Design and Performance Evaluation of an Inter-System Handover Algorithm in UMTSGSM Networks

HandoverAlgorithminUMTS/GSMNetworks

N.Saravanan†,N.Sreenivasulu†,D.Jayaram†,andA.Chockalingam‡

†Larsen&ToubroInfotechLimited,2ChurchStreet,Bangalore560001,INDIA‡DepartmentofECE,IndianInstituteofScience,Bangalore560012,INDIA

Abstract—Inthispaper,wepresenttheperformanceofa(UMTS↔GSM)inter-systemhandoveralgorithmwhichusestwopairsofUMTSandGSMHOthresholds(Ec/IothresholdforUMTSandRSSIthresholdforGSM);onepairofthresholdsformakingade-cisiontohandoverfromUMTStoGSMandanotherpairofthresh-oldsformakingadecisiontohandoverfromGSMtoUMTS.Wehavedesignedandimplementedthealgorithmbasedonasimula-tionmodelofa37-cellUMTS/GSMsystemconsistingofanislandof7UMTScellssurroundedby30GSMcells.Thesimulationmodelincludesmobilityofallusersinthesystemandspatialcorrelationofthelog-normalshadowlossesinthe2.1GHz(forUMTS)and900MHz(forGSM)frequencybands.ThetestmobileisallowedtomovealongtheboundaryoftheUMTSandGSMcellsandevalu-atedtheperformanceofthealgorithmwiththehandoverrateasafunctionofthevarioushandoverthresholdsinboththenetworks.WeshowthatbyproperlybuildingahysteresisbetweentheUMTS-to-GSMHOthresholdpairandtheGSM-to-UMTSHOthresholdpair,agoodcompromisebetweencalldroprate,andintra-andinter-systemhandoverratescanbeachieved.Further,wehaveanal-ysedtheeffectivenessofthealgorithmbyapplyingtimehysteresisvaluesandalsobymeasuringthesignalparametersincompressedmode.

analysis[2],[3]aswellassimulations[4].However,mostoftheseandotherstudiesconsidermainlyintra-systemhandovers;e.g.,soft/softerhandoversbetweencells/sectorsinCDMAsys-temslikeIS-95,WCDMA[4],orhardhandoversinTDMAsys-temslikeGSM.Somerecentpapershaveinvestigatedinter-systemhandoversinUMTS(UMTS-to-GSMandGSM-to-UMTShan-dovers)[5]-[8].However,themainfocusofthesestudiesisei-theronarchitectural/networklevelissuesinenablinginter-systemhandovers(e.g.,varioussignallingmessagesandtheirsequencerequiredtoexecuteaHO)[6],orontechniquestoenablemea-surementsforinter-systemhandover(e.g.,creatingtimegapsus-ing‘compressedmode’toenablesignalqualitymeasurementsbythemobiles)[8],[9],orondifferenteventsthattriggerHOmeasurements[5].

Ourcontributioninthispaperisthat,wefirstpresentainter-systemhandoveralgorithmwhichusestwopairsofUMTSandGSMHOthresholds(Ec/IothresholdforUMTSandRSSIthresh-oldforGSM);onepairofthresholdsformakingadecisiontohandoverfromUMTStoGSMandanotherpairofthresholdsfor

Keywords–Inter-systemhandover,UMTS,GSM,handoverrate

makingadecisiontohandoverfromGSMtoUMTS,andsec-ond,webuildasimulationmodelofa37-cellUMTS/GSMsys-I.INTRODUCTION

temconsistingofanislandof7UMTScellssurroundedby30

UniversalMobileTelecommunicationSystems(UMTS)isaGSMcells;weallowatestmobiletomovealongtheboundarythirdgeneration(3G)cellularsystemthathasbeenstandardizedoftheUMTSandGSMcellsandevaluatethehandoverrateper-bythe3rdGenerationPartnershipProject(3GPP)withanintentformanceasafunctionofthevariousthresholdsinvolved.Thetoofferenhancedmultimediaservicestomobileusersathighsimulationmodelincludesmobilityofallusersinthesystemdatarates[1].UMTSsystemsareexpectedtobeinitiallyde-andspatialcorrelationofthelog-normalshadowlossesinthe2.1ployedindensesubscriberareascoveredbyGSM/GPRS(e.g.,GHz(forUMTS)and900MHz(forGSM)bands.Weshowthatcitycenters,shoppingmalls)inordertoaugmentthecapacitybyproperlybuildingahysteresisbetweentheUMTS-to-GSMoftheseexistingGSM/GPRSnetworks.Inotherwords,initialHOthresholdpairandtheGSM-to-UMTSHOthresholdpair,aUMTSdeploymentswillresultinUMTShotspotsinaGSM/GPRSgoodcompromisebetweencalldroprate,andintra-andinter-pool.Hence,oneofthecriticalfeaturesintheinitialdeploy-systemhandoverratescanbeachieved.mentsofUMTSnetworksistheinter-systemfunctionalitybe-Therestofthispaperisorganizedasfollows.InSec.II,we

tweenUMTSandGSM/GPRS.Efficientinter-workingbetweenpresenttheprovisionsmadeinthestandardsforimplementingUMTSandGSM/GPRSsystemswillbecrucialtoprovidecon-inter-systemhandovers.InSec.III,wepresenttheproposedtinuousservicecoveragetodual-modemobiles.Amongtheinter-inter-systemhandoveralgorithm.Sec.IVdescribesthesimula-systemfunctionalities,inter-systemhandoverisprobablythemosttionmodel.PerformanceresultsanddiscussionsarepresentedincrucialonetoensureseamlessservicecoveragetoamobileuserSec.V.ConclusionsaregiveninSec.VI.ashe/shemovesfromaUMTScoverageareatoaGSM/GPRScoveragearea,andviceversa.Thefocusofthispaper,accord-II.PROVISIONSFORINTER-SYSTEMHOINUMTS

ingly,istheperformanceevaluationofinter-systemhandoveral-Inordertocarryoutsmoothinter-systemhandovers,themo-gorithmsinUMTS/GSM.

bilesintheborderareasbetweendifferentsystemsmustmake

Severalpapersintheliteraturehaveinvestigatedhandoveral-measurementsonthecarriersofbothsystemsandreportthemtogorithmsandtheirperformancesincellularsystems,boththroughthenetworktofacilitateahandoverdecision.GSM/GPRSsys-temsbeingTDMAbased(i.e.,transmitandreceiveareneither

simultaneousnorcontinuous),findingtimegapstogoandmea-sureontheothersystemcarriersusingoneRFreceivechainisnotcritical.Duringthesetimegapsthereceivercanbetunedtotheothersystemfrequency,makemeasurementsonitandcomebacktothecurrentsystemfrequencytotransmit/receive.Ontheotherhand,inUMTS(WCDMA)thetransmit/receivearecon-tinuousandsimultaneous.ThismeanseitheradifferentRFre-ceivechainmustbeprovidedinthehandsettomakemeasure-mentsontheothersystemfrequency,ortimegapshavetobecreatedinthetransmissionssothatthemobilecangoandmakeothersystemmeasurementsduringthesetimegaps.InUMTS,the‘compressedmode’operationcancreatethesetimegapsforothersystemmeasurements[10].Inaddition,the3GPPstan-dardsdefinethesignallingmechanismsandseveralhandoverpa-rametersnecessarytoperformintra-andinter-systemhandovers[11],[12].Itisnotedthatwhilethestandardsdefinetheneces-sarysignallingmechanismsandparameterswhichcanbeusedtoperformintra-andinter-systemhandovers,thedetailsoftheactualalgorithmandtheoptimizationofthehandoverparam-eters(e.g.,HOthresholdvalues)arelefttotheimplementors.Forthepurposeofusinginmakingahandoverdecision,GSMandUMTSstandardsdefinethefollowingmeasurementsandtherangeoftheirreportedvalues.

MeasurementinGSM:TherelevantmeasurementquantityforhandoverinGSMistheRXLEV,whichisalsoreferredtoastheReceivedSignalStrengthIndicator(RSSI).RSSIisdefinedastheaveragereceivedsignalpowerindBm.RSSImeasurementisperformedontheGSMBCCHcarrier.ThereportablerangeoftheGSMRSSIisfrom-48dBmto-110dBm[13].

MeasurementinUMTS:TherelevantmeasurementquantitiesforhandoverinUMTSarei)ReceivedSignalCodePower(RSCP),whichisthereceivedcodepowermeasuredontheCPICH(com-monpilotchannel),ii)UMTScarrierRSSI,andiii)CPICHEc/IowhichisessentiallytheratiobetweenRSCPandUMTScarrierRSSI.ThereportablerangeofEc/Ioisfrom-24to0dB[14].

Inadditiontotheabovemeasurements,otherparametersin-cluding,forexample,transportchannelblock/biterrorratescanbemeasuredandusedforhandoverpurposes[14].Theproposedinter-systemhandoveralgorithmthatwepresentinthefollowingsectionusesonlythereportedvaluesofGSMRSSIandUMTSCPICHEc/Io.

III.(UMTS↔GSM)INTER-SYSTEMHOALGORITHMInthissection,wepresenttheproposedinter-systemhandoveralgorithm.WeconsiderhandoversbetweenUMTSandGSMsystemswhilethesubscriberisonacircuit-switchedvoicecall.Foragivenmobilesubscriberonavoicecall,thealgorithm(tobeexecutedonthenetworkside)operatesbasedontheEc/IovaluesofallneighbouringUMTScellsandRSSIvaluesofallneighbouringGSMcellsasmeasuredandreportedbythatmo-bile.WedefinetwopairsofUMTSandGSMHOthresholds(Ec/IothresholdforUMTSandRSSIthresholdforGSM);onepairofthresholdsformakingadecisiontohandoverfromUMTS

toGSMandanotherpairofthresholdsformakingadecisiontohandoverfromGSMtoUMTS.Inaddition,wealsodefinetwothresholdhysteresisvaluesforUMTS-to-UMTSandGUMS-to-GSMintra-systemhandovers.Accordingly,wedefinethefol-lowingthresholds:

•UTHU2G:UMTSEc/Iothreshold(indB)usedformak-ingainter-systemhandoverdecisionfromUMTStoGSM•GTHU2G:GSMRSSIthreshold(indBm)usedformak-ingainter-systemhandoverdecisionfromUMTStoGSM•GTHG2U:GSMRSSIthreshold(indBm)usedformak-ingainter-systemhandoverdecisionfromGSMtoUMTS•UTHG2U:UMTSEc/Iothreshold(indB)usedformak-ingainter-systemhandoverdecisionfromGSMtoUMTSHYSU2U:Hysteresis(indB)intheUMTSEc/Iothre-sholdusedforamakingintra-systemhandoverdecisionfromUMTStoUMTS

•HYSG2G:Hysteresis(dB)intheGSMRSSIthresholdusedformakingaintra-systemhandoverdecisionfromGSMtoGSM

•THYS:Timehysteresisinmsecformakingainter-systemhandoverdecisionfromUMTS↔GSMThehandoveralgorithmworksasfollows.

•

UMTS→UMTSHandover:

•

OnceamobileisinUMTS,itwillcontinuetobeinUMTSaslongasthereportedEc/IoofthecurrentlyassignedUMTScell(basestation)isabovetheUMTSthresholdUTHU2G.Thisconditionimpliesthatthemobileisconnectedtoa‘strong-enough’UMTScell.

WhileinUMTS,intra-cellhandoverstostrongestneigh-bouringUMTScellscanoccurwhentheEc/IoofthestrongestneighbouringUMTScellgetslargerthantheEc/IoofthecurrentlyassignedUMTScellbyanamountinexcessofthehysteresisHYSU2UdB.ThishysteresishelpstoavoidfrequentU→Uintra-systemhandovers.

WhileinUMTS,i)iftheEc/IoofthecurrentlyassignedUMTScellfallsbelowtheUMTSthresholdUTHU2G(im-plyingthattheconnectedUMTScellisnot‘strong-enough’),andii)ifthereisnootherUMTSneighbouringcellwhoseEc/IoislargerthanUTHU2G,thenthealgorithmlooksforGSMcellstohandover.

Ifconditionsi)andii)abovearemet,andiftheRSSIofthestrongestGSMcellisabovetheGSMthresholdGTHU2GcontinuouslyforTHYSmsec,thenthemobileisinstructedtodoainter-systemhandoverfromU→G;otherwisethemobilecontinueswithUMTS(inwhichcasethecallqual-ity,intermsofreceivedEc/Iomaygetworse,eventuallyleadingtoacalldropifnosuitableGSMcellorbetterUMTScellisfound).

OnceamobileisinGSM,itwillcontinuetobeinGSM

UMTS→GSMHandover:

•

GSM→GSMHandover:

•

Call in ProgressUMTS callUMTSGSM call orGSMcall?YE2>U1 NNRY2>G2??YE1>U1 NNRY1 >G2??NNN(EE1>G1 E1 >U2Y>1 ?(RH2)YR?>1 RH2)N?1 YY?2Do (U −> U) HODo (G −> G) HOfrom CBS to SNBSfrom CBS to SNBSE1 :(Ec / Io)SNBSIsUEYRNIs1 :UTHU2GE2:(Ec / Io)CBSfor T_HYS1>G1 N1 :GTHU2Gmsecfor T_HYS1>U2 YGmsecR1 :(RSSI)SNBS??U2:UTHG2URDo (U −> G) HODo (G −> U) HOG2:GTHG2U2:(RSSI)CBSfrom CBS to SNBSfrom CBS to SNBSH1 :HYSU2UCBS : Currently assigned BSHSNBS : Strongest Neighbouring BS

2 :HYSG2G

T_HYS : Time Hysteresis

Fig.1.FlowchartforInter-systemHandoverAlgorithm

in msec

aslongasthereportedRSSIofthecurrentlyassignedGSMcellisabovetheGSMthresholdGTHthatthemobileisconnectedtoG2U.Thisconditionimpliesa‘strong-enough’GSMcell.

WhileinGSM,intra-cellhandoverstostrongestneighbour-ingGSMcellscanoccurwhentheRSSIofthestrongestneighbouringGSMcellgetslargerthantheRSSIofthecurrentlyassignedGSMcellbyanamountinexcessofthehysteresisHYSintra-systemG2GdB.Thishysteresishelpstoavoidfre-quentG→Ghandovers.GSM→UMTSHandover:

•

WhileinGSM,i)iftheRSSIofthecurrentlyassignedGSMcellfallsbelowtheGSMthresholdGTHGSMcellisnot‘strong-enough’),G2U(implyingthattheconnectedandiiis)largeriftherethanisnoGTHotherGSMthenneighbouringalgorithmlookscellwhoseforUMTSRSSIcellstohandover.

G2U,Ifconditionsi)andii)abovearemet,andiftheEofthestrongestUMTScellisabovetheUMTSthresholdc/IoUTHG2UcontinuouslyforTHYSmsec,thenthemo-bileisinstructedtodoainter-systemhandoverfromG→U;otherwisethemobilecontinueswithGSM(inwhichcasethecallquality,intermsofreceivedRSSImaygetworse,eventuallyleadingtoacalldropifnosuitableUMTScellorbetterGSMcellisfound).

Theflowchartdepictingtheaboveinter-systemhandoveral-gorithmisshowninFig.1.

Note:Itisnotedthattheintra-systemhandoversinGSM(i.e.,G→GHO)arehardhandoversandamobileisconnectedtoonlyonecellatanygiventime.However,theintra-systemhandoversinUMTS(i.e.,U→UHO)aretypicallysofthandoversandamobilemaybesimultaneouslyconnectedtomorethanonecell.SinceourmainfocusinthisstudyisonU↔Ginter-systemhandovers,weconsiderthataUMTSmobileisconnectedtooneUMTScellatatime.However,thealgorithmwithminormod-

UMTSthresholdsGSMthresholdsUTHG2U GTHU2G

(e.g., −14 dB)G(e.g., −90 dBm) UUUTH U2GGTHG2U

(e.g., −16 dB)G (e.g., −100 dBm)Fig.2.HysteresisbetweenUMTSandGSMHOthresholdpairs

ificationscanbemadeapplicabletoUMTSmobilesbeingcon-nectedtomultipleUMTScells.OntheChoiceofHOThresholdValues:

Itisnotedthat{UTHU2G,GTHahandoverU2G}istheUMTSandGSMthresholdpairusedtomakedecisionfromUMTStoGSM.WemaywanttodelaytheoccurrenceofaUMTStoGSMinter-systemhandoverasmuchaspossible,whichmaybeachievedbychoosinga‘weak’(lowvalue)UTHoldanda’strong’(highvalue)GTHU2Gthresh-movestoU2Gthreshold.Ontheotherhand,oncethemobileGSM,wemaywanttogivepreferencetoGSMtoGSMhandoveranddelaytheGSMtoUMTSinter-systemhandoverasmuchaspossible.Thismaybeachievedbyproperchoiceofthe{UTHpairusedtomakeahandoverdecisionG2Ufrom,GTHGSMG2U}thresholdtoUMTS,i.e.,choosea‘weak’GTHahysteresisbetweentheG2Uanda‘strong’UTHU→GandG→UthresholdG2U.Typically,pairs,asshowninFig.2,isdesiredinordertoavoidfrequentinter-systemhandoversfromUMTStoGSMandviceversa(i.e.,ping-pongeffect).Inotherwords,thethresholdscanbechosensuchthat

GTHU2G>GTHG2U,andUTHG2U>UTHU2G.

IV.SIMULATIONMODEL

Inthissection,wepresentthesimulationmodelofthemulti-cellUMTS/GSMsystemthatwedevelopedtoevaluatetheper-formanceoftheinter-systemhandoveralgorithmpresentedintheprevioussection.

Cellularsystem:Weconsidera37-cellUMTS/GSMhexago-nallayoutsystemconsistingofanislandof7UMTScellssur-roundedby30GSMcells,asshowninFig.3.GSMcellsusea7-cellfrequencyreusepatternasshown.EachcellconsistsofNmobileswhichareuniformlydistributedoverthecellarea.Propagationandchannelmodel:Thetransmittedsignalsun-dergolossesduetodistance,shadowandmultipathfading.Weassumethattheshort-termmultipathfadingisaveragedoutinthemeasurementprocess.Asinmostliterature,wetakethepathlossexponentηthatcharacterizethelossduetodistancetobe4,andassumetheshadowlosstobelog-normallydistributed[15].Sinceweareinterestedinevaluatingthehandoverperformance,usermobilityandhencethespatialcorrelationinthelog-normal

UMTS Cells

UC 1UC 2GSM Cells

GC 1GC 2Test mobile moves along this lineFig.3.37-cellmixedUMTS/GSMSystem

shadowlossmustbeincorporatedinthesimulationmodel.Ac-cordingly,weemploythespatial-correlationmodelforshadowlossproposedbyGudmundson[16]inoursimulations.Inthismodel,thespatialcorrelationinthereceivedsignalA(n)isgivenby

RA(k)=σ2(εD)|k|vT/D

(1)wherevisthevelocityofthemobile,vTisdistancemovedinT

secs,εDisthecorrelationbetweentwopointsseparatedbydis-tanceD,andvarianceσ2isusuallyintherangebetween3to10dB.From[16],forGSMmobilesinthe900MHzband,wetakethestandarddeviationofthelog-normalshadowlossσtobe7.5dB,andthecorrelationatdistanceof100m(εmobilesinthe2.1100)istakentobe0.82.Likewise,forUMTSGHzband,wetakeσtobe4.3dBandε10tobe0.3.Thesimulationofthecor-relatedlog-normalshadowlossvariableshasbeenimplementedusingtheproceduregivenin[15].

ReceivedreceivedsignalPower:powerWithPtheabovepropagation/channelgivenby[15]

model,theR(indBm)is󰀁

󰀂PR=PT+GT+GR+K−10ηlog10

dd0

+ψ,

(2)

wherePTisthetransmitpower(indBm),GTandGreceiveantennagains(indB),respectively,Rarethe

transmitandK=−20log10(4πd0/λ),d0isthereferencedistancefortheantennafar-field,λisthecarrierwavelength,disthedistancebetweenthetransmitterandreceiver,ηisthepathlossexponent,and2ψisaGaussiandistributedr.v.withmeanzeroandvarianceσthatdenotesthelog-normalshadowloss(indB).

GSMRSSI:AllGSMbasestationstransmitBCCHcarriersatafixedpilotpower.ForthepurposesofcomputingtheGSMRSSI,PTin(2)correspondstothistransmitPilotpowerontheBCCHcarrier.

UMTSEc/Io:AllUMTSbasestationstransmittheircommonpilotchannels(CPICH)atafixedpilotpower.TheRSCP(re-ceivedsignalcodepower)iscomputedbasedonthisCPICHtransmitpilotpowerandchannelparametersasper(2).TheUMTSRSSIisthetotalreceivedpowerincludingtheinterfer-encepower(intra-cellaswellasinter-cell)frommobiletraf-

ParametersUMTSGSMCelllayoutHexagonalHexagonalNumberofcells730Cellsize

5Km5KmUserspercell5050Carrierfrequency2.1GHz900MHzPilotpower

0.15W1WTx.antennagain14dB14dBRx.antennagain2dB2dBRef.distance,d100m100mPathlossexponent044SDofshadowloss4.3dB7.5dBCorrelationcoeff.0.30.82CorrelationDist.10m100mD/LPowercontrol-110dBmN/AChiprate3.84McpsN/ADatarate

12.2Kbps12.2KbpsMobilevelocity3.6,18,36,54Km/h

3.6,18,36,54Km/h

TABLEI

ficchannelsonthedownlinkandthermalnoisepower.SinceCDMAsystemsareinterference-limited,weignorethethermalnoise.AllUMTSmobilestrafficchannelsarepowercontrolledonthedownlinkbytheirassignedbasestations.Thetransmitpowerfromthebasestationonthetrafficchanneltowardseachmobileisadjustedsuchthatthesignalisreceivedataconstantdesiredpoweratthemobile.TheEc/IoatithmobilefromjUMTSthUMTSRSSI)cellasisfollows:

determined(astheratiooftheRSCPandthe󰀁Ec󰀂

PR(j,cpich)

iIo

=

ij

󰀃Ni])kP󰀃(3)

R(k∈B[i])

i

+

󰀃NPk=1=i

B[i]k=1

R(k∈B[i

wherePR(j,cpich)

iinthenumeratoristhereceivedpowerattheithmobileonthejthcell’spilotchannel,B[i]denotestheUMTScelltowhichmobileibelongs,andB[i]denotesthesetUMTScellsotherthanB[i](i.e.,othercells),andP(k∈B[i])

ofall

powerreceivedattheithmobileontheRkiistheinterferencethmobile’strafficchannelincellB[i].

Usermobilitymodel:Thehandoverstatisticsarecollectedatatestmobileoverthesimulationofseveralcalls.Eachcalllastsfor120secs.Atthebeginningofeachcall,freshlocationsforthemobiles(uniformlydistributedintheircellareas),andtheirlog-normalshadowlossesarechosen.Inaddition,avelocityanddirectionofmotionarechosenforeachmobile.Themobilevelocitiesarechosenamong3.6,18,36and54Km/hwithequalprobability.Thedirectionofmotionischosenuniformlyintherange0to2π.Eachmobilethenmovesinastraightlinefromitsstartinglocationatitschosenvelocityanddirectiontilltheendofthecall.

V.PERFORMANCERESULTSANDDISCUSSIONS

Weimplementedthesimulationmodeldescribedintheprevi-oussectioninMatlabandcarriedoutthesimulationstocharac-terizethehandoverperformanceoftheproposedalgorithm.The

00UMTS Ec/Iofrom UC2 (dB) −10−10−20)m−20−30Bd)(BGSM RSSI 1d( from GC1 (dBm) −40C G2 C−30−50mUor fm IoSrf−60S o−40R I/cME−70S dG edru−50−80ersuaseaM−90eM−60−100−110−70−120−8000.250.50.7511.251.51.7522.252.52.753−130UC1 UC2 GC1 GC2 x 104Distance (metre) Fig.4.VariationofRSSIfromGSMcellGC1(indBm)andEcellUC2(indB)asafunctionoftimeatamobile,asthemobilec/IofromUMTSmovesfromUMTScellUC1towardsGSMcellGC2showninFig.3ataspeedof54Km/h.RSSIandEc/Iovaluesaveragedover25passesofthemobilemovingfromUC1towardsGC2.

varioussystemparametersusedinthesimulationsaregiveninTableI.

A.SimulationResults

Figure4showsanillustrativeoutputfromthesimulationsthatdepictsthevariationofRSSIfromGSMcellGC1(indBm)andEc/IofromUMTScellUC2(indB)asafunctionoftimeatamobile,asthemobilemovesfromUMTScellUC1towardsGSMcellGC2showninFig.3,ataspeedof54Km/h.Thesys-temsimulationisrunatatimeresolutionof1sec(i.e.,theRSSIandEc/Iovaluesatthetestmobileiscomputedeverysecond,accountingforthetimevariationindistanceandshadowlosses,etc).TheplotsinFig.4showtheRSSIandEeragedover25passesofthemobilemovingfromc/IUC1ovaluesav-towardsGC2.Itcanbeseenthat,asexpected,theEc/IoandRSSIpeakatmobilelocationsclosetotheUMTSbasestationlocationUC2andGSMbasestationGC1,respectively.

Theperformancemeasuresofinterestinthissimulationstudyarethevarioushandoverrates(U→Uhandoverrate,U→Ghandoverrate,G→Ghandoverrate,andG→Uhandoverrate).Thesehandoverstatisticsarecollectedatatestmobileoversev-eralcalls,eachcalllasting120secs.Specifically,thetestmobileisallowedtomoveinacircularpathalongtheboundaryoftheUMTSandGSMcells(asshowninFig.3).Thispathwascho-senforitcorrespondstoaveryrepresentativescenariowhereinter-systemhandoversaremorelikelyandhencetheeffectsofthevarioushandoverthresholdsontheperformancecanbemoreclearlyobserved.

Thesystemsimulationisrunatatimeresolutionof100msec(i.e.,theRSSIandEc/Iovaluesatthetestmobileiscomputedevery100msec,accountingforthetimevariationindistanceandshadowlosses,etc).Also,thehandoveralgorithmisexecutedonce100msec.Attheendofacall,anewcallisstartedwithfreshmobilelocations,shadowlosses,velocityanddirectionsfor

101llaC reP sp100orD llaC dna srevo10−1dnaH fo rebmuN e10−2UTH (U2G) = −16 dB gaGTH (G2U) = −100 dBm(U − U) HandoversreUTH (G2U) = −14 dB (G − G) HandoversvAT_HYS = 100 msec (U − G) Handovers(G − U) HandoversCall Drops10−3−70−75−80−85−90−95−100GSM RSSI Threshold value, GTH (U2G) dBmFig.5.Intra-andinter-systemhandoverratesandcalldroprateasafunctionofGSMRSSIthreshold,GTH−U2G.UTHU2G=and14TdB,HYSGTH−16dB,UTHG2U==G2U100msec.

=−100dBm,HYSU2U=3dB,HYSG2G=3dBallmobiles.WeobservedthatthereceivedGSMRSSIvalueatthetestmobileisintherangeof-105dBmto-55dBmwhenitisconnectedwiththeGSMnetworkandifitisconnectedwiththeUMTSnetworkitreceivestheEto-7dB.

c/Iovalueintherangeof-19dBFigure5showsthevarioushandoverratesperformanceoftheproposedalgorithmintermsoftheaveragenumberofhandoverspercall(statisticscollectedover1000calls)asafunctionoftheGSMRSSIthreshold,GTH16U2G.Theotherthresholdvaluessetare:UTHU2G=−dB,UTHG2U=−14dB,andGTHG2U=−100dBm.ThehysteresisvaluesforU→UandG→GhandoversareHYSU2U=3dB,HYSG2G=3dB.Thetimehysteresis(THYS)valueistakentobe100msec.Thetestmobilespeedis54Km/h.Thenumberofuserspercellis50.FromFig.5,thefollowingobservationscanbemade.AstheGSMRSSIthresholdGTHU2Gisreducedfrom-70dBmto-100dBm,theUMTS-to-GSMinter-systemHOrateincreases(from0.025U→Ghandoverspercallto0.64handoverspercall),since,byallowingaGSMcarrierwithalowerRSSIasad-equateenoughforahandover,weallowmoreandmoreUMTScallstobehandedovertoGSM.Bydoingso(i.e.,bypreferringtohandovertoaweakGSMcarrier),theUMTS-to-UMTSintra-systemhandoverrategetsreducedfrom7.2to2,asseenfromFig.5.Fig.5alsoshowsthecalldropperformance.Anoutage(orcalldrop)isdeclaredifneitheraUMTScarrierstrongerthan-20dBEc/IonoraGSMcarrierstrongerthan-110dBmRSSIisavailableforthecall.Itcanbeenseenthat,asexpected,theoutage(intermsofaveragenumberofcalldropspercalldura-tion)getsreducedastheGSMentrythresholdGTHThesehandovervscalldropperformancetradeoffU2Gisre-duced.curvesareusefulinmakinggoodchoiceofHOalgorithmparameters.B.HandoverSimulationwithCompressedMode

Weobtainedtheperformanceresultsintheprevioussubsec-tionbyperformingEsignalc/IleveloandRSSIcalculationsfrominstanta-neousreceivedvaluessampledatevery100msectime-ticinthesimulations.Inpractice,theUTRANusescom-

GSM Carriersare measured

C1C2C3C4C5C6C7C8C1C2

C3C4C5C6C7C8C1C2C3C4

C5C6C7C8C1C2C3C4 5 6 CC

101 0

msec

TGL = 6.67 msec

10 Slots

TGPL = 40 msec

4 frames

40msec

TGL = 6.67 msec

10 Slots

80 msec

TGL = 6.67 msec

10 Slots

120 msec

Average Number of Handovers and Call Drops Per Call10010−1TGPL = 40 msecTGPL = 40 msec

4 frames4 frames

TGL: Transmission Gap Length

TGPL: Transmission Gap Pattern LengthC1C2C3. . . C8 : 8 GSM Carriers

10−2Fig.6.Compressedmodepattern

UTH (U2G) = −16 dBGTH (G2U) = −100 dBmUTH (G2U) = −14 dB (U − U) Handovers(G − G) Handovers(U − G) Handovers(G − U) HandoversCall Dropspressedmodetocreatetimegapsinthetransmission(referredtoasTransmissionTimeGap-TGL)forthepurposeofmakingGSMRSSImeasurementwhilethemobileisinUMTSnetwork.Periodictimegapsarecreatedinthetransmission.Typically,timegapofTGLsecsiscreatedeveryTimeGapPatternLength(TGPL)secs.ReceivedsignalsamplesmeasuredduringthesetimegapsareaveragedtoobtaintheRSSIvalue.Thecompressedmodefeatureandtherelevantparametersaregivenin[14].Fig.6showsthecompressedmodeparametersaswechosefortheHOsimulation.TGPL=4frames1(i.e.,40msec).TGL=10slots(i.e.,6.67msec).IneachTGLslot,oneRSSIsamplefromoneGSMcarrierismeasured.So,numberofGSMcarrierRSSIsamplesineachTGL=10.WemeasuretheRSSIonthe8strongestGSMcarriers.WeobtaintheRSSIastheaverageover3RSSIsamplesperGSMcarrier.Handoffdecisionsareexecutedevery120msectime-tic.TheHOratesandcalldroprateperformanceobtainedfromtheHOsimulationsimplement-ingthecompressedmodeforRSSImeasurementisgiveninFig.7,whichcorroborateswellwiththeresultsobtainedfromtheinstantaneoussignallevelmeasurementscarriedoutusingTimehysteresisshowninFig.5.

VI.CONCLUSIONS

Weproposedandverified(UMTS↔GSM)inter-systemhan-doveralgorithmwhichusestwopairsofUMTSandGSMHOthresholds(Ec/IothresholdforUMTSandRSSIthresholdforGSM);onepairofthresholdsformakingadecisiontohandoverfromUMTStoGSMandanotherpairofthresholdsformak-ingadecisiontohandoverfromGSMtoUMTS.Wesimulateda37-cellUMTS/GSMsystemandevaluatedintra-andinter-systemhandoverratesandcalldroprateofatestmobilemovingalongtheboundaryoftheUMTSandGSMcells.Thesimulationmodelincludedmobilityofallusersinthesystemandspatialcorrelationofthelog-normalshadowlossesinthe2.1GHz(forUMTS)and900MHz(forGSM)frequencybands.WeshowedthatbyproperlybuildingahysteresisbetweentheUMTS-to-GSMHOthresholdpairandtheGSM-to-UMTSHOthresholdpair,agoodcompromisebetweenintra-andinter-systemhan-doverratesandcalldropratecanbeachieved.Wealsointro-ducedatimehysteresisinthehandoveralgorithm,whichre-sultedinreductionofinter-systemhandoverrate.Further,wepresentedcompressedmodeparametersformeasuringGSMRSSI

1Note

10−70−3−75−80−85−90−95−100GSM RSSI Threshold value,GTH (U2G) dBmFig.7.Intra-andinter-systemhandoverratesandcalldroprateasafunctionofGSMRSSIthreshold.CompressedmodeisusedformeasuringGSMRSSI.UTHU2G=−16dB,UTHG2U=−14dB,GTHG2U=−100dBm,HYSU2U=3dB,HYSG2G=3dB.

whileinUMTSnetwork.WealsoevaluatedandpresentedtheHOperformancebyimplementingthiscompressedmodeinthesimulation.

ACKNOWLEDGEMENT

TheauthorswouldliketothankS.VijayAnandandEshanKa-mathofLarsen&ToubroInfotechLimited,Bangalorefortheirvaluablediscussionsandhelpduringthiswork.

REFERENCES

[1]H.HolmaandA.Toskala,WCDMAforUMTS-RadioAccessforThird

GenerationMobileCommunications,JohnWiley&Sons,2001.

[2]R.VijayanandJ.M.Holtzman,“Amodelforanalyzinghandoffalgo-rithms,”IEEETrans.onVeh.Tech.,vol.42,no.3,pp.351-356,August1993.

[3]N.ZhangandJ.M.Holtzman,“AnalysisofaCDMAsoft-handoffalgo-rithm,”IEEETrans.onVeh.Tech.,vol.47,no.2,pp.710-714,May1998.[4]W.K.Ahmed,“SimulationandperformanceevaluationofIS-95handoff

algorithms,”Bell.Syst.tech.Jl.,7(1),pp.165-181,2002.

[5]M.BensonandH.J.Thomas,“InvestigationoftheUMTStoGSMhan-doverprocedure,”Proc.IEEEVTC’2002,pp.1829-1833,May2002.

[6]G.Alsenmyr,J.Bergstrom,M.Hagberg,A.Milen,W.Muller,H.Palm,H.

Velde,P.Wallentin,andF.Wallgrean,“HandoverbetweenWCDMAandGSM,”EricssonReview,no.1,pp.6-11,2003.

[7]D.Lugara,J.Tartiere,andL.Girard,“PerformanceofUMTStoGSM

handoveralgorithms,”Proc.IEEEPIMRC’2004,pp.444-448,September2004.

[8]D.Lugara,L.Ammi,M.Griguer,andJ.Tartiere,“UMTStoGSMhandover

basedoncompressedmodetechnique,”Proc.IEEEICC’2004,pp.3051-3055,June2004.

[9]S.Hamalainen,T.Henttonen,J.Numminen,andJ.Vikstedt,“Networkef-fectsofWCDMAcompressedmode,”Proc.IEEEVTC’2003-Spring,pp.881-885,April2003.

[10]3GPPTS25.215,“PhysicalLayer-Measurements(FDD),”v3.11.0,2002-09.

[11]3GPPTS25.331,“RadioResourceControl(RRC)ProtocolSpecification,”

v3.14.0,2003-03.

[12]3GPPTR25.922,“RadioResourceManagementStrategies(Release6),”

v6.0.1,2004-04.

[13]3GPPTS45.008,“DigitalCellularTelecommunicationsSystem(Phase

2+);RadioSubsystemLinkControl.”

[14]3GPPTS25.133,“Requirementsforsupportofradioresourcemanagement

(FDD)(Release6),”v.6.7.0,2004-09.

[15]G.L.Stuber,PrinciplesofMobileCommunicationKluwerAcademicPub-lishers,1997.

[16]M.Gudmundson,“Correlationmodelforshadowfadinginmobileradio

systems,”ElectronicsLetters,vol.27,no.23,pp.2145-2146,November1991.

thatinUMTS,1frame=15slots=10msec.