You are on page 1of 8

Mobile&Wireless

Inthisblog,Icapturemythoughtsandanalysison
mobileandwirelesstechnologiesthataresetto
dominatethewaywelivenowandinthefuture.
2007.Pleaseseekpermissionforusingarticles.
Feeds:

Posts

Comments

LogicalChannelMultiplexinginWCDMAMAC
December19,2007byitsme
Multiplexingatanylevelisawayoflifebecausethereisgreatdealofsenseindoingit.Inprotocol
engineering,peoplehavelongrealizedthatpacketswitchednetworksoffermuchbetternetwork
capacityutilizationthancircuitswitchednetworks.InCSconnections,thelinkortrunkisreserved
fortheentiredurationofthecall.Periodsofinactivityorlessthanmaximumrateleadtowastageof
resources.InPSnetworks,anyunusedbandwidthisquicklydeallocatedandreallocatedtousers
whoneedthem.Moreimportantly,whenasingleuserhasactivatedmultipleservices,allservices
couldintelligentlysharethesameunderlyingbandwidthwithoutrequiringfullydedicated
resources.
Multiplexingisnothingbutbringingtogetherdifferentstreamsofdataintoasinglestream.While
therearedifferenttypesofmultiplexing,inthispostweareprimarilyconcernedwithtime
multiplexingofdata.Multiplexingrequiresintelligenceintermsofscheduling.Inotherwords,when
multipleservicesarewaitingtoshareasingleresource,someonehastointelligentlyschedulethem
basedonpriority,bufferoccupancy,power,interferenceandothersuchfactors.
Inthispost,IwilllookatmultiplexingdoneatMACforR99transportchannels,Rel5HSDPAand
Rel6HSUPA.Simplifyingthewholearchitecture,itissufficienttonotethatforR99,independent
datastreamsarecarriedinlogicalchannels.Theremaybesomedependencebetweenlogical
channelsthatarrivefromthesameapplication(suchasNBAMR)butthisdoesnotmattertoMAC
wheremultiplexingisdone.MACwouldhavebeenconfiguredappropriatelytohandlethese
dependentstreamsinparallel.Theselogicalchannelsaremultiplexedtotransportchannels.Once
configured,alogicalchannelcanbecarriedononlyonetransportchannelunlessRRCremapsittoa
differenttransportchannel.Multipletransportchannelsexistatthesametime.Theseareagaintime
multiplexedontheCCTrCh.ThusforR99,MACmultiplexinghappensontwolevels.
MACdecideswhattomultiplexforagivenTTI.However,ithastoobeytheconstraintsetbythe

MACdecideswhattomultiplexforagivenTTI.However,ithastoobeytheconstraintsetbythe
TFCS.TheTFCSdetailsthesetofmultiplexingpossibilities.Thissetisformedbasedonthe
underlyingphysicallayercapabilitytotransmitdataandtheaveragebitratepromisedforagiven
service.
Figure1:LogicalChannelMultiplexinginR99MAC

AnexampleofR99MACmultiplexingisinFigure1.Forthisexample,theTTIsofTrCh1,TrCh2
andTrCh3are10ms,10msand20msrespectively.Wecannotethefollowing:
Onanytransportchannel,onlyonelogicalchannelcanbecarriedatanyonetime.Logical
channelwillbeidentifiedintheMACheader.
EachlogicalchannelhasaPDUsizeofitsown.
ItispossiblethatinagivenTTI,multipleMACPDUscanbecarriedonthesametransport
channel.Thisisthecaseattime(t+20)forTrCh1.SpecificationsdonotrequirethatthesePDUs
shouldcarrythesamelogicalchannelsinceeachtransportblockhasitsownC/TfieldinMAC
headertoidentifythelogicalchannel.However,currentspecificationrequirethatallPDUs
withintheTTIforatransportchannelmustbeofthesamesize.
ItispossiblethatinsomeTTIs,nothingiscarriedforacertaintransportchannel.Thisissensible
ifalllogicalchannelsmappedtothechannelhavenodatatosend.Thiscanalsohappenifthere
arehigherprioritylogicalchannelsthatneedthebandwidth.ThisisthecasewithTrCh2attime
(t+20).
WeoughttonotethatTrCh3isofTTI20ms.Weassumeforthisexamplethattime(t)isona20
msboundary.MACdecidestotransmit2blocksofdataatthistime.Thisissentonframes
startingat(t)and(t+10).
Figure1isagrosssimplificationtodemonstratemultiplexingatMAC.Actualtransmissionfrom
PHYhappensaftermanystagesofprocessingthatincludeinterleaving,radioframe
equalizationandradioframesegmentation.ThesizeofMACheaderisvariableandnotfixedas
suggestedinthefigure.
SinceTrCh2carriesonlyonelogicalchannel,MACheaderwillgenerallynotcarryany

SinceTrCh2carriesonlyonelogicalchannel,MACheaderwillgenerallynotcarryany
identificationforthelogicalchannel.However,MACheaderwillbestillpresentbecauseithas
manyothercomponents.
Figure2:LogicalChannelMultiplexinginHSDPAMAC

LetusnowturntoHSDPAforwhichwerefertoFigure2.Letusnotethefollowing:
Thereisonlyonetransportchannel.OnlyonetransportchannelcanbecarriedontheCCTrCh.
AlthoughTFCSisdefineditisnotreallyacombinationasinR99channels.
ControlinformationformultiplexingisembeddedwithintheMAChsheaderwhileforR99this
controlinformationisinthedefinitionoftheTFCS.Becauseofthis,MAChsheaderishighly
variableinsize.
AnewconceptofMACdflowhasbeenintroducedforHSDPA.Thisismainlyusedto
differentiatetrafficontheIurandIub.ThisappliesforHSUPAaswell.
MultiplelogicalchannelscanbecarriedonthesameMACdflow.
AMACdflowcanbecarriedonmultiplepriorityqueues.Inotherwords,somelogicalchannels
oftheflowareononequeuewhileothersareonanotherqueue.ThisiscasewithFlow1and
Flow2inthisexample.
ApriorityqueuecanbeusedtohandlemultipleMACdflows.ThisisthecasewithQ2inthis

ApriorityqueuecanbeusedtohandlemultipleMACdflows.ThisisthecasewithQ2inthis
example.
ATTIcarriesonlyonetransportblock.However,theMAChsPDUcancontainmultipleMAC
hsSDUs.Theheaderindicatessuchmultiplexing.
MAChsPDUcancarryMAChsSDUsofdifferentsizes.ThisisunlikeR99.Exampleofthisis
attime(t+2).
AMAChsPDUcancarrySDUsfromonlyoneMACdflowandonlyonepriorityqueue.Thus,
althoughattime(t+2)twodifferentlogicalchannelsarecarried,theybelongtothesameMACd
flowandthesamepriorityqueue[TS25.308,6.1.4].
InR99,C/TfieldofMACheaderispresentiflogicalchannelsaremultiplexedonthesame
transportchannel.InHSDPA,MACdusesthesamefieldwhenmultiplexingisdoneonthe
sameMACdflow.
Paddingisusedtofitatransportblocksizecompletely.
Figure3:LogicalChannelMultiplexinginHSUPAMAC

EDCHmultiplexingisshowninFigure3.ThefollowingstatementscanbemadeofHSUPAMAC
multiplexing:
TwoTTIsaresupported:2msand10ms.Figure3representstheformer.
AsinHSDPA,aTTIcancarryonlyoneTB.ACCTrCHcancarryonlyoneEDCH.
MultiplexinghappensatMACe,ratherthanMACes.MACesonlyaddstheTSNfieldinits
header.ActualmultiplexinginformationisinMACeheader.
InHSDPA,multiplexinghappensattwolevels:atMACdwhichmayaddC/TfieldtoaMACd
SDUandatMAChs.InHSUPA,thereisonlyonelevelofmultiplexing.Allofithappensat
MACe.MACesSDUissimplyRLCPDU.MACddoesnotaddanyheader.Thisisbecausethe
informationinMACeheadermapsdirectlytoalogicalchannel.ThisisnotthecasewithMAC

informationinMACeheadermapsdirectlytoalogicalchannel.ThisisnotthecasewithMAC
hsheaderwhichmapsonlytothelevelofaMACdflow.
LikeHSDPA,thereisareorderingqueue.ThisisimportantintheSRNCthatdoesthe
reorderingofMACesPDUs.However,inHSUPA,thereisonereorderingqueueforevery
logicalchannel.ThismakessensebecausedemultiplexingisalreadydonebyMACebefore
forwardingdatatoMACes.ReorderingqueueisnotthatimportantattheUEandhenceitis
notexplicitlyshowninthefigure.
Headerinformationisnotcontiguous.Inparticular,MACeandMACesheadersareseparate.
ThereisanoptionalpaddingasinHSDPAtofitadefinedTBsize.Inaddition,Scheduling
Information(SI)canbesentifthereisenoughspaceorifthereisarealneedtosendit.
MultipleMACdPDUsofalogicalchannelcanbesentinthesameMACePDU.However,they
shouldbeofthesamesize.
Whenmultiplelogicalchannelsaremultiplexed,theMACdPDUsizecanbedifferentfromone
logicalchanneltoanother.
ItispossibleforaMACePDUtocarrydatafromdifferentflows[TS25.209,7.1.1].Anexample
ofthisisattime(t+4).
AnadditionalcommentisthatalogicalchannelcancarryRLCPDUsofmultiplesizes.Tomy
knowledge,NBAMRClassAbitsistheonlyonethatneedsthisfeatureatthemoment.Eveninthis
case,itisunlikelythatPDUsofdifferentsizesofthesamelogicalchannelwouldneedtobesentin
thesameTTI.ThismaybeonereasonwhyinHSUPAthereisaminorconstraintonmultiplexingas
alreadynoted.Otherwise,thedesignallowsgreatflexibility.
Inconclusion,therearesignificantdifferencesinthethreemethodsofmultiplexingdiscussedabove.
Onthesurfacetheymayappeartobesimilarbutwhenwedelveintothedetailswecanappreciate
thedifferences.Thisalsohighlightsthewayspecificationshaveevolved.WhileR99providesalayer
ofmultiplexingatCCTrCh,thisisnolongerusedinHSPA.Iseethisasanimprovementbecause
HSPAmultiplexingallowsforgreaterflexibility,quickerresponseandeasierimplementation.For
example,inHSPAweneednotdoratematchingformultipletransportchannels.InR99,ifthe
CCTrChcarriesa80mstransportchannel,theTFCSreducestoasubsetuntilthenext80ms
boundary.ThisrestrictsMACschedulingtoacertainextent.Toavoidthis,amuchlargerTFCS
needstobesignalledatRRC.BlindTFCITFdetectionisnotapplicabletoHSPA.Fixed/flexible
positioningoftransportchannelsisnotapplicableeither.Theseextracomplicationsareremovedby
usingasingletransportchannelperTTIforHSPA.
PostedinTechnicalNotes|TaggedMAC,MACdflow,multiplex|12Comments

12Responses
Asha
Hello,

onDecember1,2008at5:08pm|Reply

CanyoupleaseprovidesomenotesforRelease7logicalchannelmappingtodifferentpriority
quesues?
Thankyou.

vijay
whydontyouwriteabook?

onDecember1,2008at11:17pm|Reply

Ansa
Hello,

onDecember31,2008at8:55am|Reply

CanyoupleaseelaborateMacdismainlyusedtodifferentiatetrafficontheIurandIub.
Thankyou.
Abhishek
onJanuary14,2009at8:41pm|Reply
Hi,
Canyoupleaseexplainalittle
ThismakessensebecausedemultiplexingisalreadydonebyMACebeforeforwardingdatato
MACe.
whatisthemeaningofaboveparagraph.
Cheers,
Abhishek
arvindpadmanabhan
onJanuary14,2009at9:31pm|Reply
Abhishek,
Thanksforpointingitout.Ihavecorrectedthemistakeas:forwardingdatatoMACes.
murphy
Hi,

onApril2,2009at10:29pm|Reply

Ithinkthefollowingwordsisnotright.
ApriorityqueuecanbeusedtohandlemultipleMACdflows.
Ifitisright,howcanUEdifferantiatetheMAChsPDUsifromdifferentMACdflowsinits
priorityqueue?
TheMACdflowidisnotinMAChsPDU.
arvindpadmanabhan
Murphy,

onMay9,2009at7:55pm|Reply

QuiterightthatUEcannotdifferentiateMacdflows.Infact,theconceptofMacdflowsisused
onlybythenetworkonIubandIur.WhenUEreceivesMachsPDUsitdoesntcarewhichMac
dflowitbelongsto.
suneelkumar

onOctober9,2009at6:31pm|Reply

CanyoupleaseprovidesomenotesforRelease7logicalchannelmappingtodifferentpriority

CanyoupleaseprovidesomenotesforRelease7logicalchannelmappingtodifferentpriority
quesues?
KhasimShaheed
HelloArvind,

onApril18,2010at8:06pm|Reply

Veryusefulinformation.IthinkfollowinginformationisnottrueasMurphypointedout:
ApriorityqueuecanbeusedtohandlemultipleMACdflows.
OneMACdflowmaybemappedontomultiplequeuesbutviceversaisnotpossible.
Also,MACdflowinformationisrelevantatUEaswell.ItwillbesignaledtotheUEviaRRC.
Andinfact,UEshouldmapthecorrespondingMACdflowfromthequeueidreceivedinMAC
hsheaderasMAC.
Anycomments.
Thanks
ArunNair
HelloArvind,

onAugust13,2010at11:43am|Reply

Thanksfortheinfo.
ButthepointApriorityqueuecanbeusedtohandlemultipleMACdflows.,canbeputin
adifferentwaylike
WhenaMACdflowarrivestheMAChsentitypresentinNodeB,basedontheSPI
(schedulingpriorityindicator),theMAChsputstheMACPDUintoapriorityqueue.
AccordingtoUMTSstandardstherecouldbeupto8priorityqueues,butagainitdepends
uponimplementation.
AndaboutSPIvalues,usingthisvaluethecoren/wmentionstowhichMACdflowhastobe
givenhigherpriority.
Thanks.
Scanda
onMay18,2011at7:20am|Reply
Hi
WhatsmeantbythefollowinglinesinHSDPAOverallDescriptiondocument(ie.,Section6.1.2:
DetailsofMACdinTS25.308)?
OnereorderingbuffermapstooneC/TMUXentityandmanyreorderingbufferscanmapto
thesameC/TMUXentity.IfMACehsisconfigured,C/TMUXtowardMACehsisnotused.
jiji
onJanuary20,2012at5:10pm|Reply
Machs:Amacdflowcanhavemultiplelogicalchannels.Andthemacd
pduscangointodifferentpriorityqueuesbasedonthepriority.Atmacd
thereisaC/T,whichmultiplexesmanylogicalchannelsintooneatone

thereisaC/T,whichmultiplexesmanylogicalchannelsintooneatone
pointintimeandsendsthemacdpdudowntomachs.thelogicalchannel
tomacdflowandthemacdflowtopriorityqueuemappingispresentin
RRCsignalling.eachmacdpduhastheinformationwhichlogicalchannel
itbelongsto.
macehs:thelogicalchannelinformationisavailableinmacehsdata
blocks.thelogicalchanneltopriorityqueuemappingisavailblefrom
higherlayers.SomacdC/Tisnotneeded.

CommentsRSS

BlogatWordPress.com.
TheMistyLookTheme.

You might also like