). In these networks (the term network is used to describe a circuit or system throughout the thesis.) a decision is required to switch between the two synchronisation modes. The two layer structure of the typical pseudonoise sequence synchronisation network can increase the overall cost of spread spectrum systems. The objective of the research was therefore to find solutions to reduce the overall cost and complexity of the synchronisation network. The synchronisation structure should perform acquisition and tracking in a single structure, and thereby be low cost. To achieve the primary objective of this dissertation a. mixture of theory, simulations and practical implementation was used. The basis of the investigation was a time-variant spectral evaluation of pseudonoise sequences. It is shown that by multiplying a differentiated pseudonoise sequence with another pseudonoise sequence, useful information is obtained that can form the basis of a synchronisation network.

Bibliography: leaves 75-76., The aim of this project is to build and test a real time simulation for Trellis Coded Modulation(TCM). The tests to be performed are a comparison between Ungerboeck and pragmatic codes and varying and observing different Viterbi decoder parameters for coded 8PSK. TCM is coded modulation which means the choice of modulation scheme is linked with the encoding technique. Convolutional codes are important for an understanding of TCM. They are described by the rate (which is the number of inputs over the number of outputs) and the number of memory elements, v. For TCM schemes, soft decision decoding, based on euclidean distance, rather than hard decision decoding, based on hamming distance, is used. Ungerboeck developed a mapping of encoder bits to channel signals on a constellation diagram. The mapping is called mapping by set partitioning and aims to find the smallest free euclidean distance for a given code. The free euclidean distance is the minimum euclidean distance between output sequences in a code. The asymptotic coding gain is a measure of coding gain based on the free euclidean distance.

Includes bibliographical references (leaves 74-75)., Multi-h Coded Continuous-Phase Frequency Shift Keying (Multi-h CPFSK) has gained interest in recent years because it offers an additional degree of freedom in the coding of CPFSK. Similar to Trellis Coded Modulation (TCM), it does not use redundancy to achieve coding gain. Hence with properly chosen modulation indices, impact to spectral occupancy is kept to a minimum. While there has been less attention given of this method as compared with TCM, this method can also be used with data coding. In cases where data coding is to be implemented, simultaneous use of Multi-h coding can be implemented with very little increase in complexity. In this thesis, a thorough mathematical review of this technique is made. A multi-oscillator multih coded modulator is shown similar to one presented by Massey for MSK. A unique analytical tool called a multi-oscillator trellis is presented. This considers the phase transitions with respect to each of the signalling frequencies instead of the center frequency, f₀. The multi-oscillator trellis is used to determine the state machine that will switch a bank of oscillators. The purpose of the state machine is to maintain continuous phase at the multi-oscillator output while generating the proper signal frequencies according to the data and modulation index. The Maximum Likelihood Detection process at the receiver is shown as a partition of an uncoded CPFSK signal. Finally, an analysis is made to determine if a modulator with a non-linear frequencyvoltage characteristic is suitable in a coherent multi-h coded application. Much of the literature on this topic has been comparative to PSK. It is the intent of this work to use FSK and MSK as the baseline to determine how existing structures may be extended to realize the benefits of multi-h coding. The application of this coding to an 8 Mbps 23 GHz CPFSK point-to-point terrestrial communications system is also a topic of this thesis. It is in this context that the analysis is made.