USF Libraries
USF Digital Collections

Baseband receiver algorithms for 4G co-channel femtocells

MISSING IMAGE

Material Information

Title:
Baseband receiver algorithms for 4G co-channel femtocells
Physical Description:
Book
Language:
English
Creator:
Sahin, Mustafa ( Mustafa Emin )
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Cognitive radio
Energy detection
Inter-carrier interference
Iterative cancellation
MIMO
OFDMA
Opportunity detection
User separation
Dissertations, Academic -- Electrical Engineering -- Doctoral -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: The growing interest for high data rate wireless communications over the last few decades gave rise to the emergence of a number of wideband wireless systems. The resulting scarcity of frequency spectrum has been forcing wireless system designers to develop methods that will push the spectral ełciency to its limit. One such method is to have multiple systems utilize the same spectrum by allowing some unavoidable interference to occur between them. The idea of co-channel systems is tested in the industrial, scientific and medical (ISM) bands and it is found to be a very beneficial approach. Therefore, it can be foreseen that co-channel systems might be a potential solution to the growing spectral crowding problem. Besides the systems that are designed to be co-channel, it is sometimes also possible to encounter that multiple systems occupy the same band undesirably.This kind of unintentional co-channel system scenarios might occur especially due to the dense re-use of available frequency bands. Another reason for unwanted co-channel usage might be the coexistence of third generation (3G) and fourth generation (4G) systems. Since 4G systems will probably be targeting to use the same frequency bands as their 3G counterparts, and since the transition from 3G to 4G will take some time, unintentional co-channel scenarios might be observed between the 3G and 4G systems. This dissertation consists of baseband receiver algorithms for OFDMA-based systems that target at handling the potential co-channel interference (CCI) in various co-channel system scenarios. Three CCI avoidance and two CCI cancellation algorithms are proposed that can be applied to intentional and unintentional co-channel systems.Femtocells, which have recently been introduced as a new class of personal-use base stations that can coexist with macrocell networks in a shared spectrum manner, might constitute an appropriate example for both types of co-channel systems. Therefore, they are considered to be one of the co-existing systems in most of the algorithms presented.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2009.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Mustafa Şahin.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 155 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002069492
oclc - 608514788
usfldc doi - E14-SFE0003283
usfldc handle - e14.3283
System ID:
SFS0027599:00001


This item is only available as the following downloads:


Full Text

PAGE 3

IamgratefultoDr.IsmailGuvencnotonlyforadvisingmethroughoutourtwo-yearcollaborationwithNTTDOCOMO,butalsoforhissincerefriendship.IalsowanttoacknowledgeDr.FujioWatanabeandDr.Moo-RyongJeongfromDOCOMOUSALabsfortheirsupportandtheirdirectcontributionstothisdissertation. IowemuchtomyfriendsDr.HishamMahmoud,Dr.SerhanYarkan,Dr.HasariCelebi,Dr.TevkYucek,IbrahimDemirdo~gen,HasanBasriCelebi,AbdullahHatahet,MuradKhalid,OmarZakaria,IsmailButun,SabihGuzelgoz,SadiaAhmed,AliGorcin,EvrenTerzi,Dr.BahattinKarakaya,Dr.CelalCeken,AliRzaEkti,OzgurYurur,Ca~gatayTalay,andHazarAk.Wesharedsomanythingswiththemovertheyearsthatwespenttogether.Theyalsotaughtmesomanyvirtues.Sincerefriendshiptostartwith,unselsh-ness,tolerance,andhelpfulness.Iamgratefultothemformakingmeabetterperson. IalsowanttoexpressmygratitudetomyfriendsintheTurkishcommunityinTampa,FL,especiallytoSalihErdem,SenerGultekin,Erdo~ganBulut,LeventDa~gstanl,ErdemOnsal,andSalimErdemfortheirsupporttous,students,wheneverweneedit.Graduatestudentlifewouldbemuchlessbearableiftheywerenotthere.

PAGE 4

Last,butbynomeansleast,mydeepestgratitudegoestomywife,Muberra,forherlove,allthesacricesshemade,herrmsupport,hervastpatience,andhersteadyencour-agementformorethanveyearsnow.IfshedidnothavesuchadeepunderstandingandtoleranceforthehardshipsofbeingthewifeofaPh.D.student,Icouldnoteventrytoobtainthisdegree.Iwanttothankherfrommyheartforeverythingshehasbeendoing.

PAGE 5

LISTOFFIGURESvi LISTOFACRONYMSxi ABSTRACTxiv CHAPTER1INTRODUCTION1 1.1Co-channelInterferenceAvoidance4 1.2Co-channelInterferenceCancellation5 1.3DissertationOutline6 1.3.1Chapter2:OpportunityDetectionforOFDMA-BasedCognitiveRadioSystemswithTimingMisalignment7 1.3.2Chapter3:InterferenceScenariosandFrequencyReuseforNext-GenerationFemtocellNetworks7 1.3.3Chapter4:UplinkUserSignalSeparationforOFDMA-BasedCognitiveRadios8 1.3.4Chapter5:ReceptionofMIMO-OFDMASignalswithaSingleChannelReceiver8 1.3.5Chapter6:AnIterativeInterferenceCancellationMethodforCo-ChannelMulticarrierandNarrowbandSystems9 CHAPTER2OPPORTUNITYDETECTIONFOROFDMA-BASEDCOGNITIVERADIOSYSTEMSWITHTIMINGMISALIGNMENT10 2.1Introduction10 2.2UL-OFDMASystemModel14 2.3SpectrumSensingTechniquesandImpactofSubcarrierAssign-mentScheme16 2.3.1EnergyDetectorMethod16 2.3.2ESPRITMethod17 2.3.3SubcarrierAssignmentSchemesinDierentWirelessStandards19 2.4StatisticsoftheEnergyDetectorDecisionVariablewithTimingMisalignment20 2.4.1UserDistanceYieldingtheStrongestInterference22 2.5ReceiverOperatingCharacteristics25 2.5.1Noise-basedThreshold25i

PAGE 6

2.6DeterminingtheOptimumSynchronizationPoint28 2.7SimulationResults30 2.7.1Statisticsof(2.17)withTimingMisalignment30 2.7.2ReceiverOperatingCharacteristicsWithandWith-outICI31 2.7.3ProbabilityofOpportunityDetectionErrorwithTim-ingMisalignment33 2.8ConcludingRemarks42 CHAPTER3INTERFERENCESCENARIOSANDFREQUENCYREUSEFORNEXT-GENERATIONFEMTOCELLNETWORKS43 3.1Introduction43 3.2SystemModel46 3.3CCIandICIIssuesinFemtocellDeployments47 3.3.1DeploymentCongurations47 3.3.2CCIScenariosformMSs48 3.3.3ICIintheUplink50 3.4FrameworkforCo-ChannelFemtocellOperation54 3.4.1ObtainingSchedulingInformationfromthemBS56 3.4.1.1CommunicationoftheSchedulingInforma-tionovertheBackbone56 3.4.1.2ObtainingtheSchedulingInformationOvertheAir58 3.4.2JointlyUtilizingSchedulingInformationandSpec-trumSensingResults59 3.5SimulationResults61 3.6ConcludingRemarks65 CHAPTER4UPLINKUSERSIGNALSEPARATIONFOROFDMA-BASEDCOG-NITIVERADIOS66 4.1Introduction66 4.2UL-OFDMASystemModel69 4.3BlockSizeEstimation70 4.3.1GaussianApproximationforBlockSizeEstimation74 4.4UserSeparationMethod75 4.5UserSeparationApplicationsforOFDMA-BasedCognitiveRadios80 4.5.1ClassifyingtheSourceofCo-channelInterference80 4.5.2Hand-oBetweenMacrocell-BSandFemtocell-BS81 4.5.3DirectingSomeFemtocellUserstotheMacrocell82 4.5.4CausingCCItotheMinimumNumberofmMSsPossible83 4.5.5DeterminingtheCloseUsers83 4.6UsingBlockSizeEstimationandUserSeparationinSpectrumOpportunityDetection84 4.7SimulationResults85 4.7.1BlockSizeEstimationSimulations86ii

PAGE 7

4.7.3OpportunityDetectionSimulations88 4.8ConcludingRemarks92 CHAPTER5RECEPTIONOFMIMO-OFDMASIGNALSWITHASINGLECHAN-NELRECEIVER93 5.1Introduction93 5.2SignalModelandthePrimaryRFFront-endImpairments94 5.3EstimationandRemovalofImpairmentsintheSISOCase95 5.4ChallengesinMIMOComparedtoSISO97 5.4.1TimeOsetBetweentheBranches98 5.4.2EmployingSeparateClocks98 5.4.3UsingSeparateIQModulators99 5.4.4UsingSeparateRFComponents99 5.5DetectingtheImpairmentDierencesbyExaminingtheCon-stellationDiagram99 5.6ProceduretoHandleWiMAXMIMOSignals102 5.7Space-TimeTransmitDiversityandCombiningtheMIMOSig-nalsFromTwoTransmitterBranches108 5.8SpatialMultiplexingandJointDemodulation109 5.9Conclusion110 CHAPTER6ANITERATIVEINTERFERENCECANCELLATIONMETHODFORCO-CHANNELMULTICARRIERANDNARROWBANDSYS-TEMS111 6.1Introduction111 6.2ApplicationExamplesandSystemModel113 6.2.1ApplicationExamples113 6.2.2SystemModel115 6.2.3GaussianApproximationBasedSymbolErrorRate118 6.3JointDemodulationMethod118 6.4IterativeCCICancellationMethod120 6.5ComputationalComplexity123 6.5.1MLMethod124 6.5.2IterativeCancellation124 6.5.3ComparisonofComplexities127 6.6Simulations128 6.6.1SimulationParameters128 6.6.2AWGNChannelResults129 6.6.3MultipathChannelResults133 6.6.4EectofOverlappingBandwidth133 6.7ConcludingRemarks136 CHAPTER7CONCLUSIONANDFUTUREWORK138 7.1ListofSpecicContributions138 7.2FinalCommentsandFutureWork140iii

PAGE 8

APPENDICES150 AppendixA151 AppendixB154 ABOUTTHEAUTHOREndPageiv

PAGE 9

Table3.1Simulationparameters62 Table4.1TypicalDopplerspreadsandcoherencetimesforWiMAX77 Table4.2Simulationparameters83 Table4.3Userseparationperformanceswhenreceivedpowersdependonuserdistances92 Table5.1WiMAXMIMOsystemsettings104 Table6.1Thecomputationsrequiredformaximumlikelihoodanditer-ativecancellationalgorithms126 Table6.2CPUcyclecountsobtainedusingaXilinXDSP48slice127 Table6.3OFDMA,narrowband,andCDMAsystemparameters129v

PAGE 10

Figure2.1Cognitiveradioscenarioinconsideration.12 Figure2.2TheprimarynetworkandcognitiveradiosignalsarrivingatSU-2andthetimingmisalignmentproblem.12 Figure2.3Diagramoftheenergydetectormethod.17 Figure2.4Theoreticalversussimulatedmeanof(2.17)withtimingmis-alignment.31 Figure2.5Theoreticalversussimulatedvarianceof(2.17)withtimingmisalignment.32 Figure2.6ROCsforEsc;i=2=7dB.33 Figure2.7ROCsforEsc;i=2=10dB.34 Figure2.8Subcarrierassignmentschemesindierentstandards.35 Figure2.9Errorprobabilityversusmaxforenergydetectionwithblock-wiseandrandomizedassignments(Nsc=9;Nsymb=6).36 Figure2.10Errorprobabilityversusmaxforenergydetectionwithblock-wiseandrandomizedassignments(Nsc=4;Nsymb=3).37 Figure2.11ErrorprobabilityversusmaxfortheESPRITalgorithmwithblockwiseandrandomizedassignments(Nsc=9;Nsymb=6).38 Figure2.12ComparisonofESPRITandenergydetectionalgorithmsoverincreasingnumberofsymbols.38 Figure2.13ErrorprobabilityversusnormalizedthresholdforblockwiseassignmentwhenalluserSNRsarethesame;max=0sand60s(forSNR=20dB,10dB,and0dB).39 Figure2.14ErrorprobabilityversusnormalizedthresholdforrandomizedassignmentwhenalluserSNRsarethesame;max=0sand60s(forSNR=20dB,10dB,and0dB).39vi

PAGE 11

Figure2.16Interferenceversususerdelayanalysis.41 Figure2.17VariationofICIpowerwithrespecttosynchronizationpoint;NCP=16samples.41 Figure3.1Femtocelldeploymentathomes.44 Figure3.2Allpossibleinterferencescenariosrelatedtofemtocellcommunications.49 Figure3.3TherelationbetweenthearrivaltimesofmMSsignaldelays,CP-size,andICI.51 Figure3.4Femtocellcommunicationsintheuplink.53 Figure3.5Femtocellcommunicationsinthedownlink.54 Figure3.6Simpliedowchartforfemtocellimplementation.55 Figure3.7Combiningschedulinginformationwithspectrumsensingresults.57 Figure3.8VariationofICIpowerwrt.RTDfordierentcenterfrequen-ciesandfemtocellBSheights.63 Figure3.9Errorprobabilityversusmaxforenergydetectionwithblock-wiseandrandomizedassignments.64 Figure4.1(a)Flowchartforblocksizeestimation.(b)Flowchartforusersignalseparation.68 Figure4.26blocksinaWiMAXUL-PUSCsystem,whereeachblockisa4x3tile,i.e.,K=3andM=2.71 Figure4.3Normalizedautocorrelationsobtainedutilizinga60symbollongsignal(withFFTsize512)forablocksizeof4x3at30dBSNRinanAWGNchannel.73 Figure4.4Clustersonthe~vs.~planeina10-userscenario(30dBSNRisassumedforallusersignalsoverMPchannel).79 Figure4.5Femtocelloperatinginthemiddleof3mBSsinasystemwithfrequencyreuseof3.82 Figure4.6SimulationandGaussianapproximationresultsforestimatingthesizeofa4x3block.89 Figure4.7SimulationandGaussianapproximationresultsforestimatingthesizeofa6x6block.89vii

PAGE 12

Figure4.9PerformancesinseparatingtheusersubcarriersinAWGNandMPchannelsforvariousnumbersofusers.90 Figure4.10Errorprobabilityindetectingthespectrumopportunitiesus-ingfourdierentmethodsforaresourceblocksizeof4x3.91 Figure4.11Errorprobabilityindetectingthespectrumopportunitiesus-ingfourdierentmethodsforaresourceblocksizeof6x6.91 Figure5.1TheconstellationfortwoQPSKmodulatedSTTDsignalsbe-foreequalization.100 Figure5.2=12phasedierencebetweenthetwotransmittedsignals.101 Figure5.330%IQimbalancedierencebetweenthetwotransmittedsignals.102 Figure5.4=12quadratureerrordierencebetweenthetwotransmittedsignals.103 Figure5.50:002radianfrequencyosetdierencebetweenthetwotrans-mittedsignals.104 Figure5.6AllocationofsubcarriersindownlinkPUSCWiMAX.105 Figure5.7AllocationofsubcarriersinuplinkPUSCWiMAX.106 Figure5.8The2x1MIMOsystemsetup.107 Figure6.1AnexamplecoexistencescenarioforanLTEbasedmacrocellwithaW-CDMAbasedfemtocellduringmigrationfrom3Gto4G.115 Figure6.2DiagramoftheOFDMAandNBsymbolsintimeandfrequency.117 Figure6.3FlowchartoftheproposediterativeCCIcancellationalgorithm.122 Figure6.4FlowchartofthedemodulationandregenerationmodulesfortheNBsystem.123 Figure6.5Thespectraofthereceivedco-channelsignalsandtheOFDMAsignalalone(OFDMASNR:30dB,NBSNR:20dB).129 Figure6.6SERperformanceoftheOFDMAsystemundertheinuenceofNBinterference(AWGNchannel).131 Figure6.7SERperformanceoftheOFDMAsystemundertheinuenceofCDMAinterference(AWGNchannel).131viii

PAGE 13

Figure6.9SERperformanceoftheCDMAsystemundertheinuenceofOFDMAinterference(AWGNchannel).133 Figure6.10SERperformanceoftheOFDMAsystemundertheinuenceofNBinterference(MPchannel).135 Figure6.11SERperformanceoftheOFDMAsystemundertheinuenceofCDMAinterference(MPchannel).135 Figure6.12SERperformanceoftheNBsystemundertheinuenceofOFDMAinterference(MPchannel).136 Figure6.13SERperformanceoftheCDMAsystemundertheinuenceofOFDMAinterference(MPchannel).136 Figure6.14OFDMAsystem'sSERperformanceundertheinuenceofNBinterferenceforvariousoverlappingbandwidths(AWGNchannel).137ix

PAGE 17

Besidesthesystemsthataredesignedtobeco-channel,itissometimesalsopossibletoencounterthatmultiplesystemsoccupythesamebandundesirably.Thiskindofun-intentionalco-channelsystemscenariosmightoccurespeciallyduetothedensere-useofavailablefrequencybands.Anotherreasonforunwantedco-channelusagemightbethecoexistenceofthirdgeneration(3G)andfourthgeneration(4G)systems.Since4Gsystemswillprobablybetargetingtousethesamefrequencybandsastheir3Gcounterparts,andsincethetransitionfrom3Gto4Gwilltakesometime,unintentionalco-channelscenariosmightbeobservedbetweenthe3Gand4Gsystems.xiii

PAGE 19

Inordertoenableanumberofuserstoutilizeagivenfrequencyband,modernwirelesssystemsemploymultipleaccessingmethods.Intimedivisionmultipleaccessing(TDMA),usersareallowedtoaccessthespectrumatpre-determinedtimeslots.Infrequencydivisionmultipleaccessing(FDMA),theavailablespectrumissplitintoanumberoffrequencyblockswhereeachuserisallocatedoneormoreblock.InorthogonalFDMA(OFDMA),time-frequencyblocksareemployed,whicharemadeascompactaspossiblewhilemaintainingtheorthogonalitybetweenthem,inordertomaximizetheeciencyofspectrumusage.Incodedivisionmultipleaccessing(CDMA),eachuserisassignedadierentpseudo-noisetypeofcodethatismultipliedwiththetransmitteddatatospreadittotheavailableband.Multiplicationwiththesamecodeinthereceiverpartenablesseparationofmultipleusersignalsfromeachother.Anotherdomainthatisexploitedtoallowutilizationofthesamefrequencybandisthepowerdomain.Bylimitingthesignalpowersoftransmitterstoacertainlevel,frequencyreuseconceptcanberealized,wherethesamechunkoffrequencies1

PAGE 20

TheevergrowingneedforspectrumhasbeenforcingwirelesssystemdesignerstochoosemoreaggressivesystemparameterssuchasnarrowerOFDMAsubcarrierspacings,longerCDMAcodes,andfrequencyreusefactorscloserto1.Imposingthiskindofaggressiveparametersonasystemmightresultinvariousformsofinterferencesuchasco-channel,adjacentchannel,andmulti-userinterferences.However,practicalsystemapplicationssofarhaverevealedthatallowinginterferenceuptoacertainlevelhelpsincreasingtheoveralleciencyofspectrumusage. Inananalogytotoleratingsomeinterferenceforimprovingthespectraleciencyinmulti-userssystems,multiplesystemscanbemadeco-channeltomaximizethespectraleciencywhileallowingpotentialinterference.Thisisthecaseintheindustrial,scienticandmedical(ISM)bands,wherevariousdierenttechnologiessuchaswirelesslocalareanetwork(WLAN)routers,cordlessphones,andmicrowaveovensoperate.Anotherexampleforintentionaluseofthesamespectrumbymultiplesourcesisthemultipleinputmultipleoutput(MIMO)systems.AlthoughthiskindofasystemdescriptionmaysoundunusualforMIMOsystems,sinceeveryseparatetransmitterbranchtransmitsdierentdatausingthesameband,aMIMOsystemisindeedaco-channelsystemonitsown. Apartfromthesystemsthataredesignedtobeco-channel,therearesituationswheremorethanonesystem'ssignalsexistinthesamebandunintentionally.Thisisthecase,forinstance,incellularsystemsinwhichallcellsusethesamefrequencyband,i.e.thefrequencyreusefactorisequalto1.Insuchacase,atlocationsclosetothecellborders,itispossibletoobserveinterferingsignalsfrommultiplesources.Anotherpotentialscenarioforunwantedcoexistenceinthesamebandmightoccurduringthetransitionfromthethirdgeneration(3G)tothefourthgeneration(4G)systems.Sinceitwilltakesometimeuntil2

PAGE 21

Thebasebandreceiveralgorithmsproposedinthisdissertationaimathandlingtheco-channelinterference(CCI)thatmightoccurbothinintentionalandinunintentionalco-channelsystemscenarios.ThesealgorithmscanbeclassiedasinterferenceavoidanceandinterferencecancellationmethodsasillustratedinFig1.1.Inmostofthepresentedalgorithms,oneoftheco-channelsystemsisexempliedbyafemtocellnetwork.Femto-cellsarearecentlyemergedclassoflimitedrange,personalusehome/ocebasestations(BSs)[1]-[3].Themainpurposeoffemtocellnetworksistoimprovetheindoorcoverageofcellularnetworks.AsitwillbediscussedindetailinChapters2and3,femtocellscancoexistwithinmacrocellsinasplitspectrumorasharedspectrummanner.Thelattercaseconstitutesanappropriateexampleforco-channelsystemswhereCCIisasignicantcon-cern,andthisisthereasonwhyco-channelfemtocellsareacasestudyelementthroughout3

PAGE 22

Theorthogonalitycanbeestablishedintimedomainbyhavingthesystemstransmitatdierenttimeslots;infrequencydomainbyavoidinganyoverlapbetweenthespectraofthesystemswithinthesameband;incodedomainbyassigningdierentPNcodestodierentsystems;inspacedomainbyseparatingthetransmittedbeams;andinpowerdomainbymakingthefootprintsofthetransmitantennasnotoverlap.Itshouldbenotedthatthereisastrongsimilaritybetweentheco-channelinterferenceavoidanceapproachesandthebasicmethodsthatareusedtoimplementthemultipleaccessingschemesinwirelesssystemssuchasTDMA,CDMA,andOFDMA. Theavoidancealgorithmsthatwillbepresentedinthisdissertationaretime-frequencydomainbasedmethods.Inthesealgorithms,co-channelsystemsthatemployOFDMAareconsidered,anditisaimedtoensurethatthetwosystemsutilizemutuallyexclusivesetsofresourceblocks.InChapter2,twodierentdetectionmethodsareinvestigatedforndingtheopportunitieswithintheuplinkfrequencybandofanOFDMAbasedsystemwithtimingmisalignments.Analgorithmfordeterminingthesynchronizationpointthatminimizestheinter-carrierinterferenceduetothetimingmisalignmentsisproposed. Theremightbecaseswherebothco-channelsystemsarebeingoperatedbythesameserviceproviderasintheexampleofamacrocell-femtocellcoexistence.Insuchascenario,itmightbeconsideredthatthemacrocellsharesitsresourceblockallocationinformationwiththefemtocellsinordertohavethemavoidCCI.InChapter3,analgorithmisproposed4

PAGE 23

Incontrasttoscenarioswherethetwoco-channelsystemsareinclosecollaboration,itmightbepossiblethattheCCIavoidancehastobeperformedinasemi-blindmannerwherethesystemshaveinformationaboutonlytheverybasicsystemparametersoftheotherco-channelsystem.InChapter4,weproposetwoalgorithmsthatmightbeusedforenhancingtheperformanceinCCIavoidance.TherstalgorithmenablesdeterminingthesizeoftheresourceblocksemployedinanuplinkOFDMAsignal.Thesecondone,ontheotherhand,targetsatseparatingtheusersignalsfromeachothersothattheclose-byusers'blockscanbedeterminedandavoidedmorereliably.1.2Co-channelInterferenceCancellation ThekeyconceptthatdeterminesthefeasibilityofCCIcancellationisthesignalsepara-bility.Anydierencesinthesignalpropertiessuchasthewaveformsandspectraaswellasthedierencesthatthesignalsattaininthepropagationchannelsuchastheirdelaysmightserveasameansofsignalseparation[4]. Itmightbepossibletoseparatetwosinglecarriersignalsthatoverlapinalldomainsinthetransmittersideexploitingthedierencesinthereceivedsignalsduetotheinde-pendentchannelsthattheypropagatethrough.AgoodexampleforsignalseparationbytakingadvantageofthedierenceinthewirelesschannelsistheMIMOreceivers.Inthisdissertation,thealgorithmpresentedinChapter5takesMIMOreceiversonestepfurtherandintroducesreceptionofMIMOsignalswithasinglereceiver.Theproposedmaximum5

PAGE 24

Amorecomplicatedscenariowheresignalseparabilitymightbeattainableisthecaseoftwoco-channelsignalsoneofwhichcarriesinformationintimedomain,andtheotheroneinfrequencydomain.ThealgorithmpresentedinChapter6dealswiththiskindofascenarioandproposesaniterativecancellationmethodwhereasingleantennareceiverisconsidered.Thetwoco-channelsignalsareestimated,demodulated,andregeneratedinasuccessivemannertoyieldabetterestimateoftheothersignal.Itisshownthatafteralimitednumberofiterations,aquitesuccessfulsignalseparationcanbeobtained. Futureworkinthisdirectionincludesseparationofco-channelsignalsthatoverlapinallreadilyimaginabledomainssuchastheunintentionalco-channelsignalsthatareobservedincellularsystemswithfrequencyreusefactorequalto1.1.3DissertationOutline

PAGE 28

Thirdgeneration(3G)wirelesstechnologieshavemostlybeenusingcodedivisionmul-tipleaccess(CDMA)intheirphysicallayers.Forexample,InternationalMobileTelecom-munications2000(IMT-2000),whichistheworldwidestandardforthirdgeneration(3G)wirelesstechnologiesdenedbytheInternationalTelecommunicationUnion(ITU),denessixstandardsfor3Gnetworks,threeofwhicharebasedonCDMA:WidebandCDMA,CDMA-2000,TD-CDMA/TD-SCDMA,EDGE,DECT,andWiMAX.WhiletherstveofthesestandardswereapprovedbyITUin1999,WiMAX,whichisbasedonorthogonalfrequencydivisionmultipleaccess(OFDMA),wasapprovedbytheITUin2007.However,fourthgeneration(4G)ofwirelesscommunicationsisexpectedtobedominatedbythe10

PAGE 29

ForanOFDMAbasedcognitiveradio,spectrumopportunitycanbedenedasthesetofsubcarriersthatarenotutilizedbytheprimarysystem.Inordertotakeadvantageofthespectrumopportunities,orthogonalitytotheprimarysystemneedstobeestablished.Forsucharadio,therststepinspectrumsensingisthedetectionofthepresenceofaprimaryuser,whichcanbeachievedutilizingthecyclicprex(CP)ortheguardbandofthereceivedsignal[20]-[22].Afterdetectingthepresenceofaprimaryuser,Athresholdbaseddetectorsuchasin[23]canbeemployedfordetectingthespectrumopportunities,whereappropriateselectionofthethresholdiscriticalforagooddetectionperformance.Spectrumsensingperformanceofenergydetectorscanbequantiedbyreceiveroperatingcharacteristic(ROC)curves(seee.g.[24]).Aparticularlychallengingscenariothathasnotbeenconsideredindetailinthepriorartforcognitiveradiosystemsiswhensomeoftheuplink(UL)OFDMAusersignalsarriveatthereceiverwithdelayslargerthantheCPofthesymbol(seee.g.,[25]-[27],andFigs.2.1and2.2).Forexample,inFig.2.2,thesecondaryuserSU-1communicateswithSU-2utilizingtheavailablespectrumopportunities.However,thesignalsoftheprimaryusersarrivingatSU-2aftertheCPofSU-1(i.e.,ULsignalsofmobilestationsMS-1andMS-4)resultininter-symbolinterference(ISI)aswellasinter-carrierinterference(ICI),whichmayconsiderablydecreasethespectrumopportunities. Aninterestingcasewheresuchatimingmisalignmentproblemmayoccuristheco-existenceofafemtocellnetwork[28,29]withamacrocellnetwork,bothofwhichemployOFDMA.Asdiscussedin[30]and[31],macrocellandfemtocellmaycoexistthroughei-therasplit-spectrumapproach,wherebothnetworksareassignedorthogonalbands,orashared-spectrumapproach,whereunusedpartsofthemacrocellspectrumareutilizedbythefemtocellthatactsasacognitiveradio.Inashared-spectrumscenario,whilethemacrocell11

PAGE 30

Figure2.2TheprimarynetworkandcognitiveradiosignalsarrivingatSU-2andthetimingmisalignmentproblem.12

PAGE 31

Inthischapter,detectionofspectrumopportunitiesinUL-OFDMAisinvestigatedinthepresenceofconsiderabletimingmisalignmentbetweenusers(seee.g.[5]).TakingintoaccounttheeectsofICIthatappearasaresultoftimingmisalignments,thestatisticsoftheenergydetectorreceiverareobtained,andtherelatedROCsforspectrumsensingarederived.Moreover,aclosedformexpressionfortheprimaryuserdistancethatcausesthestrongestinterferencetothecognitiveradioisobtained.Finally,optimumUL-OFDMAsynchronizationpointthatminimizestheinterferencetothecognitiveradioiscalculated1.Throughcomputersimulations,opportunitydetectionerrorprobabilitiesusingtheenergydetectoraredeterminedforvariousscenariosandtheyarecomparedwiththedetectionperformanceoftheestimationofsignalparametersbyrotationalinvariancetechniques(ESPRIT)algorithm.Impactoftheprimarynetwork'sresourceallocationblocksizeonthecognitiveradioisalsoinvestigatedusingtheparametersspeciedinLTEandWiMAXstandards. Organizationofthischapterisasfollows.Section2.2providesthesystemmodel,whileSection2.3shortlyintroducesenergydetectionbasedandESPRITalgorithmbasedspec-trumsensingapproachesanddiscussesthesubcarrierassignmentschemesindierentwire-lessstandards.InSection2.4,thestatisticsoftheenergydetectordecisionvariableareinvestigated,andtheuserdistanceyieldingthehighestinterferenceisderived.InSec-tion2.5,ROCswithandwithouttimingmisalignmentarederivedforreceiversthatemploynoise-basedthresholdandnormalizedthreshold.Section2.6investigatestheoptimumsyn-chronizationpointforanUL-OFDMAreceiver,Section2.7presentsthesimulationresults,andSection2.8concludesthechapter.

PAGE 32

N;NcpnN1;(2.1) wheremisthesymbolindex,Ptx;iisthetransmitpowerforuseri,k2iisthesubcarrierindex,iisthesetofsubcarriersoflengthNiassignedtouserioutofNtotalsubcarriers,Ncpisthelengthofthecyclicprex,andX(m)i(k)isthedataonthekthsubcarrierandmthsymboloftheithuser. Thetimedomainaggregatereceivedsignalisthesuperpositionofsignalsfromallusers,eachofwhichpropagatesthroughadierentmultipathchannelandarrivesatthereceiverwithadelay~i=dN~i=Te,where~iisthepropagationdelayexperiencedbyuseri,andTisthedurationoftheusefulpartofthesymbol.Then,aggregatediscrete-timereceivedsignalcanbeexpressedasy(n)=NuXi=1yi(n)+w(n);(2.2) wherew(n)denotestheadditivewhiteGaussiannoise(AWGN),andyi(n)=p wherePrx;iisthereceivedpowerforuseri,Ldenotesthetotalnumberofmultipathcomponents(MPCs),(m)i(l)isthecomplexamplitudeofthelthMPCforuseri,andDl;i=dNl;i=Te+~i,wherel;iisthedelayofthelthMPCforuseri. IfDl;iNcp,itiseasytoprovethatthefrequencydomainsignalforthekthsubcarrierofuseriisgivenbyY(m)i(k)=p N;(2.4)14

PAGE 33

N+N1Xn=Dl;iNcpx(m)i(nDl;i)ej2kn N:(2.5) Afterplugging(2.1)into(2.5)andsomemanipulation,wehaveY(m)i(k)=p wherethedesiredsignal,interferencefromthesamesubcarrieroftheprevioussymbol,andthetotalinterferencefromothersubcarriersarerespectivelygivenasSd;i;l(k)=X(m)i(k)K1;i;l(k)ej2kDl;i N(2.7)I1;i;l(k)=X(m1)i(k)K2;i;l(k)ej2k(Dl;iNcp) {z }hi(p;k)X(m)i(p)ej2pDl;i N+X(m1)i(p)ej2p(NcpDl;i) {z }gi(p);(2.9) whereK1;i;l(k)=NDl;i+Ncp

PAGE 34

NotethattheinterferencetermsI1;i;l(k)andI2;i;l(k)willbothbezeroifthereceivedMPCislocatedwithintheCPduration.Theaggregatefrequency-domainsignalcanthenbewrittenasY(m)(k)=NuXi=1Y(m)i(k)+W(k);(2.12) whereW(k)CN0;2istheDFTofw(n),2=N0=2,andCN;2denotesthedistributionofacircularlysymmetriccomplexGaussianrandomvariablewithmeanandvariance2.2.3SpectrumSensingTechniquesandImpactofSubcarrierAssignmentScheme wherehypothesisH1impliesthatsubcarrierkisoccupied,andhypothesisH0impliesthatitisnot.AdiagramoftheenergydetectormethodisprovidedinFig.2.3.Statisticsofthe16

PAGE 35

TheproblemofdeterminingtheoccupiedsubcarriersinanOFDMAsignalcanbeconsideredasidentifyingthenumberandfrequenciesofasetofsinusoidsinadditivenoise.OFDMbasedsignalsaresuitableforimplementingESPRITbecausetheyaremadeshiftinvariantbytheadditionofacyclicprex,whichmeansthatatimeshiftnotexceedingtheCPdoesnotalterthestatisticalfeaturesoftheOFDMsignal.ExploitingthispropertyoftheOFDMsignal,carrierfrequencyosetestimationusingESPRITwasperformedin[41].In[42],ESPRITalgorithmwasproposedforestimatingtheoccupiedsubcarriersofanOFDMsymbol.AlthoughitisknownthatESPRITcannotbetheoptimumdetectionmethodwhenthemaximumdelayobservedinthesystemislargerthantheCP(duetothedegradationintheshiftinvarianceofOFDMsymbols),inthischapter,theESPRIT17

PAGE 36

InthepracticalESPRITimplementationfordeterminingtheoccupiedsubcarriersofanOFDMsymbol,therststepistoestimatethenumberofoccupiedsubcarriers(Nocp),whichisdoneviatheminimumdescriptivelength(MDL)algorithmintroducedin[43].Estimationofsubcarrierfrequencies,ontheotherhand,isperformedbyconstructingtheauto-andcross-covariancematricesofthereceivedsignal.Fromthereceivedsignaly(n),twosamplevectorsy(n)andz(n)oflengthareformedy(n)=[y(n);y(n+1);;y(n+1)];z(n)=[y(n+1);y(n+2);;y(n+)];(2.14) whereisequaltoM(N+Ncp),Mbeingthenumberofadjacentsymbolswiththesameoccupiedsubcarriers. Theauto-covariancematrixRyyandcross-covariancematrixRyzareobtainedasfollowsRyy=Efy(n)y(n)g;andRyz=Efy(n)z(n)g;(2.15) whereEfgdenotestheexpectationoperation.ItisimportanttonotethatthereliabilityofRyyandRyzisdirectlyproportionaltoM.Byperforminganeigen-decompositiononRyy,itseigenvaluesaredetermined,wheretheminimumeigenvalueministhenoisevariance2.NoisepowerissubtractedfromRyyandRyztoobtainCyy=RyyminIandCyz=RyzminZ;(2.16) whereIistheidentitymatrix,andZisamatrixwithonesontherstsubdiagonalandzeroselsewhere.ThefrequenciesoftheoccupiedsubcarriersareyieldedbytherstNocp

PAGE 37

WiMAX LTE PUSC1 PUSC2 ASP 3 3 4 3 9M(whereNM=6) BlockSize 84 12 9 54 largestgeneralizedeigenvaluesofthematrixpair(Cyy;Cyz)(Thereaderisreferredto[40]forastep-by-stepguideoftheESPRITalgorithm). ThematrixoperationsthatitrequiresmaketheESPRITalgorithmhighlycomputa-tionallycomplexandmayinduceanextendedprocessingdelay.ThiskindofadelayrendersESPRITlessfeasibleinareal-timeapplication.2.3.3SubcarrierAssignmentSchemesinDierentWirelessStandards InthesimulationsinSection2.7,alongwiththeblockwiseSASsusedindierentstan-dards,wealsoconsiderarandomizedassignment(RA).InRA,eachindividualsubcarriermaybeassignedtoadierentuser,i.e.,Nsc=1,andNsymbmaytakeanyappropriate

PAGE 38

{z }ZeromeanRVo+2Nu1Xi=1NuXj=i+1Y(m)i(k)Y(m)j(k);(2.17) wherethelasttermof(2.17)is0sinceY(m)i(k)andY(m)j(k)cannotbenon-zerosimultane-ously.Thestatisticsof(2.17)canbeevaluatedbyanalyzingthestatisticsoftheindividualtermsaswillbediscussedbelow.Tokeeptheexpressionsanalyticallytractable4,wecon-siderL=1in(2.6),anddropthemultipathindicesfromrelatedexpressionsin(2.6)-(2.9). First,using(2.4),wemaywriteY(m)i(k)2asY(m)i(k)2=Esc;iSd;i(k)+I1;i(k)+I2;i(k)2=Esc;iSd;i(k)2+I1;i(k)2+I2;i(k)2+2RenSd;i(k)I1;i(k)+Sd;i(k)I2;i(k)+I1;i(k)I2;i(k)o;(2.18)

PAGE 39

{z }ZeromeanRV;(2.21)Sd;i(k)I1;i(k)=X(m)i(k)X(m1)i(k)K1;i(k)K2;i(k)ej2kNcp NXp2ip6=khi(p;k)gi(p);(2.23)I1;i(k)I2;i(k)=1 with,asindicatedin(2.9),gi(p)=X(m)i(p)ej2pDl;i N+X(m1)i(p)ej2p(NcpDl;i) Notethat(2.22)-(2.24)aswellastheindicatedtermsin(2.17)and(2.21)arezero-meanrandom-variables(RVs).Then,themeanof(2.17)canbeevaluatedasEnP(m)(k)o=NuXi=1EnY(m)i(k)2o+nd2;(2.27) wherenddenotesthedegreeoffreedom(DOF)ofnoiseterms,andcalculationofEnY(m)i(k)2owillbediscussedinAppendixA.21

PAGE 40

wherethevariancesofthersttwotermsarestraight-forward,andthevarianceofthethirdtermiscalculatedasVar2RenW(k)NuXi=1Y(m)i(k)o=E4W2(k)NuXi=1Y(m)i(k)2+Nu1Xi=1NuXj=iY(m)i(k)Y(m)j(k)(2.29)=8nd4NuXi=1EnY(m)i(k)2o:(2.30) Hence,calculationofEnY(m)i(k)2oandVarnY(m)i(k)2oaresucientforobtainingthestatisticsof(2.17)asin(2.27)and(2.28),aswillbeillustratedfordierentmodulationschemesinAppendixA.2.4.1UserDistanceYieldingtheStrongestInterference Assumingasingletapchannel,userdelaysaredirectlyproportionaltotheuserdistances(di)throughD1;i=di

PAGE 41

Throughsimulations,itisdeterminedthatPk6=pI21;i(k)isnegligiblecomparedtoPk6=pI22;i(k).Hence,thederivationisbasedonndingthedistancewhereI22;i(k)(ICI)ismaximized.Thetotalinterferencepowerthatiscausedbyacertainsubcarrierp,summedoverallemptysubcarriers,isgivenbyXk6=pI22;i(k)=2 ReplacingdiwithcD1;iTs,(2.31)canberewrittenasXk6=pI22;i(k)=2 wherefisthecarrierfrequencyofthesystem.Dierentiating(2.32)withrespecttoD1;ioneobtainsdPk6=pI22;i(k)

PAGE 42

Thetrigonometrictermsin(2.34)canbeapproximatedusingTaylorseriesexpansionundertheconditionthattheinputsofcos(x)andsin(x)satisfy1
PAGE 43

whereM=nd=2isaninteger.Forcomplexnoise,wehavend=2,and(2.38)becomestheCDFofanexponentialdistribution.Then,thePFAforacertainthresholdasin(2.13)becomesPfa()=1FY()=e=22;(2.39) wherethethresholdmayalsobewrittenintermsofthePFAas=F1Y1Pfa:(2.40) Whensubcarrierkisoccupied,ontheotherhand,(2.13)followsanon-centralizedChi-squaredistribution,whoseCDFisgivenby[47]~FY(y;Esc;i)=1QMp ;(2.41)25

PAGE 44

aM1e(x2+a2)=2IM1(ax)dx;(2.42) withI(x)denotingthethordermodiedBesselfunctionoftherstkind[47].Then,using(2.39),probabilityofdetectionPdcorrespondingtoacertainPfabecomesPd(Pfa)=1~FY(;Esc;i)=QMp Therelationshipin(2.43)thatrelatesthePdtoPfaiscommonlyreferredasthereceiveroperatingcharacteristiccurves. InthepresenceofICI,sincethestatisticsofthereceivedpowerwillchange,theROCperformancewillgetworse.Inparticular,usingagaintheChi-squaredistribution6formodelingthedistributionofY(m)(k)2alongwith(2.27)and(2.28),theprobabilityoffalsealarmandprobabilityofdetection(PD)thatwillbeobservedinthepresenceofICIandusingthethresholdasin(2.40)isgivenbyPfa;ICI(Pfa)=1 ~(k);q ~(k);q

PAGE 45

wherePnandPs+naretheaveragenoiseenergyandaveragesignal+noiseenergy,respec-tively,and0Tnorm1denotesthenormalizedthreshold.Pncanpracticallybeestimatedutilizingtheguardbands(GB)oftheOFDMAsignal.Tobemorespecic,byaveragingtheenergiesmeasuredovertheoutermostsubcarriersofleftandrightGBs,anestimatethatisaectedleastfromtheICIcanbeobtained.Ps+n,ontheotherhand,canberoughlydeter-minedbyaveragingtheenergiesmeasuredoverallsubcarriersexceptthenullsubcarriersintheguardbands. Thereisatrade-obetweenprobabilityoffalsealarmsandprobabilityofmissedde-tectionsintheselectionofTnorm.AtoosmallTnormcausesmanyunusedsubcarrierstobedetectedasoccupiedandgivesrisetoahighPFA,whereasatoolargeTnormcausesPMDtoincrease.AnanalysisoftheerrorprobabilitywithrespecttotheTnormemployedisprovidedinSection2.7.27

PAGE 46

Tohaveasmanyspectrumopportunitiesaspossible,theopportunisticsystemhastominimizetheinterferencethatiscausedbythetimingmismatch.Asasolution,thesyn-chronizationpointcanbedeterminedaccordingtoD1;iofuplinkusers;maybeshiftedtowardsalaterpointthantheintuitivesynchronizationpoint,whichisthedelayoftherstarrivinguser.Inpractice,userlocationinformation(e.g.throughGPS)mightbeutilizedbythesecondarysystemtoestimatetheD1;i.Inthefollowing,aclosedformequationfortheinterferenceminimizingsynchronizationpointisderived,denotedbyopt.AsinSection2.4.1,thederivationisbasedonminimizingICI. LetSdenotethepointwheretheusefulpartofthereceivedsignalstarts,i.e.,S=+Ncp.Assumingasingleoccupiedsubcarrier7pifromeachuser,andreplacingtheNcpterminh2i(p;k)givenin(2.26)withS,thetotalICIpowerisgivenbyXk6=piI22;i(k)=NuXi=1Xk6=pi1cos2(pik)

PAGE 47

SubstitutingthecosineandsinetermswiththersttwotermsoftheirTaylorseriesexpan-sion,wherepia
PAGE 50

WhentheNTsareused,ontheotherhand,thethresholdsarenotspecicallycondi-tionedonthePFAvalues(capturedbythex-axisontheplots),buttheyaresetadaptivelybasedon(2.48).Then,thesimulationresultsareaveragedoverseveralrealizationsinordertoobtaintheaveragePFAandPDvaluesforagivennormalizedthreshold,whichareplot-tedinFig.2.6andFig.2.7.ItisobservedthatwhentheNTin(2.48)isused,thereceiveroperatessomewhereonthecorrespondingROCcurves(PDversusPFArelationcapturedbyequations(2.43)-(2.45))atthesameSNR.Notethatforlargerreceivedsignalenergies,thethresholdincreaseswhenanNTisused(i.e.,thethresholdissetadaptively),whileitisconstantforNBT.Hence,byusingNT,thePFAmaybedecreasedwithsomeacceptabledegradationinthePD.Forexample,forEsc;i=2=7dBandTnorm=0:6(withnoICI),wehave(PFA,PD)(0:140;0:920)forNT.WhentheEsc;i=2isincreased8to10dB,withNT,the(PFA,PD)(0:040;0:983).Ontheotherhand,withNBTthatusesthesamethresholdasintherstcase,wewouldhave(PFA,PD)(0:140;0:993),where0.993isonlyslightlylargerthan0.983,but0.040isconsiderablysmallercomparedto0.140.Hence,

PAGE 52

Inthesimulations,errorprobabilityinopportunitydetectioniscomputedasthesumofPMDandPFA.Forallassignmentschemesused,theoccupancyrateofthesubcarriersiskeptat50%tohaveequalcontributionfromPMDandPFAtothetotalerrorprobability.Themaximumdelaythatthelatestarrivingusersignalcanhaveismaxanditisconsideredtobebetween0sand60s,where~iU(0;max)forallusers.Notethatmaxvaluesgreaterthan11.2sexceedtheCPduration. Fig.2.9andFig.2.10demonstratetheerrorprobabilityformaxvaluesupto60sbothforRAandBA,forblocksizes43and96(showninFig.2.8(c)andFig.2.8(d)),respec-tively.ThereasonforexcludingthesimulationresultsforblocksizesgiveninFig.2.8(a)andFig.2.8(b)istheirnumericalclosenesstotheothertwo.BothinFig.2.9andinFig.2.10,anoptimumTnormisusedinallcases(see[5]foradetailedanalysisofobtainingoptimumTnormindierentscenarios).ItisobservedthatinRA,ICIhasamoredestructiveeectonthedetectionperformance.ThetworeasonsfortheerrorratesbeinghigherinFig.2.10than34

PAGE 53

TheresultsoftheerrorprobabilityversusmaxanalysisperformedfortheESPRITalgorithm(forablocksizeof96)aredisplayedinFig.2.11.ItisobservedthatthereisaconsiderableperformancedierencebetweenRAandBAinhighSNRvalues.ForlowSNR,ESPRITperformanceisconsiderablypoorregardlessofthesubcarrierassignmentschemeorthemaxvalue.AcomparisonoftheerrorprobabilitiesdemonstratedinFig.2.9andFig.2.11indicatesthattheESPRITperformanceisinferiortotheenergydetectionperformancewiththegivensetofsimulationparameters.Themainreasonforthisfactisthatthereareonly6symbolsoverwhichtheESPRITalgorithmneedstoobtainthecovariancematricesitrequires.ItisfoundthatESPRITperformancecouldcompetewithenergydetectiononlyifthesamesubcarrierassignmentwereusedoveraveryhighnumberofsymbols,sothatESPRITcancomputethecovariancesreliably.Thesimulationresultsthatcompare35

PAGE 54

AnotheranalysisisperformedonthevariationoftheerrorprobabilitywithrespecttoTnorminordertodeterminetheoptimumTnormindierentpracticalscenarios.TheerrorprobabilityversusTnormcurvesareshowninFig.2.13forBAandinFig.2.14forRA.Thecurvesthatcorrespondtothelowestandhighestmaxvaluesconsideredinthesimulations(0sand60s)aredisplayedforSNRsof0dB,10dB,and20dB,whereallreceived

PAGE 55

TheerrorprobabilityversusTnormanalysisisperformedfordierentpracticalmacrocellscenarios,aswell.Inthesepracticalsimulations,receivedusersignalpowersaredistance-dependentduetofreespacepathloss.ItisaimedtodetectsubcarriersofuserswhoseaverageSNRexceeds5dB.Fig.2.15showstheerrorprobabilitiesobtainedforBA(blocksize96),wherethedistancesof12userstothesecondaryreceiverareshowninthelegend.AnimportantobservationinFig.2.15isthattheoptimumTnormisfoundtobearound0.05inallpracticalscenariosconsidered,whichmatchedwiththehighSNR,lowmaxcaseinFig.2.14. ThevariationofinterferencepowerwithrespecttouserdelayisinvestigatedinFig.2.16forbothAWGNandMPchannels.Thedelaysconsideredareroundtripdelays(RTDs),37

PAGE 56

Figure2.12ComparisonofESPRITandenergydetectionalgorithmsoverincreasingnumberofsymbols.Tnormvaluesconsideredforenergydetectionare0.50,0.25,and0.05.38

PAGE 57

Figure2.14ErrorprobabilityversusnormalizedthresholdforrandomizedassignmentwhenalluserSNRsarethesame;max=0sand60s(forSNR=20dB,10dB,and0dB).39

PAGE 58

Simulationresultsforthevariationofinterferencepowerwithrespecttothesynchro-nizationpoint()areprovidedinFig.2.17forcase1,whereuserdistancesvaryfrom150mto1800m(instepsof150m),andforcase2,whereuserdistancesvaryfrom500mto1600m(instepsof100m).Inbothcases,allusersignalpowersareequal.Theoreticalval-uesarealsoobtainedusing(2.49)andtheyareshowntomatchwiththesimulationresults.Itisrevealedthatthepointwheretheinterferenceisminimized(opt)maybeconsider-ablylaterthanthedelayoftherstarrivingsignal,andthegainthatcanbeobtainedbyoptimizingthesynchronizationpointmaybeashighas3dB.InFig.2.17,theoptvalues40

PAGE 59

Figure2.17VariationofICIpowerwithrespecttosynchronizationpoint;NCP=16samples.Case1)Userdistances(inm):[150:150:1800].Case2)Userdistances(inm):[500:100:1600].41

PAGE 61

Besidestheirusesforenhancingcommunications,femtocellscaninteractwithsmarthomeappliancesandpersonalcomputersinsidethehouse.Thisway,anfBS-centricnetworkofhomeelectronicscanbeformed,whichwouldenabletheuserstoremotelyconnectto

PAGE 63

SomerelatedworkonavoidinginterferenceinOFDMAnetworksthroughspectrumsensing[20]andthroughintelligentradioresourceallocation[55,56]isavailableintheliteratureinthecontextofcognitiveradiosystems;however,theseworkdonotconsidersystem-specicissuesrelatedtofemtocells. Signicantimprovementsinthroughputperunitareahavebeendemonstratedin[2]whenOFDMA-basedWiMAXfemtocellsareused(ontheorderof15timesthroughputimprovementfordense-deploymentsinlargecells);however,co-channelinterferencewasstatedasanimportantfactorthatlimitstheoverallnetworkperformance.In[57],trade-osbetweenpublicaccessandprivateaccesswerecomparedforWiMAXfemtocellsthroughrealisticsystem-levelsimulations,andpublicaccesswasshowntoyieldconsiderablylarger

PAGE 64

Themajorgoalinthischapteristointroduceaninterferenceavoidanceframeworkbe-tweenafemtocellandthemMSs,whichisbasedonnotusingtheresourceblocksoccupiedbycloselylocatedmMSs.Availabilityofmacrocellfrequencyschedulinginformationisconsidered,andthisinformationiseectivelyutilizedinconjunctionwithspectrumsens-ing.Moreover,inter-carrierinterference(ICI)frommacrocellULtothefemtocellULisanalyzed.ThevariationofICIiswithrespecttomMS-to-fBSdistanceisinvestigatedviasimulations,andhowICIaectsthedecisionsaboutULspectrumopportunitiesatafemtocellisdemonstrated. Therestofthechapterisorganizedasfollows.ThesystemmodelisprovidedinSec-tion3.2.CCIandICIproblemsintheco-channelfemtocelldeploymentarediscussedinSection3.3.InSection3.4,theproposedCCIavoidanceframeworkisintroduced.Simula-tionresultsaredemonstratedinSection3.5.Section3.6concludesthechapter.3.2SystemModel N;NcpnN1;(3.1) wheremisthesymbolindex,Ptx;iisthetransmitpowerforuseri,k2iisthesubcarrierindex,iisthesetofsubcarriersoflengthNiassignedtouserioutofNtotalsubcarriers,46

PAGE 65

Thetimedomainaggregatereceivedsignalisthesuperpositionofsignalsfromallusers,eachofwhichpropagatesthroughadierentmultipathchannelandarrivesatthereceiverwithadelay~i=dN~i=Te,where~iisthepropagationdelayexperiencedbyuseri,andTisthedurationoftheusefulpartofthesymbol.Then,aggregatediscrete-timereceivedsignalcanbeexpressedasy(n)=NuXi=1yi(n)+w(n);(3.2) wherew(n)denotestheadditivewhiteGaussiannoise(AWGN),andyi(n)=p wherePrx;iisthereceivedpowerforuseri,Ldenotesthetotalnumberofmultipathcom-ponents(MPCs),(m)i(l)isthelthMPCforuseri,andDl;i=dNl;i=Te+~i,wherel;iisthedelayofthelthMPCforuseri.IfDl;iNcp,thefrequencydomainsignalforthekthsubcarrierofuseriisgivenbyY(m)i(k)=p N;(3.4) whereEsc;iistheaveragereceivedenergypersubcarrierforuseri,whichisequaltoPrx;i.Thisimpliesthatthereceivedsymbolisonlyaphaserotatedversionofthetransmittedsymbol.3.3CCIandICIIssuesinFemtocellDeployments3.3.1DeploymentCongurations

PAGE 68

ThedistancebetweenanmMSandafemtocellnetworkmaybeaslargeastwicethecellradius.Thissuggeststhatevenwithco-channeloperation,CCIbetweenafemtocellnetworkandsomefarawaymMSsmaybeinsignicant.Hence,bothinULandDL,afemtocellnetworkmayintelligentlyutilizetheresourceblocksofthesefarawaymMSs,andavoidco-channeloperationwithclose-bymMSsinordertominimizetheCCIproblem.Ontheotherhand,sincemBSislocatedatthecenterofthemacrocell,itsdistancetoanyparticularfemtocellissmallerthanthecellradius.Hence,thereceivedinterferencebythefemtocellisconsiderablystronganditneedstobecancelledatthefMSs.Somerecommendationsforreducingtheseinterferenceproblemshavebeendiscussedin[51].Forexample,intelligentpowercontroltechniquesmaybeemployedbyanfBStomanageinterferencereceivedfromthemacrocellortheneighboringfemtocells.Openaccessoperationisalsodescribedasapossiblesolutiontohandleinterferenceto/fromaclose-bymMS.DetailsofthemethodthatweproposetoaddressCCIwillbepresentedinSection3.4.3.3.3ICIintheUplink

PAGE 69

Figure3.3TherelationbetweenthearrivaltimesofmMSsignaldelays,CP-size,andICI.(a)Structureofafemtocellsymbol,(b)SignalarrivaltimesfromfourdierentmMSsatanfBSthatarewithintheCP,(c)mMSsignalarrivaltimesthatexceedtheCP,and(d)IllustrationofICIduetodelayslargerthanthefemtocellCPsize.(m-RB:Macrocellresourceblock,f-RB:Femtocellresourceblock)51

PAGE 70

N+N1Xn=Dl;iNcpx(m)i(nDl;i)ej2kn N:(3.5) Afterplugging(3.1)into(3.5)andsomemanipulation,wehaveY(m)i(k)=p wherethedesiredsignalandthetotalICIarerespectivelygivenasSd;i;l(k)=X(m)i(k)Ki;l(k)ej2kDl;i N(3.7)Ii;l(k)=1 {z }hi(p;k)X(m)i(p)ej2pDl;i N+X(m1)i(p)ej2p(NcpDl;i) {z }gi(p);(3.8) whereKi;l(k)=NDl;i+Ncp withI(i;k)denotinganindicatorfunctiongivenbyI(i;k)=8>><>>:1;ifk2i;0;ifk=2i:(3.10) Theuserdistance(di)hastwoopposingeectsontheICIpower.SinceuserdelaysaredirectlyproportionaldithroughD1;i=di=(cTs),wherecisthespeedoflight,andTsisthe52

PAGE 71

KnowledgeaboutthevariationofreceivedsignalaswellasICIpowerswithrespecttodistancecanbeespeciallyusefulinthefollowingspecicscenario.AssumethatthemacrocellBSprovidesthefBSsnotonlywiththemMSschedulinginformationbutalsowiththegeographicalcoordinatesofthemMSs.Moreover,thefBSsarenotcapableofspectrumsensing,whichmeansthat,fordeterminingthespectrumopportunities,theysolelydependontheinformationthattheyreceivefromthemBS.Inthisscenario,thefBS53

PAGE 73

ThestepsoftheCCIavoidancebasedframeworkthatweproposefortheULcanbesummarizedasfollows.First,thefBSreceivesthemMSschedulinginformationfromthemBS.Then,itperformsspectrumsensingforndingtheoccupiedpartsofthespectrum,whicharesupposedlytheresourceblocksofnearbymMSsanddeterminesthespectrumopportunitiesbycomparingthesensingresultswiththeschedulinginformation.Finally,thefBSschedulesitsfMSsoverthespectrumopportunitiesdetermined. ThemainpointsofhandlingtheCCIinthedownlink,ontheotherhand,canbesummarizedasfollows:

PAGE 74

AnexamplefortheCCIcancellationmethodsthatcanbeemployedbythefMSsisintroducedin[60],whereaLeastMeanSquare(LMS)MaximumLikelihoodEstimation(MLE)methodisproposedtoestimatethetimevaryingchannelsandthetransmittedsymbolssimultaneously.Also,anMLEbasedjointdemodulationalgorithmisprovidedin[10]forreceiverswithmultipleantennaswhereavailabilityofreliablechannelestimatesisassumed.Adetaileddiscussionofthesecancellationmethods,however,isbeyondthescopeofthischapter. Intheremainderofthissection,twocriticalstepsoftheproposedco-channelfemtocelloperationframeworkwillbediscussed.3.4.1ObtainingSchedulingInformationfromthemBS

PAGE 75

Becausethelargeaheadtimemakesitdicultforachanneldependentschedulertocaptureinstantaneouschannelvariation,eciencyofsuchaschedulerislikelytobede-graded.Alsodeliveryofschedulinginformationconsumespreciousbandwidthoftheback-haul,whichotherwisecouldbeusedtodelivertheactualdata.Onepossibleimprovementtoaddressthiseciencyproblemistomakeaheadschedulingdecision(andaccompaniedsensingandcomparison)onlywhenitisnecessary.Forexample,itcanbedoneonlyinthefollowingcases:initializationoffBSs,handoofmMSsfrom/toneighboringmBSs,signicantchangeofchannelstatus,interferencelevel,orresourcerequests,etc.Mostoftheothertimes,channeldependentschedulercouldmakeschedulingdecisionconsidering57

PAGE 76

NotethatasanalternativetoanfBSmakingschedulingdecisions(basedonitsspectrumsensingresultsandschedulinginformationfromthemBS),thespectrumsensingresultsofthefemtocellmayalsobecommunicatedtothemacrocellnetworkoverthebackhaul.Then,themacrocellnetworkmaymakeglobalschedulingdecisionsandcommunicatethesebacktothefemtocells.However,consideringlargenumberoffemtocells,communicationofsuchinformationmaybecostly,andanautonomousoperationofthefemtocelltomakeitsownschedulingdecisionsmaybemorepractical. Inordertolowertheburdenofspectrumsensingforthefemtocellandtolimittheinformationstorageandcomparisonoverheadforthenetwork(suchasinthelatterscenariodiscussedabove),thescopeofsensing,informationstorage,andcomparisonmaybelimitedtoanagreedtimeintervalthatiscommunicatedbetweenanfBSandthemacrocellnetwork.Theagreementcanbemadesuchthatthespectrumissensedforonlyacertainamountofabsolutetimeoracertainnumberofframes,slots,ortransmissiontimeintervals(TTIs)withrespecttoaparticularmBSorfBSintheneighborhood.Thescopeofspectrumsensingcanbefurtherlimitedinthefrequencydomaintoacertainsetofresourceblocks(RBs)orsubcarrierstofurtherdecreasetheburdenofspectrumsensing.3.4.1.2ObtainingtheSchedulingInformationOvertheAir

PAGE 77

TheoccupiedpartsoftheULspectrumcanbefoundbysimplemeanssuchasenergydetection.EnergydetectionisperformedbytakingtheFouriertransformofthereceivedtimedomainsignalr(t),whichyieldsa1Nfrequency-domainsamplevectorrrec.Then,itsmagnitudesquarediscomputed,andcomparedwithanenergydetectionthreshold(ED)thrs.EnergydetectionoutputsforallthereceivedsubcarrierscanbewrittenasED=Ujrrecj2(ED)thrs;(3.11) whereU(x)denotestheunitstepfunctionthatindividuallyappliestoalltheelementsofavectorx,andEDisa1Nvectorwithelements2f0;1g(a0impliesthatthesubcarrierissensedasunoccupied,whilea1impliesthatthesubcarrierissensedasoccupied). Oncethespectrumsensingresultsareavailable,thefBSmaycomparethespectrumsensingresultswiththeULschedulinginformationtodecideaboutthespectrumopportu-nities.DenotingtheschedulingvectorofamMS-jwith(j)sch(elementsoftheschedulingvectorare0sand1s),thedecisionforoccupiedresourceblockscanbemadeasfollowsused=NuXj=1UhED;(j)schi jj(j)schjj1jnt(j)sch;(3.12) whereU(x)istheunitstepfunctionthatappliestoalltheelementsofavectorx,hx;yiistheinnerproductofvectorsxandy,jjxjj1istheL1-normofavectorx,and0
PAGE 78

Spectrumsensingresultsareimpairedwithmisseddetections(MD)andfalsealarms(FA)duetoadditivenoise.Inthemacrocelluplink,anotherreasonfortheseimpairmentsistheICIthatiscausedbythetimingmisalignment.IftheinterferencelevelofanmMSwithanindex-jisstrong,itsatisesUhED;(j)schi=jj(j)schjj1jnt=1;(3.13) whichimpliesthatifthatmMS'sdelayislargerthantheCPofafemtocellsignal,itmayalsoyieldastrongICIinsomeresourceblocksleadingtofalsealarms.DenoteintasthesetofmMSsthatsatisfytheaboveequality.InordertodetectifthereareanyresourceblocksthataresubjecttostrongICI,thefemtocellcheckstheneighboringresourceblocksofmMSsthatbelongtothesetint.Typically,ICImayimpactonlyfewresourceblockswithintheneighborhoodofacertainsubcarrier.LettingNICIdenotethenumberofresourceblockswhereICImaybeconsideredasdominant,thesetofresourceblocksthatwillbecheckedforthepresenceofICIisgivenbythefollowingvector~sch(NICI)=Xj2intsign"NICIXk=NICILshift(j)sch;k#(j)sch!;(3.14) whereLshift(x;k)isalogicalright-shiftoperatorwhichshiftstheelementsofabit-vectorxinkbitpositionstowardsright,andsign(x)isasignoperatorwhichoperatesindividuallyonalltheelementsofavectorx.OncetheresourceblocksthatmaybesubjecttoICIareobtainedinvector~sch(NICI),theyarecomparedwiththeenergydetectionthreshold(ED)thrsasfollowsICI=U~sch(NICI)jrrecj2(ED)thrs;(3.15) whereisusedtoindicateaterm-by-termproduct.Then,alltheoccupiedresourceblocksarecapturedbythesumICI+used,whosecomplementarysetyieldsthespectrum60

PAGE 79

Asasimpleexample,considerthemacrocellULscenarioillustratedinFig.3.7,wherearesourceblocksizeof12subcarriersasinLTEisconsidered.ThefBScomparesthespectrumsensingresults(showninFig.3.7(c))withtheschedulinginformation(showninFig.3.7(b))todeterminetheresourceblocksthatconstitutespectrumopportunities.Energydetectionappliedtothereceivedsignal(Fig.3.7(a))yieldssomeFAsduetonoiseandduetoICIasshowninFig.3.7(c).Thesefalsealarmsmayormaynotcausearesourceblocktobeconsideredasoccupiedbasedontheirnumberwithineachresourceblock.ThespectrumopportunitiesfoundfortheillustratedscenarioareshowninFig.3.7(d),whereschedulinginformationisplottedagainforvisualcomparison.NotethatthespectrumsensingresultsindicatethatthereceivedsignalpowerformMS2isrelativelyweakbecauseitisapparentlyafar-awaymMS.ThismeansthattheresourceblocksassociatedwithmMS2maybeutilizedbythefemtocellbothintheuplinkandinthedownlink.3.5SimulationResults Thepathlossmodelusedinthesimulations,whichisderivedfromtheOkumura-Hatamodel,isobtainedfrom[52].Themodel,whichisapplicabletosmalltomediumcities,yieldsthepathloss(indB)asfollowsL=46:3+33:9log(fc)13:82log(hb)+44:96:55log(hb)log(d)F(hM);(3.16)61

PAGE 80

Parameter Value Centerfrequency(fc) 700MHz,2GHz Bandwidth(B) 5.714MHz FFTsize 512 Symbolduration 89.9sec CPsizes 1/32,1/16,1/8,1/4 BSheight(hb) 30m,50m MSheight(hM) 2m MStransmitpower 27dBm Antennagain 16dB Wallpenetrationloss 15dB NumberofWalls 1(external) NoiseooroffBS -174dBm/Hz+10log10(B)=-106.43dBm wherefcisthecenterfrequency,hbisthebasestationheightaboveground,hMisthemobilestationheightaboveground,andF(hM)isgivenbyF(hM)=1:1log(fc)0:7hM1:56log(fc)0:8:(3.17) Concerningthescenarioathand,hbshouldbeconsideredastheheightofafemtocellBS,whichisfoundinahigh-risebuilding.Theparametersrelatedtowavepropagationandpathlossusedinthesimulationsaremainlyselectedaccordingtothevaluesgivenin[3]and[52].AllsimulationparametersarelistedinTable3.1. InFig.3.8,thevariationoftheICIpowerdependingontheround-trip-delay(RTD)isplottedwherethelargestdelaycorrespondstoadistanceof5km.Toobtainthesimulatedresults,theresourceblocksallocatedtothemMS2arerandomlyspreadaroundthegivenspectrum,andICIismeasuredbydeterminingthetotalenergyintheunusedresourceblocks.Thetheoreticalcurves,ontheotherhand,areplottedutilizingtheexpressionsforICIgivenin(3.8)toverifythesimulationresults.Fromthecurvesplotted,itcanbeconcludedthat62

PAGE 81

ToinvestigatehowtheICIpowerisaectedbythechangesincertainimportantsystemparameters,theICIversusdelayanalysisisperformedfortwodierentcenterfrequenciesandtwoBSheights(employingaCPsizeof32samples).TheresultsdemonstratedinshowthatemployingalowercenterfrequencyorhavingtheBSatahigherlocationmightconsiderablyincreasethereceivedinterferencepower. AnotheranalysisisperformedontheerrorprobabilityindetectingtheoccupiedandunoccupiedsubcarriersinthereceivedULsignalviaenergydetection.Theerrorprobabilityiscomputedasthesumofprobabilityofmisseddetection(PMD)andprobabilityoffalse Figure3.8VariationofICIpowerwrt.RTDfordierentcenterfrequenciesandfemtocellBSheights.63

PAGE 82

Twodierentsubcarrierassignmentschemesareconsidered.Therstoneisablockwiseassignment,wherethetwoschemesusedinthesimulationsareanLTEresourceblockwith12subcarriersand7symbols,andaWiMAXULPUSCtilewith4subcarriersand3symbols.Theseschemeswillbeshortlydenotedas127and43,respectively.Theotherassignmentschemeconsideredisarandomizedassignment,whereeachindividualsubcarriermaybeassignedtoadierentuser.Thetworandomizedassignmentschemesemployedinthesimulationsare17and13.Althoughnotusedinanystandard,theseschemesareincludedinoursimulationstoinvestigatetheeectofusingsmallnumberofsubcarriersasanassignmentunit.ThemaximumRTDthatthelatestarrivingusersignalcanhave(max)isconsideredtobebetween0sand60s,whereallotheruserRTDsarebetween0sandmax.Notethatmaxvaluesgreaterthan11.2sexceedtheCPduration. Figure3.9Errorprobabilityversusmaxforenergydetectionwithblockwiseandrandomizedassignments.64

PAGE 84

Duetotheinvolvementofmultipleusersignals,theuplinkofOFDMAsystemsposesanumberofchallengesthatdonotexistinthedownlink(DL).Mostoftheseproblemsincludingmultiuserchannelestimation[66],carrierfrequencyoset(CFO)estimation[67],synchronizationandsymboltimingestimation[36,68],multiuserinterferencecancellation[69],andsubcarrierandpowerallocation[70]areinvestigatedextensivelyintheprior-art.However,theproblemofseparatingULusersignalswithoutaccesstothesubcarrierassignmentscheme(SAS)hasnotbeeninvestigatedindetailintheliterature.

PAGE 85

UserseparationinUL-OFDMAwasconsideredin[71]forinterleavedOFDMAsystems.In[71],subcarriersallocatedtodierentusersfollowacertainperiodicstructure,whichleadstoauserspecicCFO.Hence,byestimatingtheCFOs,dierentusersignalsareidentiedandseparated.Inthischapter,however,weproposeasemi-blinduserseparationalgorithmthatcanbeappliedtoanySAS,whichdoesnotnecessarilyinvolveanyperiodicity.TheuserseparationalgorithmconsideredinthischapterisbasedonexploitingthedierencesinuserCFOsanddelays.IntheuplinkofanOFDMAsystem,CFOsofusersvaryduetothedierencesinoscillatorfrequenciesaswellastheDopplershiftscausedbythedierentvelocitiesofusers.Userdelays,ontheotherhand,varyduetothedierentdistancesofuserstotheULreceiver. Inthischapter,weassumetimesynchronizationtotherstarrivingULusersignalaswellasavailabilityofinformationonthebasicOFDMAsystemparameterssuchasFFTsize,samplingtime,andcyclicprex(CP)duration.Consideringscenarioswhereinformationaboutblockdimensionsisnotavailable,ablocksizeestimationalgorithmisdevisedthatexploitsthecorrelationbetweenthepilotsubcarrierswithinthesameblock.AGaussianapproximationbasedapproachisthenintroducedthattriestodeterminethepotentialperformanceoftheblocksizeestimationalgorithmtheoretically. Thesecondalgorithmproposedaimsatuserseparation.ItestimatestheCFOsanddelaysforeachblockseparatelybyperformingcross-correlationsoverpilotsubcarriers,andgroupstheblocksintheULsymbolaccordingtotheirCFOsanddelaysusingthe67

PAGE 86

Theorganizationofthechapterisasfollows.Section4.2providestheUL-OFDMAsystemmodel.InSection4.3,theblocksizeestimationmethodispresentedandaGaussianapproximationapproachtoblocksizeestimationisgiven.InSection4.4,amathematicalmodeloftheproposeduserseparationalgorithmisprovided.InSection4.6,thepotentialcontributionofblocksizeestimationanduserseparationalgorithmstospectrumopportu-68

PAGE 87

whereEtx;iisthetotaltransmittedenergypersymbolforuseri,NistheFFTsize,iisthesetofsubcarrierswithNielementsassignedtouserioutofSusedsubcarriers,k2iisthesubcarrierindex,NCPisthelengthofthecyclicprex,andX(i)(k)isthedataonthekthsubcarrierofithuser. AreceivedsymbolofuseriaftertheFFToperationcanbewrittenasR(i)(k)=X(i)(k)H(i)(k)ej2ki=Nejisinc(i)ejkisinc(ki)eji+I(i)(k)+W(k);(4.2) whereiisthecarrierfrequencyoset(normalizedbythesubcarrierspacingfs=N,wherefsisthesamplingfrequency),iisthesamplingclockerror,iisthetimingosetofuseri,iistherandomphasenoisecausedbytheinstabilityofuseri'soscillator,H(i)(k)isthefrequencyselectivechannelofuseri,I(i)(k)istheinter-carrierinterference(ICI)ofuseri,andW(k)iscomplexadditivewhiteGaussiannoise(AWGN).Intheremainderofthischapter,itwillbeassumedthattherandomphasenoiseaswellasthesamplingclockerrorin(4.2)arenegligible. From(4.2),itisseenthattheCFOhastwoeectsonthereceivedsignal.First,itresultsinamplitudedegradationandaconstantphaseshift,andsecond,inICI.Anothereect,whichbecomesapparentwhenthephasesofidenticalpilotsubcarriersintwoadjacentsymbolsarecompared[72],isaphaseshiftthatchangeslinearlyoversymbols.Takingthis69

PAGE 88

wheremisthesymbolindex.4.3BlockSizeEstimation Ifthecoexistenceoftheprimarynetworkandthecognitiveradioiscooperative(whichmightbethecase,e.g.,inacognitivefemtocelldeploymentwhereboththemacrocellandfemtocellsareoperatedbythesameserviceprovider),thentheprimarynetworkmightprovideinformationaboutitsfundamentalparameterssuchasN,NCP,andfstothecognitiveradio.AlthoughtheCRmightgetinformedaboutthedimensionsofB,aswell,itispossiblethattheCRhastodeterminetheblocksizeblindly. ItisfeasibletodeterminetheblocksizeofanUL-OFDMAsysteminablindmannerutilizinganyreceivedsignalY(m;k)thatcontainsanarbitrarynumberofsymbols,giventhatthetwofollowingassumptionsarevalid

PAGE 89

Althoughthesecondconditioncausessomeslightincreaseinthepeak-to-averagepowerratio(PAPR)oftheULsignal,thisincreaseistolerableespeciallyinacooperativecoexistencescenario,wheretheprimarynetworkiswillingtofacilitatecognitivecommunications. ThepilotsineachBarecorrelatedwitheachother,whereasthedatasubcarriersareuncorrelated.Also,thereisnotaconsiderablecorrelationbetweenthepilotsindierentBs71

PAGE 90

TheverticaldimensionofBcanbefoundbyperformingautocorrelationoveranen-tiresymbol(verticalcorrelation)3.WithouttakingtheeectsofdelaysandCFOsintoconsideration,wedenetheabsolutevalueoftheverticalcorrelationasR(V)(l)=EY(m;k)Y(m;k+l)=8>>>><>>>>:2s+2nl=01 wherelisthelagindex,Efgdenotestheexpectationoperation,KistheseparationbetweenthepilotsinthesamesymbolofB,2sistheaveragesubcarrierpower,and2nisthenoisepower.Notethattheexpectationisperformedoverallsubcarriers,andthe1 Inasimilarmanner,thehorizontaldimensionofBcanbeobtainedviaanautocorrelationoverrows(horizontalcorrelation),wherearowisthesetofsubcarriersatthesamesubcarrierindexk.TheabsolutevalueofthehorizontalcorrelationisgivenbyR(H)(l)=EY(m;k)Y(m+l;k)=8>>>><>>>>:2s+2nl=01 whereMistheseparationbetweenthepilotsinthesamerowofB.Theexpectationisperformedoverallsymbolsinvolvedinthecorrelation,andthe1 Inbothverticalandhorizontalcorrelations,thedesiredpeakistheonethatisstrongestafterthepeakattheorigin.Inordertoaccentuatethedesiredpeak,noiseaveragingisperformedbyaveragingR(V)overallsymbolsavailable,andbyaveragingR(H)over

PAGE 91

AnillustrativeexampleoftheverticalandhorizontalcorrelationsisprovidedinFig.4.3,wherethemainpeaksarenormalizedto1.Theblockdimensionsthatneedtobedeterminedis4subcarriersby3symbols(43)asinFig.4.2.Hence,peaksareobservedinthe3rdlagintheverticalcorrelationandinthe2ndlaginthehorizontalcorrelation.InFig.4.3aandFig.4.3b,thetheoreticalcurvesrepresentthevaluesprovidedby(4.4)and(4.5),wherethedelaysandCFOsarenottakenintoaccount.UndertheeectofdelaysandCFOs,thesecondcurvesareobtained,wherethedesiredpeaksappearweakerthanthetheoreticalvalues.ThereasonfortheweakeningofthedesiredpeaksisthatthedelaysandCFOsintroducedierentcorrelationstothesubcarriersofeachuser,which,ineect,deteriorates73

PAGE 92

LetlandldenotethemeanandthestandarddeviationofacorrelationvalueR(l)atthel-thlag,respectively.Iflpdenotesthelagcorrespondingtothepeakofthecorrelationoutputs,wehavelp>0,andlisequaltozerootherwise.Takingintoaccountthatthepeakdetectionisperformedafterabsolutevalueoperation,theprobabilitydensityfunctionofR(lp)canbewrittenasPR(lp)=1 InorderforR(lp)tohavethelargestamplitude,allothersamplesattheothercorrelationlagsneedtohaveabsolutevaluesthataresmallerthanR(lp).Thishasaprobabilityof12QR(lp)

PAGE 93

Performing(4.7)forbothhorizontalandverticalcorrelationsyieldstheprobabilitiesofdetectingthecorrespondingpeaks.DenotingthesetwoprobabilitiesasPVandPH,theprobabilityofdetectingtheblocksizecorrectlyissimplyequaltoPVPH. Notethat(4.7)isanapproximationduetotwoprimaryreasons.First,asdiscussedbefore,noise-cross-noisetermsinthepilotcorrelationsareapproximatedusingaGA.Sec-ondly,allofthecorrelationsamplesareassumedtobeuncorrelatedrandomvariables,whichisnottrueinpractice.Theexistenceofdelaysintroducescorrelationbetweensubcarriersinthesamesymbol,andtheCFOsresultincorrelationbetweensubcarriersinadjacentsym-bols.Despitethesefactors,itwillbeshowninSection4.7thattheapproximationyieldsrelativelycloseresultstothesimulationresults,especiallywhentheblocksizeisestimatedoverlargenumberofsymbols.4.4UserSeparationMethod ThetotalenergyofeachblockBcanbecalculatedasfollows(B)=X(m;k)2BY(m;k)2:(4.8)75

PAGE 94

(K+1)(M+1)H1?H0;(4.9) wherehypothesisH1impliesthatblockBisoccupied,andhypothesisH0impliesthatitisnot.DetailsofenergydetectioninOFDMA-UL,suchasoptimizingcanbefoundin[5].LetdenotethesetofalltheoccupiedB'sthatsatisfythehypothesisH1in(4.9).Then,foreachBwithin,carrierfrequencyosetanddelayestimationsareperformed. RegardingtheCFOestimation,animportantobservationfrom(4.3)isthatthelinearphaseshiftcausedbytheCFOaectsboththedesiredsignalandICIthesameway.There-fore,areliableestimatecanbeobtainedbycorrelatingtwoidenticalpilotsymbols[72],orpilotsubcarriersindierentsymbolsasillustratedinFig.4.2.Ifjdenotestheindicesofsymbols(withinthejthblock)thatcarrypilotsubcarriers,andm;jdenotesthesubcarrierindicesofpilotsinsymbolmwithinB,aestimateforB,whichwillbedenotedas^j,canbeobtainedbyperformingpairwisecorrelationbetweenm;jindierentsymbolswithinB,separatedbyMsymbols.IgnoringtheICIandnoiseterms,thiscorrelationwouldbeasfollowsr()j(M)=Xm;kY(m;k)Y(m+M;k);m2j;k2m;j;=ej2M1+NCP wheresymbolm+MiswithinB.^jcanthenbeobtainedas^j=\r()j(M) where\r()j(M)=tan1Imr()j(M)Rer()j(M):(4.12)76

PAGE 95

CarrierFreq. Speed Max.Doppler CoherenceTime 2.5GHz 2km/h 4.6Hz 200ms 2.5GHz 45km/h 104.2Hz 10ms 2.5GHz 100km/h 231.5Hz 4ms 5.8GHz 2km/h 10.7Hz 93ms 5.8GHz 45km/h 241.7Hz 4ms 5.8GHz 100km/h 537Hz 2ms Thetimingosetcausesaphaseshiftthatchangeslinearlyoverthesubcarriers,butisindependentfromthesymbolindex.Ifpk;jdenotesindicesofrowswithpilotswithinB,aestimateforB,whichwillbedenotedas^j,canbeobtainedbycorrelatingpilotsatdierentrowsseparatedbyKsubcarriers(illustratedinFig.4.2)asr()j(K)=Xm;kY(m;k)Y(m;k+K);m2j;k2pk;j;=ej2K=NXm;kjX(m;k)j2H(m;k)H(m;k+K)sinc2();(4.13) wheresubcarrierk+KiswithinB.TheestimateforBisobtainedasfollows^j=\r()j(K) where\r()j(K)=tan1Imr()j(K)Rer()j(K):(4.15) Asseenfrom(4.10),animportantconditionnecessaryfor^jtobereliableisthatthechannelcanbeconsideredconstantduringMsymbols.TakingtheWiMAXstandardasareference,Table-4.1[73]providesinformationaboutchannelcoherencetimesfortwodierentfrequencybands.GiventhattheWiMAXsymboldurationisaround0.1ms,thechannelcoherencetimecoversupto20symbolsevenataspeedof100km/hinthe5.8GHzband.Similarly,foranytypicalOFDMAbasedstandard,itcanbeexpectedthatthischannelconstancyconditionismet.77

PAGE 96

Once^j'sand^j'sareobtainedforallelementsof,theuserseparationalgorithmre-quiresthatB'sareclusteredaccordingtotheir^j'sand^j's,takingbothvaluesintoaccountsimultaneously.Eachseparateclustergeneratedbytheclusteringalgorithmcorrespondstoadierentuseriandyieldsitssubcarrierallocationvectorestimate^i. Theclusteringmethodrstyieldsanestimateforthenumberofusers(^Nu),whichisdeterminedbyndingtheclustercentersthroughthesubtractiveclusteringalgorithmoutlinedin[74,75].Acriticalinputrequiredbythesubtractiveclusteringalgorithmistheratioofdimensionsofthepotentialclusters,whichwillbedenotedasD~andD~.Inthenextstep,utilizing^Nu,theseparationisperformedviaiterativepartitioningalgorithmdiscussedin[76,77].Iterativepartitioningsplitstheinputdatainto^Nuinitialclusters.Then,foreachcluster,itcomputesthesumofabsolutedistancesfromeachpointintheclustertotheclustercentroid,wherethecentroidisthecomponent-wisemedianofthepointsinthecluster.Byminimizingthetotalofthesesumsinaniterativemanner,theclustersaredetermined. Priortoapplyingtheclusteringmethod,thesetsof^j'sand^j's,whichwillbedenotedas^and^,respectively,needtobenormalized.Thenormalizationismandatedbythefactthattherangeofnumericalvaluesfor^iswiderthantherangeof^sbyatleasttwoordersofmagnitude.Infact,clusteringwithoutnormalizationresultsinauserseparationthatissolelybasedon^values4.Inparticular,weapplythefollowingnormalizations:~=^min(^) max(^)min(^);(4.16)

PAGE 97

and~=^min(^) max(^)min(^);(4.17) respectively,whichmapboth^and^intotheinterval[0;1].Therefore,asshowninFig.4.4,theclusteringisperformedona[0;1][0;1]plane. Asecondpointrelatedtothesubtractiveclusteringalgorithmisthatitrequirestooptimizetheratioofclusterdimensionsforthebestperformance.Thisratio(D~=D~)isproportionaltotheratioofvariancesof~jand~j,i.e.,2~j=2~j,whicharerelatedtoeachotherasfollows2~j=VarA() whereA()andA()denotethesetsofall\r()j(K)'sand\r()j(M)'s,respectively.TheD~=D~inputofthesubtractiveclusteringalgorithmissetasq

PAGE 98

AvisualexamplethatillustratestheclusteringalgorithmisprovidedinFig.4.4.Itshowstheclustersina10-userscenario,whereSNRisassumedtobe30dBforallusersignals,andamultipath(MP)channelisconsideredalongwiththedelayandCFOvaluesinTable4.2.InFig.4.4,thelargereddotsconstitutetheclustercentersfoundthroughsubtractiveclustering,andthemarkerssurroundingeachofthemindicatetheresourceblocksthatbelongtoacertainuserdeterminedthroughiterativepartitioning.4.5UserSeparationApplicationsforOFDMA-BasedCognitiveRadios TheCFO,delay,andpowermeasurementsperformedforuserseparationcanbeutilizedforclassifyingthesourceofCCI.AssumingthatthemMSswillpossiblyhaveahighmobility,whilethelocationsoffBSsarexedandthemobilityoffMSsarelimited,iftheCCIiscausedbyanmMS,theCFO,power,anddelayvaluesmeasuredwillbevaryingovertime,whereas80

PAGE 99

TobeabletotrackthechangesinpoweranddelayofacertainmMS,itisrequiredtoperformthecorrespondingmeasurementsovermultipleconsecutiveframes,whereaframeisacertainnumberofadjacentsymbols,forthesamemMS.ThatmeansthatthefBSneedstokeeptrackofthesubcarriersassignedtothesamemMSovermultipleframes.Thiscanbeachievedbybuildingachainoftheclusterswithclosedelay,power,andCFOvaluesinadjacentframes,assumingthatthedelay,power,andCFOofanmMScannotchangesharplyfromoneframetoanother.4.5.2Hand-oBetweenMacrocell-BSandFemtocell-BS InformationaboutthedirectionofmovementofmMSscanmakethehand-odecisionsbetweenfBSandmBSmorerobust.Hand-odecisionssolelybasedonpowermeasurementscanbemisleadingincaseofstrongfading,i.e.,aninstantfadinginthesignalpowercantriggeranunnecessaryhand-o.However,ifthedecreaseinpowerissupportedbyanincreasingdelaymeasurement,thenthehand-odecisionwouldbemuchmorereliable. Viauserseparationbasedontrackingthedelaysandcarrierfrequencyosetsofmacro-cellusers,theusersthatarejustpassingbyfromthevicinityofthefemtocellcanbedeterminedsothatunnecessaryhand-osbetweenthemacrocellandfemtocellareavoided. Inacasewhereamacrocelluserneedstobehandedoverfromamacrocellandmultiplefemtocellsareavailable,themostappropriatefemtocellforhand-ocanbedeterminedbythesefemtocellsbysharingtheuserseparationdatatheyhave.81

PAGE 100

NotethatanfBSmayalsorescheduleitsownuserstoadierentfrequencyband,ifanmMSusingafrequencybandreusedbythefBSmovesclosertothefBS.82

PAGE 101

Parameter Value FFTSize 512 Occupiedsubcarriers 360 1/8,11.2s 10 Samplingfrequency 5.714MHz SymbolTime 100.8s 5MHz CFOs(inHz) [-500,-400,...,0,...400,500] Userdistances(inm) [100,200,400,600,...,1800] RTDs(insamples) [4,8,15,23,30,38,46,53,61,69] 83

PAGE 102

InanOFDMAbasedprimarynetwork,thespectrumopportunitiescorrespondtotheunusedsubcarrierswithinthespectrum.Asimplemethodthatmightbeemployedforthedetectionoftheseopportunitiesbythecognitiveradiosisenergydetection,where,theunusedsubcarriersmaybesimplyidentiedthroughhytpothesistestasfollows5Y(m;k)2H1?H0:(4.19) However,withsubcarrierbasedopportunitydetectionasin(4.19),eachoftheindividualsubcarriersaresubjecttofalsealarmsandmis-detections.Asanalternative,iftheresourceblocksizeisperfectlyknown,theopportunitieswithinthespectrumofaprimarysystemcanbedeterminedviatile-basedenergydetectionusing(4.9).Sinceallthesubcarrierswithinthesametileshouldallbealiatedwiththesamehypothesis(i.e.,allsubcarriersshouldbeoccupied,or,allsubcarriersshouldbenon-occupied),probabilityofmis-detectionsandprobabilityoffalse-alarmswillbeminimizedcomparedtothesubcarrier-baseddetection.Iftheresourceblocksizeisnotknown,ontheotherhand,blocksizedetectionalgorithm

PAGE 103

Asathirdtechnique,wealsoproposeanadditionalmethodinordertodecreasethefalse-alarmprobabilityoftheblock(tile)basedopportunitydetectionwithperfectblocksizeknowledge.Inthisapproach,whichwewillcalluserseparationbasedopportunitydetection,weconsidereachresourceblockwithindexjthatisestimatedtobelongtohypothesisH1(i.e.,detectedasoccupied).Then,hypothesisfortheresourceblockjischangedtoH0ifanyofthefollowingcriteriaissatisedfortheresourceblock: f^(1)j;^(2)jg<0,i.e.,thedelayestimatesfortile-jaresmallerthan0. j^(1)j^(2)jj>thrs,i.e.,dierentdelayestimatesforthesameresourceblockhaveaconsiderablylargedierence. j^jj>max,i.e.,theabsolutevalueoftheCFOestimatefortile-jislargerthanthemaximumpossibleCFOvalue. j^(1)j^(2)jj>thrs,i.e.,dierentCFOestimatesforthesameresourceblockhaveaconsiderablylargedierence. AswillbeshowninSection4.7,theperformanceofuserseparationbasedopportunitydetectioncanbeimprovedusingtheaboveteststhatposesomeconstraintsontheoccupiedresourceblocks.4.7SimulationResults

PAGE 104

Thevariationoftheperformanceswithrespecttosignal-to-noiseratio(SNR)isplottedforbothAWGNandmultipath(MP)channelsinFig.4.6andFig.4.7,wheretheblocksizestobefoundare43and66,respectively. Theresultsshowthattheperformanceheavilydependsontheblocksize.Whilethesimulatedperformanceis100%inallcasesexaminedforthe43block,itcanbearound70%forthe66blockwhentheSNRislow.Therearetworeasonsfortherelativelylowerperformanceforthe66block.First,thenumberofsymbolsandrowswithpilotsubcarriersislower,whichweakensthedesiredcorrelationpeaks.Andsecond,thephysicalseparationbetweenthepilotsislarger,which,inaMPchannel,decreasesthecorrelationbetweenthemduetothevariationofthechannelintimeandfrequency.ItisalsoworthtonotethattheGaussianapproximationmatcheswiththesimulationresultsquitewellforthe43block.ThematchbetweenthesimulationsandtheGAisstillacceptableforthe66blockwhen120symbolsareavailable.Whentherearejust60symbols,however,thereisanapparentdierencebetweenthem.Thisisduetothefactthatlpcannotbeestimatedreliablyover60symbols,andalsothecorrelationbetweenthenon-pilotsubcarriershasanon-zerovaluethatisconsiderablylargerthanincaseof120symbols.4.7.2UserSeparationSimulations Performanceinndingthenumberofusers:PNu=1001^NuNu

PAGE 105

PiNi;(4.21) whereDistheDiracdeltafunction.TheperformancesobtainedinAWGNandMPchan-nelsusinga43blockaredemonstratedinFig.4.8.TheassumptioninthecorrespondingsimulationswasthatthereceivedSNRisthesameforallusersregardlessoftheirdistance6.TheperformanceateachSNRismaximizedbyemployingtheoptimumclusterdimensiongivenbyq InTable-4.3,additionalsimulationresultsareprovidedforapracticalscenario,wherethereceivedpowersfromdierentusersdependontheirdistancestothereceiverasspeciedinTable-4.2(freespacepathlossmodelisconsidered).Thetransmissionpowerofusersis27dBm,andthereceivedsignalSNRsdescendfrom30dBtowards5dB.Theblockswhosepowerlevelsdonotexceedacertainthresholdarediscardedasin(4.9).SimulationresultsinTable-4.3showthatPNuvaluesthatexceed80%andPvaluescloseto80%areachievable. Anotheranalysisisperformedtoinvestigatetheeectofnumberofusersontheper-formanceinndingtheusersubcarriers.PisobtainedforNuvalues5,10,and20.TheCFOsofusersareequallyspacedbetween-500Hzand500Hz,whiletheuserdistancesareequallyspacedbetween2000=Nuand2000meters.ThePcurvesobtainedforbothAWGNandMPchannelsareshowninFig.4.9.Itisobservedthatasmallerusernumbersuchas5yieldsconsiderablyhigherperformance,especiallyinAWGNchannel.ItisalsoimportanttonotethatwhentheSNRlevelishighenough,even20usersignalscanbeseparatedwithanaccuracyratethatexceeds80%.

PAGE 106

InFig.4.10,theerrorprobabilitiesforfourdierentmethodsareshownforanoptimum(=0:15)andforanon-optimum(=0:50)normalizedthresholdvalue,wheretheblocksizeoftheprimarysystemis4x3.Themethodsthatareemployedaresubcarrierbased,userseparationbased,tilebasedwithknowntilesize,andtilebasedwithestimatedtilesize.Itisobservedthatthesubcarrierbasedmethodyieldstheworstperformance,whilethetilebasedmethodperformsthebest.Therefore,ifthetilesizeisnotknown,insteadofemployingthesubcarrierbasedmethod,rsttheproposedtilesizeestimationcanbeperformedandthenthetilebaseddetectionmethodcanbeapplied.Giventhattheproposedtilesizeestimationforthissmallblocksizeisveryaccurate,thisway,thedetectionperformancecanbemadeasgoodasintheknowntilesizecase.Userseparationbasedmethodisseentointroducesomeerrorsandtodegradetheperformancewhenthethresholdisoptimum.Iftheoptimumthresholdisnotavailableandanintuitivevaluesuchas0.5isemployed,however,thentheuserseparationbasedmethodimprovestheperformance. Errorprobabilitycurvesobtainedforablocksizeof6x6aredemonstratedinFig.4.11.Beingdierentfromthe4x3case,fora6x6block,theblocksizeestimationmethoddoesnotperformverywell.Therefore,thesubcarrierbaseddetectionmethodissuperiortothetilebasedmethodwithtilesizeestimation.Itisnoteworthythattheuserseparationbasedmethodisslightlysuperiortothetilebasedmethodforbothoptimumandnon-optimumthresholds.88

PAGE 107

Figure4.7SimulationandGaussianapproximationresultsforestimatingthesizeofa6x6block.89

PAGE 108

Figure4.9PerformancesinseparatingtheusersubcarriersinAWGNandMPchannelsforvariousnumbersofusers.90

PAGE 109

Figure4.11Errorprobabilityindetectingthespectrumopportunitiesusingfourdierentmethodsforaresourceblocksizeof6x6.91

PAGE 110

AWGN MP 81:25% 77:78%

PAGE 111

MIMOaddsthemultiplexinggaintotheproventransmitorreceivediversitygainsofsingleinputmultipleoutput(SIMO)andMISOsystemsasaresultofoperatingonanumberofparallelchannels[80].Itcanachievethehighspectraleciencydesiredbyfuturebandwidth-greedywirelesssystemsattheexpenseofincreasedhardwareandcomputationalcomplexity.MIMOisespeciallyimportantforOFDMAbasedWiMAXsystemsbecauseitisapartoftheIEEE802.16and802.16estandards[81,82],whichareconsideredsuitablecandidatesfor4G[83,84]. Optimally,MIMOsignalreceptionisperformedeitherbyusingmultiplereceiversorareceiverwithmultipleRFfront-ends.Inbothcases,thehardwarecostisasignicantcon-cern.Inthischapter,anattractivesolutiontothisconcernisprovidedthatemploysasinglereceiver.ConsideringaWiMAXMIMOsystem,theprimaryRFfront-endimpairmentsareanalyzed,andaguidetoestimateeachofthemisprovided.Thepossiblereasonsfordier-entimpairmentsindierenttransmitterbranchesareaddressed.Furthermore,acomplete93

PAGE 112

Theowofthechapterisasfollows.Section5.2discussestheprimaryRFfront-endimpairments.Section5.3providesaguidetoestimateandremovetheeectsofRFfront-endimpairments.Section5.4analyzesthechallengesofMIMOsystemsincomparisontoasinglechannelsystemindetail.Section5.6describeshowtohandleWiMAXMIMOsignals.Section5.7providesthedetailsaboutcombiningthetransmittedWiMAXMIMOsignalsfromtwotransmitterbranches.Section5.9concludesthechapter.5.2SignalModelandthePrimaryRFFront-endImpairments IfXm(k)isthetransmittedOFDMAsignalinthefrequencydomain,then,ignoringtheinter-carrierinterference(ICI)eects,thereceivedsignalcanbemodeledas[87]Ym(k)=Xm(k)Hm(k)F(k)exp(j2k=N)exp(jT)sinc(T)exp(jk=fs)sinc(k=fs)exp(j2km)+Nm(k);(5.1) wheremisthesymbolindex,kisthesubcarrierindex,Tisthesymbolduration,NistheFFTsize,andfs=N=Tisthesamplingfrequency.Theremainderoftheparametersandtheireectsareasfollows:

PAGE 114

whereDisthesymbollength,andjX(n)jisthebasebandtransmittedsignal.Owingtothe(identical)pilotsubcarriersthatarerepeatedregularlyintime,Z(n)canbeutilizedtoobtainthefrequencyosetbycomputing=\Z(n)=2D. andthencomputing~=\Z(k)=2M.

PAGE 116

Incaseofatimeosetbetweenthetransmitterbranches,thetimingestimationdonebythereceiverwillnotbeaccurateforatleastoneofthebranches.5.4.2EmployingSeparateClocks Althoughitismorereasonabletoemployasingleclockfortheentiretransceiver,insomecases,thedierenttransmitterbranchesmayrunseparateclocks.Thiswillbethecaseifthesignalsaregeneratedbydierentsourcessuchastwovectorsignalgenerators,ortwocollaborativelyoperatingmobiledeviceseachwithasingleantenna.Evenifthereisasingleunitwithmultipleoutputbranches,sinceeachbranchwillhaveitsownDAC,therewillbestilltwodierentsamplingclockerrors,unlesstheDACsarerunbyacommonexternalclockinput.98

PAGE 118

Simulationshavebeenrunaimingtoseewhatkindofeectsareobservedinthecon-stellationwhentherearecertaindierencesbetweenthetwotransmittedsignals.Forthispurpose,variousdierenceswereintentionallysetbetweenthetwosignals.Whenperform-ingthecorrespondingsimulations,itwasassumedthatoneofthesignalsisnotcorrupted(doesnothaveanimpairment)buttheotheronedoes.Thatmeans,oneofthesignalshasnoIQimbalance,buttheotheronehas30%IQimbalance,etc.Itshouldbealsonotedthatonlyonetypeofimpairmentdierenceisassumedtoexistatatime;theyhavebeenexaminedonebyone,becauseitmaynotbepossibletomakeareliableguessbysimplylookingattheconstellationifmultiplesuchimpairmentdierencesexistatthesametime.100

PAGE 119

InFig.5.2,theeectof=12phasedierencebetweenthetwotransmittedsignalsisobserved.Therotationofthequadruplesaroundtheircenterisapparent.Thesameeectcanbeveriedbycheckingthepositionofthepilots.Fig.5.3showstwosignalswith30%IQimbalancedierence.InFig.5.4,theeectof=12quadratureerrordierencebetweenthesignalsisshown.Finally,inFig.5.5twosignalswith0:002radianfrequencyosetdierenceareshown.Frequencyosethasasimilareecttophasedierenceintermsofrotationofquadruples.However,sincethephaseshiftcausedbythefrequencyosetincreasesoversymbols,aclearshiftisseenintheconstellationpoints.Apparently,eachoftheseimpairmentdierenceshasadierenteectontheconstellationdiagram,andstudyingthesevisualeects,onecanmakeastrongguessaboutthepossibleproblemwiththereceivedMIMOsignalbyjustexaminingtheconstellationdiagram.101

PAGE 121

Fortestingtheproposedsolutiontwodierent21setupsareemployed.Intherstsetup,TxandRxantennasareused,whereasinthesecondonetransmittersignalsarecombinedwithanRFcombinerandfedtothereceiver.ThesecondsetupisshowninFig.5.8.TheWiMAXMIMOsystemsettings,whicharecommontobothsetups,aregiveninTable5.1.Inthese1024FFTscenarios,thereare840subcarriersleftafterremovingtheguardbands. Thesolutionthatwillbeinvestigatedinthischapterisbasedontheuseofpilotse-quences.InWiMAX,eachTxbranchistransmittingaseparatesetofpilotsthatareorthogonaltoeachotheraccordingtotheirsubcarrierallocationmaps.Basically,thisisthefeaturethatenablesseparatingtheimpairmentcontributionsfromseparatebranches. Thereceivedtimedomainsignalcontainspilotsubcarriersfrombothbranches,however,itisnotpossibletoprocessthesepilotsseparatelyintime.Therefore,thepacketdetectionandthetimedomainbasedfrequencyosetestimationcanbeapplicableonlyifthetiming103

PAGE 122

Parameter Value DL/UL Downlink Numberoftransmitters 2 Numberofreceivers 1 MIMOtype MatrixB(SMUX) Permutation PUSC FFTSize 1024 Numberofsymbols 14 Bandwidth 10MHz Modulations 16QAM&64QAM CodingRates 1/2&1/2 osetsandthefrequencyosetsfromthetwobranchesareclosetoeachother.Otherwise,onlyafterconvertingthesignaltothefrequencydomain,sincepilotsfromdierentbranchesgetseparatedfromeachother,onecanapplytheosetestimations(explainedindetailforasinglechannel)topilotsfromeachbranchseparately.104

PAGE 124

Inthesinglechannelcase,aftereachimpairmentestimationthecorrespondingeectwasbeingremovedfromthesignal.IntheMIMOcase,however,sinceeachbranchhasdierentimpairments,itisnotvalidtoremovetheeectsfromthereceivedsignal.Instead,the106

PAGE 126

IntheSTTDcase,Alamoutiencoding[88]isappliedtosubcarrierpairs,wherethesamesubcarriersoftwoconsecutiveOFDMAsymbolsconstituteapair.Inthereceiverpart,theSTTDsignalsarecombinedinaspecialwaythatwillbeexplainedshortly.AsinglereceiverisenoughforcombiningSTTDsignals,andthisisveryappropriateforthepurposeofemployingasinglereceivertokeepthehardwarecostataminimum. IntheSTTDimplementationforDL-PUSCWiMAX,whichisillustratedinFig.5.6,thesignalsofthesubcarriersx1andx2,whichconstituteasubcarrierpair,aretransmittedas[x1;x2],respectively,fromtherstantenna,andintheorderof[x2;x1]fromthesecondantenna,accordingtotheAlamouticoding.Inthereceiver,thesignalsreceivedatconsecutivesymboltimesoneachcarrierpairareY1(k)=H1(k)x1+H2(k)x2+N1;(5.5)Y2(k)=H1(k)x2+H2(k)x1+N2;(5.6) whereH1(k)andH2(k)arethechannelresponses,andNiisnoise.ThesetworeceivedsignalscanbecombinedintwodierentwaystoyieldthetransmittedsignalsasfollowsC1=^H1(k)Y1(k)+^H2(k)Y2(k)=x1(j^H1(k)j2+j^H2(k)j2)+^H1(k)N1+^H2(k)N2;(5.7)C2=^H2(k)Y1(k)^H1(k)Y2(k)=x2(j^H1(k)j2+j^H2(k)j2)+^H2(k)N1^H1(k)N2;(5.8)108

PAGE 127

InUL-PUSCWiMAX,ontheotherhand,theimplementationofSTTDisdierent.Alamouticodingisappliedtoadjacentsubcarriersinthesamesymbol(ratherthanthesamesubcarriersinadjacentsymbols),asillustratedinFig.5.7.Therefore,itismorelikespace-frequencycodingratherthanspace-timecoding.Inthereceiver,thesignalsreceivedatconsecutivesubcarrierlocationsareY1(k)=H1(k)x1H2(k)x2+N1;(5.9)Y1(k+1)=H1(k+1)x2+H2(k+1)x1+N2;(5.10) Thesesignalsarecombinedasfollows andthetransmittedsignalsx1andx2canbeobtainedasinthecaseofspace-timecoding.5.8SpatialMultiplexingandJointDemodulation Injointdemodulation,ateverysubcarriereachpossibleIQsignalpair[X1(k);X2(k)]isconsideredtobeahypothesis.Eachhypothesisissimulatedbyapplyingthechannel109

PAGE 128

whereY(k)isthereceivedsignal,X1(k)andX2(k)arethetwosignalsthatconstitutethehypothesis,and^H1(k)and^H2(k)arethecorrespondingchannelestimates. Iftwotransmitterantennas,eachtransmitting,forexample,aQPSKmodulatedsignal,areconsidered,thenthereare42hypothesestocheckforeachreceiveddatasubcarrier,whichdoesnotposeaseriouscomputationalchallenge.However,thecomplexityofthismethodincreasesproportionaltoMc,whereMisthemodulationorder,andcisthenumberoftransmitterbranches.Therefore,forMIMOapplicationsthatemployanumberoftransmittersandusehigherordermodulations,thecomputationalcomplexitymaysetapracticallimittothefeasibilityofthismethod.Aversionofjointdemodulationthatutilizesmultiplereceiverscanbeconsideredasasolutioninsuchacase.5.9Conclusion

PAGE 129

Suppressionofnarrowbandinterference(NBI)inOFDMsystemswasconsideredintheprior-art[90]-[93].In[90],linearminimummean-squareerror(LMMSE)estimatesoftheinterferenceareutilized.TheproposedalgorithmrequiresaprioriinformationaboutthepowerspectraldensityoftheNBsignal.In[91],anormalizedleastmeansquares(N-LMS)adaptivenoisecancellationalgorithmwasintroducedforsuppressingNBIinpilotsymbolassistedOFDMsystems.NBIrejectionviainterferometryspreadingcodeswasproposedin[92].In[93],anNBIcancellerforOFDMsystemsisconsidered,wheretheNBsignalisestimatedovertheunusedOFDMsubcarriers.ThefeasibilityofthismethodislimitedinpracticeduetotheveryfewnumberofunusedsubcarriersinawelldesignedOFDMbased111

PAGE 130

In[4],itisemphasizedthatbyexploitingthedierencesinsignalfeaturessuchastheirdelays,initialsignalseparationcanbeobtained,whichconsiderablyincreasestheeciencyofiterativeinterferencecancellation.Inthecurrentchapter,weexploittheinherentinitialsignalseparationthatexistsduetothemulticarriervs.singlecarriernaturesofinterfer-ingsignalsaswellasthefactthattheinformationisinfrequencydomainforMCsignalandintimedomainforNBsignal.Theproposedmethodassumesavailabilityofsignalreceptionandtransmissioncapabilitiesforbothsystems.Ateachiteration,eachsignalisdemodulatedandthenregeneratedbasedonthesymboldecisionsandthechannelimpulseresponse.Thisway,abetterestimateofthesignalisobtained.Theregeneratedsignalissubtractedfromtheaggregatesignaltoobtainanestimationoftheotherco-channelsignal.Throughextensivesimulations,itisprovedthatthismethodcanprovideafundamentalimprovementintheperformancesofbothsystemsinasfewasthreeiterations.Therela-tivelyhighcomputationalburden(associatedwithmultipletransitionsbetweentimeandfrequencydomains)aswellastheextracostcausedbytheadditionofasecondsystem'stransceiverfunctionalitiesarecompensatedbythefundamentalperformancegainobtained.Ourothercontributionsincludeadetailedcomparisonofthecomputationalcomplexityof

PAGE 131

Thechapterisorganizedasfollows:Section6.2providesapplicationexamplesandthesystemmodelsfortheMCandNBsystemsinconsideration.Also,itshortlydiscussestheGAbasedsymbolerrorrate(SER).Section6.3reviewsthejointdemodulationtechniquefortheNBandMCsignals,whileSection6.4isadetaileddescriptionoftheproposedCCIcancellationmethod.AcomplexitycomparisonofthejointdemodulationanditerativeinterferencecancellationapproachesismadeinSection6.5,simulationresultsarepresentedinSection6.6,andthelastsectionconcludesthechapter.6.2ApplicationExamplesandSystemModel6.2.1ApplicationExamples

PAGE 132

Aparticularlyimportantscenariowhereinterferencecancellationmayyieldgoodgainsforfemtocellnetworksisfortherestrictedoperationmode2offemtocells,where,themacro-cellmobilestations(mMSs)arenotallowedtomakehand-otothefemtocellnetworkevenwhenthesignalqualityissuperioratthefemtocell.AsillustratedinFig.6.1,thismayresultinsignicantuplinkinterferencefromthemMStothefemtocellMSs(fMSs),andsignicantdownlinkinterferencefromthefMStothemMSs.Asdiscussedbefore,fortheinterferencecancellationtobecomeeective,theinterferenceshouldbesucientlystrong;therefore,femtocellswithrestrictedaccessareagoodapplicationscenarioforinterferencecancellationtechniques.

PAGE 133

ThesampleddownlinkOFDMAsignalintimedomaincanbewrittenasx(n)=p

PAGE 134

wherePtxisthetransmitpower,Nisthenumberofsubcarriers,kisthesubcarrierindex,Ncpisthelengthofthecyclicprex(CP),andX(k)isthedataonthekthsubcarrier. ThereceivedtimedomainOFDMAsignalthattraversesthroughamultipathchannelh(l)withLmctapsisy(n)=p wherePrxisthereceivedsignalpower,andDlisthedelayofthelthtap.AssumingthatthemaximumtapdelaydoesnotexceedtheCPlength,thefrequencydomainOFDMAsignalcanbeshownasY(k)=p whereH(k)isthechannelfrequencyresponse. Thebasebandnarrowbandsignalcanbemodeledass(n)=Xmamg(nmT);(6.4) wheremisthesymbolindex,amdenotesthemthdatasymbol,g(n)isthepulseshapinglterwitharoll-ofactor,andTisthesymboldurationofthenarrowbandsignal.In116

PAGE 135

wherep(n)isthespreadingchipsequencewithRcchips.Sinces(n)passesthroughamultipathchannelh0(l)withLnbsymbol-spacedtaps4,thereceivedsignalbecomesz(n)=p ThediscreteFouriertransform(DFT)ofz(n)willbedenotedasZ(k).ThemainlobeofthespectrumoccupiedbyZ(k)overlapswithKsubcarriersofY(k)(seee.g.,Fig.6.2andFig.6.5).Hence,ifthecenterfrequencyofZ(k)islocatedatsubcarrier,thesubcarriersk2K Intimedomain,NBsymbolsconstitutestructuredinformationfromanitealphabet,whileOFDMAsignalbehaveslikerandomnoisespreadovermultipleNBsymbols.Infrequencydomain,ontheotherhand,OFDMAsubcarrierscarrystructuredinformation,andNBsignalcanbeconsideredlikerandomandcolorednoisecoveringmultiplesubcarriers.Thisisreadilyseenfromthereceivedsignal,whichcanbedenotedintimedomainasr(n)=NBz }| {z(n)+OFDMAz }| {N1Xk=0Y(k)ej2kn=N| {z }y(n)+w(n);(6.7)

PAGE 136

}| {Y(k)+NBz }| {1 {z }Z(k)+W(k);(6.8) whereW(k)isthefrequencydomainreciprocalw(n).6.2.3GaussianApproximationBasedSymbolErrorRate N0 2Qs N0 whereQdenotestheQ-function,Ebisthebitenergy,and2Iistheinterferencevariance,whichisequaltoPrxoftheinterferingsignal. TheGAisrathersimplebutitistypicallynotveryaccurateespeciallyathighSNRvalueswheretheresultingSERtendstobeoptimistic.Forthescenarioathand,basedon(6.7)and(6.8),theinterferenceisasumofNrandomvariables.Therefore,fromthecentrallimittheorem,thisimpliesthatGAforthespecicscenarioinconsiderationwouldbeaccurate(especiallyforlargeN).TheaccuracyoftheGAistestedinapracticalco-channelinterferencescenariobycomparingitwithactualsimulationresultsinSection6.6.6.3JointDemodulationMethod

PAGE 137

DenotingtheestimatesfortheNBandOFDMAsignalsintimedomainas^z(n)and^y(n),respectively,anMLestimateofbothsignalscanbeobtainedash^am;^X(k)i=argminam;X(k)r(mT)z(mT)y0(mT)2=argminam;X(k)(r(mT)Lnb1Xl=0h0(l)aml+K wherey0(n)isthetimedomainreciprocalofY(k)fork2K Thenumberofdierentvaluesthatz(mT)andy0(mT)cantakeshouldbelimitedinorderforthejointdemodulationalgorithmtobecomputationallyfeasible.Thisconditionissatisedforbothz(mT)andy0(mT)sincethedatasequencesamandX(k)eachbelongtoanitealphabet.ThereareMKpossibilitiesfortheOFDMAsignalintheoverlappingband,andMpossibilitiesforeachoftheCsymbolsintheNBsignal,whereMisthenumberofconstellationpointsdependingonthemodulationorder(e.g.,M=4forQPSK).Therefore,thenumberofpossibilitiesthatneedtobeconsideredforeachNBsymbolisMK+1. Implementing(6.10)requiresanexhaustivesearchthroughMK+1possiblecombinationsofz(mT)andy0(mT),whichareobtainedbyapplyingthechannelresponsestoallpossiblevaluesofamandX(k)toyieldz(mT)andY0(k),respectively,andalsobycomputingtheinverseDFT(IDFT)forallY0(k)stogety0(mT)s.Thisexhaustivesearchaswellasthecomputationsrequiredforobtainingz(mT)andy0(mT)srenderthejointdemodulationmethodprohibitivelycomplexasitwillbeclearlydemonstratedinSection6.5.119

PAGE 138

Aninitialroughestimationforz(n)canbeobtainedutilizingZ(k)ifthepowerofZ(k)ishighenoughthatitcanbesensedovertheOBthroughenergydetection.Thethresholdoftheenergydetectorissetaccordingtotheaveragesignal-to-noiseratio(SNR)leveloverk=2K IftheNBsignalistooweaktoprovideausefulestimate,orifKisunknown,then,followinganalternativeapproach,R(k)isusedasaninitialestimateforY(k). Themainideaoftheproposedmethodistodemodulatetheestimatedsignal,^z(n)or^Y(k),andthentoregeneratethesignalwaveformbasedonthesymboldecisionsmadetoobtain~z(n)or~Y(k).Notethat~z(n)and~Y(k)areexpectedtobecleanerversionsof^z(n)and^Y(k),respectively,sincetheyarefreeofAWGNandsupposedlylessaectedbyCCI. Sincetheinitialestimateemployed^z(n)or^Y(k)iscorruptedbyCCIandAWGN,thesymboldecisionsmademayincludeerrors.However,theeectofsymbolerrorsmadein^z(n)isnotlocalizedinfrequencydomain;onthecontrary,itisspreadoverKsubcarriers.Similarly,acorruptedsubcarrierin^Y(k)hasanimpactthatisspreadoverNsamplesintimedomain.Hence,subtracting~z(n)withsymbolerrorsfromr(n)doesnotnecessarily120

PAGE 139

TheowchartprovidedinFig.6.3illustratesthestepsthatneedtobefollowedafterobtainingtheinitialsignal.Therststepisdemodulation.Theinternalstagesforde-modulationareshownfortheNBsysteminaseparateowchartinFig.6.4.Itstartswithdownconvertingthesignaltothebasebandfromtheintermediatefrequency(IF)off0cfc,wherefcandf0carethecarrierfrequenciesoftheOFDMAsignalandtheNBsignal,121

PAGE 140

AfterobtainingtheIQdata,regeneration(demonstratedforNBsignalinFig.6.4)takesplace.ThestepsthatconstituteregenerationareupsamplingtheIQdata,applyingpulseshaping,(ifthesignalisaCDMAsignal)multiplyingthesignalwiththePNsequence,upconvertingit,andconvolvingitwiththebasebandchannel.Again,upsamplingandupconversionarenotperformedfortheOFDMAsignal.ThepulseshapinglterusedbytheNBsystemisassumedtobeknown.Iftheregeneratedsignalis~z(n),itsDFTistaken,andtheresultingsignal~Z(k)isremovedfromR(k)toobtainanestimatefortheOFDMAsignal,i.e.^Y(k)=R(k)~Z(k)=R(k)1

PAGE 141

AnimportantquestionthatmightberaisedabouttheproposedmethodiswhytheentireOFDMAbandishandledratherthandealingwiththeOBonly,becauseprocessingtheentirebandhasthefollowingdisadvantages: ThereasonswhywedonotdealwiththeOBonlyisthatKmaynotalwaysbeknownaccurately,andalso,subcarriersk=2K Nratiosassmallas2:5%.6.5ComputationalComplexity

PAGE 142

Afterappylingthechannelresponses,allpossibleOFDMAsignalsneedtobetransferedfromfrequencydomainintothetimedomainviaMKinversefastFouriertransform(IFFT)operationsofsizeN.EachIFFToperationrequiresN TakingintoaccountthatacomplexADDisequivalentto2realADDs,andacomplexMULisequalto4realMULsand2realADDs,thecomputationsrequiredcanbelistedintermsofrealMULs,realADDs,andCMPsasinTable6.1.6.5.2IterativeCancellation IneachoftheIiterationsdesired,todemodulatetheOFDMAsubcarriers,NcomplexMULsareneededforequalizationandNlog2MCMPsformakingsymboldecisions.For124

PAGE 143

Againineachiteration,toregeneratetheOFDMAsubcarriers,NcomplexMULsareneededtoapplythewirelesschanneleect.InNBsymbols'regeneration,ontheotherhand,aconvolutionisrequiredforapplyingthechannel,whichisequaltoCLcomplexMULsandC(L1)complexADDs,andanotherconvolutionforpulseshaping,whichisequalto2CNMULsand2(C1)NADDs.Finally,ineachiterationeachofthesubtractionsfromthereceivedsignalintimeandinfrequencyrequireNcomplexADDs.ThecomputationsrequiredforeachstepoftheiterativecancellationmethodareprovidedinthesecondpartofTable6.1.125

PAGE 144

Operation MUL ADD CMP MaximumLikelihood ApplyingChannelResponses 4MKK+4MCL TakingIFFT 2MKNlog2N r(n)^z(n)+^y(n) MK+1C TakingFFT/IFFT 4(1+I)Nlog2N 2K K K 4IN 4ICLML NBEqualization(LE) 4ICL SymbolDecision 4IN+4ICL PulseShaping 2ICN r(n)~y(n)&R(k)~Z(k) 4IN

PAGE 145

N K C L M Iter.I Iter.II 512 40 32 4 4 3 5.31028 20 16 4 4 3 4.71016 10 8 4 4 3 4.41010 40 32 4 16 3 7.01052 40 32 4 4 5 5.31028 40 32 1 4 3 5.31028 40 32 4 4 3 1.11029 InTable6.2,itisobservedthatthereisadrasticdierencebetweenthecyclenumbersrequiredforMLandIter.Ialgorithms.ThisiscausedbythefactthateverystepoftheMLestimationhasanexponentialcomplexity,whereasIter.IhasalinearcomplexityexceptfortheMLSEequalizerthatitemploys.CyclecountsforIter.IIalgorithmshowthatthecomplexityoftheiterativecancellationcanbedecreasedfurtherbyemployingalinearequalizer,especiallywhenMorLislarge. ItisseenthatparameterK(andC,whichdependsonK)actsexponentiallyonthecomplexityofMLestimationandlinearlyontheiterativecancellation.MaectsMLestimationandIter.Iexponentially,whereasithasanegligibleeectonIter.II.Nhasalineareectonallalgorithms,andIhasalineareectontheiterativeones.LhasarelativelyweakimpactonMLestimationandIter.II,whereasitaectsIter.Iexponentially.127

PAGE 146

TheSERperformancesofOFDMA,NB,andCDMAsystemsareinvestigatedbothinAWGN(Figs.6.6-6.9)andmultipath(MP)(Figs.6.10-6.13)channels.InMPsimulations,availabilityofaperfectchannelestimationisassumedforNBandCDMA,andanecient128

PAGE 147

InFigs.6.6-6.13,theuppermostcurveshowstheperformanceobtainedwithoutapplyingCCIcancellation(referredas\withoutcancellation"),whereasthelowestcurveshowstheperformancewhenCCIdoesnotexist(referredas\NoCCI").ThethreecurvesinbetweendemonstratetheSERperformancesaftereachiteration7.TheSINRvaluesonthex-axisapplyonlytothewithoutcancellationcurve.Asalastnote,thenoCCIcurveisactuallyanSERvs.SNRcurveshiftedleftwardsbytheamountofinterferenceSNR,whichis30dBinFig.6.6andFig.6.10;25dBinFig.6.7andFig.6.11;20dBinFig.6.8andFig.6.12;and15dBinFig.6.9andFig.6.13.6.6.2AWGNChannelResults Parameter OFDMA Narrowband CDMA Bandwidth 5MHz 370kHz 625kHz Samplespersymbol 512 16 32 Modulation QPSK QPSK QPSK MPchannelmodel Veh.A Out.-to-in.A Out.-to-in.A Pulseshape Rectang. Raisedcos. Raisedcos. (=0.3) (=0.3) 129

PAGE 148

Figure6.7SERperformanceoftheOFDMAsystemundertheinuenceofCDMAinter-ference(AWGNchannel).130

PAGE 149

Fig.6.6alsoshowsthetheoreticalperformancecurvethatisobtainedbyusingtheGAfortheco-channelinterference.ItisobservedthattheGAyieldsquiteaccuratevaluesupuntil0dBSINR,afterwhereityieldsoptimisticSERs.AnotherperformancecurvethatisdisplayedinFig.6.6belongstothemaximumlikelihoodreceiver,whoseSERisaslowasthe\NoCCI"caseatlowSINRvalues.TheMLreceiverissuperiortotheiterativecancelereverywhereexceptaround0dBSINR. IncaseofCDMAinterference,thegainsobtainedfortheOFDMAsystem,whicharedisplayedinFig.6.7,areconsiderablylargerthanthepreviouscase.ThereasonforthisperformancedierenceistheinvolvementofthePNsequence,whichintroducesadditionalsignalseparability.ThefactthattheCDMAsignalpowerisspreadoverawiderfrequencyband(comparedtotheNBsignal)makestheOFDMAsignalmoreaccuratelydetectable.OncethecancellationprocessstartswithareliableestimatefortheOFDMAsignal,thefollowingiterationsbecomemoresuccessful,aswell. TheNBsystemperformanceimprovementenabledbytheproposedmethodisshowninFig.6.8.ForSINRvaluessmallerthan0dB,thegainwithrespecttonoCCIcancellationcanbeashighas18dB.ForSINRgreaterthan0dB,thereisstillagainaround3dB.Fig.6.8alsoshowstheMLreceiverperformance.MLreceiverissuperiortotheiterativecanceleringeneral.However,ataround0dBSINR,ityieldsapparentlyhigherSERthantheiterativecanceler. TheimprovementoftheCDMAperformanceisagainmoresignicantasitcanbeseeninFig.6.9.TheSERvaluesaremuchclosertothenoCCIcurveatlowSINRvalues,andthereisa10dBgainevenatratherhighSINR.131

PAGE 150

Figure6.9SERperformanceoftheCDMAsystemundertheinuenceofOFDMAinter-ference(AWGNchannel).132

PAGE 151

ImprovementoftheNBperformanceisshowninFig.6.12.ThegainobtainedforSINRsmallerthan0dBismorethan12dB.Approaching0dBSINR,thisgainbecomessmaller,butevenat10dBSINRthereisstillagainofapproximately5dB.ImpactofNBchannelestimationerrorontheperformanceofiterativecancellationisalsodemonstratedinFig.6.12.ThevarianceoftheGaussiannoiseaddedtoeachchanneltapestimateissetasacertainratioofthepowerofthattap.Thetworatiosexaminedare5%and10%.Itisobservedthatthecancellationgaindecreaseswithincreasingchannelestimationerror.Still,itcanbestatedthatchannelestimationerrors,whicharelikelytooccurunderCCIeect,donothaveaverystronginuenceaterrorlevelsaslargeas5%. TheCDMAperformanceimprovement(seeFig.6.13)ismorecritical.TheperformanceisalmostasgoodasnoCCIcaseupuntil0dBSINR,afterwhereitstartstodecrease.ThedierencebetweentheNBandCDMAcurves'behaviorisagainduetotheuseofaPNsequence.6.6.4EectofOverlappingBandwidth

PAGE 152

Figure6.11SERperformanceoftheOFDMAsystemundertheinuenceofCDMAinter-ference(MPchannel).134

PAGE 153

Figure6.13SERperformanceoftheCDMAsystemundertheinuenceofOFDMAinter-ference(MPchannel).135

PAGE 156

WeanalyzedtheopportunitydetectionperformancesofenergydetectionandESPRIT(estimationofsignalparametersbyrotationalinvariancetechniques)algorithmsinthepresenceoftimingmisalignmentsinuplinkOFDMA.Fortheenergydetector,thestatisticsofsubcarrierpowerarederivedconsideringtimingmisalignments,andtheyareveriedthroughcomputersimulations.Usingthesestatistics,whichtakeinter-carrier-interference(ICI)eectsintoaccount,receiveroperatingcharacteristics(ROCs)oftheenergydetectorreceiverareobtained.Itisshownthatenergydetection138

PAGE 157

AclosedformexpressionisobtainedfortheoptimumUL-OFDMAsynchronizationpointthatminimizestheinterferenceonthecognitiveradio.Itisshownthatitmaybeatalaterpointthanthearrivaltimeoftheearliestprimaryuser'ssignal. Aco-channelframeworkisdeveloped,whichisbasedonutilizingtheresourceblocksofmacrocell-associatedmobilestations(mMSs)thatarefarawaytoafemtocellbasestation(fBS),thereforeavoidingstronginterferencethatmayoccurbetweenafem-tocellandclose-bymMSs.Anavoidancemethodthatjointlyutilizesthespectrumsensingresultsaswellasschedulinginformationobtainedfromthemacrocellbasestation(mBS)isintroduced. Weproposedsemi-blinduserseparationalgorithm,whichestimatesthecarrierfre-quencyosetsandtimedelaysofeachblockbyexploitingthecross-correlationsoverpilotsubcarriers.Atwo-dimensionalclusteringmethodisthenemployedtogrouptheestimates,whereeachgroupbelongstoadierentuser. Amaximumlikelihoodestimationbasedmethodisdevised,whichtakesMIMOre-ceiversonestepfurtherandintroducesreceptionofMIMOsignalswithasinglere-ceiver.Theproposedmethodrelaxesthehardwarerequirementwhiledemandinghigherdigitalsignalprocessingpower.

PAGE 158

Animportantrelatedresearchproblemthatwedidnottrytondasolutionforinthisdissertationistheco-channelinterferencethatoccursinthecoexistenceoftwoOFDMAsystemswiththesameparameters.Assumingthatthesignalsofthesesystemsoverlapinalldomainsthatcanbereadilyimaginedsuchastime,frequency,code,space,andpowerdomains,thereisneedforsomemeansthatwillprovidesignalseparability.Thismightbeanewdomainoracomputationallyverydemandingdigitalsignalprocessingtechnique.Indepthresearchforanelementthatwillleadtosignalseparabilityinthiskindofaco-channelscenariocanleadtoanotherPhDdissertation.140

PAGE 159

V.Chandrasekhar,J.G.Andrews,andA.Gatherer,\Femtocellnetworks:asurvey,"IEEECommun.Mag.,vol.46,no.9,pp.59{67,Sep.2008.[2] S.P.Yeh,S.Talwar,S.C.Lee,andH.Kim;,\WiMAXfemtocells:aperspectiveonnetworkarchitecture,capacity,andcoverage,"IEEECommun.Mag.,vol.46,no.10,pp.58{65,Oct.2008.[3] H.Claussen,L.Ho,andL.Samuel,\Anoverviewofthefemtocellconcept,"BellLabsTech.J.,vol.13,no.1,pp.221{245,2008.[4] H.ArslanandK.Molnar,\Cochannelinterferencesuppressionwithsuccessivecan-cellationinnarrow-bandsystems,"IEEECommun.Lett.,vol.5,no.2,pp.37{39,2001.[5] M.E.Sahin,I.Guvenc,M.R.Jeong,andH.Arslan,\OpportunitydetectionforOFDMAsystemswithtimingmisalignment,"inProc.IEEEGlobalTelecommun.Conf.(GLOBECOM),NewOrleans,LA,Nov.2008,pp.1{6.[6] M.E.Sahin,I.Guvenc,andH.Arslan,\OpportunityDetectionforOFDMA-BasedCognitiveRadioSystemswithTimingMisalignment,"IEEETrans.WirelessCom-mun.,2009,toappear.[7] M.E.Sahin,I.Guvenc,M.-R.Jeong,andH.Arslan,\HandlingCCIandICIinOFDMAFemtocellNetworksThroughFrequencyScheduling,"IEEETrans.Consum.Electron.,2009,submitted.[8] M.E.Sahin,I.Guvenc,M.R.Jeong,andH.Arslan,\UserseparationforOFDMAuplink,"inIEEEVehic.Technol.Conf.(VTC),Barcelona,Spain,Apr.2009,pp.1{5.[9] M.E.Sahin,I.Guvenc,andH.Arslan,\UplinkUserSignalSeparationforOFDMA-BasedCognitiveRadios,"EURASIPJ.onAdvancesinSignalProc.,2009,toappear.[10] M.SahinandH.Arslan,\MIMO-OFDMAMeasurements;Reception,Testing,andEvaluationofWiMAXMIMOSignalsWithaSingleChannelReceiver,"IEEETrans.Instr.Meas.,vol.58,no.3,pp.713{721,Mar.2009.[11] M.E.Sahin,H.Arslan,andD.Singh,\ReceptionandMeasurementofMIMO-OFDMSignalswithaSingleReceiver,"inIEEEVehic.Technol.Conf.(VTC),Baltimore,MD,Oct.2007,pp.666{670.141

PAGE 160

M.E.Sahin,I.Guvenc,andH.Arslan,\AnIterativeInterferenceCancellationMethodforCo-ChannelMulticarrierandNarrowbandSystems,"ElsevierPhy.Com-mun.,2009,submitted.[13] ||,\IterativeInterferenceCancellationforCo-ChannelMulticarrierandNarrow-bandSystems,"inProc.IEEEWirelessCommun.andNetwork.Conf.(WCNC),2010,submitted.[14] S.Haykin,\Cognitiveradio:brain-empoweredwirelesscommunications,"IEEEJ.Sel.AreasCommun.(JSAC),vol.23,no.2,pp.201{220,2005.[15] G.GanesanandY.G.Li,\Cooperativespectrumsensingincognitiveradio,partI:Twousernetworks,"IEEETrans.WirelessCommun.,vol.6,no.6,pp.2204{2213,June2007.[16] ||,\Cooperativespectrumsensingincognitiveradio,partII:Multiusernetworks,"IEEETrans.WirelessCommun.,vol.6,no.6,pp.2214{2222,June2007.[17] S.Mangold,Z.Zhong,K.Challapali,andC.-T.Chou,\Spectrumagileradio:ra-dioresourcemeasurementsforopportunisticspectrumusage,"inProc.IEEEGlobalTelecommun.Conf.(GLOBECOM),vol.6,Dallas,TX,Nov.2004,pp.3467{3471.[18] J.Hillenbrand,T.A.Weiss,andF.K.Jondral,\Calculationofdetectionandfalsealarmprobabilitiesinspectrumpoolingsystems,"IEEECommun.Lett.,vol.9,no.4,pp.349{351,Apr.2005.[19] Y.C.Liang,Y.Zeng,E.C.Y.Peh,andA.T.Hoang,\Sensing-throughputtradeoforcognitiveradionetworks,"IEEETrans.WirelessCommun.,vol.7,no.4,pp.1326{1337,Apr.2008.[20] S.-Y.Tu,K.-C.Chen,andR.Prasad,\SpectrumsensingofOFDMAsystemsforcog-nitiveradios,"inProc.IEEEInt.Symp.Personal,IndoorandMobileRadioCommun.(PIMRC),Athens,Greece,Sep.2007,pp.1{5.[21] N.Khambekar,L.Dong,andV.Chaudhary,\UtilizingOFDMguardintervalforspectrumsensing,"inProc.IEEEInt.Symp.Personal,IndoorandMobileRadioCommun.(PIMRC),Athens,Greece,Sep.2007,pp.1{5.[22] T.YucekandH.Arslan,\OFDMsignalidenticationandtransmissionparameterestimationforcognitiveradioapplications,"inProc.IEEEGlobalTelecommun.Conf.(GLOBECOM),Washington,DC,Nov.2007,pp.4056{4060.[23] J.MaandY.G.Li,\Softcombinationanddetectionforcooperativespectrumsens-ingincognitiveradionetworks,"inProc.IEEEGlobalTelecommun.Conf.(GLOBE-COM),Washington,DC,Nov.2007,pp.3139{3143.[24] J.-H.Baek,H.-J.Oh,andS.-H.Hwang,\Improvedreliabilityofspectrumsensingusingenergydetectorincognitiveradiosystem,"inProc.Int.Conf.Adv.Commun.Technol.(ICACT),vol.1,Gangwon-Do,Korea,Feb.2008,pp.575{578.142

PAGE 161

M.El-Tanany,Y.Wu,andL.Hazy,\OFDMuplinkforinteractivebroadbandwireless:Analysisandsimulationinthepresenceofcarrier,clock,andtimingerrors,"IEEETrans.Broadcast,vol.47,no.1,pp.3{19,Mar.2001.[26] M.Park,K.Ko,H.Yoo,andD.Hong,\PerformanceanalysisofOFDMAuplinksystemswithsymboltimingmisalignment,"IEEECommun.Lett.,vol.7,no.8,pp.376{378,Aug.2003.[27] E.BalaandL.J.Cimini,\OntheuplinksynchronizationofOFDMAsystems,"inProc.IEEEMilitaryCommun.Conf.(MILCOM),AtlanticCity,NJ,Oct.2005,pp.1133{1139.[28] L.T.W.HoandH.Claussen,\Eectsofuser-deployed,co-channelfemtocellsonthecalldropprobabilityinaresidentialscenario,"inProc.IEEEInt.Symp.Personal,IndoorandMobileRadioCommun.(PIMRC),Athens,Greece,Sep.2007,pp.1{5.[29] H.Claussen,\Performanceofmacro-andco-channelfemtocellsinahierarchicalcellstructure,"inProc.IEEEInt.Symp.Personal,IndoorandMobileRadioCommun.(PIMRC),Athens,Greece,Sep.2007,pp.1{5.[30] V.ChandrasekharandJ.G.Andrews,\Uplinkcapacityandinterferenceavoidancefortwo-tiercellularnetworks,"inProc.IEEEGlobalTelecommun.Conf.(GLOBECOM),Washington,DC,Nov.2007,pp.3322{3326.[31] V.Chandrasekhar,J.Andrews,andA.Gatherer,\Femtocellnetworks:asurvey,"IEEECommun.Mag.,vol.46,no.9,pp.59{67,Sep.2008.[32] H.Mahmoud,H.Arslan,andM.Ozdemir,\InitialRangingforWiMAX(802.16e)OFDMA,"inProc.IEEEMilitaryCommun.Conf.(MILCOM),Washington,DC,Oct.2006,pp.1{7.[33] Y.Choi,S.Park,andS.Bahk,\MultichannelrandomaccessinOFDMAwirelessnetworks,"IEEEJ.Sel.AreasCommun.(JSAC),vol.24,no.3,pp.603{613,2006.[34] I.Guvenc,\Statisticsofmacrocell-synchronousfemtocell-asynchronoususers'delaysforimprovedfemtocelluplinkreceiverdesign,"IEEECommun.Lett.,vol.13,no.4,pp.239{241,Apr.2009.[35] J.V.Beek,P.O.Borjesson,M.L.Boucheret,D.Landstrom,J.M.Arenas,P.Odling,C.Ostberg,M.Wahlqvist,andS.K.Wilson,\AtimeandfrequencysynchronizationschemeformultiuserOFDM,"IEEEJ.Sel.AreasCommun.(JSAC),vol.17,no.11,pp.1900{1914,Nov.1999.[36] M.Morelli,\TimingandfrequencysynchronizationfortheuplinkofanOFDMAsystem,"IEEETrans.Commun.,vol.52,no.2,pp.296{306,Feb.2004.[37] R.RoyandT.Kailath,\ESPRIT-estimationofsignalparametersviarotationalin-variancetechniques,"IEEETrans.Acoust.,Speech,andSignalProcess.,vol.37,no.7,pp.984{995,Jul.1989.143

PAGE 162

B.Ottersten,M.Viberg,andT.Kailath,\PerformanceanalysisofthetotalleastsquaresESPRITalgorithm,"IEEETrans.SignalProcess.,vol.39,no.5,pp.1122{1135,May1991.[39] B.Yang,K.Letaief,R.Cheng,andZ.Cao,\ChannelestimationforOFDMtrans-missioninmultipathfadingchannelsbasedonparametricchannelmodeling,"IEEETrans.Commun.,vol.49,no.3,pp.467{479,Mar.2001.[40] R.Roy,A.Paulraj,andT.Kailath,\ESPRIT{Asubspacerotationapproachtoestimationofparametersofcisoidsinnoise,"IEEETrans.Acoust.,Speech,andSignalProcess.,vol.34,no.5,pp.1340{1342,Oct.1986.[41] U.Tureli,H.Liu,andM.Zoltowski,\OFDMblindcarrierosetestimation:ES-PRIT,"IEEETrans.Commun.,vol.48,no.9,pp.1459{1461,Sep.2000.[42] T.Yucek,\Channel,Spectrum,andWaveformAwarenessinOFDM-BasedCognitiveRadioSystems,"Ph.D.dissertation,UniversityofSouthFlorida,2007.[43] M.WaxandT.Kailath,\Detectionofsignalsbyinformationtheoreticcriteria,"IEEETrans.Acoust.,Speech,andSignalProcess.,vol.33,no.2,pp.387{392,1985.[44] \3rdgenerationpartnershipproject;technicalspecicationgroupradioaccessnet-work;evolveduniversalterrestrialradioaccess(E-UTRA);physicalchannelsandmodulation(release8),"3GPP,3GPPTS36.211,April2008.[45] \IEEEstandardforlocalandmetropolitanareanetworks.part16:Airinterfaceforxedbroadbandwirelessaccesssystems,"IEEEStd802.16-2004,Oct.2004.[46] \IEEEstandardforlocalandmetropolitanareanetworks.part16:Airinterfaceforxedandmobilebroadbandwirelessaccesssystems;amendment2:Physicalandmediumaccesscontrollayersforcombinedxedandmobileoperationinlicensedbands,"IEEEStd802.16e-2005,Dec.2005.[47] J.G.Proakis,DigitalCommunications,4thed.NewYork:McGraw-Hill,2001.[48] M.Rumney,\IMT-Advanced:4GWirelessTakesShapeinanOlympicYear,"WhitePaper,Sep.2008.[Online].Available:http://cp.literature.agilent.com/litweb/pdf/5989-9793EN.pdf[49] J.Choi,D.Shin,andD.Shin,\Researchandimplementationofthecontext-awaremiddlewareforcontrollinghomeappliances,"IEEETrans.Consum.Electron.,vol.51,no.1,pp.301{306,2005.[50] FemtoForum,\Femtocellapplications."[Online].Available:http://femtoforum.org/femto/applications.php[51] \3rdGenerationPartnershipProject;TechnicalSpecicationGroupRadioAccessNetworks;3GHomeNodeBStudyItemTechnicalReport(Release8),"3GPP,3GPPTR25.820,March2008.144

PAGE 163

FemtoForum,\InterferencemanagementinUMTSfemtocells,"WhitePaper,Dec.2008.[Online].Available:http://www.femtoforum.org/femto/Files/File/InterferenceManagementinUMTSFemtocells.pdf[53] Orange,Telecom-Italia,T-Mobile,andVodafone,\RequirementsforLTEHomeeN-odeBs,"3GPPDocumentR4-070209,Lemesos,Cyprus,Mar.2007.[54] Vodafone-Group,\HomeeNodeBconsiderationsforLTE,"3GPPDocumentR4-070456,SophiaAntipolis,France,Apr.2007.[55] F.S.ChuandK.C.Chen,\RadioresourceallocationinOFDMAcognitiveradiosystems,"inProc.IEEEInt.Symp.Personal,IndoorandMobileRadioCommun.(PIMRC),Athens,Greece,Sep.2007,pp.1{5.[56] T.H.KimandT.J.Lee,\Spectrumallocationalgorithmsforuplinksub-carriersinOFDMA-basedcognitiveradionetworks,"inProc.IEEEInt.Conf.onInnovationsinInformationTechnol.,Nashville,TN,Nov.2007,pp.51{54.[57] D.L.Perez,A.Valcarce,G.D.L.Roche,E.Liu,andJ.Zhang;,\AccessmethodstoWiMAXfemtocells:Adownlinksystem-levelcasestudy,"inProc.IEEEInt.Conf.Commun.Syst.(ICCS),Guangzhou,China,Nov.2008,pp.1657{1662.[58] D.L.Perez,G.D.L.Roche,A.Valcarce,A.Juttner,andJ.Zhang;,\InterferenceavoidanceanddynamicfrequencyplanningforWiMAXfemtocellsnetworks,"inProc.IEEEInt.Conf.Commun.Syst.(ICCS),Guangzhou,China,Nov.2008,pp.1579{1584.[59] S.ChenandT.Yao,\IntercarrierinterferencesuppressionandchannelestimationforOFDMsystemsintime-varyingfrequency-selectivefadingchannels,"IEEETrans.Consum.Electron.,vol.50,no.2,pp.429{435,2004.[60] H.YoshinoandA.Czylwik,\Adaptiveco-channelinterference(CCI)cancellationforOFDMcommunicationsystems,"inProc.Internat.ZurichSeminarBroadbandCommun.,2000,pp.245{250.[61] S.Mangold,Z.Zhong,G.Hiertz,andB.Walke,\IEEE802.11e/802.11kwirelessLAN:Spectrumawarenessfordistributedresourcesharing,"WirelessCommun.andMobileComputing,vol.4,no.8,Dec.2004.[62] N.Shankar,C.Cordeiro,andK.Challapali,\Spectrumagileradios:utilizationandsensingarchitectures,"inIEEESymp.NewFrontiersDyn.SpectrumAccessNetw.(DySPAN),Nov.2005,pp.160{169.[63] S.Haykin,\Cognitiveradio:brain-empoweredwirelesscommunications,"IEEEJ.Sel.AreasCommun.,vol.23,no.2,pp.201{220,2005.[64] I.Akyildiz,W.Lee,M.Vuran,andS.Mohanty,\NeXtgeneration/dynamicspectrumaccess/cognitiveradiowirelessnetworks:asurvey,"ComputerNetworks,vol.50,no.13,pp.2127{2159,2006.145

PAGE 164

T.YucekandH.Arslan,\Spectrumcharacterizationforopportunisticcognitiveradiosystems,"inIEEEMilitaryCommun.Conf.(MILCOM),Washington,DC,Oct.2006,pp.1{6.[66] M.Pun,M.Morelli,andC.Kuo,\Maximum-likelihoodsynchronizationandchannelestimationforOFDMAuplinktransmissions,"IEEETrans.Commun.,vol.54,no.4,pp.726{736,2006.[67] Z.Cao,U.Tureli,andY.Yao,\Deterministicmultiusercarrier-frequencyosetesti-mationforinterleavedOFDMAuplink,"IEEETrans.Commun.,vol.52,no.9,pp.1585{1594,2004.[68] J.vandeBeek,M.Sandell,andP.Borjesson,\MLestimationoftimeandfrequencyosetinOFDMsystems,"IEEETrans.SignalProcess.,vol.45,no.7,pp.1800{1805,1997.[69] R.Fantacci,D.Marabissi,andS.Papini,\MultiuserinterferencecancellationreceiversforOFDMAuplinkcommunicationswithcarrierfrequencyoset."inProc.IEEEGlobalTelecommun.Conf.(GLOBECOM),vol.5,Dallas,TX,Nov.2004,pp.2808{2812.[70] K.Kim,Y.Han,andS.Kim,\JointsubcarrierandpowerallocationinuplinkOFDMAsystems,"IEEECommun.Lett.,vol.9,no.6,pp.526{528,2005.[71] Z.Cao,U.Tureli,andY.Yao,\UserSeparationandFrequency-TimeSynchronizationfortheUplinkofInterleavedOFDMA,"inProc.IEEEAsilomarConf.onSignals,SystemsandComputers,vol.2,PacicGrove,CA,Nov.2002,pp.1842{1846.[72] P.Moose,\Atechniquefororthogonalfrequencydivisionmultiplexingfrequencyosetcorrection,"IEEETrans.Commun.,vol.42,no.10,pp.2908{2914,1994.[73] J.G.Andrews,A.Ghosh,andR.Muhamed,FundamentalsofWiMAX:Understand-ingBroadbandWirelessNetworking.Prentice-Hall,Feb.2007.[74] S.Chiu,\Fuzzymodelidenticationbasedonclusterestimation,"J.IntelligentandFuzzySystems,vol.2,no.3,pp.267{278,1994.[75] R.YagerandD.Filev,\Generationoffuzzyrulesbymountainclustering,"J.Intel-ligentandFuzzySystems,vol.2,no.3,pp.209{219,1994.[76] G.Seber,MultivariateObservations.JohnWiley&Sons,1984.[77] H.Spath,ClusterDissectionandAnalysis:Theory,FORTRANPrograms,Examples.HorwoodNewYork:HalstedPress[distributor],Chichester,1985.[78] G.Stuber,J.Barry,S.Mclaughlin,Y.Li,M.Ingram,andT.Pratt,\BroadbandMIMO-OFDMWirelessCommunications,"ProceedingsoftheIEEE,vol.92,no.2,pp.271{294,2004.146

PAGE 165

H.Sampath,S.Talwar,J.Tellado,V.Erceg,andA.Paulraj,\Afourth-generationMIMO-OFDMbroadbandwirelesssystem:design,performance,andeldtrialre-sults,"CommunicationsMagazine,IEEE,vol.40,no.9,pp.143{149,2002.[80] D.Gesbert,H.Bolcskei,D.Gore,andA.Paulraj,\MIMOwirelesschannels:capacityandperformanceprediction,"inIEEEGlobalTelecommunConf.(GLOBECOM'00),vol.2,2000.[81] F.Ivanek,\ConvergenceandCompetitionontheWayToward4G:WhereareWeGoing?"inRadioandWirelessSymposium,2007IEEE,LongBeach,CA,Jan.2007,pp.265{268.[84] K.Lee,\TechnologyLeadersForum-CreatetheFuturewithMobileWiMAX,"Com-municationsMagazine,IEEE,vol.45,no.5,pp.10{14,2007.[85] H.Arslan,\ChannelFrequencyResponseEstimationUndertheEectofRFImpaire-mentsinOFDMBasedWirelessSystems,"inProc.64thVehicularTechnol.Conf.(VTC-Fall2006),Montreal,Canada,Sep.2006,toappear.[86] H.ArslanandD.Singh,\EstablishtestproceduresforWiMAXTransceivers,"Mi-crowaves&RF,vol.45,no.7,2006.[87] M.El-Tanany,Y.Wu,andL.Hazy,\OFDMuplinkforinteractivebroadbandwireless:analysisandsimulationinthepresenceofcarrier,clockandtimingerrors,"Broadcast-ing,IEEETransactionson,vol.47,no.1,pp.3{19,2001.[88] S.Alamouti,\Asimpletransmitdiversitytechniqueforwirelesscommunications,"SelectedAreasinCommunications,IEEEJournalon,vol.16,no.8,pp.1451{1458,1998.[89] A.Hafeez,D.Hui,andH.Arslan,\InterferencecancellationforEDGEviatwo-userjointdemodulation,"inProc.58thVehicularTechnol.Conf.(VTC-Fall2003),vol.2,Oct.2003,pp.1025{1029.[90] R.Nilsson,F.Sjoberg,andJ.LeBlanc,\Arank-reducedLMMSEcancellerfornar-rowbandinterferencesuppressioninOFDM-basedsystems,"IEEETrans.Commun.,vol.51,no.12,pp.2126{2140,Dec.2003.[91] A.Coulson,\NarrowbandinterferenceinpilotsymbolassistedOFDMsystems,"IEEETrans.Commun.,vol.3,no.6,pp.2277{2287,Nov.2004.[92] Z.WuandC.Nassar,\NarrowbandinterferencerejectioninOFDMviacarrierin-terferometryspreadingcodes,"IEEETrans.Commun.,vol.4,no.4,pp.1491{1505,July2005.147

PAGE 166

D.Zhang,P.Fan,andZ.Cao,\AnovelnarrowbandinterferencecancellerforOFDMsystems,"inProc.IEEEWirelessCommun.andNetwork.Conf.(WCNC),vol.3,Mar.2004,pp.1426{1430.[94] X.G.DoukopoulosandR.Legouable,\IntercellinterferencecancellationforMC-CDMAsystems,"inProc.IEEEVehic.Technol.Conf.(VTC),Dublin,Ireland,Apr.2007,pp.1612{1616.[95] M.MohaisenandK.H.Chang,\Maximum-likelihoodco-channelinterferencecan-cellationwithpowercontrolforcellularOFDMnetworks,"inProc.Int.Symp.onCommun.Inform.Technol.(ISCIT),Sydney,Australia,Oct.2007,pp.198{202.[96] H.YoshinoandA.Czylwik,\Adaptiveco-channelinterference(CCI)cancellationforOFDMcommunicationsystems,"inProc.Int.ZurichSeminaronBroadbandCom-mun.,Zurich,Switzerland,Feb.2000,pp.245{250.[97] P.A.Hoeher,S.B.Hoeher,W.Xu,andC.Krakowski,\Single-antennaco-channelinterferencecancellationforTDMAcellularradiosystems,"IEEEWirelessCommun.Mag.,vol.12,no.2,pp.30{37,Apr.2005.[98] M.Shibahara,T.Fujii,I.Sasase,andT.Saba,\Performanceevaluationofadaptiveco-channelinterferencecancelingreceiverusingfrequencyspreadcodingandfrequencyinterleavingforOFDMsystems,"EuropeanTrans.Telecommun.,vol.14,pp.15{24,Jan.2003.[99] H.SchoeneichandP.Hoeher,\Iterativesemi-blindsingle-antennacochannelinter-ferencecancellationandtightlowerboundforjointmaximum-likelihoodsequenceestimation,"ElsevierSignalProc.,vol.84,no.11,pp.1991{2004,2004.[100] P.KaczorekandD.Rutkowski,\AcomparisonofnarrowbandandbroadbandoverlayofcellularCDMAonGSM,"inProc.IEEEVehic.Technol.Conf.(VTC),vol.3,May2000,pp.1859{1863.[101] R.SahotaandP.Whiting,\OnthefeasibilityofspectrumsharingbetweenGSMandIS-95,"inProc.IEEEPersonalWirelessCommun.,Mumbai,India,Dec.1997,pp.439{443.[102] I.Guvenc,M.-R.Jeong,F.Watanabe,andH.Inamura,\Ahybridfrequencyas-signmentforfemtocellsandcoverageareaanalysisforco-channeloperation,"IEEECommun.Lett.,vol.12,no.12,pp.880{882,Dec.2008.[103] ECMAInternational,\HighrateultrawidebandPHYandMAC,"ECMA-368Standard,Dec.2008.[Online].Available:http://www.ecma-international.org/publications/les/ECMA-ST/ECMA-368.pdf[104] ||,\Highrate60GHzPHY,MAC,andHDMIPAL,"ECMA-387Standard,Dec.2008.[Online].Available:http://www.ecma-international.org/publications/les/ECMA-ST/ECMA-387.pdf148

PAGE 167

A.Nasri,R.Schober,andL.Lampe,\AnalysisofnarrowbandcommunicationsystemsimpairedbyMB-OFDMUWBinterference,"IEEETrans.WirelessCommun.,vol.6,no.11,pp.4090{4100,Nov.2007.[106] S.M.Mishra,R.W.Brodersen,S.T.Brink,andR.Mahadevappa,\Detectandavoid:anultra-wideband/WiMAXcoexistencemechanism,"IEEECommun.Mag.,vol.45,no.6,pp.68{75,June2007.[107] P.Ranta,A.Hottinen,andZ.-C.Honkasalo,\Co-channelinterferencecancellingre-ceiverforTDMAmobilesystems,"inProc.IEEEInt.Conf.Commun.(ICC),vol.1,Seattle,WA,Jun.1995,pp.17{21.[108] H.SchoeneichandP.Hoeher,\Singleantennainterferencecancellation:iterativesemiblindalgorithmandperformanceboundforjointmaximum-likelihoodinterfer-encecancellation,"inProc.IEEEGlobalTelecommun.Conf.(GLOBECOM),vol.3,Dec.2003,pp.1716{1720.[109] Y.KopsinisandS.Theodoridis,\Anecientlow-complexitytechniqueforMLSEequalizersforlinearandnonlinearchannels,"IEEETrans.SignalProcess.,vol.51,no.12,pp.3236{3248,Dec.2003.149

PAGE 169

where,EnS2d;i(k)o=K21;i(k);EnI21;i(k)o=K22;i(k);EnI22;i(k)o=2 Ontheotherhand,againbasedon(2.18)-(2.26)(notethat(2.19),(2.20)arezero-varianceRVs),thevarianceofY(m)i(k)2canbeevaluatedasVarnY(m)i(k)2o=8>>>>>>>>>><>>>>>>>>>>:E2sc;iVarI22;i(k);ifk=2iE2sc;iVarI22;i(k)+4VarnReSd;i(k)I1;i(k)o+4VarnReSd;i(k)I2;i(k)o+4VarnReI1;i(k)I2;i(k)o;ifk2i;(A.3)151

PAGE 170

Similaranalysiscanbeappliedtohigher-ordermodulationschemesthathavesymmetricconstellationpointswithrespecttotheorigin,namelytheQPSK,16-QAM,and64-QAM,wheretheconstellationsarerespectivelygivenbyM(QPSK)=n[(21)+(21)j]p Then,using(2.18)-(2.26),itcaneasilybederivedthatthemeanandthevarianceofY(m)i(k)2areidenticalwith(A.1)and(A.3)foralltheabovethreeconstellationsofhigher-152

PAGE 172

{z }Cl1;l2:(B.67) Inamultipathchannel,despitesomeanalogieswithwiththemultiuserchannel,thestatisticsof(2.17)willbedierentthaninanAWGNchannel.Wemayre-write(2.17)inamultipathchannelasP(m)(k)=Y(m)(k)2=Y(m)i(k)+W(k)2=p whereasingle-userscenarioisconsideredforanalyticaltractability.Aftersomemanipu-lation,(B.1)canbewrittenasin(B.67),whereotherthanthelastsetoftermsinvolvingCl1;l2,theearliertermsareanalogoustotherstthreetermsin(2.17)forthemultiuserAWGNchannelscenario(i.e.,thedierentMPCsinthemultipathchannelmaybeconsid-eredasmultiusersignalswithdierentdelaysandattenuations),andtheirstatisticshavealreadybeencapturedthroughequations(2.18)-(2.30).However,asopposedtothemul-tiuserAWGNchannelscenarioanalogy,theCl1;l2termwillbenon-zerointhemultipathchannel,sincetheMPCscorrespondingtothesameuserwillbeusingthesameSASaswell154

PAGE 173

In(B.3),onlytheSd;i;l1(k)Sd;i;l2(k),I1;i;l1(k)I1;i;l2(k),andI2;i;l1(k)I2;i;l2(k)termshaveanon-zeromean.Forexample,thersttermisequaltoSd;i;l1(k)Sd;i;l2(k)=X(m)i(k)2K1;i;l1(k)K1;i;l2(k)ej2k(Dl1;i+Dl2;i) IfDl1;iandDl2;iareconsideredasknown,(B.4)becomesadeterministicvariable1.Similarly,itmaybeshownthattheothertwotermshavenon-zeromeans,anditisalsostraightforwardtoderivethat(B.3)hasanon-zerovariance.Insummary,sincetheCl1;l2termsarenon-zeroinamultipathchannel,thereisnotaone-to-oneanalogybetweensingle-usermultipathandmulti-userAWGNchannels,andtheformerscenario(consideringexactlysamedelaysandfadingcoecientsasinamultiuserAWGNchannel)resultsinworsespectrumopportunitiesduetolargerICI.


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200433Ka 4500
controlfield tag 001 002069492
005 20100423082050.0
007 cr mnu|||uuuuu
008 100422s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003283
035
(OCoLC)608514788
040
FHM
c FHM
049
FHMM
090
TK145 (Online)
1 100
Sahin, Mustafa
q (Mustafa Emin)
0 245
Baseband receiver algorithms for 4G co-channel femtocells
h [electronic resource] /
by Mustafa ahin.
260
[Tampa, Fla] :
b University of South Florida,
2009.
500
Title from PDF of title page.
Document formatted into pages; contains 155 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2009.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: The growing interest for high data rate wireless communications over the last few decades gave rise to the emergence of a number of wideband wireless systems. The resulting scarcity of frequency spectrum has been forcing wireless system designers to develop methods that will push the spectral eciency to its limit. One such method is to have multiple systems utilize the same spectrum by allowing some unavoidable interference to occur between them. The idea of co-channel systems is tested in the industrial, scientific and medical (ISM) bands and it is found to be a very beneficial approach. Therefore, it can be foreseen that co-channel systems might be a potential solution to the growing spectral crowding problem. Besides the systems that are designed to be co-channel, it is sometimes also possible to encounter that multiple systems occupy the same band undesirably.This kind of unintentional co-channel system scenarios might occur especially due to the dense re-use of available frequency bands. Another reason for unwanted co-channel usage might be the coexistence of third generation (3G) and fourth generation (4G) systems. Since 4G systems will probably be targeting to use the same frequency bands as their 3G counterparts, and since the transition from 3G to 4G will take some time, unintentional co-channel scenarios might be observed between the 3G and 4G systems. This dissertation consists of baseband receiver algorithms for OFDMA-based systems that target at handling the potential co-channel interference (CCI) in various co-channel system scenarios. Three CCI avoidance and two CCI cancellation algorithms are proposed that can be applied to intentional and unintentional co-channel systems.Femtocells, which have recently been introduced as a new class of personal-use base stations that can coexist with macrocell networks in a shared spectrum manner, might constitute an appropriate example for both types of co-channel systems. Therefore, they are considered to be one of the co-existing systems in most of the algorithms presented.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor: Hseyin Arslan, Ph.D.
653
Cognitive radio
Energy detection
Inter-carrier interference
Iterative cancellation
MIMO
OFDMA
Opportunity detection
User separation.
690
Dissertations, Academic
z USF
x Electrical Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.3283