1000 900 800

Clipped signal at threshold A=500

Time
Amplitude

700 600 500

SCFDMSybmol
400 300 200 100 0

CP

100

200

300

400 500 600 Input signal x(n)

700

800

900

1000

Subcarrier Bandwidth

Frequency

UplinkPowerControlforIMTAdvanced
MastersThesisintheMasterDegreeProgramme,CommunicationEngineering

MujeebUrRehman FahadMuslim

DepartmentofSignalsandSystems CHALMERSUNIVERSITYOFTECHNOLOGY Gteborg,Sweden,2010

Authorscontributedequallytothiswork.

Ex018/2010

UplinkPowerControlforIMTAdvanced
ThesisforthedegreeofMasterofScience

Coverpage:Thefirstfigurerepresentathresholdclippingdeviceusedtomodelthenon linearamplifierwhilethesecondoneshowstheSCFDMAschemeusedintheuplinkto lowerthePAPRratiothatincreasestheamplifierefficiency. All rights reserved. This publication is protected by law in accordance with "Lagen om Upphovsrtt,1960:729".Nopartofthispublicationmaybereproduced,storedinaretrieval system,ortransmitted,inanyformorbyanymeans,electronic,mechanical,photocopying, recording,orotherwise,withoutthepriorpermissionoftheauthors.Pleasecontactforany commentsorfeedbackatitzmujeeb@gmail.com. CopyrightMujeebUrRehman,FahadMuslim,Gteborg,Sweden2010.

II

ABSTRACT
The challenges in the wireless world are basically creating motivations to bring advanced techniques to improve not only the data rates but also the quality of service for users in mobilecommunication.Developmentisanongoingprocessthathasenhancedthedifferent chainofstandardslikeGlobalSystemsforMobilecommunications(GSM),UniversalMobile Telecommunications System (UMTS), Code Division Multiple Access (CDMA) and Long TermEvolution(LTE)andnowInternationalMobileTelecommunication(IMT)Advancedis of our interest after LTE that has made a promise to give more services, capabilities, and better performance to the end users than any other radio interface technology has been abletogivetilldate. ThisthesisstudyinvolvestheimplementationofanIMTAdvanceduplinkmodelfor thebiterrorrate(BER)performanceanalysisofTurbocoding,differentsubcarriermapping schemes e.g. Localized Frequency Division Multiple Access (LFDMA), Interleaved FDMA (IFDMA), Block Interleaved FDMA (BIFDMA) and Discrete Fourier Transform (DFT) pre coding. These techniques have been implemented in linear and nonlinear environments. Since pulse shaping has not been used so IFDMA performance analysis in nonlinear environment was not possible with the nonlinearity model used in this thesis. Hence the BER performance analysis in nonlinear environments has been limited to BIFDMA and LFDMA schemes only. Nonlinearities have been introduced by using a threshold clipping model using three different clipping percentages i.e. 1%, 10% and 30%. Moreover simulations have been performed for single user and multiuser environments, where we alsoanalyzedthenearfareffect. IthasbeenfoundasexpectedthatTurbocodeperformsbetterintheAdditiveWhite Gaussian Noise (AWGN) channel than in the Rayleighfading channel. DFT precoding is mainly intended to keep the Peak to Average Power Ratio (PAPR) value small, which is necessary for efficient performance of High Power Amplifiers (HPA), but it also gives slightly improved BER performance as compared to the non DFT precoded system. The singleuserBERperformanceforallthethreemappingschemesisidenticaltothelongterm average BER performance of multiple users in linear environments. The performance in nonlinear environments is however distorted in single user case mainly by Intercarrier Interference(ICI)andinmultiusercasebymultiaccessinterference(MAI).

III

ACKNOWLEDGEMENTS
FirstofallwewouldliketothankAlmightyALLAH,themostmercifulandbeneficent,whoalwaysshows usthepathofsuccessandhelpsusineverymomentoflifeandtowhomwehavetoreturnoneday. Special thanks to our Examiner and Supervisor Ass. Prof Tommy Svensson, who is not only a good researcher but also a devoted supervisor who helped us at every stage and taught us the way the research, is done, you are really a nice person to work with. We are thankful to all those persons that helped us during our Masters Program. We are also grateful to Prof Erik Strm and Assoc. Professor Erik Agrell who not only supported us in different subjects but always provided a good guidance. We are also thankful to Johnny Karout, who always helped us and provided us guidance whenever we knockedhisdoor. Mujeeb:IamespeciallythankfultoyouMother,youradvices,wisdom,guidanceandlovehavenotonly helped me at every step of life but also have given me a confidence to deal and solve every problem coming in my way and specialthanksto my father who is not in this world but this was all his company and advices that I learned from. I am also thankful to my great sisters, brother in laws and my elder brother who is always there to guide me and help me. I am also thankful to all my friends in native country & in Sweden, room mates who were there to spread cheers around and help me, especially, Talha,TabishandUmair.FinallyIamthankfultoFahadfornotonlybeingagoodthesismatebutalsoa cooperative and helping friend. I really enjoyed working with him. We together faced many difficult situationsandweresourceofmotivationsforeachother.Iwishhimgoodluckforhisfuturelife. Fahad: First of all, I would like to thank my parents for all they have done for me. It was only due to theirprayersandtheirencouragementthatmadeitpossibleformetoachievewhateverIhaveachieved up to now. I would also take this opportunity to thank my brother and my sisters whose support and affectionhas been very crucial for me. Very special thanks to myauntwho is oneofmy best friends and hasdonealotformeandallmysiblings.IalwaysgotwordsofencouragementwheneverItalkedtoher. I would also like to thank all my friends here in Sweden. Without them my stay in Sweden would have beenaveryboringexperience.TheyaddedflavortomylifehereinSweden.Myspecialgratitudeformy flatmatesAfaqandSalman.Mytimespentwiththemhasreallybeenamemorableone.Iwouldalsolike to thank Mujeeb. Working with him has been a nice experience. We both come from same country and that helped us totune ourselvesaccordingto each others style of work.Throughoutour thesis work we have been through some very tough times where we apparently had no clue as to what is the right way out. It was his motivation towards the work that kept me motivated as well. I would like to wish him goodluckinallhisfutureendeavors.

IV

Contents
ListofFigures ListofTables ListofAcronyms&Notations

1.INTRODUCTION&BACKGROUND
1.1 TheEvolutionof3GMobilesystems..12 1.2 EmergenceofLTE..12 1.3 RequirementsandTargetsfortheLTE.13 1.4 LTEtoIMTAdvanced..14 1.5 ThesisObjectives....15 1.5.1Scopeofthethesis15 1.5.2OrganizationofThesis..15 1.5.3TheAssessmentmethodology.16

2.SYSTEMMODEL
2.1ShannonsDigitalCommunicationModel18 2.1.1Sourceencoder.18 2.1.2ChannelencoderandModulator19 2.1.3Channel..19 2.1.4Demodulatorandchanneldecoder..19 2.1.5Sourcedecoder..19 2.2IMTAdvanceduplinkPowerControlSystemModel.20 2.2.1ChannelCoding.22 2.2.1.1Turboencoder........22 2.2.1.2Turbodecoder.23
V

2.2.1.3PerformanceofTurbocodes.24 2.2.2Modulation&Pulseshaping25 2.2.2.1BasebandModulation25 2.2.2.2Pulseshaping..26 2.2.3DFTPrecoding&Deprecoding.29 2.2.4SubcarrierMapping..29 2.2.4.1TimeDomainRepresentation....................32 2.2.4.1.1IFDMA..33 2.2.4.1.2LFDMA.33 2.2.5OFDM,OFDMA&SCFDMA34 2.2.5.1OFDMandOFDMA..34 2.2.5.2AdvantagesofUsingSCFDMA...39 2.2.6Nonlinearamplifiers...40 2.2.6.1Amplifiercharacteristics.40 2.2.6.1.1Cutoffregion..41 2.2.6.1.2Saturationregion.41 2.2.6.1.3Activeregion...41 2.2.6.1.4Breakdownregion..42 2.2.6.2NonlinearityModeling42 2.2.6.3SystemdegradationbyHPAnonlinearity..43 2.2.6.4PowercontrolinUplink..43 2.2.6.4.1PowerConsumption..43 2.2.6.4.2InterandIntraCellinterference44 2.2.6.4.3PowercontrolEquations45 2.2.6.4.4TypesofPowercontrol46 2.2.7Channel&Equalization47 2.2.7.1ChannelCharacteristics...47 2.2.7.1.1PathLoss...47 2.2.7.1.2ShadowFading...48 2.2.7.1.3NarrowbandFading...48

VI

2.2.7.2Widebandfading..49 2.2.7.2.1Powerdelayprofile.50 2.2.7.2.2Coherencebandwidth51 2.2.7.2.3DopplerPowerspectrum&CoherenceBandwidth52 2.2.7.3ChannelEqualization52

3.SIMULATIONS&RESULTS
3.1SystemSpecifications55 3.2Delimitations......58 3.3AWGN&RayleighFadingChannel58 3.4PerformanceevaluationofTurboCoding.60 3.4.1SingleUser60 3.4.2MultiUsers(2,4,8)61 3.5PerformanceevaluationofDFTprecoding..63 3.5.1SingleUser63 3.5.2MultiUsers(2,4,8)....64 3.6ModelingAmplifiernonlinearities..66 3.6.1InterCarrierInterference(ICI)...66 3.6.2PerformanceofDFTandNonDFTinnonlinearenvironment...68 3.6.3MultiAccessInterference(MAI)69 3.7MultiAccessInterferencewithNearFareffect71

4.CONCLUSIONS&FUTUREWORK
4.1Conclusions73 4.2FutureWork..74

AppendixA AppendixB References

VII

ListofFigures
Figure2.1 Figure2.2 Figure2.3 Figure2.4 Figure2.5 Figure2.6 Figure2.7 Figure2.8 Figure2.9 Figure2.10 Figure2.11 Figure2.12 Figure2.13 Figure2.14 Figure2.15 Figure2.16 Figure2.17 ShannonCommunicationModel TheIMTAdvancedmodel RSCencoderwithgeneratormatrixG=[7,5]octal,constraintlength3 TheIterativeTurbodecoderwithtwoSISOdecodingmodules TurbocodeBERPerformancewithIterations[1,2and5] QPSKconstellationdiagram Frequencyandimpulseresponseofsincpulseshapingfilter FrequencyandImpulseresponseofRCfilteratdifferentvaluesof LocalizedmappingvsDistributedmappingofSCFDMA BIFDMAwith4subcarriersand3OFDMsymbolsperblockincomparison withIFDMAandLFDMA ExampleofmappingfoursubcarriersinIFDMA,LFDMAandBIFDMA TimeDomainRepresentationofIFDMAmapping TimedomainrepresentationofLFDMAmapping ThestructureofFDMandorthogonalsubcarriersinOFDM OFDMTransceiver SubcarriersinUplinkinOFDM SuperpositionofallsubcarriersinOFDMA

Figure2.18 OFDMModulationusingmanysubcarriers Figure2.19 SCFDMAusingsinglecarrierintheuplink Figure2.20 BJTIVcharacteristicscurve

Figure2.21 Thresholdclippingfornonlinearitiesmodeling Figure2.22 Figure2.23 Figure2.24 Figure2.25 Figure2.26 BJTIVcharacteristicscurve Thresholdclippingfornonlinearitiesmodeling Transmitterimperfectionmodel:PSDwithISSL=25dBs Powercontrolschemesbasedonvaluesof CombinedPathloss,shadowingandmultipathpropagationeffect

VIII

Figure2.27 Figure2.28 Figure3.1 Figure3.2 Figure3.3 Figure3.4 Figure3.5 Figure3.6 Figure3.7 Figure3.8 Figure3.9 Figure3.10 Figure3.11 Figure3.12 Figure3.13 Figure3.14 Figure3.15 Figure3.16 FigureA.1 FigureA.2 FigureB.1 FigureB.2 FigureB.3 Table3.1

TimeVaryingDiscreteTimeImpulseResponseModelforaMulti PathChannel NormalizedPowerDelayProfileofWINNERC2NLOS BIFDMA&LFDMAPerformanceonAWGNChannel WithrespecttoTurbo[Iterations=1] BIFDMA&LFDMAPerformanceonRayleighFadingChannel WithrespecttoTurbo[Iterations=1] UncodedVsTurboCodedPerformanceonRayleighFadingChannel UncodedVsTurboCodedPerformancefor2Users UncodedVsTurboCodedPerformancefor4User UncodedVsTurboCodedPerformancefor8Users DFTprecodedVsNonDFTprecodedperformanceforSingleUser DFTprecodedVsNonDFTprecodedperformancefor2Users DFTprecodedVsNonDFTprecodedperformancefor4Users DFTprecodedVsNonDFTprecodedperformancefor8Users InterCarrierInterference(10%,30%andnoClipping) DFT&NonDFTprecodingSystemwith10%,30%Clipping MultiAccessInterferencewith2Users MultiAccessInterferencewith4Users MultiAccessInterferencewith8Users MultiAccessInterferencewithNearFarEffect8Users Channelgainastheoutputofalinearfilter Discretetimemodelofatimeandfrequencyvaryingchannel MultiAccessInterferencewith2Users(10%Clipping) MultiAccessInterferencewith4Users(10%Clipping) MultiAccessInterferencewith8Users(10%Clipping) SimulationParameters

FigureA.3TimeandFrequencyResponseofWINNERC2NLOSChannel

IX

ListofAcronyms
AWGN BER BIFDMA CCDF CDMA CSIR DFT EGPRS FDM/FDMA FPLMTS GPRS GSM HPA HSPA ICI IFDMA IMT IP ISI ITUR ISSL LFDMA LOS LTE Mbps MMSE MAI OFDM OFDMA PAPR PDF PS PCCC SCCC SCFDE SCFDMA TDMA TWT UMTS WCDMA WINNER AdditiveWhiteGaussianNoise BitErrorRate BlockInterleavedFrequencyDivisionMultipleAccess ComplementaryCumulativeDistributionFunction CodeDivisionMultipleAccess ChannelStateInformationattheReceiver DiscreteFourierTransform EnhancedGeneralPacketRadioService FrequencyDivisionMultiplexing/MultipleAccess FuturePublicLandMobileTelecommunicationsSystems GeneralPacketRadioService GlobalSystemsforMobileCommunications HighPowerAmplifier HighSpeedPacketAccess IntercarrierInterference InterleavedFrequencyDivisionMultipleAccess InternationalMobileTelecommunications InternetProtocol IntersymbolInterference InternationalTelecommunicationUnionRadiocommunicationSector InbandSubcarrierSetLeakage LocalizedFrequencyDivisionMultipleAccess LineOfSight LongTermEvolution MegaBitPerSecond MinimumMeanSquareErrorEqualizer MultiAccessInterference OrthogonalFrequencyDivisionMultiplexing OrthogonalFrequencyDivisionMultipleAccess PeakToAveragePowerRatio ProbabilityDensityFunction PulseShaping Parallelconcatenatedconvolutionalcode Serialconcatenatedconvolutionalcodes SingleCarrierwithFrequencyDomainEqualization SingleCarrierFrequencyDivisionMultipleAccess TimeDivisionMultipleAccess Travellingwavetube UniversalMobileTelecommunicationsSystem WidebandCodeDivisionMultipleAccess WirelessWorldInitiativeNewRadio


CHAPTER1 Introduction&Background

In this chapter the historical background regarding advancement in wireless communicationhasbeendescribed.Italsoincludesthedetailsabouttheobjectives, scopeandorganizationofthethesis.

Chapter1

Introduction&Background

1.1

Evolutionof3rdGenerationMobileSystems
It is said that necessity is the mother of invention and this was the reason that

manyrapidadvancementswerebeingmadeintheworldofwirelesscommunicationsover theyears.Higherdataratesandabetterqualityofservicearealwaysrequiredbytheusers andthisiswhateverythingcirculatesaround.Thestorybeginswiththeinventionofthe1st telephonebyAlexanderGrahamBellandsincethen,manyrapidadvancementstookplace till the 1st International mobile communications system Nordic Mobile Telephony (NMT) wasintroducedin1981atthesametimewhenAdvanceMobilePhoneService(AMPS)was introducedinNorthAmerica.DuetosomeserviceissuesintheFirstGenerationSystemthe SecondGenerationSystem(2G)emerged. ThesecondGenerationMobileSystemcamewithagoodprospectofincreasingthe capacityandconsistentqualityofserviceusingthedigitalcommunicationsystem.Theshift fromanalogtodigitaldomaingaveagreatboostinthecapacity,improvedcost,speedand power efficiency of the digital hardware [1]. The primary services introduced in 2G were textSMSandcircuitswitchedservicesenablingemailandotherdataapplications.Multiple timeslotswereassignedtotheusersandcodingschemesweremodified.Lateron2.5Gwas introduced with some more latest services like GPRS and packet data over the cellular technologies[2]. Wirelesscommunicationisafieldwhererapidchangesaretakingplaceduetothe needforhighspeed,highdataratesandgoodqualityofservicewhicharethemeasuresthat bringadvancementsinthewirelessworld.Thiswasthereasonthat3GorUMTSemerged onthescene.Theinitialnamefor3GwasFuturePublicLandMobileTelecommunications Systems(FPLMTS),whichwaslateronchangedtoInternationalMobileTelecommunication (IMT2000).ITUstartedthestandardizationprocessanddefinedthestandardfor3G.The 3G system was intended to provide high speeds but also to provide a global mobility and various kinds of services including broadband and internet. As the time passed different types of improvements took place and the role of ThirdGeneration Partnership Project (3GPP)isveryimportanttodiscusshere. 3GPPisacollaborationbetweengroupsoftelecommunicationsassociationsthatis developingthespecificationsfor3GUniversalTerrestrialRadioAccess(UTRA)andGlobal system for mobile communication (GSM). 3GPP is a partnership project formed by the different standard bodies [2]. 3GPP has published many releases among which in the

12

Chapter1

Introduction&Background

release 8 and 9 it has specified the specifications and enhancements of LTE. The latest releaseistherelease10.

1.2 EmergenceofLTE
LTE is an abbreviation of Long Term Evolution. This is an upgraded standard of UMTS.ThestandardizationprocessofLTEwasinitiatedinaworkshopinNovember2004. Differentcompaniesofbroadbandandmobilecommunicationsparticipatedandpresented theirperceptionsandrequirementswhichneededtobesatisfied.Inthefollowing,wehave illustratedsomeofthemajorrequirementsandtargetssetfortheLTE.

1.3 RequirementsandTargetsfortheLTE
Therearealwayssomefactorsandneedsforwhichanewtechnologyarisesandfills thegapleftbytheoldertechnologies.ThesameisthecasewithLTEthatcomeswithsome objectives and targets. The requirements and targets were refined and finalized in June 2005.Thekeytargetswerelike, Reducingthedelaysintermsofconnectionestablishmentandtransmission Increasingthedataratesandincreasingthecelledgebitrate Reducingcostperbitinordertoimprovethespectralefficiency Creatingagreaterflexibilityforspectrumusage,inbothnewandpreviouslyexisting bands Creatingsimplifiednetworkarchitecture Seamlessmobility,includingmobilitybetweendifferentradioaccesstechnologies Workingonreasonablepowerconsumptionforthemobileterminals The different releases of 3GPP have shown a tremendous development to achieve the requirementsandtargetsforLTE.

13

Chapter1

Introduction&Background

1.4 LTEtoIMTAdvanced
The 3GPP partners compiled a formal documentation and submitted it to ITU and proposed that LTE release 10 and beyond must be evaluated as a candidate for IMT Advanced.TheIMTAdvancedcapabilitiesgobeyondthoseofIMT2000.Themajorfeatures and characteristics of IMTAdvanced includes worldwide functionality, roaming, its compatibility of services with previous technologies, internetworking with other radio access systems, enhancing the data rates, services along with applications, also improving thedeficiencieslyinginthepreviousversionsofLTE.Inadditiontotheabovefeatures,one of the main reasons for aligning LTE with the call for IMTAdvanced is to make it a candidateforfuturenewspectrumbandstobeidentifiedatWRC07. If we take a look at the overall research areas and the community involved then therearemanyprojectsrunningintheareaofIMTAdvancedandinthenextgenerationof radioaccess.ItwillbeworthwhiletodiscusstheWINNERprojectwhichwaspartlyfunded by the European Union [2]. The WINNER concept is very close to LTE, however it is concentratingmoreonthehigherdataratesandthatswhyithasbeendesignedforawider bandwidth than 20MHz. The WINNER is also working with the multihop modes and relaying concept. This is also true that the major stability and enhancements have been broughtandsettledbythestandardizationorganizations[2]. So as a summary of the above discussion, we can say that the development is an ongoingprocessthatalwaysenhancesthepreviousversionsoftechnologyandperformance. Theresearchisgoingonbutthedeploymentandfeedbacksfromcustomerswillmakethis process even faster. We have to admit that IMTAdvanced is still several years away whereas deployment of HSPA Evolution and LTE are in the process. High Speed Packet Access (HSPA) is a continuing evolution of the existing standards like WCDMA/HSPA technology whereas LTE is a new entrant and radio access optimized purely for IP based traffic.Thesetwotechnologieshavemadeapromisetogivemoreservices,capabilities,and performancetotheendusersthananyotherradiointerfacetechnologyhasbeenabletotill date.

14

Chapter1

Introduction&Background

1.5 ThesisObjectives
Thebasicobjectiveofthethesisistodosomestudy,researchandthenwritingout resultsinapresentableformatthroughareport.Inthispartwehavediscussedbrieflywhat wehavedoneindifferentpartsofthereport.

1.5.1 SCOPEOFTHETHESIS
Inthisthesisprojectfirstlyweobservetheperformanceofdifferentmultipleaccess schemesonLTEUplink.Thisrequiresthesimulationofatimevaryingfrequencyselective channelandthesimulatedsystemshouldovercometheshortcomingsofsuchchannellike fading,InterSymbol Interference(ISI)andInterCarrierInterference(ICI).Thenamplifier nonlinearities are introduced by clipping the modulated signal to different amplitude clippinglevelsandanalyzingitseffectontheoverallperformanceofthesystem.Firstofalla singleusersystemissimulatedfollowedbyintroducingmoreusersandstudyingtheeffects on the overall system behavior. Finally the near far problem inherent in mobile cellular communicationsisconsideredinthesystemunderstudy.

1.5.2 ORGANIZATIONOFTHETHESIS
In chapter 2 of this thesis report, readers are given an over view of the different building blocks present in the transceiver of the simulated system. Additionally all the blocks are explained briefly to give the readers an idea of what this thesis study was all about.Thechaptercoversdifferentblocksinorderastheyoccurinthesystem.Thechannel encoder is explained first, which is followed by modulation and the importance of pulse shaping.ThisisfollowedbyabriefintroductionofDFTprecodinganddifferentsubcarrier mappingschemes.OFDMmodulationandamplifiernonlinearitiesareintroducedfollowed by some theory about power control. Finally we describe the fading channel and equalizationtomitigatetheadverseeffectsofwirelesschannels. In chapter 3, readers are given details about the simulations and results obtained from those simulations. This includes presenting the various system parameters and describingwhatshouldthedifferentsystemparametersbetokeepthesystemsafefromthe wireless channel impairments i.e. ICI and ISI. Besides simulation parameters, the results deducedfromthesimulationsarealsobeingpresentedwithdiscussionsonthoseresults.

15

Chapter1

Introduction&Background

Inthefinalchapterofthisthesisreportweconcludeourworkandtrytosummarize what we have learnt from this thesis work and also we shed some light on what possible futureworkcanbedoneintheareascoveredinthismastersthesis.

1.5.3 ASSESSMENTMETHODOLOGY
The simulations have been performed in MATLAB. The channel model used is WINNERC2NLOSchannel.Ithasbeensimulatedusingthespectrummethod(seeappendix A). First of all, all the blocks have been constructed individually and tested with an ideal channel to see if they are working properly. This is followed by introducing an AWGN channelandseeingifourBERcurveresemblesthetheoreticalBERcurve.Finallythefading channel has been introduced and various results have been obtained thereafter by modifyingtherelevantsystemparametersforeachaspectbeingtested.Allthesimulations havebeencarriedoutusingtheC3SeLinuxPlatformatChalmers.

16

CHAPTER2 SYSTEMMODEL
ThisChapterincludestheblockbyblockdescriptionoftheShannonModelandthe IMTAdvancedUplinkSystemModel.Beforegoingtothesimulationschapter,here the theoretical background is sketched related to channel coding, modulation, subcarriermappingschemes,nonlinearamplifierandpowercontrol.

Chapter2

SystemModel

2.1ShannonDigitalCommunicationModel
Inabroaderperspectiveacommunicationsystemconsistsofthreemainpartsi.e. the transmitter, the receiver and the channel. In a digital communication system the information is transmitted in the form of binary digits or bits. The information source inputsbitsintothetransmitter,whichprocessestheinformationandthenitistransmitted over the channel. The channel distorts the signal through sources such as noise or interference and the role of the receiver is to retrieve the signal sent by the information sourcewithcertainprecision. Claude Shannon is rightly called the Father of modern communications as his communication model is the basis of modern communications. In this chapter we briefly describe the different blocks of Shannons communication model which is shown in the figurebelow: Receiver Figure:Shannondigitalcommunicationmodel Sink Source Decoder Channel Decoder Figure2.1:ShannonCommunicationModel Demodulator Channel Source Source Encoder Transmitter Channel Encoder Modulator

2.1.1 SOURCEENCODER
Informationtheorysaysthatanyinformationcanberepresentedindigitalformbya number of bits depending on required fidelity. The source encoder here does this by representing the information from source into number of bits depending on end users requirements[3].

18

Chapter2

SystemModel

2.1.2 CHANNELENCODERANDMODULATOR
Unlikethesourceencoderwhichremovesredundancy,thechannelencoderactually adds redundancy to the data to mitigate the channel distortions. The modulator then translatestheencodeddataintoananalogwaveformthatcanthenbetransmittedoverthe channel.

2.1.3 CHANNEL
The physical characteristics of the channel can vary widely and good channel models are necessary for designing efficient communication systems. The channel would corruptthesignalwithdistortionssuchasnoise.Thechannelmaybeawirelineorwireless oneandithastobeestimatedsomehowforefficienttransmissionacrossit.

2.1.4 DEMODULATORANDCHANNELDECODER
Thedemodulatorthenprocessesthereceivedanalogwaveformwhichisactuallya distorted version of the transmitted signal. The main tasks of the demodulator is to take careofsynchronizationbetweentheclocksandoscillatorsattransmitterandreceiver,and channel equalization to compensate for ISI introduced by a dispersive channel. Thus the demodulatorsultimategoalistoproducehardorsoftdecisionsonthetransmittedsymbols to be fed into the channel decoder which then makes use of the redundancy in the coded data to improve upon the estimates from demodulator and to produce estimates of the sequenceofinformationsymbolsthatwereinputtothechannelencoder[3].

2.1.5 SOURCEDECODER
The source decoder then converts the estimated information bits produced by the channeldecoderintoaformthatcanbeusedbytheenduseri.e.thesink[3].

19

Chapter2

SystemModel

2.2

IMTAdvancedUplinkPowerControlSystemModel
In this section of the report we have discussed the different blocks of the IMT

AdvancedUplinkSystemmodel,whichwehavedevelopedandused ittoanalysetheBER performanceofdifferentsubcarriermappingschemes(BIFDMA,IFDMAandLFDMA)and toseetheimpactofamplifiernonlinearities.

Information Bits

Decoded Bits
^

d (q)

d (q )

Channel Encoder (Turbo Encoder)

Channel Decoder (Turbo Decoder)

Modulation (QPSK)

Demodulation (QPSK)

Fk

DFT Precoding (K-Point DFT)

DFT De-Precoding (K-Point IDFT)

FHk

(q)

Sub-Carrier Mapping

Channel Equalization (MMSE) Sub-Carrier

C (q)

T ( q )

HM F

OFDM Modulation (M-Point IFFT)

De-Mapping OFDM Modulation (M-Point FFT)

FM

Cyclic Prefix Remove Cyclic Prefix

H
Threshold Clipper

(q)

+ WM

Channel (WINNER-C2 NLOS)

Figure2.2:TheIMTAdvancedmodel

20

Chapter2

SystemModel

We have used the IMTAdvanced model with different variations that are briefly describedbelow,

1.

UncodedSystem

InthismodelwehaveusedthesamesystemwithouttheTurbocodingblockstosee the performance of the system. Later on we have compared uncoded system results with codedsystemresults.

2. WithoutDFTPrecodedSystem In this model we have removed the DFT precoding and DFT deprecoding blocks in
ordertoseetheeffectofDFTprecodingonthesystemundertest.

3.

Thresholdclippingmodel

Thresholdclippingmodelhasbeenusedtoseetheimpactofamplifiernonlinearities
byusingdifferentclippinglevels.

21

Chapter2

SystemModel

2.2.1 CHANNELCODING
The basic objective of channel coding is to introduce redundant bits to save the information from channel impairments and to minimize the errors. There are different channelcodingtechniquesusedtomitigatetheadverseeffectsofthewirelesschannelsfor exampleBlockcodes,ConvolutionalcodesandTurbocodes.Inthisprojectwehaveusedthe Turbocodeduetoitsefficientperformancewhendealingwithlargeblocksofdata. Before going in to the details of Turbo codes here it is very necessary to mention aboutthehistoryofTurbocodes.TurbocodesweredevelopedbyC.BerrouandA.Glavieux attheEcoleNationaleSuperieuredesTelecommunicationsdeBretagneinBrest,France,as anexerciseinVLSIdesign[4]. ThefirstproposedTurbocodewasaparallelconcatenated convolutional code (PCCC) with two convolutional codes with an interleaver between the two codes. The concatenated codes are basically divided into two types, one is parallel concatenated convolutional code (PCCC) and the other one is serial concatenated convolutionalcodes(SCCC)[4].AverygoodperformancefromTurbocodecanbeachieved bypassinglargerblocksofdata.Inthefollowingwehavediscussedaboutthestructureand functionality of the Turbo encoder and decoder to understand the necessary background, applicationsandtheperformanceofTurbocodes. 2.2.1.1Turboencoder The Turbo encoder consists of two or more essential recursive systematic convolutional(RSC)encodersseparatedbyapseudorandominterleaver.Itisnotnecessary that the encoders will be identical but normally they are identical. As an example a block diagramofaTurboencoderisshownwithsystematicfeedbackconvolutionalencodersand aninterleaver.Theconstraintlengthis3withgeneratormatrixG=[7,5]octal.

bk

Yk
(Systematic bits)

D Yk

(Parity bits) Figure2.3:RSCencoderwithgeneratormatrixG=[7,5]octal,constraintlength3

22

Chapter2

SystemModel

In the figure the data bits bk are generated and fed into the first encoder that generates a set of systematic and parity bits. The data bits are passed in to the second encoder after being altered by the pseudorandom interleaver. The second encoder also generatesasimilarsetofbitswhiletheoverallcodeispuncturedbydeletingthesecondset of systematic bits. So the resulting bit stream consists of systematic bits from the first encoder followed by the parity bits from the first and second encoders, respectively. The overall code rate in this model is 1/3 while code rate can be varied to 1/2 by alternately puncturingtheparitybitsfromeachoftheencodersbeforethetransmissiontakesplace. By increasing the code rate, the bandwidth efficiency is increased while the performanceofthesystemisruinedintermsofpowerefficiency[4].Anotheressentialand fundamentalpartoftheTurbocodeistheinterleaverthatisnecessarytobedescribedhere withtheencoder.Therearedifferenttypesofinterleaversincludingblock,pseudoRandom andconvolutionalInterleavers.TheTurbocodeperformanceishighlyaffectedbythetype ofinterleaver.Withafewexceptions,randominterleaverprovidesgoodperformance[4]. 2.2.1.2Turbodecoder A very effectivefeature ofthe Turbocodeis the iterativedecodingprocess.Asan example in the following figure an iterative Turbo decoder has been shown that basically comprisesoftwosoftinputandsoftoutput(SISO)decodingmoduleswhichareseparated byaninterleaverandadeinterleaver. Figure2.4:TheIterativeTurbodecoderwithtwoSISOdecodingmodules
DeInterleaver Decoder1 Interleaver Decoder2 De Interleaver Interleaver Decision

23

Chapter2

SystemModel

ThedecodingprocessfollowsadecodingalgorithmusedbytheSISOmodules.There are two major types of trellis based algorithms that are used to decode the Turbo codes. One is the Viterbi algorithm and the other one is the Maximum A Posteriori (MAP) algorithm.TheViterbialgorithmacceptssoftinputsandproduceshardoutputswhilethe MAPacceptssoftinputsandproducessoftoutputs.TheSISOalgorithmsarenecessaryfor Turbo decoding as they share their extrinsic information with each other while they are computationally more complex. They also allow iterative sharing of results between decoders,whichpermitstheuseofpowerfulconcatenatedcodingstructures[4]. 2.2.1.3PerformanceofTurboCodes TheTurbocodesareveryefficientforprocessinglongblocksofdatabutresultsin longerdelays.TherearesomefactorsthataffecttheTurboencoderperformance.Themajor onesincludethesizeofinterleaverandthenumberofiterationsonthedecoderside.The figure 2.5 shows the performance of the Turbo code on the AWGN channel with different numberofIterations.AsthenumberofIterationsisincreasedtheperformancegetsbetter buttheprocessingtimeisalsoincreased[5]. 10
B E R (B it E rror Rate)
-1

10

Turbo codes Performance

10

-2

10

-3

10

-4

Turbo 5-iterations Turbo 1-iteration Turbo 2-iterations uncoded

0 5 6 7 4 Eb/No (dB) Figure2.5:TurbocodeBERPerformancewithIterations(1,2,and5) 1 2 3 8

24

Chapter2

SystemModel

2.2.2 MODULATIONANDPULSESHAPING
Inthissectionwewilldiscussaboutthebasebandmodulationschemes(QPSK)and abriefdescriptionofpulseshapingwillalsobegiven. 2.2.2.1Basebandmodulation Modulation is a technique in which one or more properties of a high frequency carriersignalisvariedwithrespecttothemodulatingsignal.Thethreemainpropertiesof carriersignalsareitsfrequency,phaseanditsamplitudeandhencethemodulationcanbe frequency,phaseandamplituderespectivelydependingonwhichpropertyofthecarrieris beingvaried.Quadratureamplitudemodulation(QAM)isatechniqueinwhichthedatais modulatedbothonthephaseandamplitudeofthecarriersignal.Modulatingdigitaldataon an analog signal is called Keying. So we have phase shift keying (PSK), frequency shift keying(FSK)andamplitudeshiftykeying(ASK)techniquestomodulatedigitaldataonthe phase,frequencyandamplitudeofananalogsignalrespectively. Here in this thesis the baseband modulation technique used is Quadrature Phase Shift Keying (QPSK) in which two bits are assigned to each symbol and which consists of fourpointsinitsconstellationplot.Thefourconstellationpointsareplacedineachofthe fourquadrantsofthesignalspace.Theyareequallyspacedfromtheoriginandcanbeseen as lying along the circumference of a circle with the origin representing the centre of the circle.ThusalltheQPSKsymbolshavethesameenergybuthavedifferentphases.Theaxis lines thus become decision boundaries while each of the quadrants becomes the decision regions.ThefigurebelowshowsatypicalQPSKconstellationplotconsistingoffourpoints alongthecircumferenceofacircle 01 11 Figure2.6:QPSKconstellationdiagram 10 00

25

Chapter2

SystemModel

It should be noted that the figure 2.6 depicts gray mapping in which each next symbol bit allocation differs from the neighbor symbols by a single bit. Since in QPSK we havetwobitsrepresentedpersymbolsoasinglesymbolinerrorwouldcorrespondtotwo bits in error at the maximum. Grey mapping ensures that the most likely symbol errors would only cause a single bit in error. The theoretical bit error probability for QPSK modulationwiththeAWGNchannelisgivenby

Pb = Q(

2 Eb ) N o (2.1)

Where Eb istheaveragebitenergy, N o isthenoisepowerspectraldensityand QrepresentstheQfunction. 2.2.2.2Pulseshaping With continuous advancements in communications and with the need of accommodating multiple users in a limited bandwidth, the bandwidth has become one scarceresource.Thuseffortsaremadetomakeanefficientuseofthisscarceresource.Pulse shapingfiltersareusedtonarrowbandwidthandtoimprove bandwidthusage.However, pulse shaping can introduce distortions and can increase the risk of inter symbol interference(ISI).Thesedistortionsmakethedesignofanoptimalreceiverdifficult[7].The choiceofpulseshapingfilteristhusatradeoffbetweenbandwidthefficiencyinfrequency domainandISIintroductionintimedomain.Hencepulseshapingfilterischosentolimitthe bandwidthofthesignalwhileminimizingtheISIinthesystem. Inordertobetterexplaintheconceptofpulseshapingwewouldconsiderthemost basicunitofdigitaltransmissionwhichisasimplerectangularpulse.Ideallytherectangular pulsehasaninfinitebandwidthbutsinceinpracticalcaseswehavebandlimitedchannels hencethetransmittedpulsestendtobespreadduringtransmission.Thispulsespreading wouldcausethepulsestooverlapinadjacenttimeslotsthuscausingISI[8]. One method of controlling ISI is to shape the transmitted pulses properly and one pulseshapethatproduceszeroISIisthesincpulsewhichisshowninthefigurebelowin timeandfrequencydomain.Thefrequencyresponseisjustarectangularwaveform,band limitedbetween5 f o and+5 f o ,where f o isthepulserate.

26

Chapter2
Frequency response H(f) 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -5 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -5

SystemModel

0 Frequency

Impulse response h(t)

0 Time

Figure2.7:Frequencyandimpulseresponseofsincpulseshapingfilter Thisfiltershapeishoweverpracticallyunrealizablesinceanideallowpassfilteris notcausalandalsobecausethiswaveformcriticallydependsontimingprecision[8].Sofor practicalpurposesanexcessbandwidthfactorisusedtospreadthesincpulsethrough some extra bandwidth which solves the problems with the sinc pulse. This new pulse is called the raised cosine pulse and the factor is called roll off factor. The frequency responseofthisraisedcosinefilterisgivenintheequationbelow
1 , | f |< 2 1+ 1 HRC ( f ) = | f || |; [1 + cos( (| f |) (1 ) / 2 ))], 2 2 2 2 (2.2) 0; otherwise

Wherefisthefrequencyandisthepulseinterval.Theimpulseresponseisgivenas

t cos( ) t hRC (t ) = sin c( ) (2.3) 2t 2 T (1 ( ) )

Wheretisthetime. 27

Chapter2

SystemModel

As the rolloff factor is increased from 0, at which the raised cosine pulse is nothing but a sinc pulse, the response characteristics of the raised cosine filter would change and we would move further and further away from the ideal rectangular shaped frequency response. The figures below show the impulse and frequency response of the raisedcosinefilterwithdifferentvaluesoftherollofffactor. Figure2.8:FrequencyandImpulseresponseofRCfilteratdifferentvaluesof[9] So we see that as the value of increases from 0 to 1 the oscillations in the impulseresponsealsoincreaseswhiletheimpactinthefrequencyresponsecanbeseenin theformofmoresmoothtransitionsinthemagnitudeoffrequencyresponseasthevalueof increases from 0 to 1. So the rectangular pulse after pulse shaping doesnt remain perfectlyrectangularbutisasmoothlyroundedpulseconsistingofripplesbeforeandafter theoriginalperiod,andtheamplitudeoftheripplesactuallyindicateswhatvalueofis being used. So in short designing a pulse shaping filter requires a tradeoff between the bandwidth efficiencyinfrequencydomainandrippleattenuationinthetimedomain. The tighterthebandwidthinfrequencydomainthehighertheamplitudeofsidelobeswouldbe inthetimedomain. 28

Chapter2

SystemModel

2.2.3DFTPRECODING&DEPRECODING
The DFT precoding & Deprecoding blocks have been introduced in the uplink transmissioninLTEandIMTAdvanced.TheprocessofDFTprecodingandDeprecodingis verysimple.InDFTprecodingwetaketheKpointFFTofthemodulatedsymbolswhileon the other side we need to take Kpoint IFFT of demapped symbols. This type of transmissionschemebasedonDFTprecodedOFDMAisalsocalledsinglecarrierfrequency divisionmultipleaccess(SCFDMA). Some of the major advantages of SCFDMA are to enable power efficient transmissionintheuplink,improvingthecoverage,reducingtheterminalcostandresulting in low envelop fluctuations of OFDMA signals [10]. DFT precoded OFDMA has much smaller peak to average power ratio (PAPR) than regular OFDM thus enabling very less complex and higher power terminals [11]. In this project we have seen the effect of DFT andNonDFTprecodingoverthedifferentcarriermappingschemesthatwillbediscussed inthecomingsectionsofthereport.

2.2.4 SUBCARRIERMAPPING
The DFT precoded symbols coming out of the DFT precoding block are then mapped to a set of subcarriers and the process is called subcarrier mapping. The sub carriermappingblockactuallyassignstheDFTcomplexvaluesasamplitudestosomeofthe selectedsubcarriers[12].Usuallya setofNsubcarriersismappedto asetofM sub carriers where M>N. Subcarrier mapping is of two main types: localized (LFDMA) and distributed(DFDMA)mapping.Inlocalizedmappingconsecutivesubcarriersoracomplete chunk(atimefrequencyunitinwhichthesubcarriersincludedessentiallyexperiencesflat fading) is assigned to users while in distributed mapping subcarriers separated by some frequencyoffsetareassignedtousers. Henceinlocalizedmappingthesubcarriersareconfinedtosomepartof thetotal bandwidth while in distributed mapping the subcarriers are assigned noncontinuously over the whole bandwidth thus resulting in zero values for the remaining subcarriers. A specialcaseofthedistributedmappingwherethesubcarriersareseparatedequidistantly overtheentirebandwidthiscalledinterleavedSCFDMA(IFDMA)[12].Itshouldbenoted thatDFTfollowedbythesubcarriermappingschemeseffectivelyimplementsinterpolation [13].Thefigurebelowshowsageneralizedviewofboththemappingschemes 29

Chapter2

SystemModel

Figure2.9:LocalizedmappingvsDistributedmappingofSCFDMA It is clear from the figure above that the localized mapping consists of N consecutivesubcarrierswhileincaseofdistributedmappingeachsetofNsubcarriersis separated by L1 zeros [13]. SCFDMA inherently gives diversity gain over OFDM as the information data is spread over multiple subcarriers by the DFT mapper. In comparison however the distributed mapping scheme collects more frequency diversity than the localized mapping scheme, but the pilot overhead for channel estimation is increased comparedtothelocalizedmappingscheme,LFDMAhoweverincombinationwithchannel dependant scheduling can offer multiuser diversity in frequency selective channel conditions[12]. InadditiontoLFDMAandIFDMAanothermappingschemecalledblockinterleaved FDMAorBIFDMAwasintroducedin[14]whichisactuallyageneralizationofIFDMAand LFDMA.UnlikeIFDMAinBIFDMAequidistantblocksconsistingofmorethanonesubcarrier areassignedtousersandunlikeLFDMAnonconsecutivesubcarriersareassignedtousers [15]. If the DFT precoding step is removed before BIFDMA then it converts to block equidistant IFDMA or BEFDMA. The figure below shows the subcarriers allocation in the threeschemes

30

Chapter2

SystemModel

Figure2.10:BIFDMAwith4subcarriersand3OFDMsymbolsperblockincomparison withIFDMAandLFDMA HereinthisresearchstudythesubcarrierallocationmatrixTisanMKmatrix where M is the total number of subcarriers e.g. 1024 and K is the number of sub carriersperusere.g.128.Thesubcarrierallocationmatrixinallthethreecasescanbegiven bythefollowingequations: IFDMA:

1, m = k .Q + q; (q) ( m, k ) = TIFDMA 0, otherwise;

(2.4)

LFDMA:

1, m = q.K + k ; (q) (m, k ) = TLFDMA 0, otherwise;

(q) = TB EFDMA ( m, k )

(2.5)

(q) BIFDMAandBEFDMA: TB IFDMA ( m, k )

M + l + q.L; 1, m = p = (2.6) P 0, otherwise;

31

Chapter2

SystemModel

WherePisthenumberofblocksassignedtoaspecificuserandListhenumber of subcarriers per block. The index variables in the equation are m=0,,M1, k=0,,K1, p=0,,P1andl=0,,L1.Thefigurebelowshowsexactlyhowfoursubcarriersaremapped inthethreemappingschemes. Figure2.11:ExampleofmappingfoursubcarriersinIFDMA,LFDMAandBIFDMA 2.2.4.1TimeDomainRepresentation Thesymbolsafterbeingmappedarethentransformedintotimedomainsymbolsby taking their IFFT on the transmitter side. Thus it is useful to study the time domain representationofthethreemappingschemes.Forthispurposewewouldusetherunning examplethatisbeingusedinthepreviousfigure.

32

Chapter2

SystemModel

2.2.4.1.1IFDMA TheIFDMAschemecanbevisualizedbyupsamplingasignal.Theexampleabove

canbeseenasequivalenttoupsamplingtheoriginalDFTprecodedsequenceXnbyfactor of3andasaresultweget2zerosbetweeneachconsecutivesample.Thus,iftransformed intotimedomainitwouldgiverepetitionsoftheoriginaltimedomainsymbolsasshownin thefigurebelow {xIFDMA}

x0

x1

x2

x3

x0

x1

x2

x3

x0

x1

x2

x3 Time

Figure2.12:TimeDomainRepresentationofIFDMAmapping HencewegetQrepetitionsoftheoriginalsignalwhichisscaledbyafactorof1/Q whereQisthenumberofusersinthesystem. 2.2.4.1.2LFDMA The figure below shows the time domain representation of LFDMA which is similarforthetimedomainrepresentationofBIFDMAinourrunningexample x0 *0 *1 {xLFDMA} x1 *2 *3 x2 *4 *5 x3 *6 *7 Time

*=

c
k =0

k ,m .x k , c k , m

= ComplexWeight

Figure2.13:TimedomainrepresentationofLFDMAmapping
Hence in LFDMA, the input symbols are repeated with a scaling factor of 1/Q at samplepositionswhichareactuallytheintegermultiplesofQwhereQisthenumberof users. The values indicated by * are actually combinations of all the input symbols with different complex weights. It should be noted that the time domain notations for LFDMA andBIFDMAarethesameinourrunningexample,buttheintermediatetermsbetweenthe inputsymbolsaredifferentinthecaseofBIFDMAthaninthecaseofLFDMA. 33

Chapter2

SystemModel

2.2.5 OFDM,OFDMA&SCFDMA
Withthepassageoftime,toaccommodateincreasednumberofusersandtohandle thechannelconditions,differentmodulationschemesandmultipleaccesstechniqueshave been developed. The next block in the IMTAdvanced simulation model is the OFDM modulationanddemodulation.InthissectionwediscussabouttheOrthogonalFrequency DivisionMultiplexing(OFDM),itsmultiuserversionthatisOrthogonalFrequencyDivision Multiple Access (OFDMA) and Single CarrierFrequency Division Multiple Access (SC FDMA)incombinationwiththeSingleCarrierFrequencyDomainEqualization(SCFDE)to giveourreaderabriefandinterestingpictureofthesebasicterms.

2.2.5.1OFDMandOFDMA OFDM is a multicarrier modulation scheme that is based on frequency division multiplexing (FDM). FDM uses multiple frequencies simultaneously to transmit multiple signals in parallel. FDM and OFDM are not similar as there are few differences between themlikeOFDMismuchmorespectrallyefficientbyspacingthecarriersorchannelsmuch closer together and saving the bandwidth by excluding the guard bands and allowing signalstooverlap.FDMconstitutesguardbandssoitcoversmorebandwidth. Another advantage of OFDM over FDM is the orthogonality of subcarriers due to whichtheycanbereceivedwithoutinterference.Thefigure2.14belowshowsthestructure ofFDMandOFDMsignalinfrequencydomain Figure2.14:ThestructureofFDMandorthogonalsubcarriersinOFDM 34

Chapter2

SystemModel

A very important objective to bring OFDM into the game is to mitigate the fading channel effects. If a signal wants to experience flat channel then it must occupy less bandwidththanthecoherencebandwidthofthechannel.OFDMis averygoodsolutionto defend the hindrances created by a fading channel as each of its subchannels or sub carriersexperienceaflatfadingduetolesserbandwidth.
S0

S(k)

Transmitter

S(k)

Receiver

S/P Sk1 IFFT S0

P/S Sk1 FFT

P/S S/P

AddCP RemCP

D/A

A/D

Channel

Figure 2.15: OFDM Transceiver OFDM is also good from the time domain perspective as having slim and narrow band signals will cause the symbol duration of each one to be long enough. As symbol durationgetslongerthanthemaximumdelayspreadintime,ithelpstoeliminatetheInter Symbol Interference (ISI) [9]. In the figure above a simple OFDM transceiver has been shown with its major blocks. The S (k) and S (k) are the modulated and demodulated symbolsrespectively.Onthetransmittersidethesymbolsarefirstconvertedfromserialto parallelandarepassedthroughtheIFFTblockthatbringsthesignalintothetimedomain. Afterthat,itispassedthroughtheparalleltoserialconverterandfinallyacyclicprefixis addedtoitandwegetanOFDMsymbol.ThisOFDMsymbolisthensentonthechannelafter applyingthepulseshaping(thatreducestheoutofbandinterference) andisthenpassed throughD/Aconverter.

35

Chapter2

SystemModel

Thecyclicprefixsizeshouldalsobeadjustedanditmustbesameasorgreaterthan thedelayspreadofthechannelsothattheoriginalinformationmustbeprotectedfromthe channel transients. On the receiver side the inverse process takes place and we get the demodulated symbol. In this simple model, there is no equalization but normally equalization is applied to mitigate the adverse effects of the channel. In OFDM, there are differenttypesofsubcarriers;mostareusedtocarrydata.Fewareusedaspilotandsome donthaveanyinformationandtheyarecallednullcarriers.Thepilotsubcarriersprovides a reference to minimize frequency and phase shift during transmission and null sub carriersserveasguardbandandDCcarrier[17].
D/A A/D AddCP RemCP P/S P/S S0 IFFT S/P S0 Sk1 FFT S/P Sk1

S(k)

S(k)

Transmitter
SubcarrierMapping SubcarrierDeMapping

Receiver

Channel

Figure 2.16: OFDMA Transceiver Thefigure2.16aboveshowsatypicalOFDMAtransceiver.Thebasicconstructionof OFDMA transceiver is almost similar but the difference comes at the carrier mapping and DeMapping stage. OFDMA is nothing but multiple user version of OFDM. OFDMA is a multipleuseraccesstechniqueinwhichtheinformationissenttomanyusersondifferent subcarriers assigned to each user. This technique is used in the downlink when information is sent to a lot of users. Each user extracts the information from the sub carriersassignedtoit.Intheuplink,thewayhowthecarriersareassignedisdifferent.In OFDM, all the subcarriers or the full frequency band is assigned to a single user for a specific time interval. While in case of OFDMA the subcarriers are allocated to users and 36

Chapter2

SystemModel

theusersreceiveinformationonthesesubcarriers.Thistechniqueimprovesnotonlythe bandwidth utilization but also brings power gain over OFDM system in the uplink, as it allowstheuserstotransmitonlythroughthesubchannelsthatareassignedtothem[17].
Subcarrier

User1
FFTSymbol

User2

User3

User4

User1 User2 User3

User1 User1 User1 User2 User4 User3 User3 User4 User2 User3 User3 User3 User4
Time

Figure2.17:SubcarriersinUplinkinOFDM

User4
Subcarrierss

User2 User2

Time

Figure2.18:SubcarriersinUplinkinOFDMA In the figure 2.17 and figure 2.18, the subcarrier allocation has been shown for 4 usersforOFDMandOFDMAinuplinkrespectively.OFDMAhasmanyadvantageslike;ithas broughtreductionintheinterferenceamongtheusersaswellas,hasimprovedtheNonline ofsight(NLOS)communicationinamobileenvironment[12]. TherearemanyadvantagesofOFDMlike;itcreatesimmunityagainstthefadingand interference, provides high spectral efficiency, has no ICI due to orthognality among the subcarriers (assuming linear environments) and is better in scenarios where channel conditionsareconstant.OFDMAalsooffersimpressivefeatureslikebetteruseoffrequency resources and power gain and is suitable in the scenarios where the channel is time

37

Chapter2

SystemModel

frequencyselective.TherearealsosomedrawbacksofOFDM/OFDMAthatareneededtobe mentioned atthispoint ofdiscussion.Thefirst andmajorone isthehighpeakto average powerratio(PAPR)thatistheresultofsuperpositionofallthesubcarrierswithdifferent carrierfrequencies.Duetothesuperpositionofallsubcarriers,itisverydifficulttoavoid highpeaksofthesignal.Thisisindicatedinthefigurebelow Figure2.19:SuperpositionofallsubcarriersinOFDMA

Subcarrier Bandwidth
O F D M S Y M B O l

Time Amplitude

O F D M S Y M B O l

CyclicPrefix

Frequency

Figure2.20:OFDMModulationusingmanysubcarriers

38

Chapter2

SystemModel

Time
Amplitude

SCFDMSybmol

CyclicPrefix

Subcarrier Bandwidth

Frequency

Figure2.21:SCFDMAusingsinglecarrierintheuplink
2.2.5.2AdvantagesofUsingSCFDMA SCFDMAisaSinglecarrierFrequencydivisionmultipleaccesstechniquethathas been used in the uplink in IMTAdvanced and LTE. This modulation technique is used in combinationwithfrequencydomainequalizers(FDE)duetoincreasedcomplexityoftime domainequalizers.AmajordifferencebetweentheSCFDMAandtheOFDMAtransceiveris just an addition of a DFT block on the transmitter side and IDFT detection block on the receiver side. Hence on the transmitter side, the symbols are transmitted as time domain symbols while on the receiver side, they are transformed into frequency domain symbols beforeequalizationandafterthattheyaredetected. ThemajoradvantageofusingSCFDEorSCFDMAintheuplinkislowPAPRratioof thetransmittedsignalcausedbyDFTprecodingdoneatthetransmitter.PAPRisbasically theratiobetweenthepeakpowertoaveragepowerofthetransmittedsignal.PAPRisvery importantinrelationwiththeamplifierefficiencyatthetransmitterandintheamplifierthe maximum efficiency is achieved when it operates at the saturation point. The low PAPR valuewillallowtheamplifiertooperateclosetothesaturationregion.

39

Chapter2

SystemModel

SCFDMA has subcarrier bandwidth common to all subcarriers, so there is no superposition of subcarriers like in OFDMA in figure 2.19 so the peak of the transmitted signal is not so high. Interleaved SCFDMA is a preferred modulation technique for lower PAPR.PulseshapefilteringofSCFDMAinfactdegradesthePAPRlevelofinterleavedSC FDMAwhereasitshowsnoeffectwithlocalizedSCFDMA[12].Nowifwelookatfigure2.20 andfigure2.21wecanseethatSCFDMAusesthewholebandwidthforeachsymbolinthe frequency domain hence, it is more immune to frequency selective fading. But in time domain it is very short and will face more ISI as compared to OFDMA. So it becomes necessaryforSCFDMAtouseanequalizertomitigatetheISIeffects[9].

2.2.6 NONLINEARAMPLIFIERS
It is a well known fact that OFDM comprises of large envelope fluctuations as compared to single carrier systems with same modulation and same spectral characteristics.ThisisbecauseOFDMisnothingbutsumofmultitudeofdigitallymodulated signals with Gaussian like probability density function (PDF). Hence the peak to average powerratio(PAPR)ofOFDMisquitelargewhichresultsinaninefficientusageofthehigh poweramplifier(HPA)atthetransmitter[18].ThehighPAPRcausestheHPAtoenterinto thenonlinearregionandthuswegetdistortionsduetononlinearitiesattheoutputofthe HPA. Normally nonlinear regions are prevented by backing off the amplifier from the saturation region thus keeping it in the active or linear region of operation. This backoff wouldhoweveraffecttheefficiencyoftheHPA.Henceitisdesiredtomaintainlowbackoff values but this would cause an additional interference as nonlinear distortions and peak value limitations produced as a result would result in intermodulation between different carriers[19].Itisthusdesiredtokeepthebackoffvalueslowbutatthesametimeprevent thenonlinearitiescausedbytheHPA. 2.2.6.1Amplifiercharacteristics To better understand the impact of nonlinearities on our multicarrier modulated signalitisusefultostudybrieflythebasicsofabipolarjunctiontransistor(BJT)amplifier.A BJT basically consists of three alternatively doped semiconductor regions and two PN junctions.Thethreelayersarecalledemitter,baseandcollectorandthejunctionbetween emitterandbaseistheemitterbasejunctionwhilethejunctionbetweencollectorandbase iscalledthecollectorbasejunction.

40

Chapter2

SystemModel

FornormaloperationoftheBJTtheemitterbasejunctionisforwardbiasedandthe collectorbasejunctionisreversebiased.Therearefourdifferentmodesofoperationsofa BJTwhicharebrieflyexplainedbelow: 2.2.6.1.1Cutoffregion Inthecutoffregionboththeemitterbasejunctionandthecollectorbasejunctions are reverse biased and hence there is a negligible current flow due to minority carriers. HencetheamplifierisessentiallyinOFFstate. 2.2.6.1.2Saturationregion If both the PN junctions of the BJT amplifier are forward biased, it will be in saturationregionofitsoperation.Thesaturationregionoperationisoppositeofthecutoff regionoperationandisessentiallyalogicalONstate.Thecollectoremittercurrentinthe saturationregionreachesamaximumwhichisindependentofthebasecurrent[20]. 2.2.6.1.3Activeregion In this case the emitter base junction is forward biased while the collector base junctionisreversebiasedandtheoutputcollectoremittercurrentincreasesasafunctionof the base current in this region. Moreover the collector emitter current has a very small slopewithVCE[20]. Figure2.22:BJTIVcharacteristicscurve 41

Chapter2

SystemModel

2.2.6.1.4Breakdownregion In the breakdown region the input collector to emitter voltage would increase

beyond the breakdown value as a result of which the transistor breaks down with a tremendous increase in the collector emitter current. The figure 2.22 shows IV characteristics of a typical BJT indicating the four regions of its operations that are mentionedabove. 2.2.6.2NonlinearityModeling Nonlinearities in multicarrier systems can be modeled in many different ways. Some of the techniques used for modeling HPA nonlinearities are travelling wave tube (TWT) amplifier with strong AM/PM conversion, solid state amplifier (SSA) model and idealizedamplifiermodelcalledenvelopelimiter(LIM)[19].Forouranalysisinthisthesis weusenonlinearamplifieratthetransmittersideandweimplementitusingsimplenon lineardistortionclippingorsoftlimiter(SL).Howeverthesameanalysiscanbeappliedto more sophisticated nonlinear amplifiers. The clipped outputs xc(n) of our time domain signalsx(n)aregivenby

x(n), | x(n) | A x c (n) = i arg{ x ( n )} , | x( n) |> A Ae

WhereAistheclippingthreshold[21].

(2.7)

ThefigurebelowshowsatypicalthresholdclipperwiththresholdvalueequaltoA=500.
1000 900 800 Clipped signal at threshold A =500 700 600 500 400 300 200 100 0

0 100 200 300 400 500 600 Input signal x(n)

700

800

900

1000

Figure2.23:ThresholdclippingforNonlinearitiesmodeling 42

Chapter2 2.2.6.3SystemdegradationbyHPANonlinearity

SystemModel

The distortions produced by a nonlinear amplifier are dependent on envelope fluctuations. A non constant envelope signal when modulated by a nonlinear amplifier wouldbedistortedindifferentways.ThenonlinearitiesintheHPAwouldcauseadditional interference in the receiver. We would get interference between the inphase and quadraturecomponentsduetheAM/PMconversion.Moreoverthepowerspectraldensity (PSD) of the modulated signal with fluctuating envelope would be broadened by the non linear effects of the HPA that would cause spectral spreading of the signal and hence adjacentchannelinterferencewouldbeintroducedintheamplifieroutput.Finallyitshould benotedthatwhenseveralchannelsareamplifiedinthesameHPA,therewouldbeinter modulationeffectsintheamplifieroutputsignal.InotherwordstheorthognalityinOFDM wouldbecompletelydestroyedbythenonlineardistortionsintheHPA[19]. As mentioned earlier one way of preventing these effects is to make the amplifier operateinthelinearregionwhichwouldrequireintroductionofoutputbackoff(OBO)of theHPAwhichisgivenby

OBO = 10 log10 (

Po ) indBs Ps

(2.8)

WherethereferencepowerPoisthesaturatingpowerofHPAandPsisthemean outputpowerofthetransmittedsignal.Itishoweverunfortunatethattheefficiencyofthe HPA is very low at high values of OBO so there is a tradeoff between maximizing output power on one side and avoiding degradations due to nonlinear distortions on the other side[19]. 2.2.6.4PowerControlinUplink InthispartofthereportwediscusstheroleofpowercontrolintheLTEuplinkand

itsmathematicsfollowedbyitstypes.

2.2.6.4.1PowerConsumption
As discussed already, power consumption is a key factor when we are dealing with the LTE uplink. The user equipment (UE) is actually energy limited by its energy sourceusuallyabattery.TheSCFDMAapproachmentionedearlierisactuallyintendedto keepthePAPRtoalowerlevel.Powercontrolinuplinkcanalsobeseenasamethodtosave

43

Chapter2

SystemModel

unwantedpowerconsumptionintheuserequipment(UE)andhenceincreasingthebattery life[22].

2.2.6.4.2InterandIntraCellinterference
SecondlytheSCFDMAsignalsfromdifferentUEsareorthogonalandhencethey do not interfere at the receiver in ideal circumstances but in practice the orthognality is disturbed by certain impairments in transmitter and receiver. These transmitter imperfections would cause the signals from different UEs to leak beyond their intended frequency bands and interfere with each other. The transmitter impairments can be expressed by the Inband Subcarrier Set Leakage (ISSL) which describes the allowed disturbancefromoneUEtransmitterlocatedononesubcarriersetontoanothersubcarrier set within the regulatory bandwidth of the UE. ISSL is thus a measure of the leakage between the subcarriers. It should be noted that ISSL of about 25dBs or more seems reasonable to achieve [23]. The figure below shows a typical received PSD affected by transmitterimpairmentwithanISSLof25dBs Figure2.24:Transmitterimperfectionmodel:PSDwithISSL=25dBs The transmitter impairments may include phase noise or IQ imbalance. Moreover, theHPAnonlinearitiesbeingdiscussedearlierinthischapterisalsooneofthetransmitter impairments that would cause leakage between subcarriers assigned to different UEs (intra cell multiple access interference). In addition to intracell interference due to

44

Chapter2

SystemModel

transmitterandreceiverimpairmentstherecanbeinterferencefromtheneighboringcells whichiscalledintercellinterferenceandwhichcanlimitthesystemperformanceinterms of its capacity [24]. These forms of interference can be controlled by a power control schemewhichwouldessentiallycontrolthereceivedPSDsandwouldrestrictthesignalsto theirdesiredfrequencybands[25]. 2.2.6.4.3PowercontrolEquation Power control is a technique that varies the radiolink transmit power in accordancewiththeinstantaneousvariationsinthechannelconditions.Insimplewordsthe transmitpowercontrolwouldincreasethetransmitpowerwhentheradiolinkexperiences poorradioconditionsandwouldreducepowerincaseofgoodradioconditions.Hencethe transmit power is essentially inverse of the channel quality [2]. The agreed upon power controlformulaforLTEuplinkis P = min{Pmax ,10. log10 M + Po + .PL + mcs + f ( i )} [dBm](2.9) Where PmaxisthemaximumpowerallowedanditdependsontheUEpowerclass. Misthenumberofphysicalresourceblocks(PRBs).

Po iscell/UEspecificparametersignaledbyradioresourcecontrol(RRC).
ItismeasuredindBmandrepresentspowercontainedinonePRB.

ispathlosscompensationfactorwithvaluesrangingbetween0to1.
PListhedownlinkpathlossestimateattheUE. mcsiscell/UEspecificmodulationandcodingscheme. f(i)isUEspecificanditpermitstouseabsoluteoraccumulatecorrectionvalue.

Po intheaboveequationcanbecalculatedas

Po = .( SNRo + Pn ) + (1 )( Pmax 10. log10 M o ) [dBm](2.10)

Where

SNRo istheopenloopSNRtarget. Pn isthenoisepowerperPRB. M o isthenumberofPRBsforwhichthetargetSNRisachievedwithfullpower[24].

45

Chapter2

SystemModel

Theparameters Po andarebroadcastbythebasestation(alsocalledevolved nodeB(eNB)in3GPPterms)toalltheUEsandtogetherwithpathlossaresufficienttoset initial transmit power. These are thus called open loop terms. The mcs and i" are however signaled to UE as its grant from the eNB after the initial power has been set accordingtoopenlooptermsandfeedbackissenttoeNB. These are thus called the closed loop terms [26]. The power control equation given abovecanbesimplifiedbyneglectingtheunnecessarytermstogetthesimplifiedequation givenbelow

P = Po + .PL [dBm/PRB](2.11)
2.2.6.4.4TypesofPowercontrol Basedonspeedofchannelvariationpowercontrolisoftwotypes SlowPowerControl:Designedtocompensateforslowchannelvariations. FastPowerControl:Designedtocompensateforfastchannelvariations.

BasedonthetypeofinformationsenttotheUEtoadjustitstransmitpower,powercontrol isoftwotypes OpenLoopPowerControl:ThetransmitpowerismeasuredattheUEusingthe openloopparameterssentbytheeNBandnoinformationissenttoeNB. Closed Loop Power Control: The UE send feedback to the eNB in this kind of methodtocorrectitstransmitpowervalue[26]. Basedonthevalueofwehavethreetypesofpowercontrolschemes NoPathLossCompensation:Whenis0wewonthaveanypowercontroland alluserswouldusemaximumallowedtransmissionpower. Fractional Path Loss Compensation: For 0< <1 the type of power control is called fractional power control scheme as it compensates fraction of the path loss. Since Po and are same for all UEs so its the path loss PL which would makeeachusertobereceivedwithdifferentvaluesofPSDs. ConventionalPathLossCompensation:When=1thepowercontrolschemeis called conventional power control scheme as it compensates for the whole of path loss. This scheme would essentially equalize the power of all UEs before

46

Chapter2

SystemModel

beingreceivedattheeNBandthushasanadvantageofeliminatingthenearfarproblemas oftypicalCDMAsystems[24]. Fractionalcompensation NocompensationFullcompensation =0 =1 Figure2.25:Powercontrolschemesbasedonvaluesof

2.2.7 Channel&Equalization
Inthispartofthetextwehavediscussedaboutthekeychannelcharacteristicsand laterontheequalizationmethodstoreducetheeffectsofthechannelimpairments.

2.2.7.1ChannelCharacteristics
The wireless channel consists of many severe challenges like noise, interference and many other impairments that effect mobile communication over time and frequency duetothehighmobilityofusers.Wehavediscussedsomekeyimpairmentingeneralthat affectthewirelesscommunicationsignal.

2.2.7.1.1PathLoss
Pathlossisoneofthechannelcharacteristicsthatiscausedbydissipationofthe powerradiatedbythetransmitteraswellaseffectsofthepropagationchannel.Thereare differentpathlossmodelsincludingthefreepathlossandempiricalpathlossmodelsthat describe the path loss. In case of free space model no obstructions are considered in between the transmitter and receiver and the propagation is assumed to be line of sight (LOS). The channel model associated with this transmission is called a lineofsight (LOS) channel,andthecorrespondingreceivedsignaliscalledtheLOSsignalorray[1]. Let Ptbethetransmittedpowerofthesignalandlet Prbethereceivedpowerofthe signalthenpathlossPLisdefinedas, PL =

Pt Pr

(2.11)

WhilethepathlosscanalsobedefinedasindB, PL dB = 10 log10
Pt dB (2.12) Pr

Themoderndaycomplexmobileenvironmentscannotbemodeledbyusingsimple free space path loss models so some more complex models that are more near to mobile 47

Chapter2

SystemModel

communicationenvironmentsarerequired.Theempiricalmodelsareoneofthemthatare usedtopredictpathlossintypicalwirelessenvironmentssuchaslargeurbanmacrocells, urban microcells, with more distance in a given frequency range and a particular geographicalareaorbuilding.TheexamplesofsuchkindofmodelsaretheOkumuraModel andHataModel[1].

2.2.7.1.2ShadowFading
A signal that is transmitted through a wireless channel experiences different random variations due to the hurdles and different obstructions or objects in the environmentsthatcomesinthesignalpath.Suchvariationsarealsocausedbychangesin reflecting surfaces and scattering objects and they result in the variation of power of received signal. This type of effect on the received power is called shadowing or shadow fading.Hereinthefigurebelowwecanseethecombinedeffectofthethreedifferentfactors thatcausesvariationinsignalstrength.

Figure2.26:CombinedPathloss,shadowingandmultipathpropagationeffect 2.2.7.1.3Narrowbandfading
Beforemovingtonarrowbandfadingherewewouldliketodiscussbrieflyabout multipath propagation that creates ISI at the receiver. The Multipath propagation results due to the multiple copies of the same signal coming from different paths at the receiver with different delays in time. The multiple copies are also called multi path components. 48

Chapter2

SystemModel

Here it is also necessary to mention about the delay spread that is the time difference betweenthearrivalofthe1stmultipathcomponentandthelastmultipathcomponentofthe transmittedsignal.ThisisdenotedbyTm.Thereisadifferencebetweenthedelayspread and Doppler spread. The Doppler spread occurs due to the Doppler shift and the Doppler shift is basically the change of frequency of wave as observed by the receiver when it is movingtowardorawayfromthesourcewithsomevelocity[9]. Whenmanymultipathcomponentsaregeneratedthenduetotheconstructiveand destructive addition of multipath signal components, the received signal strength varies very rapidly over a small distance usually in the order of signal wavelength. Since the propagation channel varies over small distance it is called small scale fading. Small scale fadinghasbeenmodeledasrandomtimevaryingimpulseresponse. Narrowbandfadingisoneofthetypesofthesmallscalefadingthatresultsfromthe multipath propagation and is based upon the relationship between signal bandwidth and maximumdelayspreadofthechannelanditfulfillsthisrelation Tm << B 1 . TmisthedelayspreadwhileBisthebandwidthofthetransmittedsignal.Sothis meansthatinnarrowbandfadingthedelayspreadorthedifferencebetweenthedelaysof multipath components will be very small as compared to the bandwidth of the signal. Narrowband fading can be modeled practically by Clarke's or Jake's model that assumes that all the multipath components arrive uniformly with same average power and with sameangleofarrival[9].

2.2.7.2 Widebandfading
This is another type of small scale fading where we observe distortion due to multipath delay spread when the signal bandwidth is approximately equal to inverse of delay spread B 1 / Tm . In the wideband fading the delay in between the multipath componentsismuchlowerthanthenarrowbandfading.Thisresultsinlongerdurationof received pulses that creates interferences with the neighbor pulses thus causing the ISI. OFDM is a kind of mitigation technique that is used to change the characteristics of the transmitted signal by extending the symbol duration mainly to overcome ISI [9]. The Widebandchannelsareusuallycharacterizedbytimevaryingimpulseresponseandtime varying frequency response. Timevarying impulse response c( ;t) is modeled as a two dimensionalrandomprocessandtimevaryingfrequencyresponse C(f;t)isobtainedfrom

49

Chapter2

SystemModel

c ( ; t) by taking the Fourier transform with respect to i.e., C (f; t) = F [c ( ; t)]. An exampleofatwodimensionaldiscretetimevaryingimpulseresponse c( ;t)isshownin thefigurebelow[9]. Figure2.27:TimeVaryingDiscreteTimeImpulseResponseModelforaMulti PathChannel In the following some key properties of a wideband fading channel has been discussed including the power delay profile, coherence bandwidth, and Doppler power spectrum.Thesecharacteristicsarebasedonchannelautocorrelationorscatteringfunction.

2.2.7.2.1PowerDelayProfile Power delay profile (PDP) is one of the important characteristics of wideband
fading.ThePDPrepresentstheaveragepowerofeachmultipathdelaycomponent.ThePDP basicallyshowsthestrengthofmultipathcomponentsandtheircorrespondingdelaysinthe channel so it is also referred to as multipath intensity profile. It is important to mention here about the channel delay spread that can be either average delay spread ( Tm ) or the rootmeansquare(RMS)delayspread( Tm ).Bothofthemcanbeanalyzedfromthepower delayprofile.

Tm =

Ac( )
0 0

Ac( )

, Tm =

(( Tm ) Ac ( ))

Ac( )

(2.13)

50

Chapter2

SystemModel

ThepowerdelayprofileoftheWINNERC2NLOScanbeseenshownbelow, Figure2.28:NormalizedPowerDelayProfileofWINNERC2NLOS We can describe the delay spread as the maximum delay Tm after which the power of all multipath components becomes negligible. The PDP can be used to assess whether the system has ISI or not. In the channels that have many scatterers the delay spreadisapproximatelyequaltotheRMSdelayspread.

2.2.7.2.2 Coherencebandwidth
ThecoherencebandwidthBcaccordingto[1]iswhere Ac(f)0forallf>Bc. Coherencebandwidthisactuallytherangeoffrequenciesoverwhichthechannelpassesall spectral components with almost equal gain and linear phase [9]. It is also the band of frequencies over which the channel is highly correlated in frequency. There is a high relationship between the fading and the coherence bandwidth of the channel. If the bandwidthofthesignalismuchlessthanthecoherencebandwidth,thenthissignalisgoing to experience fading equal over its entire bandwidth and this type of fading is called flat

51

Chapter2

SystemModel

fading. If the bandwidth of the signal is more than the coherence bandwidth of the signal thenthesignalwillexperiencefrequencyselectivefading.

2.2.7.2.3DopplerPowerSpectrum&CoherenceBandwidth
Doppler power spectrum is also one of the important characteristics of the wideband fading. The Doppler power spectrum basically describes the variations of the channelwhichoccurduetothemovementoftransmitterorthemovementofreceiver.As thetransmitterorthereceivermovesrapidlythereisachangeorshiftinthefrequencyof thereceivedsignalthatiscalledtheDopplershift.ThisDopplerEffectcanbemeasuredby taking the Fourier transform of the autocorrelation Ac ( f ; t ) ) with respect to t so
Sc( f ; ) = Ft [ Ac (f ; t )].Sc ( ) = Sc (0; ) characterizes Doppler shift for a single frequency

and Ac(t ) = Ac(f = 0; t ) describesthebehavioroftheautocorrelationfunctionwithtime andisusedindefiningtheCoherencetime[1].ThecoherencetimeTcdescribeshowthe channel decorrelates over time. It defines the range of values over which Ac ( t ) is approximatelynonzero.Thefastortheslowfadingdependson thecoherencetime.Ifthe symboldurationisverysmallascomparedtothecoherencetime(Ts<<Tc)thenthechannel impulseresponsewouldbetimeinvariantoveronesymbolperiodandthistypeoffadingis calledslowfadingwhileonotherhandifTc<<TsthentheImpulseresponsevariesoverone symbolperiodandthefadingisknownasFastfading.

2.2.7.3

ChannelEqualization

Equalizationis atechniquethat isusedtorecoverthesignalbymitigatingthe adverseeffectsofthechannel.ThebasicobjectiveoftheequalizeristotypicallybalanceISI mitigation with noise enhancement. There are various types of the equalizers including linear and nonlinear equalizers. The nonlinear equalizers suffer less from noise enhancement than linear equalizers, but they are more complex. Equalizers estimate the channelimpulseorfrequencyresponsetomitigatetheresultingISI.Asthewirelesschannel variesovertime,thentheequalizermustalsodiscoverthefrequencyorimpulseresponse of the channel (training) and then update its estimate of the frequency response as the channelchanges(tracking)[1].Thistypeofprocessinwhichtheequalizersadoptthemself with the change of channel by training and tracking process is often called as adaptive equalization, as the equalizers adapt to the changing channels. The process of the

52

Chapter2

SystemModel

equalization might be very different and the process of tracking and training becomes complexwhenthechannelischangingrapidly. Anequalizercanbeimplementedatbaseband,RadioFrequencies(RF),orIF.Most equalizers are implemented digitally after A/D conversion, since such filters are small, cheap, easily tunable, and very power efficient [1]. The OFDM symbol, with flat fading experiencedbyeachofitssubcarriers,offersanadvantageinequalization.Thisisbecause; asingletapequalizercanbeusedforcompensation.Asingletapequalizerisnothingbuta simpledelaylinemeaningthatapreviousdetectedsymbolcanbeusedforequalizationof flatfadingchannels. There are different types of equalizers but in this thesis project we have used an MMSEequalizer.InMMSEequalizationthegoaloftheequalizer designistominimizethe averagemeansquareerror(MSE)betweenthetransmittedsymbol anditsestimateatthe output of the equalizer. Here is the equation that describes the parameters of an MMSE equalizer

C MMSE =

HH

2 |H | + w s2
2

(2.14)

53

Chapter3 SIMULATIONS&RESULTS
In this chapter we describe the simulations parameters as well as the results acquired from these simulations that are based upon different cases related to performanceanalysisofTurbocoding.

Chapter3

Simulations&Results

3.1 SystemSpecifications
Thebasicsystemthatisbeingsimulatedlooksliketheoneshownearlierinfigure 2.2. Here we try to present the signal flow from the transmitter to receiver and the dimensionsofeachblockusedtoprocessthesignal.Allthesignalsarerepresentedbytheir discretetimeequivalentsinthecomplexbaseband.Thematricesarerepresentedbycapital letterswhilethevectorsarerepresentedbysmallletters.Superscriptsandsubscriptsare usedtoindicatethenumberofusersandthenumberofsubcarriersrelatedtoeachsignal respectively.(.)T,(.)Hand(.)areusedtorepresenttranspose,hermitionandpseudoinverse respectively. A system having Q users is considered throughout the derivation with the user indexqrangingfrom0,1,Q1.Rawdataforeachuserisrepresentedby d(q).Thedata from each user is first encoded by the Turbo encoder and the coded bits in one chunk duration are fed into the QPSK modulator. The energy of each modulated symbol is normalized to make it equal to unity. The modulated symbol for each user is denoted by sK(q)whereKisthenumberofsubcarriersassignedtoauser. In the DFT precoded linear system the modulated symbols are converted into frequency domain by taking a Kpoint FFT. The DFT precoding matrix is a KK matrix denoted by Fk . The DFT precoded symbols are them mapped to K out of M sub carriersbythesubcarriermappingmatrixdenotedby T ( q ) whichisanMKmatrix.The matrix T ( q ) cantakeanyoftheformmentionedearlierinthereportinequations2.4,2.5 and2.6dependingonthetypeofsubcarriermappingschemebeingused.MpointIFFTof the signal is then taken and cyclic prefix of suitable length is added. The signal is then transmitted over the channel. The channel being used is the WINNERC2 NLOS whose multipath propagation channel coefficient matrix for qth user is denoted by H ( q ) . The signal is also corrupted by AWGN WM as it is passed through the channel. The mathematicalrepresentationofreceivedsignalinDFTprecodedsystemisthusgivenas
(q) (q) H rM = H ( q ) .FM .T ( q ) .Fk .s k + WM

(3.1)

TheprocessinthenonDFTprecodedlinearsystemisthesameexceptthattheDFT precodingmatrix Fk inequation3.1isnotused.ThisimpliesthatinnonDFTsystemthe modulated symbols can be seen as time domain symbols unlike in DFT precoded where theyareinfrequencydomain. 55

Chapter3

Simulations&Results

The mathematical representation of the received signal in a nonDFT precoded systemisgivenbelow

(q) (q) H rM = H ( q ) .FM .T ( q ) .s k + WM

(3.2)

Inthenonlinearsystemathresholdclipperisaddedintothesystemwhichisseen shaded in the figure 2.2. This makes the system nonlinear by clipping the signal above a certain predetermined threshold. Moreover it is worth mentioning that the cyclic prefix and the clipping are not included in the mathematical equations. This is because cyclic prefix is just addition and subtraction of some bits and has no real impact on the mathematicalmodelingofthesystemasfarasitslengthisgreaterthanorequaltothedelay spread of the channel [9]. The clipping on the other hand will change the mathematical modeling of the system and can be approximated by mathematical models but they are ignoredsincetheyareoutsidethescopeofthisthesis. Onethereceiversidethecyclicprefixisfirstremovedwhich isfollowedbytaking an Mpoint FFT. This is followed by subcarrier demapping. The subcarrier demapping matrix T(q)isaKMmatrixandispseudoinverseofthesubcarriermappingmatrix T(q). Theadverseeffectsofthechannelaremitigatedbyusinganequalizer.Theequalizerbeing used in this thesis is MMSE frequency domain equalizer (FDE). The equalizer coefficient matrixisrepresentedbyaKKdiagonalmatrixdenotedby C(q).ThisisfollowedbyDFTpre decoding the signal if DFT precoding has been employed at the transmitter. Then the received signal is demodulated and decoded to get estimates of the transmitted bits. The mathematicalrepresentationofthereceivedsignalbeingprocessedatthereceiverisgiven below
^

Sk

(q)

(q) = FkH .C ( q ) .T ( q ) .FM . rM

(3.3)

FornonDFTprecodedsystemtheequationremainsthesamewith theremovalof theDFTpredecodingmatrixFHK.

^

Sk

(q)

(q) = C ( q ) .T ( q ) .FM . rM

(3.4)

56

Chapter3

Simulations&Results

Thetablebelowshowsthesimulationparametersofthesimulatedsystem, Bandwidth CarrierFrequency SamplingFrequency SamplingTime GuardInterval TotalNumberofSCs(M) NoofSCsperUser(K) SCsperBlock(BIFDMAOnly) ChunkWidth Modulation Coding Equalizer ChannelEstimation 80MHz 3.7GHz 1/12.5ns 12.5ns 1.47ns 1024 128 4 12 QPSK TurboCoding,Coderate, [G1,G2]=[11111,10001] MMSE Perfect WINNERC2NLOS Channel UserVelocity=50Km/h CoherenceBandwidth=680.27KHz CoherenceTime=5.8ms

Table3.1:SimulationParameters

57

Chapter3

Simulations&Results

3.2 Delimitations
The cases with amplifier nonlinearities have been analyzed excluding the IFDMA mapping scheme. The reason behind this is that we have not used pulse shaping in this thesis. At a later point in the thesis it was realized that IFDMA mapped signal cannot be clipped with the clipping method that we have employed. In our threshold clipper the absolute value of the input signal to the threshold clipper is first sorted to find the appropriate clipping threshold which is then used to clip the original input signal to the thresholdclipper.IncaseofIFDMAhoweverifweobserveitstimedomainrepresentation as shown in figure 2.12, its nothing but Q repetitions of the original signal scaled by a factor 1/Q. Thus sorting it and then clipping it will not clip the IFDMA signal by the percentage intended but would rather clip each repetition by some amount which is not what is required. A more practical approach for analyzing IFDMA in nonlinear environments will be to include pulse shaping in the simulations which we have not includedduetotimeconstraintsinthisthesis.

3.3 AWGN&RayleighFadingChannel

10
-14

10

10

-2

10

-4

BER values

10

-6

10

-8

10

-10

10

-12

Simulated BIFDMA Simulated LFDMA IDEAL AWGN BER

0 1 2 3 4 5 6

Eb/No in dBs

Figure3.1:BIFDMA&LFDMAPerformanceonAWGNChannel WithrespecttoTurbo[Iterations=1] 58

Chapter3
10
-1

Simulations&Results

10

BER values

10

-2

Simulated BIFDMA Simulated LFDMA IDEAL AWGN BER
0 1 2 3

10

-3

Eb/No in dBs
4 5 6 7

10

-4

Figure3.2:BIFDMA&LFDMAPerformanceonRayleighFadingChannel WithrespecttoTurbo[Iterations=1] Figures 3.1 and 3.2 above show the performance of Turbo codes over AWGN and Fading channel respectively with 1% clipping. As expected the performance with AWGN channelisbetterthanwithfadingchannel.TheperformanceofbothBIFDMAandLFDMAon AWGN channel is similar with some deviations above approximately 4dBs. The reason is that both the schemes collect the same frequency diversity over AWGN channel with no fading, while the deviations are a result of the interuser interference caused by the non linearitieswhichwouldcausethepowerfromoneusertoleakintotheotheruser.Here1% clippingmeansclippingonly10samplesoutoftotal1024sampleswhichwillresultinvery smalldeviationsasseeninboththeabovefigures.ConsideringtheRayleighFadingchannel the BER is higher for LFDMA beyond 5dBs due to the increased frequency diversity collected by BIFDMA compared to LFDMA. Moreover BIFDMA is more robust against amplifier nonlinearities due to relatively small envelope fluctuations compared to the LFDMAscheme[10].

59

Chapter3

Simulations&Results

3.4 PerformanceevaluationofTurboCoding
In this section we analyze the performance of Turbo code over Rayleigh Fading ChannelinLinearenvironmentswithsingleuserandmultiusers.

3.4.1 SINGLEUSER
10
-1

B E R values
10
-2

10 10
10

-3

10
-4

-5

-6

Simulated BIFDMA Simulated IFDMA Simulated LFDMA BIFDMA-Uncoded IFDMA-Uncoded LFDMA-Uncoded IDEAL AWGN BER
4 5 6

7 Eb/No in dBs
8 9 10

Figure3.3:UncodedVsTurboCodedPerformanceonRayleighFadingChannel The figure 3.3 represents the BER performance of the three subcarrier mapping schemes with and without the Turbo code. The performance of the three schemes in uncodedcaseisalmostsimilarandmuchworsethanthecodedperformance.IntheTurbo coded case the BER performance of IFDMA is better in comparison to the other two schemes.ThisisduetothefactthatIFDMAcollectsmorefrequencydiversitycomparedto BIFDMA which in turn collects more frequency diversity than the LFDMA scheme. This is visibleinthefigureaboveasBERperformanceofIFDMAisfollowedbyBIFDMAandthen LFDMA.

60

Chapter3

Simulations&Results

3.4.2 MULTIUSERS(2,4,8) Case1:2Users:

10
-1

10

-2

BIFDMA-Uncoded IFDMA-Uncoded LFDMA-Uncoded BIFDMA-Coded IFDMA-Coded LFDMA-Coded IDEAL AWGN BER
4 5 6 7

BER values

10

-3

10

-4

8 9 10

10

-5

10

-6

Eb/No in dBs

Figure3.4:UncodedVsTurboCodedPerformancefor2Users

Case2:4Users: 10 10
BER values
-2 -1

10
10 10 10

-3

-4

-5

-6

Simulated BIFDMA Uncoded Simulated IFDMA Uncoded Simulated LFDMA Uncoded Simulated BIFDMA Coded Simulated IFDMA Coded Simulated LFDMA Coded IDEAL AWGN BER
4 5 6 7

Eb/No in dBs

10

Figure3.5:UncodedVsTurboCodedPerformancefor4User

61

Chapter3

Simulations&Results

Case3:8Users:
10
-1

10

-2

BER values

10

-3

10

-4

BIFDMA - Uncoded
IFDMA - Uncoded

10

-5

LFDMA - Uncoded BIFDMA - Coded

IDEAL AWGN BER IFDMA - Coded
5

10

-6

LFDMA - Coded
4 6 7

Eb/No in dBs

10

Figure3.6:UncodedVsTurboCodedPerformancefor8Users

InMultiUserLinearEnvironmentasweincreasethenumberofusersfromtwoto eight,thelongtermaverageBERperformanceresemblesthesingleuserBERperformance asshowninfigures3.4,3.5and3.6above.Thisisbecausewedonothaveanyinterference in a linear environment. In a nonlinear environment the deviation in performance, as we increase the number of users, is more visible due to effect of ICI and MAI. This will be discussedinsomemoredetaillateroninthereport.

62

Chapter3

Simulations&Results

3.5 PerformanceevaluationofDFTPrecoding
InthissectionweanalyzetheperformanceofDFTprecodedsystemcomparedtoa NonDFTprecodedsystemhavingasingleuserandmultiusersinlinearenvironments.

3.5.1 SINGLEUSER
10
-1

10
-5

10

-2

BER values

10

-3

10

-4

10

-6

BIFDMA-With DFT IFDMA-With DFT LFDMA-With DFT BIFDMA-Without DFT IFDMA-Without DFT LFDMA-Without DFT IDEAL AWGN BER
4 5 6 7

8 9 10

Eb/No in dBs

Figure3.7:DFTprecodedVsNonDFTprecodedperformanceforSingleUser The figures 3.73.10 show the BER performance analysis of the three subcarrier mapping schemes for a single user and multiple users with respect to DFT precoded and NonDFTprecodedsystem.InallthecasestheBERforDFTprecodedsystemisbetterthan the system without DFT precoding. The change in BER beyond approximately 7dBs with DFTprecodingcomparedwithNonDFTprecodingSystemisalmostsameforallthethree mapping schemes. The reason behind the lower BER for DFT precoded system is the spreading of each data symbol over multiple subcarriers that also causes the spreading gainorfrequencydiversitygaininafrequencyselectivechannel[12].ThedifferenceinBER duetoDFTprecodingismorevisibleathigherSNRs.

63

Chapter3

Simulations&Results

3.5.2 MULTIUSERS(2,4,8) Case1:2Users:

10
-1

10
-2

BER values

10

-3

10
-4

10

-5

10

-6

BIFDMA-With DFT IFDMA-With DFT LFDMA-With DFT BIFDMA-Without DFT IFDMA-Without DFT LFDMA-Without DFT IDEAL AWGN BER
4 5 6 7 8 9 10

Eb/No in dBs

Figure3.8:DFTprecodedVsNonDFTprecodedperformancefor2Users

Case2:4Users:
10
-1

10
-2

BER values

10
10

-3

10

-4

-5

10 4
-6

Simulated BIFDMA DFT Prec Simulated IFDMA DFT Prec Simulated LFDMA DFT Prec Simulated BIFDMA NDFT Simulated IFDMA NDFT Simulated LFDMA NDFT IDEAL AWGN BER
5 6 7 8 9 10

Eb/No in dBs

Figure3.9:DFTprecodedVsNonDFTprecodedperformancefor4Users 64

Chapter3

Simulations&Results

Case3:8Users: 10
BER values
10
-3 -1

10

-2

BIFDMA-With DFT IFDMA-With DFT LFDMA-With DFT BIFDMA-Without DFT IFDMA-Without DFT LFDMA-Without DFT IDEAL AWGN BER
4 5 6 7

10

-4

Eb/No in dBs

10

-5

10

-6

10

Figure3.10:DFTprecodedVsNonDFTprecodedperformancefor8Users It is thus clear from the figures above that as was the case in performance evaluationofTurbocoding,hereaswell,theBERperformanceforsingleuseristhesameas thelongtermaverageBERperformanceformultiusersystem.Againthiscanbemotivated bythelackofnonlinearitiesinthesystemwhichmakessurethatthereisnooutofband interference in the system. The Performance impact of DFT precoding in nonlinear environmentwillbediscussedlateroninthereport.

65

Chapter3

Simulations&Results

3.6 ModelingAmplifierNonLinearities

The results up to now (except figures 3.1 and 3.2) have been achieved in linear environmentswherethepowerineachsubcarrierandeachuser,incaseofmultipleusers, islimitedtoitselfanddoesnotleakintoothersubcarriersandusersrespectively.Beyond this point, nonlinearities are introduced in the system. This is done by employing a threshold clipping device which would clip the signal above certain predetermined threshold.Resultsincludeclippinglevelsof10%and30%.Itshouldbenotedthatclipping the signal will reduce the average energy of the clipped signal which would also result in some difference in BER performance due to the reduction in energy of clipped signal but thisisnotaninterestingresult.Sowepreventthisbyrescalingtheclippedsignaltokeep theaverageenergyoftheclippedsignalthesameastheaverageenergyoftheinputsignal totheclippingdevice.Thiswouldenableustoseetheeffectofonlytheinterferenceinthe BERcurves.

3.6.1INTERCARRIERINTERFERENCE(ICI) This section of the report deals with a single user simulation done in linear
environmentandinnonlinearenvironmentwithdifferentpercentageclippinglevelstosee theeffectofICI.
10
0

BER values

10

-1

10

-2

10
-6

10

-3

10

-4

10

-5

BIFDMA No Clip LFDMA No Clip BIFDMA 10% Clipping LFDMA 10% Clipping BIFDMA 30% Clipping LFDMA 30% Clipping IDEAL AWGN BER
4 5 6 7

Eb/No in dBs

10

Figure3.11:InterCarrierInterference(10%,30%andnoClipping) 66

Chapter3

Simulations&Results

Thefigure3.11indicatesthatasthepercentageclippinglevelisincreased,theBER performance becomes worse. This is expected as increasing the clipping level will cause increasednonlinearitiesandhenceincreasedpowerleakagebetweenadjacentsubcarriers i.e.ICI.RescalingensuresthatthevariationinBERwithincreasednonlinearitiesisonlydue to ICI and not due to any power variation of clipped signal with respect to the unclipped signal.Itisnoteworthyherethat10%clippingmeansclippingabout102samplesand30% clippingmeansclippingabout307samplesofthetotal1024samplesofSCFDMAsymbol. Moreover the figure above also shows that the BIFDMA performs better than the LFDMA whether we have 10% or 30% clipping. This can be explained as the out of band radiation, caused due to nonlinear distortions produced by nonlinear amplifier, is dependent on the envelope of the input signal to the amplifier [15]. The envelope fluctuationsofBIFDMAinturnareconsiderablylowerthanthoseofLFDMAwhenwehave no oversampling and no windowing of the signal [10]. Thus these lower envelope fluctuationsforBIFDMAmayaccountfortherelativelybetterBERperformanceofBIFDMA comparedtoLFDMAinnonlinearenvironments.

67

Chapter3

Simulations&Results

3.6.2 PERFORMANCEOFDFTANDNONDFTINNONLINEARENVIRONMENT
The Performance of DFT precoded system is still better even in nonlinear environmentascomparedtotheNonDFTprecodedsystem.Inthissectionweanalyzethe performance of BIFDMA and LFDMA by introducing nonlinearities using two different clipping levels (10% & 30%). The figure 3.12 shows the change in BER of BIFDMA and LFDMAat10%and30%clippinginpresenceofDFTandNonDFTprecoding.TheBERof theNonDFTprecodedsystemishigherthanthatofDFTprecodedsystemduetothefact that the absence of DFT precoding results in higher envelope fluctuations that would ultimatelycausehigheroutofbandinterference.FurthermoretheBERforBIFDMAislower as compared to LFDMA as it has comparatively lower envelope fluctuations [10]. Another reasonforthelowBERforDFTprecodedsystemisthespreadingofeachdatasymbolover multiplesubcarriersthusgivingfrequencydiversityasmentionedbefore.TheBERforthe twoschemesishigherfor30%clippingthanfor10%clippingduetomoresamplesbeing clippedwith30%clipping.
10
0

10
B E R v a lu e s
10

-1

-2

10 10
10
-6 -5 -4

10

-3

BIFDMA 10% Clipping With DFT LFDMA 10% Clipping With DFT BIFDMA 10% Clipping Without DFT LFDMA 10% Clipping Without DFT BIFDMA 30% Clipping With DFT LFDMA 30% Clipping With DFT BIFDMA 30% Clipping Without DFT LFDMA 30% Clipping Without DFT IDEAL AWGN BER
4

10

Eb/Noin dBs
Figure3.12:DFT&NonDFTprecodingSystemwith10%,30%Clipping

68

Chapter3

Simulations&Results

3.6.3 MULTIACCESSINTERFERENCE(MAI) Inthispartofthereportwediscussthemultiaccessinterferencebetween multipleusersinnonlinearenvironment.Testcaseswithtwo,fourandeightusers havebeenusedand30%clippingisappliedtointroducenonlinearities. 10

0

10
-1

BER values
10
-2

BIFDMA 30%Clipping 2Users BIFDMA 30%Clipping Single User
5 6 7 8 9 10

LFDMA 30%Clipping 2Users

10
-3

LFDMA 30% Clipping Single User

4

Eb/No in dBs

10

Figure3.13:MultiAccessInterferencewith2Users

10
BER values
-1

10
10
-3 -2

BIFDMA 30%Clipping 4Users LFDMA 30%Clipping 4Users BIFDMA 30%Clipping Single User LFDMA 30% Clipping Single User
4 5 6 7 8 9 10

Figure3.14:MultiAccessInterferencewith4Users

Eb/No in dBs

69

Chapter3

Simulations&Results

10
0

10
B E R values
-1

10
10
-3 -2

BIFDMA 30%Clipping 8Users LFDMA 30%Clipping 8Users BIFDMA 30%Clipping Single User LFDMA 30% Clipping Single User
4 5 6 7

Eb/No in dBs
8 9 10

Figure3.15:MultiAccessInterferencewith8Users

Thefigures3.13,3.14and3.15aboveclearlyindicatethatincreasingthenumberof users would worsen the BER performance for both BIFDMA and LFDMA. Again rescaling makes sure that any variation in the long term BER of multiple users relative to BER of single user is purely due to multiaccess interference. Increasing the number of users in multiuser nonlinear system should increase the multiaccess interference which in our case holds true but the variation in BER for different number of users and that for single userisnotthatevidentfromthefiguresabove.Moreoverthe degradationinBERismore noticeable at higher values of SNRs beyond approximately 8.5dBs. The multiaccess interference with 10% clipping for two, four and eight users is depicted in figures in appendixBofthisreport.

70

Chapter3

Simulations&Results

3.7 MultiAccessInterference(MAI)withNearFarEffect
10
0

10

-1

BER values

10

-2

10

-3

10

-4

10

-5

BIFDMA Affected User LFDMA Affected User BIFDMA Long term avg of rem users LFDMA Long term avg of rem users
4 5 6

10

Eb/No indBs

Figure3.16:MultiAccessInterferencewithNearFarEffect8Users Near far effect is a phenomenon that is introduced when one mobile user is more near to the base station than the others. In this situation the user that is near to the base station will be receiving higher power than the ones far from it. In figure 3.16, we have analyzedthenearfareffectinasystemhavingeightusers.Inthisscenariowehaveoneuser thatisfarawayfrombasestationwhiletherestofothersareneartoit.Theaffecteduser (fartherfromtheBS)hasbeenassigned5dBslesspowerascomparedtotherestofseven users while we have introduced nonlinearities in our system by clipping 30% of the samples in the signal. The seven users are rescaled with respect to the affected or weak user so as to eliminate the effect of power change due to clipping while analyzing BER of affecteduserduetothenearfareffect.

71

Chapter4 Conclusions&FutureWork

The chapter includes the summary in terms of our conclusions made from simulations and results. The possible future work related to our thesis is also discussedhere.

Chapter4

Conclusions&FutureWork

4.1

Conclusions
Inthispartofourreport,wehavedescribedourkeyfindings&conclusionsbased

uponthesimulationswedidinthisthesis.

TheperformanceofTurbocodeoverAWGNchannelismuchbetterascompared to the fading channel. Moreover owing to the fact that there is no frequency selectivity in AWGN channel both the BIFDMA and LFDMA collects same frequencydiversityandhenceBERcurvesforbothalmostfallovereachother. IncaseoffadingchannelhoweverBIFDMAperformsbetterthanLFDMAowing to its ability to collect more frequency diversity. Nonlinearities on the other hand cause slight deviations in the BER curves for the two schemes in both AWGN and fading channel due to out of band interference resulting from non linearities.

In linear environments, the BER performance of single user is identical to the long term BER performance of multiple users because of lack of out of band interferenceasisthecaseinnonlinearenvironments.IFDMAperformsthebest intermsofBERwhichisfollowedbyBIFDMAandthenLFDMA.Thisisbecause oftheabilityofIFDMAtoachievemorefrequencydiversitythantheothertwo schemes.

DFTprecodingismainlyintendedtokeepthePAPRofthesignaltoalowerlevel and hence to ensure efficient use of HPA. However DFT precoded system also shows an improvement in BER performance compared to non DFT precoded system owing to the frequency diversity gain in frequency selective channels causedbyspreadingofeachdatasymbolovermultiplesubcarriers.

BIFDMAandLFDMAperformdifferentlyintermsoftheirBERperformancesfor a single user, in linear environments and in nonlinear environments. The reasonforthisdifferenceinBERperformanceistheICIthatiscausedduetothe nonlinearities in the system. The ICI increases as the clipping percentage is increased that result in higher nonlinearities. This is evident from increased differenceintheBERcomparedtolinearcase.

In nonlinear environments, DFT precoded systems again have better BER performancecomparedtononDFTprecodedsystems.Thereasonforthisisthe reduced PAPR caused by DFT precoding which reduces not only envelope 73

Chapter4

Conclusions&FutureWork fluctuationsbutalsoreducesinfluenceofoutofbandradiationscausedbynon linearities.SimilarlyBIFDMAalsoperformsbettercomparedtoLFDMAinnon linear environments, again due to lower envelope fluctuations in BIFDMA comparedtoLFDMA.

In nonlinear environments with multiple users, the system suffers from MAI which degrades the BER performance of multiuser system compared to single usersystem.AsthenumberofusersisincreasedtheBERperformanceworsens duetoincreasedMAI.

Inthemultiusersystemwithnonlinearities,thenearfareffectisevidentwhen few users with high power (near to base station) creates interference for the usersthatarefarawayfromthebasestation,resultingincomparativelyhigher BERfortheeffectedusers.

4.2 FutureWork
Thisthesisresearchworkdoesnotincludepulseshapingduetowhichwehavent beenabletoassesstheperformanceofIFDMAinnonlinearenvironments.Infuturethe simulations can be carried out in presence of pulse shaping to have more realistic comparisons of the three mapping schemes in linear and nonlinear environments. Threshold clipping has been used to model the amplifier nonlinearities in this thesis workwhichcanbereplacedbymoresophisticatednonlinearity modelsliketheRapp model.Theimpactofcarrierfrequencyoffset(CFO)andDopplercanalsobeincludedto inthefutureresearchwork.Finallythechannelscenariobeingusedinthisthesiswork istheWINNERC2NLOS.Itcanbeaninterestingoptiontouseotherchannelscenarios.

74

AppendixA

AppendixA
SIMULATIONOFRAYLEIGHFADINGCHANNEL
In this Appendix we have described spectrum method by which we have simulated our channel.TherearetwomethodstosimulateRayleighfadingchannel,oneisFiltermethod andtheotheroneisSpectrummethod.Filtermethodismoreaccuratebutmorecomplex as compared to the spectrum method, this is due to the convolution process which is replacedbythemultiplicationinspectrummethod.

A.1SpectrumMethod
In the spectrum method, the complex gain of the channel c(t) is calculated in the frequencydomainthathasbeenshownintheequationbelow.

C ( f ) = X ( f )G ( f )

(1)

Where C(f)istheFouriertransformofthechannelgain c(t), G(f)isthefrequencyresponse ofthe filter while X(f)isthe Fourier transformoftheinputtothe filter x(t).Where x(t)is complexwideGaussianNoisewithunitpowerspectraldensity.Thesamplesofc(t)thatis c(nTs)isthencomputedasinverseFouriertransformofthesamplesofC(f)[6].

FigureA.1:Channelgainastheoutputofalinearfilter Inspectrummethod, Tsischosensuchthat, fs = 1 / Ts > 2 f D .Thismakessurethat no aliasing occurs. After that, G ( f ) is formed from G(f) as G ( f ) = G ( f ) + G ( f fs ) . Then Nssamplesof G ( f ) areformedas G (kfs / Ns ) where k=0,1,.....,Ns1.Finallyc(nTs) is computedasan NspointinverseFFTof G (kfs / Ns ) Z(k),where Z(k)fork=o,1,...., Ns1
^

75

AppendixA

are independent, zero mean, complex Gaussian random variables. The variance of Z(k) is chosensuchthatc(nTs)willhaveunitvariance[6].

A.2Time&FrequencyVaryingRayleighFadingChannel
Inthissectionwehaveexplainedhowtosimulateatimefrequencyvaryingchannel using the complex channel gains obtained from the spectrum method. A discretetime modelofsuchachannelwithLtapsandsampleintervalTsisshowninfigurebelow, FigureA.2:Discretetimemodelofatimeandfrequencyvaryingchannel The C l ( nTs) represents the tap gains for each channel tap. The tap gains would amplitude scale any signal passing through any particular channel tap. The time and frequencyvaryingresponse, C(f,nTs)istheamplitudescalingacomplexexponentialwith frequencyfexperiences,whenitistransmittedoverthechannel[6].Sowehavethat

C ( f , nTs ) =

c (nTs ) exp( j 2f T ) (2)

l l s l =1

Where l isthetapdelayforeachchanneltap.Forfrequenciesf=k/(NTs)where k=0,1,. ....,N1.Theequation2canbesimplifiedas

C (k /( NTs ), nTs ) =

c (nTs) exp( j 2 k / N )
l l l =1

= C[k , n]

(3)

76

AppendixA ThetimefrequencyvaryingresponseforWINNERC2NLOSwith1000timesamples

and128Frequencysamplesisshowninthefigurebelow, FigureA.3:TimeandFrequencyResponseofWINNERC2NLOSChannel

77

AppendixB

AppendixB
AdditionalResults
10
-1

10
-2

10 10 10

BER values

-3

-4

-5

BIFDMA 10%Clipping 2Users LFDMA 10%Clipping 2Users BIFDMA 10%Clipping Single User LFDMA 10% Clipping Single User
4 5 6 7 8 9 10

Eb/No in dBs

FigureB.1:MultiAccessInterferencewith2Users(10%Clipping)
10
-1

10

-2

BER values

10

-3

10

-4

10

-5

BIFDMA -10%Clip-4Users LFDMA -10%Clip-4Users BIFDMA -10%Clip-Single User LFDMA -10%Clip-Single User
4 5 6 7 8 9 10

FigureB.2:MultiAccessInterferencewith4Users(10%Clipping) 78

Eb/No in dBs

AppendixB

10

10 10 10

-2

B E R values

-3

-4

10

-5

BIFDMA 10%Clipping 8Users LFDMA 10%Clipping 8Users BIFDMA 10%Clipping Single User LFDMA 10% Clipping Single User
4 5 6 7

Eb/No in dBs
8 9 10

FigureB.3:MultiAccessInterferencewith8Users(10%Clipping)

79

References

[1]Goldsmith,Andrea.WirelessCommunications.NewYork:CambridgeUniversity Press;2005. [2]DahlmanE,ParkvallS,SkldJ,BemingP.3GEvolutionHSPAandLTEforMobile Broadband.London:AcademicPressInc;2007. [3]MadhowUpamanyu.FundamentalsofDigitalCommunications.NewYork:Cambridge UniversityPress;2008. [4]PuckettW.Bruce.ImplementationandPerformanceofanImprovedTurboDecoderona ConfigurableComputingMachine.MastersThesisReport.Virginia:VirginiaPolytechnic InstituteandStateUniversity;2000. [5]RehmanMujeebUr,JlmbrantUlf,AhmedAmeer.ChannelCoding.GroupProjectreport. Project2.DigitalcommunicationCourse.Gteborg:ChalmersUniversity;2008. [6]StrmErik,SeifiNimaSimulationoffadingwirelesschannel.Technicalreport.Project1. WirelessCommunicationCourse.Gteborg:ChalmersUniversity;2009. [7]RamanaD.Venkata,PalSurendra,PrasadA.P.Shiva.Simulationandrealizationof BasebandpulseshapingfilterforBPSKmodulator;2004. [8]HsuHweiP.Schaum'soutlineoftheoryandproblemsofanaloganddigital Communications.Montville:McGrawHillCompanies;2003. [9]SrikarLeelaM,KarroutJohnny.UplinkMultipleAccessForIMTAdvanced.Masters ThesisReport.Gteborg:ChalmersUniversity;2009. [10]FrankTobias,KleinAnja,HausteinThomas.ASurveyontheEnvelopeFluctuationsof DFTPrecodedOFDMASignals.InternationalConferenceonCommunicationsICC. Beijing;2008. [11]AstlyDavid,DahlmanErik,FuruskrAnders,JadingYlva,LindstrmMagnus,Parkvall Stefan.LTE:TheEvolutionofMobileBroadband.IEEECommunicationsMagazine. EricssonResearch;2009. [12]SCFDMASingleCarrierFDMAinLTE.IXIAWorldwideHeadquarters.Calabasas,CA; 2009. [13]ScheimJacob.SCFDMAfor3GPPLTE.CommunicationandSignalProcessing Limited;2008.

References

80

References

[14]SvenssonT,FrankT,FalconerD,SternadM,CostaE,KleinA.BIFDMAAPower EfficientMultipleAccessSchemeforNonFrequencyAdaptiveTransmission.Mobile& WirelessCommunicationsSummit.Budapest;2007. [15]SvenssonT,FrankT,ErikssonT,AronssonD,SternadM,KlienA.BlockInterleaved FrequencyDivisionMultipleAccessforPowerEfficiency,Robustness,Flexibility,and Scalability.EURASIPjournalonWirelessCommunicationsandNetworking.Hindawi PublishingCorporation;2009. [16]MyungHyungG.SinglecarrierFDMA.PresentationSlides; 2008. (hgmyung.googlepages.com/scfdma.pdf) [17]MachPavel,BekRobert.ImplementationofOFDMintoBroadbandWireless Networks.DepartmentofTelecommunicationEngineering.Prague:CzechTechnical UniversityinPrague. [18]SariHikmat.OFDMPeakPowerReductionTechniquesPerformanceAnalysisfor WiMAXSystems.PresentationSlides.4thAnnualWirelessBroadbandForum. Cambridge;2005. [19]RappChristoph.EffectsofHPAnonlinearityona4DPSK/OFDMsignalforadigital soundbroadcastingsystem.2ndEuropeanConferenceonSatelliteCommunications. Liege;1991. [20]ChaniotakisM,CoryD.Transistors:BipolarJunctionTransistors(BJT);2006. [21]MaXiaoqiang,KobayashiH,SchwartzStuartC,ZhangJ,GuDaqing.IterativeChannel estimationforOFDMwithClipping.5thInternationalSymposiumonWireless PersonalMultimediaCommunications.Hawaii;2002. [22]SvenssonTommy.ULPowerControl,EfficiencyandPerformanceAssessmentCriteria. CELTICWirelessWorldInitiativeNewRadio+(WINNER+)Seminar.Gteborg;2007. [23]R4060867.UplinkPowerControlforEUTRA.Tallinn;2006. [24]MuhammadBilal.ClosedlooppowercontrolforLTEuplink.MastersThesisReport. Karlskrona:BlekingeTekniskaHogskola;2008. [25]SkillermarkPer.UplinkPowerControl.CELTICWirelessWorldInitiativeNewRadio+. [26]QuinteroNestorJ.AdvancedPowerControlforUTRANLTEUplink.MastersThesis Report.Aalborg:AalborgUniversity;2008.

81