Uppsala University Signals and Systems SPACE-TIME PROCESSING AND EQUALIZATION FOR WIRELESS COMMUNICATIONS Erik Lindskog Dissertation in Signal Processingto be publicly examined in room K23, Magistern, Dag Hammarskjo(cid:127)lds v(cid:127)ag 31, Uppsala, on June 7, 1999, at 10.15 a.m., for the degree of Doctor of Philosophy. The discussion will be held in Swedish. ABSTRACT Erik Lindskog, 1999. Space-time processing and equalization for wireless commu- nications. Uppsala, 318pp. ISBN 91-506-1350-2. In this thesis several aspects of space-time processing and equalization for wire- less communications are treated. We discuss several di(cid:11)erent methods of improv- ing estimates of space-time channels, such as temporal parametrization, spatial parametrization, reduced rank channel estimation, bootstrap channel estimation, andjointestimationofanFIRchannelandanARnoisemodel. Inwirelesscommu- nication the signal is often subject to intersymbol interference as well as interfer- encefromotherusers. Weherediscussspace-timedecisionfeedbackequalizersand space-timemaximumlikelihoodsequenceestimators,whichcanalleviatetheimpact of these factors. In case the wireless channel does not experience a large amount of coupled delay and angle spread, su(cid:14)cient performance may be obtained by an equalizerwith alesscomplexstructure. We thereforediscussvariousreducedcom- plexity equalizers and symbol sequence estimators. We also discuss re-estimating the channel and/or re-tuning the equalizer with a bootstrap method using esti- mated symbols. With this method we can improvethe performanceof the channel estimation, the equalization, and the interferer suppression. This method can also be used to suppress asynchronous interferers. When equalizers and symbol detec- tion algorithms aredesigned based on estimated channels we need to consider how errorsin the estimated channels,orerrorsdue totime variations,a(cid:11)ect the perfor- manceoftheequalizerorsymboldetector. Weshowthatequalizerstunedbasedon ordinary least squares estimated channels exhibit a degree of self-robusti(cid:12)cation, which automatically compensates for potential errors in the channel estimates. Key-words: Space-time processing, channel estimation, equalization, maximum likelihood sequence estimation, decision feedback equalization, interference sup- pression, robustness. Erik Lindskog, Signals and Systems, Uppsala University, PO Box 528, SE-751 20 Uppsala, Sweden. Uppsala University Signals and Systems SPACE-TIME PROCESSING AND EQUALIZATION FOR WIRELESS COMMUNICATIONS Erik Lindskog UPPSALA UNIVERSITY 1999 Dissertation for the degree of Doctor of Philosophy in Signal Processing at Uppsala University, 1999 ABSTRACT Erik Lindskog, 1999. Space-time processing and equalization for wireless commu- nications. Uppsala, 318pp. ISBN 91-506-1350-2. In this thesis several aspects of space-time processing and equalization for wire- less communications are treated. We discuss several di(cid:11)erent methods of improv- ing estimates of space-time channels, such as temporal parametrization, spatial parametrization, reduced rank channel estimation, bootstrap channel estimation, andjointestimationofanFIRchannelandanARnoisemodel. Inwirelesscommu- nication the signal is often subject to intersymbol interference as well as interfer- encefromotherusers. Weherediscuss space-timedecisionfeedbackequalizersand space-timemaximumlikelihoodsequenceestimators,whichcanalleviatetheimpact of these factors. In case the wireless channel does not experience a large amount of coupled delay and angle spread, su(cid:14)cient performance may be obtained by an equalizerwith alesscomplex structure. Wethereforediscussvariousreducedcom- plexity equalizers and symbol sequence estimators. We also discuss re-estimating the channel and/or re-tuning the equalizer with a bootstrap method using esti- mated symbols. With this method we can improvethe performance of the channel estimation, the equalization, and the interferer suppression. This method can also be used to suppress asynchronous interferers. When equalizers and symbol detec- tion algorithmsaredesigned based on estimated channels we need to consider how errorsin the estimated channels,or errorsdue to time variations,a(cid:11)ect the perfor- manceoftheequalizerorsymboldetector. Weshowthatequalizerstunedbasedon ordinary least squares estimated channels exhibit a degree of self-robusti(cid:12)cation, which automatically compensates for potential errors in the channel estimates. Key-words: Space-time processing, channel estimation, equalization, maximum likelihood sequence estimation, decision feedback equalization, interference sup- pression, robustness. Erik Lindskog, Signals and Systems, Uppsala University, PO Box 528, SE-751 20 Uppsala, Sweden. (cid:13)c Erik Lindskog 1999 ISBN 91-506-1350-2 Printed in Sweden by Elanders Graphic Systems AB, Angered 1999 Distributed by Signals and Systems, Uppsala University, Uppsala, Sweden To my parents iv v Preface This thesis consists of seven chapters. The (cid:12)rst chapter is an overview of space-space time processing for wireless communications and serves as an introduction to the subject. The remaining chapters treat various aspects of channel estimation, equalization and interferer suppression. Parts of the material in the respective chapters has been published in the following papers: Chapter 1: Arogyaswami Paulraj and Erik Lindskog, \A Taxonomyofspace-timeprocessingforwirelessnetworks", IEEProceed- ings, Radar, Sonar and Navigation, vol. 145, no. 1, February 1998. Chapter 2: Erik Lindskog, \Array Channel Identi(cid:12)cation Using Directional of Arrival Parametriza- tion". Proceedings of IEEE International Conference on Universal Personal Communications, Cambridge, Massachusetts, U.S.A., Sept 29 - Oct 2 1996, vol 2, pp 999-1003. Erik Lindskog, \ChannelEstimationExploitingPulseShapingInformation-AChannelIn- terpolationApproach",UPTECReport97138R,UppsalaUniversity,Signals and Systems, 1997, PO Box 528, SE-751 20 Uppsala, Sweden. Erik Lindskog and Jonas Strandell, \Multi-user Channel Estimation Exploiting Pulse Shaping Information". Proceedings of European Signal Processing Conference, Rhodos, Greece, Sept 8-11 1998. vi Erik Lindskog and Claes Tidestav, "Reduced rank channel estimation", IEEE Vehicular Technology Confer- ence, Houston, Texas, USA, May 16-20 1999. The methods in Section 2.3 and 2.5 has also been combined and published in: Jonas Strandell and Erik Lindskog, \Channelestimationbymaximumlikelihoodprojectiononto aparametrized subspace". ProceedingsofEuropeanSignalProcessingConference, Rhodos, Greece, Sept 8-11 1998, Chapter 3: Erik Lindskog, Anders Ahl(cid:19)en and Mikael Sternad, \Combined Spatial and Temporal Equalization using an Adaptive Antenna Array and a Decision Feedback Equalization Scheme", Proceedings of Int. Conf. onAcoustics,SpeechandSignalProcessing,Detroit,Michigan,U.S.A., May 8-12 1995, vol. 2, pp 1189-1192. Erik Lindskog, Anders Ahl(cid:19)en and Mikael Sternad, \Spatio-TemporalEqualizationforMultipathEnvironmentsinMobileRadio Applications". Proceedings of 45h IEEE Vehicular Technology Conference, Rosemont, Illinois, USA, vol. 1, July 26-29 1995, pp 399-403. Chapter 4: Erik Lindskog, \Multi-channel maximum likelihood sequence estimation". Proceedings of the 47th IEEE Vehicular Technology Conference, Phoenix, Arizona, USA, May 5-7 1997, vol. 2, pp 715-719. Chapter 5: Erik Lindskog, \Making SMI-beamforming insensitive to the sampling timing for GSM sig- nals", Proceedings of the Sixth International Symposium on Personal, In- doorandMobileRadioCommunications,Toronto,Canada,September27-29 1995, vol. 2, pp 664-668 vii Erik Lindskog and Claes Tidestav, \Reduced rankequalization", InProceedings of IEEEInternational Sympo- sium on Personal, Indoor and Mobile Radio Communication, Boston, Mas- sachusetts, USA, September 8-11 1998, Vol 3, pp 1081-1085. Chapter 6: Erik Lindskog, \Combatting Co-Channel Interference in a TDMA System Using Interfer- enceEstimatesFromAdjacentFrames". Proceedingsof29th AsilomarCon- ference onSignals,Systems & Computers, Paci(cid:12)c Grove, California,U.S.A., Oct 30 - Nov 1 1995, vol. 1, pp 367-371. Claes Tidestav and Erik Lindskog, \BootstrapEqualization",ProceedingsofIEEEInternationalConferenceon Universal Personal Communications, Florence, Italy, October 5-9 1998. Chapter 7: Mikael Sternad and Anders Ahl(cid:19)en and Erik Lindskog, \Robust Decision Feedback Equalizers", Proceedings of Int. Conf. on Acoustics, Speech and Signal Processing, Minneapolis, Minnesota, U.S.A., April 1993, vol. 3, pp 555-558. Erik Lindskog, Mikael Sternad and Anders Ahl(cid:19)en, \DesigningDecisionFeedbackEqualizerstobeRobustwithrespecttoChan- nel Time Variations", Proceedings of Nordic Radio Symposium seminar, Uppsala, Sweden, November 10-11 1993. viii ix Acknowledgments First, I wish to express my gratitude to my supervisors Professor Anders Ahl(cid:19)en and Dr. Mikael Sternad for their valuable guidance and support during my time as a graduate student. It has been very stimulating and rewarding to work in their group. Anders always has a positive outlook on things, and Mikael will always (cid:12)nd interest in every problem you approach him with. I also want to give my thanks to them for their careful reading of various manuscripts, including this thesis. I would also like to thank my other co-authors on various papers, Professor ArogyaswamiPaulraj,ClaesTidestav,JonasStrandell,MattiasWennstr(cid:127)om, Dr. Tommy O(cid:127)berg Dr. Anders Rydberg and Yulianto Naserudin for inter- estingandrewardingco-operation. Furthermore, I wouldliketo thank Mats Cedervall and Boon Ng for sharing some of their insights into the problem of temporal parametrization of a channel. IamalsogratefultoDrS(cid:127)orenAndersson,DrUlfForss(cid:19)en, andHenrikDamm at Ericsson Radio Systems for the experimental data that has been used in some of the evaluations in this thesis. My thanks also to the Swedish Research Council for Engineering Sciences (TFR), who has been the major (cid:12)nancial supporter of this work under con- tracts 281-95-800 and 281-98-654. During my time as a graduate student I have belonged to two di(cid:11)erent departments, System and Control in \House 8" and Signals and Systems in \Magistern". It has been joyful and rewarding to meet and work with the members of both departments and I would like express my warmest gratitude to all of them. I would also like to thank our shared system administrator, Ove Ewerlid, for valuable help with computer problems and for all the interesting discussions we have had at our favorite \hang-out". A special mentioning also of my (cid:12)rst o(cid:14)ce mate Kenth O(cid:127)hrn for the good company and to Fredrik Lingvallfor introducingme to climbingin Uppsala.