NETWORKENERGYEFFICIENCY

byNGMNAlliance

Version:

1.0

Date:

03.10.2021

DocumentType:

FinalDeliverable(approved)

ConfidentialityClass:

P-Public

Project:

GreenFutureNetworks

Editor/Submitter:

JohanvonPerner(Huawei),VasilisFriderikos(KCL)

Contributors:

JavanErfanian,BellCanada

JianhuaLiu,ChinaMobile

SaimaAnsari,DeutscheTelekom

DanielDianat,Ericsson

DavidLópez-Pérez,Huawei

JohanvonPerner,Huawei

VasislisFriderikos,MischaDohler,King'sCollegeLondon(KCL)

LjupcoJorguseski,TNONL

EmreBilgehanGedik,KorhanYaman,G?khanKALEM,

Turkcell

AnaGalindoSerrano,MariaOikonomakou,Orange

GaryLi,WilliamRedmond,Intel

Marie-PauleOdini,HPE

Approvedby/Date:

NGMNBoard,4thNovember2021

NGMNe.V.

Gro?erHasenpfad30?60598Frankfurt?Germany

Phone+4969/9074998-0?Fax+4969/9074998-41

TheinformationcontainedinthisdocumentrepresentsthecurrentviewheldbyNGMNe.V.ontheissuesdiscussedasofthedateofpublication.Thisdocumentisprovided“asis”withnowarrantieswhatsoeverincludinganywarrantyofmerchantability,non-infringement,orfitnessforanyparticularpurpose.Allliability(includingliabilityforinfringementofanypropertyrights)relatingtotheuseofinformationinthisdocumentisdisclaimed.Nolicense,expressorimplied,toanyintellectualpropertyrightsaregrantedherein.Thisdocumentisdistributedforinformationalpurposesonlyandissubjecttochangewithoutnotice.Readersshouldnotdesignproductsbasedonthisdocument.

?2021NextGenerationMobileNetworkse.V.Allrightsreserved.NopartofthisdocumentmaybereproducedortransmittedinanyformorbyanymeanswithoutpriorwrittenpermissionfromNGMNe.V.

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page2(55)

CONTENTS

1

ExecutiveSummary

5

2IntroductionandPurposeofDocument

8

2.1

Introduction

8

3

Definitions

10

4

EnergyEfficiency

11

4.1

RANSiteOverviewfromanEnergyPerspective

12

5HowtoDecreaseEnergyConsumption

14

5.1

BaseStationHardware

15

5.1.1

VirtualizationofRAN

17

5.2

ProcessorandNetworkServerPowerEfficiencyImprovement

18

5.2.1

CPUPowerManagement

19

5.2.2

VirtualizationTechnology

21

5.2.3

AcceleratorUse

21

5.2.4

InstructionSetArchitectureImprovements

21

5.2.5

WorkloadProfilingandOptimization

22

5.3

Software&Functions

22

5.3.1

SleepModeFunctions

22

5.3.2

SymbolShutdown

28

5.3.3

ChannelShutdown

29

5.3.4

SparseAntennaArrays

29

5.3.5

CarrierShutdown

30

5.3.6

NetworkEnergySavingUsingArtificialIntelligence

30

5.3.7

NetworkDesign

31

5.4

ArtificialIntelligence

32

5.4.1

AIinMobileTelecommunications

33

5.4.2

EnergySavingsThroughArtificialIntelligence

35

5.4.3

EnergyConsumptionofAI

36

5.5

Terminal’sImpactonNetworkEfficiency

41

5.6

Sunsetof2G/3G

42

6EnergyEfficiencyinTechnicalSite

44

6.1

TechnicalSiteCooling

44

6.1.1

FreeCooling

45

6.1.2

LiquidCooling

45

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021

Page3(55)

6.1.3

AIforTechnicalSiteManagement

46

6.1.4

HeatReuse

47

6.2

NextGenerationUninterruptiblePowerSupply

47

7

ConclusionsandRecommendations

49

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page4(55)

1EXECUTIVESUMMARY

EnergyconsumptionofmobilenetworksisakeyconcernfortheoperatorsasitnotonlyleadstoanincreaseintheOPEXbuthasanimpactonCO?emissionsaswell.Therefore,MobileNetworkOperators(MNOs)arefocusingonfindingthebestpossiblewaystoreducetheenergyconsumptionoftheirnetworkseitherbyusinglatesttechnologyorbyoptimizingtheuseoftheactiveandpassivecomponents.Introductionof5Gbringsmoreenergyconsumptionduetothedeploymentofadditionalradiosinnewfrequencylayersbutontheotherhandthe5Gtechnologyismoreenergyefficientthanitspredecessors,thankstotheimprovedspectrumefficiencywhichcomesalongwithhigherMultipleInputMultipleOutput(MIMO)schemesanditsultra-leandesign.Tosupporttherisinguseofcellularconnectivityinthe5Gera,whilereducingenergyconsumptionandemissionsonaper-bitbasisinthecontextofanabsolutereductioninemissions,themobilenetworksneedtobemoreefficient.OnewaytoreducetheseemissionscouldbetheuseofrenewableenergysourceswhichiscoveredintheNGMN“SustainabilityChallengesandInitiativesinMobileNetworks”WhitePaper

[1].

Tomaximizetheend-to-endenergyperformance,itwillbecrucialforMNOstoadoptadifferentapproachtowardsnetworkplanning,deployment,andmanagement.

Leveragingthespectralefficiencyofthe5Gairinterfaceanditsmoreadvancedsleepmodesisimportant,butfurtherefficienciescanbecombinedacrossthreelevels–thebasestationequipmentlevel,sitelevelandnetworklevel.Thereisawiderangeoftechniquesthatcanbeusedacrossthesethreelevelsofnextgenerationnetworkoperation,whichareexploredinthisdocument.Sincethebasestationscoverthelargestpartoftheenergyconsumptioninamobilenetwork,thisWhitePaperdetailsvarioustechniquesforautomaticwake-up/sleepmodesincludingshutdownonsymbol,channelorcarrierbasisandusageofefficientpoweramplifierscombinedwithmassiveMIMO.Sincethetrafficloadvariesduringtheday,itisimportanttodeploysleepmodefunctionsthatshutoffhardwarewhentheloadislow.Symbolshutdownisthemostimportantfunctiontoaddressthisasitsavesapproximate10%energyinalessloadedscenario.Completecarriershutdowncanalsobeusedbuthaslessgainandmighthaveimpactonuserexperience.ThisWhitePaperexploresdifferentsleepmodefunctionsindetail.

Virtualizationhasbeengainingalotofattentioninthemobileindustry,whichmeansthedecouplingofsoftwarefromhardwaretherebyenablingmobileoperatorstodevelopanddeployservicesquickly.Thisalsohelpsintheagiledeploymentofthenetworksandreduces

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page5(55)

thedependencyonproprietaryhardware.VirtualRadioAccessNetwork(RAN)isbasedonCommercialOff-The-Shelf(COTS)hardware,suchasGeneral-PurposeProcessors(GPP)andstandardEthernetNetworkInterfaceCards(NICs).Withtheadvancementsdeliveredinmodernserverplatforms,workloadswithinthewirelessandwirelinenetworkscanbeservedwith(COTS)serverscombinedwithhardwareacceleratorsforoff-loadingheavybasebandprocessing,whileatthesametime,powersavingmodescanbeachievedwithoutcompromisingthestricttelecomgradedeterminismrequirements.

Forthesitelevel,techniquesrangingfromusingrenewableenergyforon-gridandoff-gridsites,smartbatteries,powerefficientpowersuppliesareexplained.FreeandliquidcoolingareseenasanimportantsolutionintechnicalsitesasITequipmentinindoorsitesrequirecoolingsolutionswhichcurrentlycanrepresentupto50%ofthenetworkenergyconsumption.Atnetworklevel,flexiblecooperationbetween5GandLTEisnecessarytodelivertherightamountofcapacityatthelowestpracticalpowerlevelinordertoplanthenetworkoperationbasedonenergyperformance.

MassiveMIMOisakeytechnologyforimprovingenergyperformanceasitallowsantennaarraystofocusnarrowbeamstowardstheuserstherebyincreasingthespectralefficiency.5GtogetherwithmassiveMIMOisthreetofivetimesmorespectrumefficientthan4Gdeployedwithtraditionalradiosolution.Furthermore,itisimportanttoselectthebestantennaconfigurationdependingonthescenarioinordertoachievebestenergyperformance.Betterscalingofenergyconsumptioncanbeachievedbyswitchingoffpartsofthetransmittersatlowertrafficload.

Thedevicesidewillbeimportanttoosincetheperformanceofthedevicewillimpacttheoverallnetworkperformance.Thedevicesignalstrengthreceiversensitivityhasanimpactontheenergyperformance.Withbettersensitivity,higherthroughputandlessre-transmissionareneededwhichmeansthatthebasestationcanuselesspower.

Manyenergysavingsolutionsmentionedinthisdocumenthavealreadybeenimplementedinmobilenetworks.However,theseadvancementsinenergyefficiencywillnotbesufficientasforecastspointtoasignificantriseinenergyconsumptionoverthenextcoupleofyearsduetoconsiderableincreaseintrafficacrossavastrangeofusecases,newtechnologiesandspectrum,greatdealofconnections,andnetworkdensification.Here,ArtificialIntelligence(AI)couldplayanimportantrole.Bypredictingandlearningthetrafficbehaviour,AIalgorithms

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page6(55)

definetheactivation/deactivationofsleepmodefunctionalityandsiteenergymanagementwithoutimpactingtheoverallperformanceincludingQualityofExperience(QoE).AIisstillinanearlyphaseandmoredevelopmentandresearchisneededtoreachitsfullpotential.AIbasedenergysavingsolutionscangreatlyincreasetheenergyperformanceofcellularnetworks.

ThisWhitePaperalsopresentsamethodologyoncalculatingtheenergyconsumptionofAI.ItisbasedonanapproximationwhichcanberefineddependingonthespecificAIarchitectureandusecase.Also,theanalysisisonlyconductedforacanonicalconfigurationofaCNN,i.e.,oneconvolutional,onepoolingandonefullyconnectedlayer.Thechainingofseveralconvolutionallayersscalestheenergyconsumptionlinearly.SimilarenergycalculationmethodologiesareapplicabletoRecurrentNeuralNetworks(RNNs),GenerativeAdversarialNetworks(GANs)andDeepReinforcementLearningbutomittedhereforthesakeofbrevity.AdetailedanalysiscontainingthebenefitsofusingAI,itsimpactonthenetwork,applicableareas,standardization,maturity,andchallengesarealsolisted.

Withoutthesebestpracticesmentionedabove,a5Gnetworkdespiteimprovedbit/joulepowerconsumption–couldtypicallyusemorepowerthana4Gonewithsimilarcoveragearea,becauseofthegreaterdensityofbasestationswhenhigherfrequencyisused.Byadoptingthefullrangeofpowerandsiteoptimisationtechniquesincurrentandfuturenetworks,mobileoperatorscanreduceoratleastkeeptheenergyconsumptionstableeveninadensernetwork.

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page7(55)

2INTRODUCTIONANDPURPOSEOFDOCUMENT

2.1Introduction

InNGMN5GWhitePaper1

[2],

publishedatthebeginningof2015,wesetagoalofanimprovementinenergyefficiencybyafactorof2000within10years,suchthata1000timesprojectedincreaseintrafficcanbecarried,usinghalftheamountofenergyconsumptionrequiredatthetime.Otherssetatargetofatleast100timesimprovementofcapacity,comparedto4G.The5Gsystemissignificantlymoreenergyefficientthanthepreviousgenerations,thoughstillinanearlyphaseonitspathtoachievetherequiredtargets.Furthermore,theend-to-endincreaseinenergyefficiencyshouldfocusonsustainabilityandenvironmentaltargets,towardscarbonneutrality.

Manytrendsarecontributingtotherisinguseofdatatransmittedoverfixedandmobilenetworks.Changesinthewaypeoplework,playandcommunicatehavebeenenabledbymoderntechnology,andcontinuetodriveitsusage.Cellulardatatrafficisprojectedtogrowby6timesbetween2018and2024inemergingeconomies,andby3timesindevelopedmarketsoverthesameperiod

[3].

By2025thereareforecasttobe100billionconnections,including40billionsmartdevices.

Figure1:TrafficGrowth,AnalysisMason2020

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page8(55)

Inordertomeettheseaggressivetargets,astepchangeinenergyefficiencyisneeded.5Gwillplayacriticalroleinaddressingthischallenge.Themostchallengingelementsofamobilenetwork,inenergyconsumptionterms,aretheradiobasestationswhichrepresentabout57percentoftheconsumedenergy

[3].

LongTermEvolution(LTE)networksthatarehighlyloadedhavea20to30percentaverageuseoftheairinterface.Furthermore,approximately20percentofthebasestationscarry80percentofthetraffic,thereforethereareseveralopportunitiestoimprovetheenergyefficiencyofthebasestations.Thiscanbedonebyreducingthepowerconsumptionwhennodataisneededtobetransmittedandincreasehardwareefficiency,especiallywhentransmissionpowerisbelowmaximum.

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page9(55)

3DEFINITIONS

ThroughoutthisWhitePaper,thefollowingpowerandenergyrelateddefinitionsareused:

Powerconsumption:AmountofenergythatistransferredorconvertedperunittimeiscalledpowerconsumptionanditisexpressedinSIunitasWatt[W].Poweriscalculatedbymultiplyingvoltage[V]andcurrent[I].OtherunitsthanWattareJoules/sorVolt-Ampere[V.A](1V.A=1W).

EnergyConsumption:AmountofpowerusedoveratimeperiodiscalledenergyconsumptionanditisexpressedinSIunitasWattSecond[Ws]orJoules[J].Normallyinelectricitybillskilowatthour[kWh]isused(1Ws=1J,1Wh=3600Ws=3600J).

PowerEfficiency:Theproducedpowerofaunitvstheconsumedpowerforproducingtheoutputpowerperunittimeiscalledpowerefficiencyandnormallyitispresentedaspercentage[%].PowerEfficiency=(OutputPower/InputPower)x100[%].

EnergyPerformance:Theratiobetweentheproducedtaskorworkandtheconsumedpowerforproducingthistaskorworkoveratimeperiodiscalledenergyperformance.ThetaskorworkcouldbeanythingandintelecommunicationitcanforexamplebethedeliveredbitstoaUserEquipment(UE).Inthiscasetheunitcouldbeforexample[Mbits/kWh]or[bits

Wh]or[Mbits/Joules].SincetheelectricitybillsforoperatorsarenormallypresentedinKWhandtheworkcanbeexpressedasdeliveringMbitstoauseritwouldbemoreconvenienttoexpresstheunitas[Mbits/KWh].

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page10(55)

4ENERGYEFFICIENCY

Theenergyefficiencyin5Gsystemscanbeattributedtomultipleadvancements.Theenhancementsindatatransmission,efficiencyincontrolmessagingandsignalling,andabilitytousesleepmodebasedontrafficandloadconditions,areamongthecapabilitiesbydesign.Furthermore,thegranulararchitecture,andincreasingdisaggregationandcloudnativearchitectureandoperation,withvirtualizationandsoftwarization,increaseagilityandreducefootprint.Thispathtowardsintelligentanddynamicorchestration,programmability,lifecyclemanagement,andfullautomation,hasgreatpotentialtobeleveragedtowardsourgoals.Inaddition,products,anddeploymentandoperationalstrategies,havealreadyfocussedoninnovativewaystoadvanceenergyefficiency.

Despitetheseadvancementsinenergyefficiency,theforecastspointtoasignificantincreaseinenergyconsumption,andtherebyinCO?emissions,intheabsenceofactiveintervention,overthenextseveralyears.Thisisduetoaconsiderableincreaseintrafficacrossavastrangeofusecases,newtechnologiesandspectrum,greatdealofconnections,anddensification.Someofthesefactorsneedtobeevaluatedandoptimized,andultimatelyleadtotheneedforintelligentMachineLearning(ML)-basedoperation.Forexample,smallcells,bynature,introduceefficiencythroughcarryingtrafficatlowerenergyconsumption.However,thiscanhaveareverseeffectwithincreasingdensificationandinterference,withoutintelligentdynamicplanningandallocation.Similarly,massiveMultipleInputMultipleOutput(MIMO)hasthepotentialtoincreaseefficiency,intermsoftrafficperunitofenergyconsumption,ifbalancedagainstthecomplexityandconsumptionitintroduces.Trade-offsmayneedtobeconsideredtooptimizeandmaintaindesigngoals.

Advancementinequipmentpowerefficiencyandnetworkenergyperformancemustbeconsideredfromanend-to-endperspective.Thesecanbroadlyinclude:

Specificationsanddesign,suchasthoserelatedtodatatransmission,signalling,sleepmodes,distributedarchitecture,cloud,andre-configurableRadioAccessNetwork(RAN)andEdge.

AI-drivencognitiveandautonomousarchitectureandoperation,thatsupportenergyefficientdynamicplanning,deployment,resourceallocation,monitoringandoptimization,shutdowns,etc.

Powerefficiencyinequipment,devices,boards,andsiteordatacentrecooling,etc.

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page11(55)

Energyharvestingandtransfer,andefficiencyachievedfromcirculareconomy.

Indirectresourcessuchasexternallogistics,andendofequipmentlife.

Ecosystem,regulatory,andsupply-chaincollaborativefocus,awareness,andorganizationalalignment.

MobileNetworkOperators(MNOs),theirpartners,andtheentireecosystemwillnotachievethesustainabilitytargets,withoutanend-to-endcollaborative,committedandorchestratedeffort.Atitshighestlevel,thegoalinvolvestheinter-connectedpillarsofachievingdigitaltransformationwithfull5Gandsubsequently6Grealization,sustainability,andsocialresponsibility.Withthewideandgrowingrangeofusecases,particularlyforautomatedindustries,agreatimpactonothersectorsiswellexpectedandarticulated.

4.1RANSiteOverviewfromanEnergyPerspective

Aradioaccessnetwork(RAN)consistsofalargenumberofsites,whichtypicallyaccommodatetwotypesofequipment:siteinfrastructureequipment(site)andmainequipment(basestation).SiteinfrastructuretypicallyconsistsofrectifiersforconvertingACpowertoDCpower,powerbackupequipment(e.g.,batteries)andotherequipmentsuchasairconditionersforcooling.Themainequipmentisthebasestationequipmentinthecabinet,typicallybasebandunitandradiounits.Fromthebasestation,dataistransmittedovertheairinterface(viaradio)totheuserterminal.

Figure2

providesanillustrationofthisandthetypicalequipmentinvolved,anditalsoillustratestheenergyflowfromthemainACinputfromthegrid,viaDCpowerconversion,deliverytothemainequipment(basestation),conversiontocabinet-toppowerbythebasestation,andfinallytransmissionovertheairinterfacetotheuserterminal.Asshowninthefigure,theenergyflowcanbedividedintothreestages:

Siteinfrastructure:fromtheACmainsupplytotheDCpowersupplyforthebasestation.Theenergyefficiencyofthesiteinfrastructure,SiteEE,canbemeasuredbydividingtheDCinputpowerofthebasestation(PBS)bytheACinputpowerofthesite(PAC).Thismeasureistheinverseofthewell-knownPUE(PowerUsageEffectiveness),whichiscommonlyusedfordatacentres.

Basestation:fromtheDCpowerinput(PBS)tothecabinet-toppoweroutputofthebasestationantenna(Poutput).Thepowerefficiencyofabasestationcanbemeasuredbydividingthecabinet-toppowerPoutputbytheDCinputpowerPBSofthebasestation.

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page12(55)

Airinterface:indicatesthelinkfromtheoutputoftheantennaonthetopofthecabinettotheradiotransmissionreceivedbytheuserterminalovertheairinterface.Theenergyperformanceoftheairinterface,RadioEP,canbemeasuredbydividingtheserviceprovidedbythebasestation(e.g.,deliveredbits,coverage,ornumberofsubscribersservedbythebasestation),Spi,bytheoutputpoweratthetopofthecabinet(Poutput).

Figure2:IllustrationofatypicalRANsite,andenergyflowfrommainACinputtoreceptionattheuserterminal.

Theoverallenergyefficiencyisaffectedbythesethreefactors:powerefficiencyofthesiteinfrastructure,powerefficiencyofthebasestationequipment,andenergyperformanceoftheairinterface.Bymultiplyingthesethreefactors:theSitePowerEfficiency(SitePE),theBaseStation(BSPE),andtheRadioEnergyPerformance(RadioEP),weobtaintheoverallenergyperformance(EP)as:

Whenintegratedoveratimeperiod,wegettheEnergyPerformanceasdefinedinchapter

3,

measuredin,e.g.,bits/JouleormorecommonusedMbits/Wh.

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page13(55)

5HOWTODECREASEENERGYCONSUMPTION

Thebasestationsinamobilenetworkrepresentthelargestpartoftheenergyconsumption,about57%oftotalpowerusageofatypicalcellularnetworkisreportedbyAnalysisMason

By2025,thatfigurecanbelowerwhen5Gbecomesmorewidelydeployed,butstillthebasestationswillstillbethebiggestconsumerofenergy.Foraconvergedoperator,theRANcouldaccountforaround50%ofitstotalnetworkenergyconsumptionacrossfixedandmobilenetworksin2025,accordingtoastudybythreeEuropeanuniversities

[4],

see

Figure3.

Figure3:Energyconsumptionbreakdownbynetworkelement2025,UniversityofSplit.

5Gcanbeusedtoimproveenergyefficiency.Moreefficientpoweramplifiershavebeendeveloped,renewableenergysourcesforpoweringon-gridandoff-gridsites,includingsolarpower,arestartingtobewidelyadopted.Moreover,newgenerationofbatteriesarebecominganintegralpartofany5Gsitetoenhanceenergymanagement,andliquidcoolingisbeingimplementedtoreducetheneedforairconditioning,seechapter

6.

3GPPNRspecificationhasenabledanumberofnewtechnologiesthatcanhelptoimproveenergyperformanceonnetworklevel.Theunique3rdGenerationPartnershipProject(3GPP)NewRadio(NR)powersavingtechnologiesare:

Massivemultiple-inputmultiple-output(massiveMIMO)

Leancarrierdesign

ImprovedSleepModes

Artificialintelligence(AI)

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page14(55)

5.1BaseStationHardware

Figure4:Powerconsumptionofbasestation.

Thelargestpowerconsumerofasiteisthemainequipmentconsistingoftheradiounit,basebandandmaincontrol,whichaccountsforapproximately50%.Thesecondistheairconditioning,accountingfor40%.Whenlookingatthebasestation,theradioprocessingaccountsfor40%.Thisunitconvertsthedigitalsignalfromthebasebandintoamplifiedradiowaves.Sincethepoweramplifieroftheradiounitconsumesmostofthepower,itdefinestheefficiencyofaradiounit.

TheseenergyconsumptionpercentagesmayvarydependingontheTelecomequipmentpowerefficiency,thetechnologyandcapacityofairconditioningunits,theclimateandthelocationofthebasestationetc.Operatorsshouldevaluatetheirenergyconsumptionpercentagesbymeasuringenergyconsumptionfortheirnetworkbeforedecidingwheretofocusonforenergysavings.

Thepowerconsumptionof5GbasestationusingmassiveMIMOcanbedividedintotwomajorparts:theantennaunitandthebasebandunit(BBU).Thepowerconsumptionofantennaunitaccountsforabout90%ofthetotalconsumptionofthebasestation,anditisthemaincomponentofthepowerconsumptionofthebasestation.Theantennaunitpowerconsumptioncanbedividedintopoweramplifier,smallsignal,digitalintermediatefrequencyandpowersupply.

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page15(55)

Figure5:Powerconsumptionofantennaunit.

When2Gwasintroduced,thepowerefficiencyofthepoweramplifierwasbelow20%.Overtimeefficiencyofpoweramplifiershasbeengraduallyimprovedduetotheimprovementofpoweramplifierarchitecturedesignandsemiconductormaterialtechnology.MoreadvancedDohertyarchitecturescombinedwithenvelopetrackingandGalliumNitride(GaN)materialshavebeenappliedinthepoweramplifiers,increasingtheefficiencyofpoweramplifiersmorethan50%.GalliumNitridepoweramplifiersarewidelyusedin5Gequipment,sothepowerefficiencyismuchhigherthan2G/3G/4Gequipmentrunninginthecurrentnetwork.

Thepowerconsumptionofthebasestationchangeswiththetrafficload,andthepowerconsumptionratioofeachfunctionalmodulealsochangesaccordingly.Underthefullloadcondition,thepowerconsumptionofthepoweramplifieraccountsforthehighestproportion,about59%;underno-loadconditions,thedigitalintermediatefrequencyparthasthehighestproportionofpowerconsumption,about46%.Therefore,inthefieldofequipment-levelenergy-savingtechnology,itisnotonlynecessarytoimprovetheefficiencyofthepoweramplifier,butalsotoreducethebasicpowerconsumptionofsmallsignalanddigitalintermediatefrequencymodulesundertheconditionoflowinitialloadof5G.

Foratraditionalradiounit,thepoweramplifieraccountsforthelargestpartofthepowerconsumption.However,withthedeploymentofmassiveMIMO(e.g.64T64R),powerconsumptionforthedigitalcomponentswillincreaseduetoincreasednumberoftransmitters,sincemorebasebandprocessing,e-CPRIprocessingandsupportofalgorithmslikeDigitalPre-Distortion(DPD)andCrestFactorReduction(CFR)isrequired.Therefore,itisimportanttousepowerefficienthardwaresuchasApplicationSpecificStandardProducts(ASSPs)andApplication-SpecificIntegratedCircuits(ASICs)andtomakesurethatthesilicon

GreenFutureNetworks–NetworkEnergyEfficiency

Version1.0,03-October-2021 Page16(55)

designornextgenerationprogrammableradioscanpoweroffmodulesatlowormediumload.

5.1.1VirtualizationofRAN

DecouplingofsoftwarefromthehardwareintheICTindustryandmorespecificallythemobilecorenetworkhasbeengoingonforseveralyears.Datacentresarebe