EEMCS EPrints Service


Korevaar, C.W.
(2016)
Hermitebased communications; a timefrequency perspective.
PhD thesis, Univ. of Twente.
CTIT Ph.D.thesis series No. 16397
ISBN 9789036541367
Full text available as:
Official URL: http://dx.doi.org/10.3990/1.9789036541367 AbstractThe radio spectrum is like a crowded house. Everyone is welcome, but when more and more guests are entering and the space is limited, there might not be enough room for everyone. Entry fees may be charged and you need to speak loud to make yourself heard in the overcrowded setting.The radio spectrum is such a crowded house. Wireless communications has become increasingly popular over the last century and wireless devices popup literally verywhere. As there is limited bandwidth available for radio communications, a fierce competition for bandwidth takes place between the wireless devices. Bandwidth auctions have become billion dollar affairs. Furthermore, it is costing more and more energy to transmit a message without errors in the overcrowded spectrum. To enable a future with even more wireless devices and datahungry applications, spectrum and energyefficient communications is of everincreasing importance. Since the early days of communications we have relied upon the modulation of (timelimited) Fourier basis signals, better known as sinusoidal signals. Timelimited signals are frequencyunlimited which leads to spectral leakage and a degradation of the spectrumefficiency when there are multiple, unsynchronized spectrum users. Bandlimited signals would eliminate spectral leakage, but are unlimited in time. Ideally, we would like to design signals which are both bandlimited and timelimited. However, there is a fundamental bound, the uncertainty principle, which limits the extent to which one can design such signals. This thesis studies optimally timefrequency localized signals: Hermite functions. Hermite functions form an orthogonal signal basis and are optimally localized in timefrequency. Their mathematical properties make Hermite functions – from a theoretical point of view – an interesting candidate for wireless communications; in particular for unsynchronized, multiuser communications. Besides mentioned characteristics of Hermite functions, more criteria play a role in the design of an effective and efficient communication system. Effective indicates the extent to which the transmitter is able to send a message to the receiver which is wellinterpreted, whereas efficient means that the messages are transmitted with the least amount of time, bandwidth, energy, power and (computational) complexity. A list of criteria is formulated in this thesis for effective and efficient (multiuser) communications. Among the criteria are the energy efficiency, power constraints, synchronization, equalization and the sensitivity for timefrequency dispersion/offsets. As most of the signal criteria impose similar constraints on the time representation as well as the frequency representation, a more equal treatment of these domains seems justified. Therefore, timefrequency analysis, the twodimensional description and analysis of signals in timefrequency, plays a prominent role throughout this thesis. Compared to the Fourier basis and Fourierbased communications, relatively little is known about Hermite functions and in particular Hermitebased communications. To assist signal analysis for Hermite functions, in this thesis closedform expressions are derived for popular mathematical operators involving Hermite functions, including the product, convolution, correlation, Wigner distribution function (WDF) and ambiguity function (AF). It was already known that these mathematical operations performed on Gaussian functions (Hermite functions of the zerothorder) lead to a result which can be expressed as a Gaussian function again. We generalize this reciprocity to Hermite functions of arbitrary order. The product, convolution, correlation, WDF and AF operations performed on two Hermite functions of order n and m lead to remarkably similar closedformexpressions, which can be expressed as a bounded sum of n + m Hermite functions.The connection and difference between all these mathematical operations is primarily determined by distinct phase changes of the weights of the Hermite functions in the result. In addition to spectral leakage, another fundamental problem in (multicarrier) communications is the high peaktoaverage power ratio (PAPR) of the transmit signals. Both Fourierbased and Hermitebased communications are characterized by transmit signals which can lead to high peak powers. These peak powers impose a demanding constraint on the dynamic range of analog building blocks like the power amplifier and lead to a degradation of the overall energyefficiency of the communication system. In this thesis a novel PAPR reduction technique is proposed which is based on rotating the timefrequency presentation of the transmit signal. It effectively shifts the peak (powers) manifesting in the time domain to the frequency domain. Simulations and comparisons with existing techniques show that PAPR reduction by timefrequency rotations is an intuitive and effective method to reduce peak powers. Hermitebasis signals are neither stationary in time nor in frequency. Therefore, synchronization needs to be established in both domains. A novel correlation procedure is proposed called spiral correlation. Instead of correlating either in time or in frequency, spiral correlation is a onedimensional formulation of correlation in timefrequency, starting at one point and following a trajectory according to the turns of the spiral. Spiral correlation, using the pattern of a sunflower, is simulated and evaluated. The proposed method can lead to a significant reduction in the computational complexity necessary for synchronization. In addition, it is shown that fractional delay filters – necessary for synchronization – can be omitted by using fractional Fourier transform (FrFT) identities. When designing a Hermitebased communication system, more criteria are relevant and require attention, including numerical evaluation, sampling, sensitivity for timefrequency offsets, eliminating the DC component and stacking multiple symbols ofHermite functions.These criteria have been assessed by theoretical analysis, simulations and actual measurements on the first Hermitebased transceivers (for as far as known by the author).The conclusion is that Hermite functions are an ideal candidate as a signal basis for tomorrow’s transceivers. However, two potential barriers are identified that block the widespread adoption today. A first concern are the signal characteristics of Hermite functions, which are dependent on the order of the Hermite function. E.g., both the timebandwidth product and the timefrequency localization of Hermite functions are linearly dependent on the order of the Hermite function, causing a different robustness against timefrequency offsets and dispersion. It has been partly overcome by the introduction of a new set of signals; FourierHermite functions. FourierHermite functions average out the properties of individual Hermite functions, but not to the full extent. A second concern which remains is the nonorthogonality of timeshifted symbols of Hermite functions.The interference caused by overlapping Hermitebased symbols is insignificant in many scenarios, but can cause a degradation of the bit error rate (BER) in some use cases and lownoise regimes.
Export this item as: To correct this item please ask your editor Repository Staff Only: edit this item 
