DPCCH Gating Gain for Voice over IP on HSUPA

Transcription

1 DPCCH Gating Gain for Voic ovr IP on HSUPA Oscar Frsan, Tao Chn, Esa Malkamäki, Tapani Ristanimi Institut of Communications Enginring, Tampr Univrsity of Tchnology P.O. Box 553, FIN-33101, Tampr, Finland {oscar.frsan, tut.fi Abstract In this papr, th concpt of DPCCH (Ddicatd Physical Control Channl) gating for HSUPA (High Spd Uplink Packt Accss) is analyzd. Gating tchniqu has rcntly bn undr considration within WCDMA to inactivat control channls during silnt priods of th usrs, and hnc, not to misus capacity. In ordr to study th concpt bnfit, a concrt ral-tim srvic is slctd in this papr: Voic ovr IP (VoIP) for mobil communications. It will b shown that VoIP would b highly bnficd by DPCCH gating inclusion in 3GPP spcifications. Both analytical and simulation studis wr run to confirm th gain xpctations. Kywords- DPCCH gating, VoIP, HSUPA Nokia Rsarch Cntr P.O. Box 45, FIN Nokia Group Hlsinki, Finland {tao.t.chn, nokia.com As on of th ral-tim srvics considrd in th continuous connctivity work itm, VoIP would b highly bnficd by th gating tchniqu inclusion. What is mor, th charactristics of VoIP srvic maks it asir to implmnt th gating procss as w will xplain latr. Th papr is organizd as follows: in sction II, th DPCCH gating tchniqu is dscribd. Sction III dtails th VoIP concpt. Sction IV covrs th analytical study prvious to th simulation procss. Through sction V th modling and assumptions considrd in th simulator ar prsntd. Th actual simulation rsults ar analyzd in sction VI and final conclusions ar drawn in Sction VII. I. INTRODUCTION High spd uplink packt accss (HSUPA) concpt was laboratd in th rlas 6 of th 3GPP spcifications for th third mobil gnration. It is th uplink countrpart of th high spd downlink channl (HSDPA) that was alrady introducd in rlas 5. Both high spd channls wr concivd in ordr to achiv highr capacitis and covrag for high transmission rats by th mans of innovativ tchniqus such as HARQ (Hybrid-ARQ), short fram siz and Nod B controlld packt schduling. Rlas 7, currntly undr dvlopmnt, continus to nabl vn highr data rats and capacity improvmnts in addition to improvd support for ral-tim srvics.g. Voic ovr IP (VoIP). At th radio accss ntwork sid, continuous connctivity is on of th nhancmnts. Th main problm to support continuous connction in WCDMA is th high numbr of UE s not transmitting, whil kping th control channls in us which ovrwhlms th cll capacity. On of th proposals to ovrcom this situation is th discontinuous transmission of th DPCCH channl whn th UE is in idl mod. This is rfrrd to as th DPCCH gating tchniqu. II. DPCCH GATING CONCEPT Prviously to rlas 7 (figur 1), DPCCH channl was transmittd continuously rgardlss whthr thr is actual usr data to b transmittd or not, thus highly loading th cll. An idal solution for incrasing th uplink capacity would b to kp th UE silnt during th priods that is not transmitting any data and activat th control channls just for th transmissions priods. Thrfor, th intrfrnc to othr usrs would b rducd. This is th basic ida bhind th DPCCH gating concpt prsntd in [1] and dpictd in figur 2. Du to DPCCH gating, it might b difficult for th ntwork to distinguish an inactiv priod from a loos connction. That is why th DPCCH transmission is dsirabl to not b totally silnt but transmittd priodically following a prdfind pattrn. Powr E-DPDCH E-DPCCH HS-ACK HS-CQI Voic ovr IP srvic has bn proposd thinking on th volution of UMTS to an all-ip ntwork. It is mant to substitut th prsnt circuit switchd voic connctions, thus it should show a clar gain ovr thm in ordr to b worth implmnting. DPCCH DPCCH DPCCH Figur 1. HSUPA transmission in Rl 6 Tim

2 Powr DPCCH E-DPDCH E-DPCCH Gating DPCCH HS-ACK DPCCH HS-CQI Figur 2. HSUPA transmission with gating DPCCH Anothr issu to considr is th transmission powr. Aftr a gating priod without tracking th rcivd SIR, th channl rspons variations could dgrad th transmission and incras th intrfrnc btwn th UEs. In th 3GPP spcifications for DPCCH gating [1], optional powr control prambls ar dfind to b snt prvious to th data channl ractivation. It is intrsting to point out hr that th HARQ protocol spcifid for HSUPA hlps to dcras th impact of inaccuracis in th powr control algorithm, as it was illustratd in [2], and so it will rduc th gating tchniqu on th powr control prformanc. III. VOIP OVER HSUPA A. VoIP packt dlay rquirmnts VoIP is dfind as a convrsational class srvic and thus th packt dlay should b maintaind strictly undr som rasonabl limits. According to [3], th maximum mouth-to-ar dlay accptabl for a good quality communication should b lowr than 250ms. Supposing that th ntwork dlays, such as cor ntwork, RNC procssing and Iub transport is approximatly 100ms, thr is a 150ms margin for th air intrfac dlays. Supposing also that both UE s ar communicating through HSDPA/HSUPA channls, thn th maximum accptabl dlay of on way transmission must b undr a 80ms dlay budgt. VoIP follows a prdictabl pattrn of data packts and silnc priods, which ar gnratd by th Adaptiv Multi-Rat (AMR) spch codc. In our study w suppos that th AMR spch codc producs on packt vry 20ms, compoundd of 320 information bits, during th activity priods. B. VoIP trasnmission ovr E-DCH with HARQ Two possibl TTI sizs hav bn spcifid for HSUPA: 2ms and 10ms. Evry third gnration-compatibl usr s quipmnt must support 10ms TTI whil 2ms is optional. 2 ms will transmit at highr pak rats, with lowr dlays, but rquiring highr transmission powr from th UE than 10ms TTI and thrfor prforming worst in th cll dg and soft handovr situations. Th synchronous HARQ protocol is usd in HSUPA, i.., rtransmissions hav to tak plac at a fixd duration aftr th Tim prvious transmission). For th 10ms TTI fram siz, 4 paralll procsss ar spcifid in [4], thus a packt rtransmission would b snt 40ms aftr th first transmission. Thrfor, th maximum allowd transmission numbr would b two, sinc a third rtransmission would imply packt dlays byond th 80ms dlay budgt. Th transmission procss is sktchd in th following figur. As it can b sn, th rtransmission of packt #1 implis an xtra dlay for packt #3, stting th maximum packt dlay to 60ms. VoIP packts from application vry 20 ms HARQ procss id #1 #2 #1 #3 #4 #5 #3 #4 #6 #7 #8 10 ms #1 dlay 50 ms 1 VoIP packt transmittd vry 20 ms #3 dlay 60 ms Figur 3 VoIP ovr HSUPA with 10ms TTI siz For a 2ms TTI siz, [4] dfins th HARQ protocol with 8 paralll procsss, dlaying a rtransmission by 16 ms from th first transmission. In this cas, up to 4 transmissions could b contmplatd without xcding th 80ms margin. Figur 4 shows a possibl transmission procss with a maximum of 3 transmissions. VoIP packts from application vry 20 ms HARQ procss id ms 1 VoIP packt transmittd vry 20 ms #1 dlay 50 ms Figur 4 VoIP ovr HSUPA with 2ms TTI siz C. Schduling schm Two schduling schms ar dfind for HSUPA: Nod B controlld schduling transmission and RNC controlld nonschduling transmission. Howvr, th formr schm dos not guarant a minimum bit rat sinc th schduling algorithm dcids whthr to allocat any powr to th usr or not. On th othr hand non-schduld transmission dfins a minimum data rat that th UE can transmit at without any prvious rqust. Thrfor, NRC controlld non-schduling transmission is th most suitabl choic for VoIP traffic. VoIP can b clarly bnficd by th gating tchniqu sinc usually a man of 50% of th tim a usr kps silnt during a voic convrsation. Drawing on this fact, only silnc dscriptors (SID) ar snt vry 160ms as soon as a silnc is dtctd. Thus, during th usr s silnc priods, th UE could

3 transmit discontinuously on th DPCCH. On th othr hand, during th activ priods, th voic codr producs on packt vry 20ms. As th transmission is not continuous, a DPCCH gating transmission could b also considrd during voic activity priods. Howvr, it should b takn into considration th rtransmissions du to th HARQ algorithm, which would dcras th gating opportunitis during activity priods as it can b sn in figur 1. IV. SEMI-ANALYTICAL STUDY W should answr now to th qustion of how much gain w can xpct from th DPCCH gating tchniqu. W can tak as initial point for th analysis th tim prcnt that DPCCH should b on, i.. th DPCCH activity factor that w will callν c. From th gating concpt dfinition, ν c will b quivalnt to know th prcnt of th tim that data is snt in any channl dpndnt on DPCCH transmission, ν ν + ν + ν c =. (1) ACK CQI Hrν is th HSUPA activity factor and ν ACK and ν CQI corrspond to th activity factors of HS-DPCCH and fdback channl for HSDPA, rspctivly, considring that HSDPA holds th VoIP downlink transmission. On th othr hand, ν is qual to th product of th voic DTX activity factor ( ν i ), th frams arrival rat to th physical layr ( R F ) and th avrag transmission numbr du to th HARQ schm (AvgTx). That is, Whr N is th amount of UEs, v is th activity factor and finally i rprsnts th intrfrnc lvl from othr clls. EcNo is th signal to nois targt pr chip aftr th antnna combining. Four physical channls tak part in VoIP communications, ach on with diffrnt EcNo rquirmnts and activity factor. Hnc, w can divid th formula as [ NRdB = 10log10 1 ( EcNoc vc + EcNo v +. (4) + EcNo ν ) N h h ( 1+ i) ] Whr EcNo c and ν c corrspond to DPCCH, EcNo and ν to HSUPA rlatd physical channls (E-DPCCH and E-DPDCH togthr) and finally EcNo h and ν h to HS-DPCCH. ν h is th sum of v ack and v cqi supposing thy shar th sam targt EcNo h. At this point w ar abl to draw th cll nois ris figur with rspct to th numbr of UEs in th cll in figur 5. Th rquird EcNo valus will b obtaind from th link lvl simulations. W can first focus on continuous DPCCH. It is noticabl that VoIP shows similar or wors prformanc than 12.2 Kbps on DCH channls using 10ms TTI siz: th rduction on channl activity (0.67 for 12.2 Kbps on DCH to for 10 ms) dos not compnsat th wors channl condition for highr data rats ndd in VoIP. 2ms TTI alrady improvs 12.2kbps on DCH thanks to a farthr rduction on th activity factor down to ν =ν AvgTx. (2) i R F In quation (1) w hav assumd that HSUPA and HS-DPCCH nvr ovrlap in tim, so th DPCCH activity quals th sum of activitis. On th othr hand, w ar considring th idal DPCCH gating, i.. DPCCH is not transmittd if no dpndnt channl is transmittd and thus no powr control prambls ar in considration. W prsum that both assumptions somwhat compnsat ach othr. Onc w got th prcnt of tim th DPCCH will b on, w can masur th systm capacity through th nois ris formula [5], ( 1 EcNo N v ( i) ) NR db = 10 log (3) Figur 5 Smi-analytical cll nois ris for two diffrnt throughput valus Th DPCCH gating tchniqu will prmit VoIP to finally bat voic 12.2 kbps circuit switchd capacitis. With 10ms TTI siz and considring a maximum allowd nois ris of 6dB, w

4 obtain gating gains of 50% and 64% rspcting th continuous DPCCH simulations. In th othr hand, 2ms gts a 40% gating gain. With this improvmnt in prformanc, 10 ms TTI prsnts capacitis around 50% highr than circuit switchd transmissions, manwhil 2ms TTI gos up to 64% gain. Th throughput is dfind as th invrs of th avrag transmission numbr. In continuous DPCCH transmission, a lowr throughput (highr transmission numbr) prsnts lowr nois ris (s figur 5) du to th fact that lowr powr is ndd as th rcivd SIR for ach transmission will b finally combind. Howvr, Gating DPCCH bhavs in th opposit way with th avrag transmission numbr. In (2), a smallr AvgTx valu rsults in lowr DPCCH activity factor and thrfor, longr gating priods ar b possibl. This is why w obtain bttr gating gains: highr throughput mans smallr amount of rtransmissions, which rsults in lowr DPCCH activity. Consquntly w obsrv in figur 5 that both throughput valus gt th sam rsults with th Gating DPCCH tchniqu: thr is a tradoff btwn th gating DPCCH and HARQ. V. MODELING AND ASSUMPTIONS A. VoIP modlling Th VoIP call duration is fixd to 60s. Considring discontinuous transmission, th on and off priods durations ar both modld as a dcaying xponntial with 3s man. During th on phass, VoIP packts ar gnratd vry 20 ms, containing 31 byts from th voic codc and 9 byts of hadr from layrs on top of MAC. 31 byts vry 20 ms corrsponds to a sourc rat of 12.2 Kbps. During th off priods no SID transmission is implmntd. B. DPCCH gating modlling Th gating concpt has bn modld in th simplst mannr. At any TTI that a packt is not rcivd in th MAC layr, DPCCH is not transmittd at all, i.. no activity pattrn is considrd whn nithr HSUPA nor HS-DPCCH is transmittd. Rgarding th fast powr control, aftr a gating priod th UE transmits with xactly th sam powr as bfor th off priod, as DPCCH powr control prambl is nithr implmntd. C. HSUPA modling Thr diffrnt physical channls, I/Q cod multiplxd [6], ar nd for th uplink transmission of VoIP traffic: DPCCH, Enhancd ddicatd physical control channl (E-DPCCH) and Enhancd ddicatd physical data channl (E-DPDCH). Th high spd downlink physical control channl (HS-DPCCH), fdback channl for HSDPA, has also bn includd in th study, supposing that th downlink VoIP frams ar dlivrd through HSDPA. As pointd out in th smi-analytical study in sction IV, HS-DPCCH should b contmplatd for th DPCCH activity factor calculations in gating scnario. D. VoIP quality of srvic critria In our study, two diffrnt critria hav bn considrd to masur th systm capacity: th cll outag prcnt and th cll nois ris. Th cll outag dals with th prcnt of discardd frams. W stat that an UE is in outag situation if 5% of th frams ar rronous or discardd ovr a 10ms priod. Thrfor, th systm will rach th maximum allowd capacity whn up to 5% of th UE in th cll ar in outag. Th scond critria, th cll nois ris lvl, provid us a link with th smi-analytical study, as th latst is basd on th rcivd signal to nois ratio at th bas station. Th capacity will b givn by th numbr of UEs in th cll whn a 6dB man nois ris is masurd at th bas station. E. Environmnt In ordr to analyz systm lvl prformanc, a quasi-static simulator is usd basd on dscription in [7]. Th protocol layrs implmntd ar MAC and Physical layrs. Layrs abov MAC ar only considrd in th MAC SDU siz. All th ndd RRM algorithms wr implmntd. Th main simulation paramtrs ar summarizd in tabl 1. TABLE I. SIMULATION PARAMETERS SETTINGS Paramtr Valu Fram Siz 10ms and 2ms TTI Intr sit distanc 2.8 km Cll configuration ITU Vh-A, Macrocll Voic call lngth 60 sconds Voic on/off man lngth 3 sconds Payload siz 31 byts VoIP packt arrival intrval 20 ms (with 10ms, 2 packts transmittd vry 40ms) Comprssd hadr siz 9 byts UE spd 3kmh Voic Activity 0.5 Numbr of HARQ channls 4 (10ms) / 8 (2ms) Max numbr of L1 transmissions 2 (10ms) / 4 (2ms) (β c /β c ) db (10ms) / 3 db (2ms) (β d /β c ) 2 8 db Schduling algorithm Non-schduld DPCCH gating patrn No transmission patrn DPCCH gating powr prambl lngth 0 slots Discardd timr 80 ms Dlay budgt 80 ms Outag obsrvation window lngth 10 sconds Cll outag thrshold 5%

5 Notic that 10ms is implmntd to snd two packts pr fram, aiming to giv mor chancs for DPCCH gating kping still th packt dlay undr 80ms but rquiring to doubl th data rat in ordr to snd two packts at onc. In 2ms TTI transmission th possibl prformanc improvmnt dos not justify th incras in data rat. VI. SYSTEM LEVEL SIMULATION RESULTS Fig. 6 dpicts th nois ris bhavior from th systm lvl simulation. Considring sttld a 6dB cll nois ris limit, a gating gain around 42% and 50% is obtaind for th 10ms cass and 41% for 2ms. Th systm lvl prformanc is undr th smi-analytical rsults in Fig. 2 whr 50% and 64% gains for 10ms and 40% for 2ms wr rachd. Th smi-analytical study assums that thr is not limit for th UE tx powr, whras in th actual systm simulations th UE powr shortag limits prformanc, as xplaind in [8] for th 2ms continuous scnario. Figur 7 Cll outag for diffrnt throughput valus VII. CONCLUSIONS In this papr w hav considrd th prformanc of DPCCH gating tchniqu for HSUPA. Gating tchniqu has rcntly bn proposd within WCDMA to rlas unusd rsourcs by inactivating control channls during silnt priods of th usrs. In ordr to study th concpt bnfit, Voic ovr IP was slctd as th srvic undr considration. Both analytical and simulation studis wr run to confirm th gain xpctations. It was shown that VoIP would b clarly bnficd by th DPCCH gating inclusion in 3GPP spcifications. Farthr study concrning th us of powr control prambls sms to b ncssary to diminish th powr control variability that th DPCCH gating tchniqu introducs. Figur 6 Man nois ris for diffrnt throughput valus A bttr gating prformanc was obsrvd rgarding th outag capacity in figur 7: 50% and 77% gating gains can b obsrvd for th throughput valus of 60% and 80%, rspctivly. 2 ms shows lowr gating gain (47%) gtting th 10ms gating rsults quit clos to th 2ms prformanc. Th main drawback concrning gating is th ffct that its implmntation might caus to th powr control prformanc. As th DPCCH would not b continuously transmittd, th fast powr control algorithm will not b abl to follow th changs in th channl charactristics during th inactivity priods. That is why th us of a powr control prambl, prvious to th ractivation of th data channls is contmplatd in [1]. In our study, no prambls ar considrd to analyz th ffct of gating on powr control. REFERENCES [1] 3GPP TR V1.0.0 ( ), Continuous Connctivity For Packt Data Usrs. [2] Y. Kim; J.-K Han; H. Kwon; D. Kim Effct of imprfct powr control on th prformanc of HARQ in CDMA2000 rvrs link, 15 th PIMRC, vol.3,pp ,Spt2004. [3] ITU-T rcommndation G.114, On way transmission tim [4] 3GPP TR V6.4.0 ( ), FDD Enhancd Uplink; Ovrall dscription. [5] H. Holma, A. Toskala (Editors), WCDMA for UMTS - Radio Accss for Third Gnration Mobil Communications, John Wily and Sons, Third Edition, [6] 3GPP TR V6.4.0 ( ), Sprading and modulation (FDD). [7] 3GPP2 Tchnical Spcification Group C, 3GPP2/TSG-C.1002: 1X EV- DV Evaluation Mthodology. [8] T. Chn; M. Kuusla; E. Malkamäki Uplink Capacity of VoIP on HSUPA, VTC Spring 2006.