averagewcdma联通

EURASIPJournalonAppliedSignalProcessing2005:11,1725–1735c2005BoHagermanetal.
WCDMAUplinkParallelInterferenceCancellation—SystemSimulationsandPrototypeFieldTrialsBoHagermanEricssonResearch,EricssonAB,16480Stockholm,SwedenEmail:bo.
hagerman@ericsson.
comFredrikGunnarssonEricssonResearch,EricssonAB,58117Link¨oping,SwedenEmail:fredrik.
gunnarsson@ericsson.
comHideshiMuraiNipponEricssonK.
K.
,Tokyo112-0004,JapanEmail:hideshi.
murai@ericsson.
comMiokoTadenumaNipponEricssonK.
K.
,Tokyo112-0004,JapanEmail:mioko.
tadenuma@ericsson.
comJonasKarlssonEricssonResearch,EricssonAB,16480Stockholm,SwedenEmail:jonas.
b.
karlsson@ericsson.
comReceived1March2004;Revised20September2004Interferencecancellation(IC)isoneidentiedkeytechnologytoenhanceWCDMAuplinkperformance.
Thegoalofthiscontri-butionistohighlighttherelativeuplinksystemcapacityimprovementavailableforWCDMA,especiallyinrealistictypicalurbanradioenvironmentswhenemployingreceiverimplementationsincludingrealisticchannelestimation,searcher,andsoforth.
Theperformanceoftheselectedlimited-complexityparallelICreceiverisrstevaluatedwithlink-levelsimulationsinordertopro-videinputtosystem-levelsimulations.
Thesystem-levelmethodologyisexplainedanda40%system-leveluplinkcapacityincreasecomparedtoutilizingtheconventionalRAKEreceiverisfound.
Thelimited-complexityparallelICreceiveristhenevaluatedinasingle-celleldtrial.
Thetrialsshowthatboththemeanandthevarianceoftheouter-looppowercontrolisreduced,whichimpliesanoverallincreasedcapacityandanincreasedbatterylifeoftheterminals.
Furthermore,theobservedcapacitygainsareinaccordancewithsystemsimulations.
Keywordsandphrases:CDMA,eldtrials,interferencecancellation,link-levelsimulations,system-levelsimulations.
1.
INTRODUCTIONInterferencecanceling(IC)isregardedasoneofthekeytechnologiesforenhancingCDMAuplinkperformance.
ThegeneralinterestinICstartedafterVerdupublishedhisThisisanopenaccessarticledistributedundertheCreativeCommonsAttributionLicense,whichpermitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycited.
resultsontheoptimalreceiverforGaussianmultiple-accesschannels[1].
Severalstudieshavebeenpublishedonsuboptimalandlesscomplexreceivers,forexample,aclassoflinearreceivers[2],successivecancellationschemes[3],andparticularlyonparallelIC(PIC)schemes[4,5,6,7,8,9].
However,thesestudiesaremainlyfocusedonlinkperformanceorusinganidealanalyticalsystemapproach.
Realisticchannelmodels[10]anddynamicbehaviorinasystemenvironmenthave1726EURASIPJournalonAppliedSignalProcessingReceivedsignal0000ICUICUICUICUrep(1)1rep(1)2rep(1)3rep(1)4ICUICUICUICUrep(2)1rep(2)2rep(2)3rep(2)4ICUICUICUICUSymbolSymbolSymbolSymbolFigure1:PIClayout.
ReceivedsignalChannelestimationRAKETentativedecisionRespreadWeightrep(n+1)kSymbolrep(n)j(j=k)+Figure2:ICUlayout.
beenlessthoroughlycoveredintheliterature.
Thesetypesofcomponentsareimportanttomodel,analyze,andunder-standhowtoutilizeICinsystemstoensuretheUL/DLca-pacitybalanceandimprovetherobustnessofrealisticsystemimplementations.
Thispaperinvestigatessystem-levelperformanceofin-terferencecancellationfortheWCDMAuplink[11]bymeansofsimulationsandmeasurementsfromaprototypePICimplementation.
TheprototypePICtestsystemdevel-opmentandtheeldtrialwereperformedincollaborationbetweenEricsson,ChinaAcademyofTelecommunicationsTechnology(CATT),andDatangTelecomTechnologyCo.
Ltd.
Theeldmeasurementswereperformedatthenorth-westpartofurbanBeijing,China.
Thelink-levelsimulatoremploysarealisticCOST259channelmodel[10]andacompleteWCDMAreceiverin-cludingsearcher,channelestimation,coding,andsoforthinadditiontoICfunctionality.
Thesystem-levelsimulatormodelsuserdynamicssuchaspowercontrol,mobility,softhandover,andsoforth.
Thetestsystemutilizesaproto-typePICmultiuserreceiverimplementationintegratedintoacommercialEricssonRBS3202WCDMAradiobasestation.
ThePICsimulatormodelhasbeencarefullydesignedtobeequivalenttotheprototypeimplementation.
Thispaperhasbeenorganizedasfollows.
Firstanalgo-rithmdescriptionsegmentwithPICalgorithmdetailsandadescriptionofICinteractionwithamulticellsystemarepresented.
Then,thesimulationworkowisdescribedasaniterativeprocedureoflinkandsystemsimulations,togetherwithsimulatedcapacityresults.
Finally,testsystemdetailsandeldmeasurementsprerequisitesarepresented,andthetrialsystemperformanceisevaluated,followedbysomecon-clusiveremarks.
2.
PARALLELICALGORITHMAPICconsistsmainlyofseveralcascadeddetectionunits(e.
g.
,RAKEreceivers)foreachuser,seeFigure1,whereeachdetectionunitafterdetectionregeneratesareplicaofthesignalbasedonthedetectedsymbols,estimatedchan-nelresponses,andtheuser'sspreadingcodes.
Thesedetec-tionunits,denotedbyinterferencecancellationunits(ICUs)showninFigure2,receiveasinputalltheotheractiveusers'signalreplicasfromthepreviousstageinthecascadedchainofICUs.
AsisvisibleinFigure1,theoriginaltotalreceivedsignalisalsooneinputtotheICU.
WithintheICU,thereplicasaresubtractedfromtheorig-inaltotalreceivedsignalandatentativesymboldecisionisformedusingastandardRAKEreceiverandchannelre-sponseestimator.
ThechannelresponseestimatoraveragesthepilotsymbolsfromtwoconsecutiveslotsoftheWCDMAuplinksignalinordertoformanestimate.
TheICUout-putconsistsofthetentativesymboldecisiontogetherwithaweightedformofthereplicasignal.
ThreecascadedICUstageshavebeenselectedforthePICalgorithminvestigatedinthispaperaswellasfortheproto-typeimplementation.
Theweightingfactorutilizedforeachstageshouldreectthecondenceinthetentativedecision[12,13].
Noreplicaisgeneratedinthethirdstage,andthere-forethereisnoweightingfactor.
A2Dsearch,usinglink-levelWCDMAULPIC:SystemSimulationsandPrototypeFieldTrials172721*102BLER0.
20.
30.
40.
50.
60.
70.
8Weightingfactor1(a)21*102BLER0.
20.
30.
40.
50.
60.
70.
8Weightingfactor2(b)Figure3:Twoslicesofthe2Dsearchforoptimalweightingfactorsarepresented.
(a)showstheblockerrorrate(BLER)asfunctionofweightingfactor1whenweightingfactor2issetto0.
625.
(b)showstheBLERasafunctionofweightingfactor2whenweightingfactor1issetto0.
375.
simulations,wasdoneinordertondtheoptimalweightingfactors.
Theoptimalweightingfactorsfortherstandsec-ondstageswerefoundtobe0.
375and0.
625,respectively;seeFigure3.
3.
PICINTERACTIONWITHSYSTEMPERFORMANCEAlluserswill,dependentontheirservice,haveaqual-ityrequirementonthededicatedcommunicationlink.
InWCDMA,thecontrolmechanismtoensurethefulllmentofthequalityrequirementismainlypowercontrol(PC).
TheWCDMAPCisperformedontwolevels,theinner-loopandouter-loopPC.
Theinner-loopPCoperatesat1500Hztofol-lowfastchannelvariations.
Theouter-loopPCevaluatestheservicequalityonhigherlayer(i.
e.
,onalongertimescale)andsetsthetargetfortheinner-loopPCaccordingly.
Inamultiple-celldeployment,interferenceisgeneratedbothfromusersinsurroundingcells(intercellinterference)aswellasfromusersintheowncell(intracellinterference).
ThePICalgorithmisofcourseeectiveonlytowardsintra-cellinterference.
Thus,thePICwillreducetheimpactoftheintracellinterferencecomparedtoaconventionalRAKEre-ceiverandtherebyviathePCreducetherequiredoutputpoweroftheUEstomaintaintheservicequality.
Thetypicaluplinkradiointerfaceloadmeasureisthenoiserise,thatis,theratiobetweentotalreceivedwidebandpoweratthebasestationPandthenoisepowerN.
ApopularmodelistointroducethefractionalloadL[14]asPN=11L=11M/Mp,(1)whereMisthenumberofusersandMpisthepolecapac-ity—essentiallytheupperlimitoftheusersthenetworkcanaccommodate.
Inthesingle-cellcase,itisstraightforwardtoderiveanexpressionforL.
ThereceivedwidebandpowercanbeexpressedasP=Mm=1pmgm+N,(2)wheremdenotesthemthuserequipment(UE)inthecell,pmistheUEpower,andgmthepowergainfromUEtothebasestation.
Furthermore,thecarrier-to-interferencera-tio,CIR(denedhereasdedicatedphysicalcontrolchannel(DPCCH)receivedsignalcodepowerrelativetothereceivedwidebandpower),isgivenbyγm=pmgmPpmgm=Pγm.
(3)Hence,(2)and(3)yieldP=PMm=1γm+NPN=11Mm=1γm,(4)whichprovidesthefractionalloadLin(1),seealso[15].
Thededicatedphysicaldatachannel(DPDCH)istransmittedataxedpowerosetβtotheDPCCHpower.
Therefore,boththesechannelscontributetothefractionalload:L=Mm=1γm(1+β).
(5)Theinterferencecancellationcapacitygaincanbedescribedbycomparingtheaveragefractionalloadforthesamenum-berofUEs,M,usingeitherRAKEorPIC.
Utilizingtheaver-ageloadforthetwodierentreceiverschemes,LRAKE=MMRAKEP,LPIC=MMPICP,(6)theaveragepolecapacitygainisrelatedvia(6)bytheaverageloadforequalnumberofUE'sasMPICP=LRAKELPICMRAKEP.
(7)Figure4illustratestherelationbetweenthenumberofusersandnoiseraiseonaverageforbothconventionalRAKEandPIC.
Furthermore,itisemphasizedhowtheperfor-mancegaincanbothbeseenasacapacityimprovementgivenaxednoiserise,oracoverageimprovementduetoalowernoiserisewithPICgivenaxednumberofusers.
1728EURASIPJournalonAppliedSignalProcessing12111098765432100MRAKEpMPICpNo.
ofusersNoiserise(dB)RAKEPIC(0.
75Mp,6dB)CapacitygainCoveragegainFigure4:IllustrationoftherelationbetweennoiseriseandnumberofusersforconventionalRAKEandPIC.
4.
PARALLELICSIMULATIONMODELINGAdynamicsystemsimulatorisusedtoevaluatesystemcapac-ityforadenedoutageprobabilitylevel.
Anoutageoccurswhenausers'averageservicequalityrequirementisnotsup-ported.
Theoutageprobabilityisthereforeestimatedasthefractionofunsatisedusersrelativetothetotalnumberofusersintheentiresimulation.
Thesystemsimulatormodelsthepropagationenvironment,mobility,andtracservices.
Radioresourcecontrolalgorithmssuchaspowercontrol,softhandover,cellselection,andsofortharealsoincluded[16].
ThePICfunctionalityismodeledasanintracellinterferenceadjustmentreectingtheeectivereceivedCIRasγm,b=pm,bi=mRIFi,b·pi,b+i,j=bpi,j+N,(8)whereγm,b,pm,b,andNaretheCIRforthemthuserinthebthcell,thereceivedpowerfromthemthuserinthebthcell,andthebackgroundnoisepower(AWGN),respectively.
Theparameterresidualinterferencefactor(RIFm,b)denotesthera-tiobetweentheequivalentintracellinterferenceforthemthuserinthebthcellafterandbeforePICexecutionandreectsthePICperformance.
RIFm,bisdenedaccordingtoRIFm,b=LFl=1hm,b,l·dm,bhm,b,l·dm,b2LFl=1hm,b,l·dm,b2,(9)wherehm,b,l,dm,b,hm,b,l,dm,b,andLFarethechannelcoef-cientofthelthpath,thedatasymbol,theestimatedchan-nelcoecient,theestimateddatasymbol,andthenumberofchanneldelaypaths,respectively.
Inthesequel,thesequanti-tieswillbediscussedgenerically,andtheindividual-userin-dicesmandbareomittedforclarity.
Thetypicalsystemsimulationworkowistorstrunde-tailedlinksimulationstodeterminerelationsbetweenCIRandblockerrorrate(BLER)statistics,andthenusethesemodelsinsystemsimulationstoobtainthesystemcapacity.
Single-usersimulationMultiusersimulationSimulationformulticellenvironmentSimulationforcapacityYesNoEstimatedcapacityConvergedsystemoperationpointResidualinterferencemappingInter-to-intracellinterferenceCIRvs.
servicequalityLinklevelSystemlevelFigure5:WorkowforPICsystemevaluation.
Thedivisionbetweenlink-andsystem-levelsimulationsistheselectedapproachtoovercometheoverwhelmingcom-plexitytosimulatealltheradiocommunicationlinkdetailswithinamultiuserandmulticelldynamicsystemenviron-ment,especiallyunderrealisticconditionswithrealisticalgo-rithms.
However,inthiscase,linksimulationstoobtainCIRandRIFmappingsdependontheintercellinterferencesitu-ation,whichisobtainedfromsystemsimulations.
Thesys-temsimulationsinturndependonRIFtoCIRmappings.
Therefore,aniterativeworkowasillustratedbyFigure5isadopted,whereiteratedlinkandsystemsimulationsareper-formedtoconvergetothecorrectoperationpointofboththereceiverperformanceandtheimpactthrougheectiveinter-ferencebetweenconnections.
ThiswillsecurethePICsysteminteraction.
4.
1.
Link-levelsimulationmodelingThedetailedlink-levelsimulatorcanevaluatePICperfor-manceunderrealisticconditions.
Multipleusersaresimu-latedinasinglecell,andintercellinterferenceismodeledwithanintercell-to-intracellinterferenceratioF.
ThemainpurposeistoprovideaservicequalityandRIFmappingtotheCIR.
RIFisamultidimensionalfunctionofchannelen-vironment,F(intercell-to-intracellinterferenceratio)value,targetquality,numberofactiveusers,andPICparameterssuchasweightingfactors,DPCCH/DPDCHpowerratio,andsoforth.
Thelink-levelsimulatorcalculatestheRIFasin(9).
Realisticpropagationchannelconditionshavebeensim-ulatedusingtheCOST259channelmodel[10],usingthesettingscorrespondingtoatypicalurbanenvironment.
TheCOST259channelmodelconsiderslocation-dependentchannelvariationssuchthatausermovingcontinuouslywithinacellwillexperiencedierentchannelconditions—pathloss,shadowfading,timedispersion,fastfadingstatis-tics,andsoforth—indierentpartsofthecell.
EachofthesechannelcharacteristicsisdescribedbyphysicalparametersWCDMAULPIC:SystemSimulationsandPrototypeFieldTrials1729Table1:KeycharacteristicsoftheCOST259channelmodel.
CharacteristicModelapproachTypicalbehaviorPathlossCOST231Walsch-Ikegamimodelfornon-line-of-sight40log(d)fornon-line-of-sightDistance-dependentprobabilityforline-of-sight20log(d)forline-of-sightClusteringofmultipathcomponentsMultipleclusterscausedbyreectionsfromdistantbuildingsaremodeled13%probabilityofmorethanoneclusterShadowfadingLognormal,exponentialautocorrelationfunction6dBFastfadingRayleighfadingofchanneltaps,rsttapmaybeRicefadingClassicalDopplerspectrum;higherRicefactorsinline-of-sightTimedispersionExponentiallydecayingpower-delayprole;delayspreadismodeledbyalognormaldistribution0.
2–1.
0sRMSdelayspreadDiversityAngularspreadandpolarizationcross-couplinggiveaveragepowerratioandcorrelationLowcross-correlation,equalpowerondiversitybranches100101102103104BLER23.
52322.
52221.
52120.
520TargetCIR(dB)UE=1UE=25UE=50F=0.
59F=0Figure6:Link-levelsimulationsofPICperformance.
Multipleusersaresimulatedinasinglecellwithintercellinterferencemod-eledasAWGNwiththepowerFtimestheintracellinterferencepower.
NotethatUE=1correspondstoconventionalRAKEper-formanceinasinglecell.
thataremodeledusingstatisticaldistributions.
Correlationsbetweenlinkstodierentbasestationshavebeenaddedtothemodelforthesystemsimulations.
FurtherdetailsonthecharacteristicsofthemodelcanbefoundinTable1.
Link-levelsimulationsareconductedfollowingthe3GPPspecication[11]regardingphysicalparametersandproto-cols.
TheULclosed-looppowercontrolmodelisCIR-basedwitha1dBstepsize,butincluding1-slotfeedbackdelaywithoutfeedbackerrors.
Theradiobasestation(RBS)re-ceiverisconguredwitha2-branchantennadiversity.
ThePICsimulatormodelhasalsobeencarefullyde-signedtobeequivalenttotheprototypeimplementation(i.
e.
,functionalitylikepathsearcher,channelestimation,andsofortharemodeledrealistically).
TheintercellinterferenceisemulatedbyadditivewhiteGaussiannoisewithapowerlevelaccordingtotheFvalue.
Forlink-levelperformanceex-amplesseeFigure6and[17].
Thepresentedlink-levelperformanceresultsinFigure6aresimulatedinasingle-cellenvironmentwithanF-factortomodelintercellinterference.
Itrepresentsexamplesofsimu-lationresultsfrombothsingle-andmultiple-userlink-levelsimulations,whichwerepointedoutinthesimulationwork-owinFigure5.
InFigure6,exempliedforanAMR(adap-tivemultirate)speechuserscenario,theBLERperformanceisshownasafunctionofthepowercontroltargetCIR,thatis,theouter-looppowercontroltargetforthefastinner-loopPChandlingthechannelvariations.
ObservethatthetargetCIRrequirementforPICdecreaseswhentheintracellinterferencetobackgroundnoiseratioincreases,thatis,inamultiple-userenvironment.
ThiscanbeseeninFigure6whenthenumberofusersincreasesfrom1to50withanequivalentincreasingFvalueof0and0.
59,respectively.
Notethatthesingle-usercasewithPICisequivalenttousingaconventionalRAKEreceiver.
Theoutputfromeachbatchoflink-levelsimulationsisaRIFtoCIRmapping,attheopera-tionpointsofinterest,asexempliedinFigure7a.
4.
2.
System-levelsimulationmodelingSystem-levelsimulationsareperformedwithinacellularstructureconguredwith3cellsitesdeploymentwithcellradiusof1km.
Toavoidsystembordereects,amethodof7-siteclusterwraparoundtechniqueisutilized.
Thesimula-tionsareusingasingleWCDMAfrequencycarrier,andthesofthandoverfunctionalityislimitedtoanactivesetequalto3(i.
e.
,connectionshavemaximumthreelegs).
TheUEtransmitpowerdynamicsisintheregion44to24dBm.
Axednumberofusersareconnectedpersimulationsetup,andnoloadcontrolsuchasadmission/congestioncontrolisactivated.
TheRIFmappingmodelsPICperformanceinthesystemsimulationstodeterminetheequivalentCIRafterinterfer-encecancellation.
TraditionalCIRtoblockerrorprobability(BLEP)models(see,e.
g.
,[18])arethenusedtodetermine1730EURASIPJournalonAppliedSignalProcessing10.
80.
60.
40.
2030282624222018161412InstantaneousCIR(dB)RIF(a)0.
120.
10.
080.
060.
040.
0200.
500.
511.
52F(Iintercell/Iintracell)Probability(b)Figure7:(a)RIFmappingand(b)Fdistributionoverallcells.
transportblockerrors.
(NotethatBLEPisanerrorproba-bility,whileBLERistheerrorrateoftheactualrealization.
)Outer-looppowercontrolactsontheblockerrorsequencestodetermineanappropriatetargetCIRforinner-looppowercontroltomeettheBLERrequirements[11,TS25.
214].
Forexample,therequirementoftheconsideredAMRspeechserviceisBLER1%.
Themainoutputfromabatchofsys-temsimulationsisanFvaluedistributionasexempliedinFigure7b.
ThesimulationworkowiterationsinFigure5arehaltedwhentheRIFmappingandtheFdistributionhaveconverged,andthesystemcapacitycanbeevaluated.
5.
SYSTEMPERFORMANCEEXAMPLETheuplinksystemperformanceisstudiedfromtheperspec-tiveofthecapacitythatoneWCDMAcarriercansupportforeachcellinthedeployment.
Thesystemcapacityisdenedforacertainsystemqualitylevel,thatis,outageprobabil-ity.
IntheperformanceexamplesshowninFigure8,allusersutilizeAMR(12.
2kbps)speechserviceinatypicalurban(COST259)environmentandslowlymovearound.
Themo-bilitymodelisarandomwalkwithameanvelocityof3km/h.
AnAMRspeechuserisconsideredunsatised(isinoutage)whentherequiredaverageservicequalityofmaximum102BLERisnotsatised.
ReasonsbehindoutagescouldtypicallybeinsucientUEpowertoovercometheuplinkinterfer-ence.
Anoutageprobability(fractionofunsatisedusers)at5%isconsideredacceptable,andthecorrespondingnumberofusersinthesystemisthereforethesystemcapacity.
ThesystemcapacityinFigure8isnormalizedtothesystemca-pacitywhentheconventionalRAKEreceiverwithrealisticchannelestimatesisusedinthesystem(Figures8aand8bleftmostcurve).
Usingtheabovedescribedqualitymeasures,itcanbeconcludedfromFigure8bthatPICissupportingapproxi-matelya40%uplinksystemcapacityincreasecomparedtoutilizingtheconventionalRAKEreceiverinthissimulatedrealisticmultiple-celltypicalurbanradioenvironment.
NotethatthePICiscarefullymodeledinthesimulationsreectingthelimited-complexityimplementationofrealisticsearcher,channelestimation,andsoforth.
Inanequivalentsingle-cellenvironmentsetup,therealisticreceiverimplementationshowsthatPICsupportsanimprovementofthecapacityintheorderof70%asindicatedbyFigure8a.
Theanticipatedgainincapacityisexpectedtodecreaseforhigh-speedusers[6].
Asanupperboundregardedasreferenceofthetechnol-ogylimit,theoutageperformanceofanidealinterferencecanceller(idealIC)isalsoshowninFigure8b(rightmostcurve).
AnidealICreceiverremovesperfectlyallintracellin-terference,equivalenttoaRIFequaltozerointheCIRmodelin(8).
FortheidealIC,thecapacityincreaseishenceapprox-imately180%inthemultiple-cellnetworkenvironmentwithrealisticchannelestimates.
6.
SINGLE-CELLSYSTEMEVALUATIONInacooperativeprojectbetweenEricsson,ChinaAcademyofTelecommunicationsTechnology(CATT),andDatangTele-comTechnologyCo.
Ltd.
,aPICtestsystemwasdevelopedandaeldtrialwasperformedatthenorthwestpartofurbanBeijing,China.
Theon-aireldtrialwasperformedduringtheperiodfromDecember2002toMay2003.
Thetrialsys-temconceptwasbasedonaradionetworkcontroller(RNC)simulatorconnectedtoamodiedcommercialEricssonRBS3202WCDMAradiobasestationprovidingPICmultiuserreceiverfunctionality.
ThenonoptimizedimplementationofthePICdemodulatorhasabout5timesthecomplexitycom-paredtoaRAKEdemodulator.
However,notethatthisiscomparabletoatotalreceivercomplexity,basebandpart,in-creaseofslightlylessthan2times.
PicturesofthemodiedradiobasestationhardwareemployedforthetestscanbefoundinFigure9andamapofthemeasurementareainFigure10.
WCDMAULPIC:SystemSimulationsandPrototypeFieldTrials1731100101102103OutageProb.
0.
511.
522.
533.
5NormalizedcapacityRAKEPIC(a)100101102103OutageProb.
0.
511.
522.
533.
5NormalizedcapacityRAKEPICIdealIC(b)Figure8:SimulatedsystemperformancerelativetotheperformancewithaconventionalRAKEfor(a)asingle-cellscenarioand(b)amultiple-cellscenario.
Figure9:ModiedEricssonRBS3202hardwareintegratingmul-tiuserPICfunctionality.
Figure10:DenedmeasurementareaswithinRBScellcoverage.
TheradiobasestationswereduringallpartsofthetestsfullyintegratedintotheWCDMARANinfrastructure(i.
e.
,connectedtoanRNCsimulatorusingthe3GPPNBAP,NodeBApplicationProtocol)supportingalllayers2and3func-tionalitywithforexamplesysteminformation(BCCH),pag-ingandcallsetuphandling,aswellastheouter-looppowercontrol.
Theeldtrialwasperformedinasingle-cellsystemenvironment(Figure10),whichimpliesthatnohandoverfunctionalitywasactivatedduringthetrial.
Theusedpro-totypehandsetUEsweresupporting3GPPbaseline,Release99December2000,withAMRspeechand64kbpsUDIdata[11].
Figure11:AntennainstallationnorthwestBeijing,China.
TheeldtrialmeasurementmethodologyisbaseduponthefactthatUEsmove/drivearoundinaforeachUEcon-nedmeasurementarea;seeFigure10.
TheUEswillfollowthenormaltracowonapredenedroute(i.
e.
,mimicnormal-useroperation/behavior).
AsexempliedinFigure10,withinthetestsystemcov-erage,asetofconnedmeasurementareaswasdenedwithasizeintherange50500m*50500m.
Ineachcon-nedarea,ameasurementroutewasdenedwheretheim-portantparametersaretheradioenvironment,angulardi-rection,anddistance.
Thedierentconnedareasandroutesareinsomecasesoverlapping,andinaddition,multipleUEscanbeallowedwithinthesameareaduringatest.
Theradiobasestationantenna(17dBigain)wasin-stalledattherooftopofathree-oorbuilding(seeFigure11)innorthwestBeijingatapproximately25mheight.
The65-degreeantennaboresightwaswest-north-westwithase-lectedsectorradiiof3kmoverlookinganurbanareain-cludingopenareasandhousesofbothequivalentheightastheinstallationaswellasverymuchhigherbuildings.
TheradioenvironmentinthetestsystemcoverageareaisexempliedinFigure12,includingmeasurementsfromtheconnedareasnumber2and3indicatedinFigure10.
ShowninFigure12aretheimpulseresponsesasafunctionoftimewhenmovingonthededicatedrouteinthetwoar-eas.
Inconjunctiontotheimpulseresponses,themeasured1732EURASIPJournalonAppliedSignalProcessing1086420Timedelay(s)020406080100120Time(s)(a)121086420Timedelay(s)050100150Time(s)(b)80859095100105110115120RSCP(dBm)020406080100120Time(s)(c)80859095100105110115120RSCP(dBm)050100150Time(s)(d)Figure12:(a)and(b)Impulseresponsesand(c)and(d)receivedcodepoweralongthemeasurementroutesinareas2and3,respectively.
receivedcodepowersarepresentedwhentransmittingataxedUEpowerlevel(24dBm).
Theimpulseresponseforarea2(Figure12a)hasmainlyonedominatingpathwhileforarea3,multiplepathsaredominating.
Themeasuredimpulseresponsesfortheselectedtestsystemcoverageareashowthatthetrialenvironmentcanbeclassiedastypicalurbanfromaradiopropagationmodelingperspective.
Thenetworkintegratedsingle-cellsystemperformancetestscenarioswereconductedwithdierentsetsofmultiple-usertracandtheresultsindicatedthattheimplementedprototypePICsystemprovidesperformance(servicequal-ity,requiredUEpowerlevel,andsoforth)andbehaviorim-provementsinaccordancewithexpectations.
Toexemplifythebehaviorimprovement,resultsareshowninFigure13toFigure16fromliveon-the-airtestsce-narioswithwalkingusersinconnedarea3.
InFigure13,ascenariowithtwocommunicatingUEswasused,andinFigure14toFigure16,ascenariowithfourcommunicatingUEswereused.
Intheexecutionforthedierentscenarios,theenvironmentandservicequalityrequirementsresultedindierentqualitysettings(TargetCIR)fromthelayer3outer-looppowercontrolalgorithmusedperUEinthetests.
Eachtestscenariowassetupandexecutedinanidenti-calfashiontwice.
Betweentheexecutions,theonlydierencebeingthatthePICprototypetestsystemwasconguredforutilizingeithertheconventionalRAKEorthePICreceiver.
Figure13shows,forbothtestoccasions,theouter-looppowercontroltargetCIRcommands(theRNClayer3algo-rithmdecisions)forthetwoactiveUEswhenmovingalongthetestroute.
Observethatapoint-by-pointcomparisonofthemeasurementvaluesmaynotbevalideventhoughthetestsetupandcompletionhasbeenperformedassimilaraspossibleforthetwotestoccasions.
However,comparingtheoverallstatistics,relativedierencesinthebehaviorcanbeWCDMAULPIC:SystemSimulationsandPrototypeFieldTrials173310121416182022TargetCIR(dB)0510152025303540Time(s)ConventionalRAKEPICMS1MS2Figure13:RNC(layer3)targetCIRcommandswheneithercon-ventionalRAKEorPICreceiverisutilizedforusers1and2.
found.
Whenexaminingandcomparingthespecicmea-suredscenarioinFigure13,themeasurementresultsimplythatthePICreceiverimprovesthecontrolprocessstabilitycomparedtoasystemcongurationutilizingtheconven-tionalRAKEreceiver.
ThestatisticsforthetwocasesshowareductioninmeanandstandarddeviationoftheSIRtar-get.
ForUE1,themeanvaluesarereducedfrom3.
43dBto3.
3dBandthestandarddeviationfrom0.
938dBto0.
601dBwhenthePICisactive.
ForUE2,themeanvaluesarere-ducedfrom1.
08dBto0.
499dBandthestandarddeviationfrom0.
802dBto0.
677dBwhenthePICisactive.
Forthismeasuredscenario,thisimpliesanoverallincreasedcapacity(lessinterferencegenerated)andanincreasedbatterylifeoftheUEterminal.
Sincetheouter-looppowercontrolismea-suringandcontrollingtheservicequalityoverall3layersintheradioaccessnetwork(RAN)andadjuststhetargets,thedecreaseinstandarddeviationindicatesastabilizingeectonthesystem.
Figures14and15illustratestheouter-looppowercontroltargetCIRcommands(theRNClayer3algorithmdecisions)fortheactiveUEswhenmovingalongthetestrouteduringcallsetupwithfourUEs.
Thesetupphaseisselectedtoillus-tratetheimprovedpowercontrolstabilityduetoPIC.
Essen-tially,thetargetCIRvariationsarerelatedtotheuplinkload,andtherefore,theloadreductionduetoPICalsoimprovessystemstability.
ThisisevidentfromFigure16,whichpro-videscumulativedistributionfunctionsoftheuplinkloadforRAKEandPIC,respectively,forthecaseoffouractiveUEs.
Clearly,PICimpliesamorestablesystem.
Correspondingsystemloadasdenedby(1)andcom-putedin(5)forsingle-cellnetworksisusedtocomparetheresultingloadinthetwotests(RAKEandPIC,respectively)whenallUEsareconnected.
Thentheestimatedcapacity101520UE1020406080100120140160180200Time(s)(a)101520UE2020406080100120140160180200Time(s)(b)101520UE3020406080100120140160180200Time(s)(c)101520UE4020406080100120140160180200Time(s)(d)Figure14:RNC(layer3)targetCIRcommandswithconventionalRAKE.
gaincanbeevaluatedasin(7):Capacitygain:meanLRAKEmeanLPIC≈1.
65.
(10)ThisisinaccordancewithobservationsfromtherealisticsimulationsinSection5.
7.
CONCLUSIONThemainconclusionthatcanbedrawnfromtheextensiveworkthathasbeencarriedout,involvinglink-andsystem-levelsimulationsandeldtrials,isthattherearemajorup-linkperformancegainsachievableevensoforinterferencecancellationbasestationarchitecturesofratherlimitedcom-plexity.
Thisisprimarilyduetotheknownnatureofthein-herentintracellinterferencegeneratedinWCDMAnetworks,whichcanbeexploitedbytheICtechnologytooeralargetheoreticaluplinkgain.
Thelimited-complexityPICreceiversupportsa40%up-linknetworkcapacityincreaseforwalkingspeechusersinthesimulatedrealistictypicalurbanradioenvironment.
Inthesimulations,thePICiscarefullymodeledtoreectre-alisticimplementationsofsearcher,channelestimation,and1734EURASIPJournalonAppliedSignalProcessing101520UE1020406080100120140160180200Time(s)(a)101520UE2020406080100120140160180200Time(s)(b)101520UE3020406080100120140160180200Time(s)(c)101520UE4020406080100120140160180200Time(s)(d)Figure15:RNC(layer3)targetCIRcommandswithPIC.
soforthandtheon-the-airtrialresultsindicatethattheim-plementedprototypePICsystemprovidesperformanceandbehaviorimprovementsinaccordancewithexpectations.
Observationsoftheon-the-airtrialmeasurementdy-namicsimplythatthePICreceiverstabilizesthecontrolpro-cessviaareductioninstandarddeviationoftheRNCouter-loopPCtargetCIRcommands.
Thesingle-cellestimatedca-pacitygainintheorderof70%mightpotentiallybeexpe-riencedininhomogeneousdeploymentswithisolatedhigh-demandinghotspotcells.
Moreover,theestimatedcapacitygainforsingle-cellsystemsisinaccordancewithresultsfromrealisticnetworksimulations.
Thisalsoindicatesthatthesimulationmodelsarerelevantandrepresentative.
ACKNOWLEDGMENTSNeedlesstosay,allthe"background"workthathasbeencar-riedouttodesignandbuildaprototypesystem,andtosetupandperformtheeldmeasurements,hasinvolvedquitealotofpeople.
TheauthorswouldinparticularliketothankthepersonsofallcollaborationpartnersinBeijing,Erics-sonRadioNetworkR&DCenterBeijing,ChinaAcademyofTelecommunicationsTechnology(CATT),andDatangTele-comTechnologyCo.
Ltd.
,fortheirexcellenteortandthe10.
80.
60.
40.
20CDF(LRAKE)00.
10.
20.
30.
40.
5UplinkloadL(a)10.
80.
60.
40.
20CDF(LPIC)00.
10.
20.
30.
40.
5UplinkloadL(b)Figure16:CDFcomparisonoftheuplinkloadfor(a)RAKEand(b)PIC.
magnicentfruitfulcooperation.
Furthermore,wearealsoverygratefultoalltheotherpersonsthathasbeeninvolvedforcriticalsupportfromvariousorganizationsandplaces:EricssonResearchinTokyo,Link¨oping,Jorvas,Budapest,andKista;EricssonR&DinBeijing,M¨olndal,Kista,andEn-schede;TietoEnatorinG¨oteborg,Karlstad,andUmea.
REFERENCES[1]S.
Verdu,"MinimumprobabilityoferrorforasynchronousGaussianmultiple-accesschannels,"IEEETrans.
Inform.
The-ory,vol.
32,no.
1,pp.
85–96,1986.
[2]R.
LupasandS.
Verdu,"Linearmultiuserdetectorsforsynchronouscode-divisionmultiple-accesschannels,"IEEETrans.
Inform.
Theory,vol.
35,no.
1,pp.
123–136,1989.
[3]P.
PatelandJ.
Holtzman,"AnalysisofaDS/CDMAsucces-siveinterferencecancellationschemeusingcorrelations,"inProc.
IEEEGlobalTelecommunicationsConference(GLOBE-COM'93),pp.
76–80,Houston,Tex,USA,December1993.
[4]M.
K.
VaranasiandB.
Aazhang,"Multistagedetectioninasyn-chronouscode-divisionmultiple-accesscommunications,"IEEETrans.
Commun.
,vol.
38,no.
4,pp.
509–519,1990.
[5]T.
Shima,R.
Esmailzadeh,J.
Karlsson,andT.
Yamauchi,"UsercapacityofasinglecellWCDMAsystemwithdierentI-Qpowerratioswithandwithoutinterferencecancellation,"inProc.
IEEECDMAInternationalConferenceandExhibition(CIC'99),pp.
36–39,Seoul,Korea,September1999.
[6]M.
Ariyoshi,T.
Shima,J.
Han,J.
Karlsson,andK.
Urabe,"Ontheeectofforward-backwardlteringchannelestimationinW-CDMAmulti-stageparallelinterferencecancellationre-ceiver,"IEICETransactionsonCommunications,vol.
E85-B,no.
10,pp.
1898–1905,2002.
[7]M.
Sawahashi,K.
Higuchi,H.
Andoh,andF.
Adachi,"Exper-imentsonpilotsymbol-assistedcoherentmultistageinterfer-encecancellerforDS-CDMAmobileradio,"IEEEJ.
Select.
Ar-easCommun.
,vol.
20,no.
2,pp.
433–449,2002.
WCDMAULPIC:SystemSimulationsandPrototypeFieldTrials1735[8]T.
Suzuki,Y.
Takeuchi,F.
Watanabe,andS.
Niida,"Implemen-tationandeldtestresultofabasestationcancelingCDMAinterferencefortheIMT-2000airinterface,"inIEEE50thVe-hicularTechnologyConference(VTC'99),vol.
2,pp.
1003–1007,Amsterdam,TheNetherlands,September1999.
[9]T.
Futami,H.
Seki,andY.
Tanaka,"PerformanceevaluationofDS-CDMAmultistageinterferencecancellerinmulti-cellenvironments,"inProc.
IEEEGlobalTelecommunicationsCon-ference(GLOBECOM'98),vol.
1,pp.
247–252,Sydney,NSW,Australia,November1998.
[10]L.
M.
Correia,Ed.
,WirelessFlexiblePersonalizedCommunications—COST259:EuropeanCo-operationinMobileRadioResearch,JohnWiley&Sons,Chichester,England,UK,2001.
[11]3GPPTechnicalSpecicationGroupRadioAccessNetwork,Release99,December2000,availablefromhttp://www.
3gpp.
org/ftp/Specs/2000-12/R1999/,specically23-series,24-seriesand25-series.
[12]D.
Divsalar,M.
K.
Simon,andD.
Raphaeli,"ImprovedparallelinterferencecancellationforCDMA,"IEEETrans.
Commun.
,vol.
46,no.
2,pp.
258–268,1998.
[13]N.
S.
Correal,R.
M.
Buehrer,andB.
D.
Woerner,"ADSP-basedDS-CDMAmultiuserreceiveremployingpartialparal-lelinterferencecancellation,"IEEEJ.
Select.
AreasCommun.
,vol.
17,no.
4,pp.
613–630,1999.
[14]H.
HolmaandJ.
Laakso,"UplinkadmissioncontrolandsoftcapacitywithMUDinCDMA,"inIEEE50thVehicularTech-nologyConference(VTC'99),vol.
1,pp.
431–435,Amsterdam,TheNetherlands,September1999.
[15]E.
GeijerLundin,F.
Gunnarsson,andF.
Gustafsson,"UplinkloadestimationinWCDMA,"inProc.
IEEEWirelessCommu-nicationsandNetworkingConference(WCNC'03),vol.
3,pp.
1669–1674,NewOrleans,La,USA,March2003.
[16]M.
Karlsson,M.
Almgren,S.
Bruhn,K.
Larsson,andM.
Sun-delin,"JointcapacityandqualityevaluationforAMRtele-phonyspeechinWCDMAsystems,"inIEEE56thVehicularTechnologyConference(VTC'02),vol.
4,pp.
2046–2050,Van-couver,Canada,September2002.
[17]H.
Murai,B.
Hagerman,M.
Tadenuma,etal.
,"Systemperfor-manceforWCDMAUp-Linkinterferencecancellation-sim-ulatedresultsandeldmeasurements,"inProc.
IEEEInterna-tionalSymposiumonWirelessPersonalMultimediaCommuni-cations,Yokosuka,Kanagawa,Japan,October2003.
[18]H.
Olofsson,"Improvedqualityestimationforuseinsimula-tionofwirelesssystems,"inProc.
IEEEInternationalConfer-enceonUniversalPersonalCommunications,SanDiego,Calif,USA,October1997.
BoHagermanreceivedtheM.
S.
E.
E.
,Lic.
Tech.
E.
E.
,andPh.
D.
degreesinradiocom-municationsystemsfromtheRoyalIn-stituteofTechnology(KTH),Stockholm,Sweden,in1987,1993,and1995,respec-tively.
From1987to1990,hewasamem-berofthetechnicalstaattheErics-sonRadioSystemsResearchandDevel-opmentDepartment,whereheworkedintheareaofsignalprocessingwithapplica-tionstoGSMreceivers.
HejoinedtheDepartmentforRadioAc-cessandAntennaSystemsResearchatEricssonResearch,Stock-holm,in1995,wherehehasbeenworkinginresearchonadap-tiveantennasincellularsystems.
In1999,hewasappointedastheSeniorSpecialistintheareaofadvancedantennasystems.
Hiscurrentresearchinterestisintheareasofstatisticalsig-nalandarrayprocessingforwirelesscommunicationsandofradioresourcemanagementforcellularsystems.
FredrikGunnarssonisaSeniorResearchEngineeratEricssonResearch,andaRe-searchAssociateatLink¨opingUniversity,Sweden.
HeobtainedhisPh.
D.
degreeinelectricalengineeringfromLink¨opingUni-versityin2000.
Hisresearchinterestsin-cluderadioresourcemanagementandsig-nalprocessingforwirelesscommunica-tions.
HideshiMuraireceivedtheB.
E.
degreeandtheM.
S.
degreefromtheUniver-sityofElectro-Communications,Tokyo,Japan,in1983and1985,respectively.
HealsoreceivedthePh.
D.
degreefromOs-akaUniversity,Osaka,Japan,in1988.
Dur-ing1988–2000,heworkedforMitsubishiElectricCorporationintheareaofdigitalsatellitecommunications,spread-spectrumcommunications,andradioaccesstech-nologiessuchasPDCandCDMA.
Since2000,hehasbeenworkingforEricssonResearchinradioaccesstechnologiesfocusedon3Gandbeyond.
During2000–2001,heworkedinSwedenasaVisitingResearcherfromNipponEricsson.
HeisaMemberofIEICEandIEEE.
MiokoTadenumawasborninKanagawa,Japan,onJanuary27,1976.
ShereceivedtheB.
E.
andM.
E.
degreesinelectricalen-gineeringfromKeioUniversity,Japan,in1998and2000.
In2000,shejoinedEricssonResearchinJapan.
Herresearchinterestsin-cluderadioaccesstechnologiesandinterfer-encecancellation.
JonasKarlssonwasbornin¨Orebro,Swe-den,onJune30,1969.
HereceivedtheM.
S.
degreefromLink¨opingUniversity,Sweden,in1993,theLicentiatedegreefromRoyalInstituteofTechnology,Sweden,in1998,andthePh.
D.
degreefromtheUniversityofTokyoin2003,allinelectricalengineer-ing.
From1993to1998,heworkedforEric-ssonResearchinSwedenasaResearchEn-gineer.
In1998,hejoinedEricssonResearchinJapan,wherehehasworkedasaResearchManagersince2000.
In2004,herejoinedEricssonResearchinSwedenasaSeniorResearchEngineer.
Hisresearchinterestsincluderadioaccesstechnologies,advancedantennas,andinterferencecancellation.