Sunteți pe pagina 1din 11

LTEBasicProcedureGoBackToIndexHome:www.sharetechnote.

com

MIMO(MultipleInputMultipleOutput)

IthinkeverybodywouldknowwhatMIMOis.Simplyput,itisatechniquetoincreasethedatathroughputbyusing
multipletransmitterantennaandmultiplerecieverantenna.

Followingsaretopicsthatwillbecoveredinthispage.
Basic..Basic..Concept
OverallConfigurationandProcedure
MathematicalModelingofMIMO
ConsiderationsforReality
RankIndicator
ConditionNumber

<Basic..Basic..Concept>

FollowingillustrationisshowingaverybasicandoversimplifiedconceptofMIMO.Inverystrictsense,itmaybea
littlemisleadingbutthiscanshowthemostcommonsenseofMIMOconcept.
Asyouseeintheillustration,withMIMOmutipleindependentdatastreamsaretransmittedsimulteneouslyandasa
resultitcanachievemuchhigherdatathroughputincomparisontoSISO(SingleInputSingleOutput)case.However,
thereisalmostnoadvantageintermsofreliabilityofdatatransfer(e.g,lesserror)comparingtoSISOcase.

ThereisanotherconceptwhichisoftenconfusedwithMIMO.ItisDiversity.Whatwouldbethedifferencebetween
MIMOandDiversity?
FollowingillustrationwouldshowtheconceptofDiversity.TherearebasicallytwotypesofDiversitycalledReciever
DiversityandTransmitterDiversity.

Asyouseeintheillustration,inthisDiversityconfigurationSinglecopyofonebitstreamisbeingtransmittedand
reachestomultiplerecieverantennaviaalittledifferentpath,itmeanstherecievercanhavemultipleversionsof
samedata.Outofthemultipleversion,therecievercanselectthebestoneorcombinethemalltogetherinsucha
waytoimprovedataquality.Bydoingthis,communicationreliability(lesserror)canbeincreased..butno
advantageintermsdatathroughput.

HeregoesanothertypeofDiversity.InthisDiversityconfiguration,Multiplecopiesofone(single)bitstreamisbeing
transmittedviamultipleTxantennaandreachestosinglerecieverantennaviaalittledifferentpath,itmeansthe
recievercanhavemultipleversionsofsamedata.Outofthemultipleversion,therecievercanselectthebestoneor
combinethemalltogetherinsuchawaytoimprovedataquality.Bydoingthis,communicationreliability(lesserror)
canbeincreased..butnoadvantageintermsdatathroughput.

So..overallconceptissimple,butdetailedprocessandimplementationisnotassimpleasyoumaythink.Ifyoujust
wanttogetacommonsenseofMIMO.Thiswouldbeenough.

However,ifyouwanttoknowverydetailsofMIMO/Diversityespeciallyaboutwhat'sgoingoninphysicallayer,it
wouldbealmostimpossibletounderstandthedetailsofMIMO/Diversityconceptwithoutgoingthroughunderlying
mathmatics.Actuallymathitselfisnotthatcomplicatedbutinterpretingthemathandtranslatethemathintoreal
implementationisdifficult.Ihopemymathsectioninthissitewouldhelpfulonthis.(PleaseseetheEngineering
MathMatrixasyougoalongthissection).

HoweverdescribingandexplainthefulldetailsofMIMOmathwouldbehugetaskanditwouldtakeseveralmonth..
meaningthispagewillgetupdatedforawhile.

<OverallConfigurationandProcedure>

Beforejumpintomathematicaldetails,IwanttosummrizeoverallMIMOconfigurationsthataremostcommonly
acceptedordiscussed,andthePHY/MACproceduresthatareusedtoimplementeachconfiguration.Thisispurelymy
personalclassification.Youmayagreeintermsofbigpicture,butmaynotagreeindetailedlevel.Thepointisto
showseveralkeywordsandkeytechnologiesrelatedtoMIMO.
Idon'tthinkIcandescribe/explainalloftheseatmathematicallevel(atleastinnearfuture),IwillbesatisfiedifI
cansuccessfullydescribeatleast"SUMIMOwithOpenLoop".

<MathematicalModelingofMIMO>

Incaseof2x2,overalldatatransmissionprocesscanbeillustratedasfollows.Theredarrowandfourblocks
(h11,h12,h21,h22)betweenthetwoantennaistoillustratethepossibledatapathbetweenthetwoTxandtwoRx
antenna.

h11,h12,h21,h22arespecialnumbers(coefficient)toshowhowmuchofthedataisgoingthrougheachofthe
possiblepath.Thegreaterthevalueis,thelargerportionsofdataisbeingtransmittedinthatpath.Amatrixwhichis
madeupofthesechannelpathcoefficientiscalled"ChannelInformationMatrix".Therecieverandtransmitter
relationshipisrepresentedasfollows.

AsIsaidthemathitselfisverysimple.Itwouldbelikeafirstexampleatthefirstchapterofanylinearalgebra
book.Theimportantthisishowtointerpretthisequationintorealimplementation.IfIinterpretitinanillustration,
itwouldbeasthepathrepresentedingreenarrows.Bythisway,wecanmathematicallyrepresenttherecieveddata
(datadistortedbytheairpathbetweentransmitterandrecieverantenna),butjustcalculatingtherecievedisnotour
goal(notthefunctionofthereciever).Ourgoalistoextract/figureoutthe'transmitteddata(originaldata)'.
Mathematicallythisisalsosimpleanditisrepresentedingreenarrows.

Asyouseeintheprocessillustratedingreenarrow,torecoverthetransmitteddata(originaldata)fromthe
receiveddataweneedtotakeinverseofthechannelinformationmatrix.Unfortunatelythereareacoupleofissues
withthismethod.
i)Notallmatrixisinvertable.Therearesomematrixforwhichinversematrixdoesnotexists.(Pleasereferto
Matrixsectionandseewhatkindofmatrixistheonewhichisnotinvertable).
ii)Calculatingtheinversematrixisnotthesimpleprocess.
Toworkaroundtheseproblems,wechangethechannelinformationmatrixintothreematricesbythemethodcalled
SVD(SingularValueDecomposition).SeetheSVDsectioninMarixpageandclearlyunderstandthemeaningofthis
process.ProbablymypagewouldnotbeenoughtogiveyoufullunderstandingaboutSVD.Googleasmuchmaterials
aboutthisaspossibleandtrytohavesome"Intuitive"understandingoftheconcept.
Whenweapplyanymathematicaltechniqueinengineeringarea,itshouldbemeaningfulintheengineeringsense
andshouldbeimplementable.ThemeaningandimplementationofSVDcanbeillustratedasfollows.

Theimportantpointisthatbyimplementingthismethod,wecanexpressthechannelinformationmatrixasasimple
'diagonalmatrix'.Thatdiagonalmatrixisexpressedasfollowsandcanbeillustratedasfollows.Isn'titlooksimple
andclear?:)

ForsomeMIMOimplementation(e.g,TM4inLTE=ClosedLoopMIMO),youestimatethechannelandselecta
specificprecodingmatrixandsenditbacktothereceiverasshownbelow.

ThealgorithmbywhichUEselectthecodebookwhichisbestfitforthechannelatspecificmomentisasfollows.

Onceaspecificcodebookelement(precodingmatrix)isselectedasshownhere,thatprecodingmatrixisusedto
transformtheincomingbitsasexplainedinPrecodingsection.

<ConsiderationsforReality>

NowyouknowaboutvariousaspectofMIMOchannelmodelandconceptualmathematicalrepresentation.Ihopeit
makesensetoyouatleastintermsofmathematically.Butasyouknow,whenyoutrytoimplementthe
mathematicalmodelintoarealsystem,almostalwaysyouwouldmeetsomethingyouhaven'timaginedwhenyou
areplayingonlyinmathematicalworld.

Let'slookintofolloingmathematicalmodelagain.

Canyouguesswhatisknownvariableandwhatisunknownvariableforeachside(recieverandtransmitterside)?
Wecanthinkoffollowingsasakindofknownvariables.
Thethreematrixatthecentercanbeaknownatleasttotherecieversincetherecievercanestimatechannel
MatrixHfromthereceivedsignaandcalculatethesematrixfromH.
Thevectorxisknowntotransmittersinceitisjustthedatathat'stransmitted.
ThematrixUcanbeaknowntotherecieversinceitcanbederivedfromH.
TheissueishowtofigureoutthematrixV.Aboutthismatrix,therearetwomainproblems.
Thisistheonethatshouldbeusedbytransmitter,buttransmitterdoesnothaveanyinformationaboutH.Soit
cannotbedirectlycalculatableontransmitterside.
ThismatrixisusedintransmittersideBEFOREthedata(thevectorx)istransmitted.Howcantransmitter
knowsomethingbeforeithappens?
Oneofthemostcommonsolutionforthisisthattherecieveranalyzechannelmatrix(H)andinformthetransmitter
ofthebestmatrixV.Thisiscalled'ClosedloopMIMO'andLTETM4(TransmissionMode4)belongstothiscategory.


Nowthiskindofclosedloopmethodcansolvealltheproblemslistedabove?UnfortunatelyNo.
Whataretheproblemsnow?
Theproblemsthatyoucaneasilythinkofwouldbeasfollows:
EventhoughUEcanfigureoutthematrixV,itisfromthedataitjustrecieved.Ifitsendthisinformationto
transmitter,thetransmitterwillusetheinformationforNEXTtransmission.Ifthereisnotmuchdifferencein
channelconditionbetweenthecurrenttransmissionandnexttransmission,thisinformationfromrecievercan
helpalotfornexttransmissionontransmitter.Butifthechannelconditionchangessofast,theinformation
maynotbeuseful.
EventhoughUEfiguredoutcorrectmatrixV,itishugeamountofdataandoverheadtosendthewholematrix
itselftothetransmitter.InsomecaselikeLTE,theyuseakindofaworkaround.Intheworkaround,weuse
onlyafew/severalpredefinedmatricesforVandthereceiverevaluateeachofthecandidatesandpickthe
bestoneandinformthetransmitteroftheindexofthecandidatematrixinsteadofthewholematrixcontents.
Forexample,inLTETM4theydefinedonly4candidateVmatrix(precodingmatrix)for2x2MIMOand16
candidatesfor4x4MIMO.Theselectedcandidatemaynotbethebestfitintermsofmathematics,butitcan
beabesteffortintermsoftradeoffbetweenmathematicalaccuracyandreportoverhead.
InLTETM8,9,10theyintroducedamethodbywhichthetransmitercanuseanyprecodingmatrix(V).Inthis
case,thetransmitterisusingspecialreferencesignalcalledUEspecific(recieverspecific)referencesignalto
helpbetterdecodingontherecieverside.

<RankIndicator>

ThetermRankIndicatorcamefromthepuremathematicaltermRankIndex.So,simplyput,RankIndicatorisa
specialkindofUEmeasurementthatreportsRankIndex.

Puttingitinmorepracticalaspect,RankIndicatorisakindofnumber(indicator)thatrepresentshowwellaMIMO
communicationworks.Forexample,let'sassumethataUEiscommunicatingwithaeNBwith2x2MIMO
configurationandUEreportsRI(RankIndicator)of'2',itimpliesthatthe2x2MIMOisperformingasarealMIMO
fashion.However,ifUEreportsRIof'1',itimpliesthatthecommunicationisgoingonasifitiswithsingleantenna.
Itmeansthatthe2x2MIMOisnotperformingwiththebestefficiency.

Explainingthesamethinginamoreformalway,itwouldgoasfollos:

Let'sremindusofthedefinitionofSingularMatrix.Itisdefinedasfollows.Itisadiagonalmatrix.Onceyougetthe
singularmatrixofachannelyoucanextractacoupleofveryimportantinformationfromit.

OneofthemostimportantinformationyoucangetfromthesingularmatrixisRankIndicator.RankIndicatoris
definedasNumberofthediagonalelementswhichisnotzero.PhysicalmeaningofRankIndicatoristhenumberof
independentcommunicationchannel.Incaseof2x2Singularmatrix(for2x2MIMO)asanexample,iftheRank
Indictoris2.Bothdiagonalelementsasshownbelowisnonzerovalueandtwoseparatecommunicationpipeis
established.IftheRankIndicatorofthematrixis1,oneofthediagonalelement(Lamda1orLamda2)iszero.It
meanseventhoughwesetuptwoantenna,thereallyworkingdatastreamisonly1,implyingthatthesignalfrom
oneoftheantennaistotallygotlostorburriedundernoisewhichcannotbedecoded.

Example1>RIReportfromadrivetestonaHighWayunder2x2TM3

FollowingplotisfromthedatacapturedbyadrivetesttoolAzenqosDriveTesttool(AZQAndroid).Igotthelog
capturedbythetoolandexportedthedataascsvfileandthenplotitonMicrosoftExcel.Themapdisplayingthe
pathofthemeasurementshownatthebottomistheoneautomaticallycreatedbyAZQreportingtool.
Inthisspecificexample,youwouldseeaconsiderableamountofreportwithRI=1,itimpliesthatinmanycase
2x2MIMOdoesnotworkwiththefullefficiency.

<ConditionNumber>

Thenyoumighthavequestion.RankIndictorcaresonlyonthenumberofnonzerovaluesinthesingularmatrix.
Doesthismeanthattheamountofthesenumberdoesn'tmatter?Forexample,isn'tthereanydifferenceintermsof
realcommunicationperformancebetween(lamda1=1,lamda2=1)and(lamda1=2,lamda2=0.1)?Inboth
case,RankIndicatorissamebutrealcommunicationperformanceisdifferent.Soinordertoproperlyestimatethe
realperformanceofMIMOchannelwemayneedanotherindicator,thatindicatorisConditionNumberasdefinedas
below(accordingtoreference[1].Istronglyrecommendyoutoreadthisreference)

Asstatedabove,asthedifferenceamongthediagonalnumbersarelessandless,themoreoptimaltheMIMO
channelworks.Followinggraphfrom[1]showsthisproperty.HorizonalaxisindicatesConditionNumberindBunit
andVerticalAxisindicatestheSNRindBunit.Eachlineshowsthecasewherethesameamountofdatarateis
achieved.Eachpointsonthelineindicate'Toachieveacertainperformance(datarate)atacertainconditionnumber,
howmuchSNRisrequired?".oryoucaninterpretitas"Howmuchperformance(datarate)youcangetwhenyou
weregivenacertainconditionnumberandacertainSNR".
Forexample,let'slookatthepointlabed(B).Thispointtells"Inthisspecificcommunicationsystem,whenthe
conditionnumberis10dBandSNRis20dB,wecanachievemax12bits/s/Hzdatarate".Inorderwords,"to
achieve12Bits/s/Hzdatarateattheconditionnumber10dB,weneedatleast20dBSNR".
Nowlet'slookatpoint(A).Ittells"Inthisspecificcommunicationsystem,whentheconditionnumberis0dBand
SNRis16dB,wecanachievemax12bits/s/Hzdatarate".Inorderwords,"toachieve12Bits/s/Hzdatarateatthe
conditionnumber0dB,weneedatleast16dBSNR".
Whatwouldyoutellifyoucomparepoint(A)andpoint(B)?Atbothpoints,wecanachivethesameperformance
(datarate),butatpoint(A)wecouldachievethedatarateevenwithpoorerSNRcomparingtopoint(B).Inother
words,asConditionNumbergetcloserto0dB,wecanachievethedataratewithpoorerSNR.
TrytopicksomeotherplacesanddescribethesituationasIdiduntilyougetfamiliarwiththeinterpretationofthe
graphandyouwillgetthepracticalunderstandingofConditionNumber.

Nowlet'slookintoacoupleofexampleofchannelmatrixandseehow/whatinformationwecangetoutofthem.The
examplematrixitselfcomesfromreference[1]andthetoolthatIusedtoperformSVDisreference[2](Wolfram
Alpha).

Firstexamplegoesasfollows.Canyouvisualizethephysicalsetupoftransmitterandrecieverfromthismatrix?
Sinceitis2x2,youwouldhavetwotransmitterantennaandtworecieverantenna.Itis2x2matrix.Theelements
exceptthoseondiagonallineareallzero.Itmeansthatthereisnocrosstalkbetweeneachantennapath.Allthe
elementsinthediagonallineareall'1'.Itmeans100%energyfromtransmitterantenna1isdeliveredtoreciever
antenna1withoutanylossand100%energyfromtransmitterantenna2isdeliveredtorecieverantenna2without
anyloss.Ofcourse,youwouldneverhavethiskindofconditioninrealwirelesscommunication.Butyoucanhave
thiskindofconditionwhenyoudoalabtestconnecting2DUDantennato2communicationportonatestequipment.

Nowlet'sdoSVDtogetSingularMatrix.UsingWolframAlpha,Igotthefollowingresult.

Nowlet'sextractsomeinformationfromthesingularmatrix.First..whatisRankIndicatorvalue?Itis2sincethe
numberofNonzeroelementsondiagonallineis2.
Thenlet'staketheconditionnumber.Itis'1'asshownbelow.Itmeansitisperfectconditionforspatialmultiplexing
andyoushouldhavemaximumperformancefromthischannel.

Secondexamplegoesasfollows.Canyouvisualizethephysicalsetupoftransmitterandrecieverfromthismatrix?
Sinceitis2x2,youwouldhavetwotransmitterantennaandtworecieverantenna.Itis2x2matrix.Allthe
elementsevennondiagonalelementsarenonezero.Itmeansthatthereiscrosstalkinallthepossibleways.All
theelementsinthematrixarearesamevaluewhichis1/sqrt(2).Itmeans50%energyfromtransmitterantenna1
isdeliveredtorecieverantenna1andanother50%energyfromtransmitterantenna1isdeliveredtoreciever
antenna2.anditalsomeans50%energyfromtransmitterantenna2isdeliveredtorecieverantenna1and
another50%energyfromtransmitterantenna2isdeliveredtorecieverantenna2.

Nowlet'sdoSVDtogetSingularMatrix.UsingWolframAlpha,Igotthefollowingresult.

Nowlet'sextractsomeinformationfromthesingularmatrix.First..whatisRankIndicatorvalue?Itis1sincethe
numberofNonzeroelementsondiagonallineisonly1.
Thenlet'staketheconditionnumber.Inpuremathematicalsense,itcannotbecalculatedsincethedenominatoris
zero.Butinpracticalsense,youcanthinkof'0'asaverysmallnumber.Inthatcase,youcangetthecondition
numberasshownbelowwhichisverylargenumber.ItmeansthischannelistheworstconditionforMIMO.

Thirdexamplegoesasfollows.Canyouvisualizethephysicalsetupoftransmitterandrecieverfromthismatrix?
Sinceitis2x2,youwouldhavetwotransmitterantennaandtworecieverantenna.Itis2x2matrix.Allthe
elementsevennondiagonalelementsarenonezero.Itmeansthatthereiscrosstalkinallthepossibleways,but
thevaluebetweendiagonalelementandnondiagonalelementshas90degreephaseshift.Itmeans50%energy
fromtransmitterantenna1isdeliveredtorecieverantenna1andanother50%energyfromtransmitterantenna1is
deliveredtorecieverantenna2with90degreephaseshift.anditalsomeans50%energyfromtransmitterantenna
2isdeliveredtorecieverantenna1with90degreephaseshiftandanother50%energyfromtransmitterantenna
2isdeliveredtorecieverantenna2.Itmeanstwosignalstransmittedfromtwotransmitterantennaandcomming
intoonerecieverantennadoesnotinterfereatall.

Nowlet'sdoSVDtogetSingularMatrix.UsingWolframAlpha,Igotthefollowingresult.

Nowlet'sextractsomeinformationfromthesingularmatrix.First..whatisRankIndicatorvalue?Itis2sincethe
numberofNonzeroelementsondiagonallineis2.
Thenlet'staketheconditionnumber.Itiscalculatedtobe1whichmeanstheperfectconditionforMIMO.Asyousee
here,eventhoughthereiscrosstalkbetweentwocommunicationpath.ItcanbeaverygoodconditionforMIMO

dependingonthecharacteristicsofthecrosstalk.Wouldtherebeanywaytoimplementthiskindofchannelcondition
?Yes..thereis.Ifyoucrosspolarizetheantenna,youcanhavethiskindofchannelcondition.

Heregoesanotherexamplewhichcanbemorerealistic.Canyouvisualizethephysicalsetupoftransmitterand
recieverfromthismatrix?
Sinceitis2x2,youwouldhavetwotransmitterantennaandtworecieverantenna.Itis2x2matrix.Allthe
elementsevennondiagonalelementsarenonezero.Itmeansthatthereiscrosstalkinallthepossibleways,butit
isnoteasytointuitivelyknowexactlyhowmuchengergygoestowhichantenna.Ofcourseyoucancalculatethisby
takingthemagnitueandangleofallofthesecomplexnumbers..butlet'sjustthinkofintuitivemeaning.Sincethe
magnitudeandangleofalltheelementsarenotthesame,thecrosstalkbetweenthecommunicationpathswouldnot
go100%destructiveway.

Inthiskindofsituation,SVDcanplayimportantrole.IgottheresultasbelowusingWolframAlpha.

Nowlet'sextractsomeinformationfromthesingularmatrix.First..whatisRankIndicatorvalue?Itis2sincethe
numberofNonzeroelementsondiagonallineis2.
Thenlet'staketheconditionnumber.Itiscalculatedtobe1.6563whichisnotthebestcaseforMIMOandnotthe
worstcaseeither.Itislessthan3indBunit.Ifyouplugthisnumberintothegraphyousawabove,itisprettygood
conditionforMIMO.

Reference
[1]AssessingaMIMOChannelRohde&Schwarz
[2]WolframAlpha

S-ar putea să vă placă și