This thesis is missing page 155, the other copies of the thesis are also missing this page. -Digitization Centre / This thesis develops a two dimensional Viterbi Algorithm for the maximum likelihood sequence estimation over band limited baseband channels with intersymbol interference. Degradation, decision depth, 99% energy bandwidth and the channel cost are used as the performance measures for the comparisons of different channels. The four measures are extensively evaluated for channels with length up to four signalling intervals. The results of each measure are presented in contour form. Error events analysis shows that the degradation contours are governed by elliptical equations. Maximum degradation results from state path merge at a depth equal to the channel length plus one. By analysing periodic state sequences, we found that catastrophic error propagation contours are mainly governed by linear equations. Generally, channels with longer length have narrower minimum bandwidth but higher degradation. A channel cost similar to Shannon capacity equation is proposed to jointly minimize both degradation suffered and bandwidth required for signalling over a channel. According to the equation, the channel cost is influenced more by the bandwidth than by the degradation and thus the regions of low channel cost lie on the regions of narrow bandwidth. Also low channel cost regions are found to be on the regions of long decision depth and thus require higher complexity for maximum likelihood sequence estimation. In addition, it is found that minimum channel cost decreases with increasing channel length. / Thesis / Master of Engineering (ME)

International Telemetering Conference Proceedings / October 21, 2002 / Town & Country Hotel and Conference Center, San Diego, California / This paper presents noncoherent limiter-discriminator detection and differential detection of FQPSK (Feher quadrature phase-shift-keying) with maximum-likelihood sequence estimation (MLSE) techniques. Noncoherent FQPSK systems are suitable for fast fading and cochannel interference channels and channels with strong phase noise, and they can offer faster synchronization and reduce outage events compared with conventional coherent systems. In this paper, both differential detection and limiter-discriminator detection of FQPSK are discussed. We use MLSE with lookup tables to exploit the memory in noncoherently detected FQPSK signals and thus significantly improve the bit error rate (BER) performance in an additive white Gaussian noise (AWGN) channel.

Cooperative relaying systems aim to improve weak communication links by exploiting the spatial diversity obtained by the statistically independent channels between relays and the destination. In this thesis a cooperative relaying scheme called the Opportunistic Multiple Relaying (OMR) is proposed with its special receiver structure. Unlike most relaying schemes in the literature, multiple relay nodes are allowed to transmit in nonorthogonal channels in OMR without requiring any control overhead for relay coordination. OMR is compared to a benchmark scheme called the Selection Relaying (SR) in which the relay node is preselected by the source before transmission according to the average channel quality information. It is observed that OMR performs significantly better than SR in terms of error performance.

Enhanced Data Rates for Global Evolution (EDGE) is one of the 3G wireless communication standards, which provides higher data rates by adopting 8-PSK modulation in TDMA system infrastructure of GSM. In this thesis, a turbo detection receiver for Enhanced General Packet Radio Services (EGPRS) system, which is the packet switching mode of EDGE, is studied. In turbo detection, equalization and channel decoding are performed iteratively. Due to 8-ary alphabet of EGPRS modulation, full state trellis based equalization, as usually performed in GSM, is too complex not only for turbo detection but even for conventional equalization / so suboptimum schemes have to be considered. The Delayed Decision Feedback Sequence Estimation (DDFSE) is chosen as suboptimal and less complex trellis based scheme and it is examined as a conventional equalization technique firstly. It is shown that the DDFSE has a fine tradeoff between performance and complexity and can be a promising candidate for EGPRS. Then it is employed to reduce the number of the trellis state in turbo detection. Max-log-MAP algorithm is used for soft output calculations of both SISO equalizer and SISO decoder. Simulation results illustrate that proposed turbo detection structure improves bit error rate and block error rate performance of the receiver with respect to the conventional equalization and decoding scheme. The iteration gain varies depending on modulation type and coding rate of Modulation Coding Scheme (MCS) employed in EGPRS.

Recently, a study on obtaining better Euclidean distance for CPM (Continuous Phase Modulation) signals that fit well-known GSM spectral envelope has been carried out, and significant performance improvements were obtained. Two new pulse shapes, which are represented using 8th degree polynomials, were optimized to give the best error performance under the constraint that the PSD stays below GSM spectral standards. However, the approach uses parameters that cause the number of states to increase considerably, and thus yielding high complexity for receiver implementation. In this thesis, a study on finding a feasible receiver design that can provide a performance with acceptable degradation but affordable complexity is carried out for those new pulse shapes. After a survey about complexity reduction techniques, a decision is made to go on with a receiver structure based on Laurent Decomposition (LD) of phase modulated signals. Unlike other complexity reduction techniques, usage of LD based receivers permits reduction in both the number of matched filters and trellis states. Throughout the study, different numbers of matched filters and trellis states were used in LD based receivers for the new pulse shapes, and good results are obtained. For the pulse shape with pulse length L = 3, about a gain of 0.93dB in power is achieved by only 2 matched filters and 14 trellis states. For the case where L = 7, approximately a gain of 2.25dB is achieved with only 8 matched filters and 56 states, whereas 896 matched filters and 448 states are needed in the optimum case without complexity reduction.