Design and analysis of digital receivers

Sammanfattning: This thesis consists of a summary and nine included papers, grouped into three parts. There is one journal paper, three reports written in the style of articles, four conference papers and one paper submitted to a conference. The thesis proposes and investigates a number of digital receivers, especially receivers based on the maximum a posteriori and maximum likelihood criteria. Signal processing methods and models are developed and applied to a number of estimation and detection problems in systems with time dispersion and additive Gaussian noise. Digital receivers in two application areas are investigated: telecommunications and ultrasonic distance measurements. Within telecommunications, particular attention is given to block transmission systems, where digital data is transmitted in independent blocks. With a geometric approach and reflecting on properties of the binary hypercube, it is shown that the minimum bit-error probability receiver (OBER) becomes the maximum likelihood sequence detector (MLSD) when the expected SNR used for designing the OBER goes to infinity. Likewise, the OBER reduces to the whitened matched filter with hard decisions in the limit when the expected SNR decreases. Furthermore, a novel detector is developed that makes MLSD-decisions on scattered bits in a block. This low-complexity detector can, if combined with a sub-optimal receiver such as a linear or decision- feedback equalizer, substantially reduce the system bit-error rate. Finally, using the geometric approach, the genie-aided detector, a device proposed by Forney for deriving performance bounds, is reconsidered and augmented with an explicit statistical description of the side information. This renders a more flexible tool, new performance bounds, and gives an instructive view on earlier work. Reduced complexity Viterbi detection is addressed by means of combined linearViterbi equalizers. These equalizers reduce the complexity of the Viterbi detector, a structure for implementing the MLSD, by linear pre-equalization of received data and by giving the Viterbi detector a truncated channel model. Three receivers in this class are introduced, one of which is intended for multiple-antenna reception in block transmission systems. In the field of ultrasonics, the problem of estimating the time-of-flight of an ultrasonic pulse is addressed under the assumption that the pulse has been distorted by an unknown, linear and time-dispersive system and by additive Gaussian noise. Two approaches for taking the linear distortion into account are presented. Both assume that the transmitted pulse is known and narrowband. Although the intended application is distance estimation using the ultrasonic pulse-echo method, the assumed basis of the time-of-flight estimation problem is more general: a known narrowband waveform is transmitted through a dispersive, linear system with additive Gaussian noise.