2010 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media,

Transcription

1 2 IEEE. Personal use of his maerial is permied. Permission from IEEE mus be obained for all oher uses, in any curren or fuure media, including reprining/republishing his maerial for adverising or promoional purposes, creaing new collecive works, for resale or redisribuion o servers or liss, or reuse of any copyrighed componen of his work in oher works.

2 Impac of Sof Handover and Pilo Polluion on Video Telephony in a Commercial Nework Lu Chen, Faisal Mohd. Madani, Kumbesan Sandrasegaran, Riyaj Basukala Cenre of Real-ime Informaion Neworks, Faculy of Engineering and Informaion Technology Universiy of Technology Sydney, Ausralia Lu.chen-4@suden.us.edu.au Absrac Sof handover (SHO) provides beer suppor for user mobiliy. The qualiy of service (QoS) o he mobile wireless subscribers can also benefi from SHO. To analyze he impac of SHO on he circui swiched (CS) video elephony (VT); real-ime measuremens have been conduced on a commercial WCDMA mobile nework in he CBD of Sydney, Ausralia. Live measuremen provides a pracical approach o evaluae performance of commercial nework. The obained SHO gain implies ha beer QoS can be achieved. Furhermore, he negaive impac of pilo polluion on radio link qualiy has been demonsraed in his paper. Keywords-Mobile neworks; 3G WCDMA; video elehpony (VT); sof handover (SHO); qualiy of service (QoS); pilo polluion I. INTRODUCTION One of he mos aracive feaures of mobile communicaion is service coninuiy wih user mobiliy. However, his feaure is largely resriced by coverage and capaciy issues in cellular nework. Handover (HO) plays a fundamenal role for supporing service coninuiy wih user mobiliy wihin coverage limis while balancing raffic load beween adjacen or overlapped cells []. In radiional handover (hard HO), e.g. handover in GSM, a session is inerruped [2], i.e. he original connecion from he source base saion (BS) is released before esablishing a new connecion a he arge base saion (BS). This is referred o as break before make. The inerrupion inroduced by hard HO, hough small, could have an impac on service coninuiy. In wideband code division muliple access (WCDMA), which is he mos popular air inerface for he 3rd generaion mobile wireless communicaion (3G) nework, a new handover mechanism has been inroduced and named sof handover (SHO). One key difference of SHO, when compared o hard HO, is he no inerrupion characerisic of handover procedures, which allows he user equipmen (UE) o mainain muliple acive connecions simulaneously wih differen BS s. This handover mechanism is also referred o as make before break. Sof handovers are needed for CDMA based sysems because CDMA based sysems are inerference limied sysems. Wihou SHO s, a mobile moving from one cell o an adjacen cell will have o ransmi a a high power o communicae wih he source BS resuling in inerference in he arge cell. In WCDMA, a UE is coninuously measuring all received signals on Primary Common Pilo Channel (P-CPICH; simply referenced as CPICH). An Acive Se is a lis of cells o which UE is simulaneously conneced o [3]. If a CPICH signal has been measured and is above a hreshold value (se by he nework) for a period of ime ( Time-To-Trigger ), his CPICH would be included ino he Acive Se, and he BS, which his CPICH belongs o can be used for he UE ransmission. BS wih CPICH, which has been measured wih reasonably srong signal bu no saisfied he above crieria o be included in he Acive Se, will be regisered in Moniored Se, as candidaes for being added ino Acive Se [3]. In his manner, UE is able o connec o cells, which are capable of providing higher radio link (RL) qualiy. Therefore, he advanages of SHO can be, in urn, helpful for improving he qualiy of service (QoS) for he mobile wireless subscribers []. Sof handover resuls in a diversiy gain called sof handover gain or SHO gain. A large number of invesigaions have been carried ou on analyzing and improving SHO gain based eiher on heory or on simulaion resuls, e.g. in [2,4]. However, oher publicaions have sudied he SHO gain and SHO performance in operaional WCDMA neworks. An early effor has been carried ou by OrangeUK [5]. They conduced measuremen on he rial WCDMA nework hen verified ha SHO gains in downlink were achieved by receiving ransmissions from wo BS (in wo-way handover) wih spaial diversiy as well as by using RAKE anenna on he receiving side a UE. A more recen work, [6] gave evaluaions of SHO algorihm implemened on a pre-commercial WCDMA nework by measuremens resuls. Ye, neiher [5] nor [6] have carried ou measuremens on a commercial nework i.e. in he same environmen as experienced by live subscribers. The auhor of [8] has conduced measuremens in live scenario and assessed he qualiy of mulimedia services on WCDMA nework. However, he sudies in [8] were focusing on he opimizaion of nework planning procedure. On anoher hand, pilo polluion issue commonly exiss in WCDMA nework. The polluing siuaion arises due o he severe inerference among muliple CPICH s in cell overlapping area. In Reference [9], CPICH inerference is named as Muliple Access Inerference (MAI), which has

3 imporan impac on capaciy loss. However, only he severe condiion of MAI-pilo polluion has been sudied in [6] as well as a focus in our paper. In he sudy of [6], pilo polluion had impac on he performance of SHO algorihm. However, he work in [6] has only considered he polluing effec caused by he sronges CPICH in Moniored Se. In our work, he analysis of he impac of SHO gain and of pilo polluion on mobile nework has been carried ou by performing live measuremens on a commercially deployed WCDMA nework in he CBD of Sydney, Ausralia. A circui swiched (CS) video elephony (VT) session has been used during measuremens. The focus of his paper is on analyzing he impacs of SHO and of pilo polluion on mainaining QoS during SHO procedure in real-ime scenario. In Secion II, he seup informaion of our measuremens is described. Measuremen resuls and analyses are presened in Secion III. Finally, Secion IV summarizes our sudy. II. A. Measuremen environmen MEASUREMENTS Nemo Handy (Anie), has been embedded ino he handse. Nemo Handy is used o display and record daa of each measuremen. The lock o scrambling code funcion in Nemo Handy is used when conducing measuremens for he no-sho scenarios. An independen GPS device has been conneced o he handse via Blueooh inerface during measuremens. The geographic informaion from he GPS can be inegraed ino he log files produced by Nemo Handy during measuremens. For pos processing of he measuremens, Nemo Analyze, MS Excel, MATLAB and Google Earh were used. C. Measured service The CS video elephony (VT) service was in use when performing live measuremens. In WCDMA, he ypical CS VT raffic is ransferred on a symmeric-bidirecional dedicaed channel (DCH) wih 64kbps fixed bi rae []. A he UE used in his work, he CS VT has been compressed ino 3GPP MPEG-4 forma wih a video resoluion of 64*48 pixels per frame, and sreams a he frame rae of 5fps. VT was seleced o be he service used in he measuremens for hree reasons. Firs, CS RT services are inherenly sensiive o inerrupions and errors, which may occur during handover processes. Second, VT is a mulimedia service ha emerged wih 3G. Third, comparing o voice service (i.e. speech), VT has higher requiremen of QoS (due o e.g. live video sreams a higher bi rae and compressed picure frames are less robus o errors). D. Measuremen cases Two measuremen cases have been conduced in his paper. One is SHO case (Case ) when UE is allowed o have SHO connecion on he measured nework. The oher case (Case 2) is No-SHO case, when Nemo Handy has been configured o perform he funcionaliy of Scrambling code locking hus he UE can only be conneced o a specific cell (Cell_) (Table I.): Figure. Downlink block error rae in es roue. The upper picure is measured in live user case (SHO enabled); he lower one is measured when UE locked o Cell_. The color se is shown on righ-hand side. The measuremens are conduced on a commercial WCDMA mobile nework implemened on he 2MHz band in he CBD of Sydney, Ausralia. Specifically, he esing area was seleced along he sideway of a main road of he ciy, while a lile ouside he ciy cenre of high-rise core (surrounding by all buildings) hus smaller densiy of user raffic and all buildings. The sar locaion of he measuremen pah was seleced o be approximaely a he edge of a 3G macro-cell and he end locaion was seleced o be approximaely a he middle of he cell. The specific cell will be referenced as Cell_ or Scr. (corresponding scrambling code of Cell_ ), in he res of his paper. B. Measurmen ools The measuremens were carried ou by a commercial Nokia mobile handse of model 672US (i.e. User Equipmen (UE)), and UE Power Class 3. The real-ime monioring sofware, TABLE I. MEASUREMENT CASES Cases Descripions Measuremens Case Case 2 Live user scenario (SHO) Cell locked scenario (No-SHO) Case s run; Case 2nd run Case 2 s run; Case 2 2nd run The pedesrian velociy measured by GPS is approximaely 3.km/h, on average. E. KPI selecion In SHO, he service coninuiy should be mainained, and he QoS is mainly affeced by he radio frequency condiion [8]. Thus, key performance indicaors (KPI s) in he analysis are seleced for indicaing he radio frequency condiions and radio link qualiy of he UE: Call drop;

4 CDF CDF Downlink block error rae (BLER DL); Primary Common Pilo Channel (CPICH) Ec/No; Primary Common Pilo Channel (CPICH) RSCP; Downlink power up command and uplink ransmission power of UE. To illusrae he poenial influence of pilo polluion, wo addiional KPIs are specified: The number of polluing pilos; and Toal level of pilo polluion. III. ANALYSES RESULTS In his secion, an analysis of he resuls is presened. A definiion of he KPIs used in his work is presened. A. Sof handover performance ) Call drop Call drop is a primary KPI of QoS o user. Possible causes of call drops can be from nework s operaional acions or due o poor RL qualiy. Technically, a number of facors may decrease RL qualiy hence cause call drops, such as handover failure, poor coverage, insufficien capaciy, ec. []. In our measuremens, call drop has no been obained in eiher Case or Case 2. 2) Downlink block error rae (BLER DL) For an acive call session, he analysis of downlink block error rae (BLER) is used for evaluaing he overall qualiy of recepion for UE, given ha any physical layer errors occurred in ransmission can be refleced by BLER. The ransmission errors will mainly affec he qualiy of video raher han voice when he received daa is exraced and fed o he user applicaions []. An erroneous block in VT can be perceived by user as image disorion or inconinuous image []. References [2] and [] have specified he BLER DL in VT should be lower han % a he receiver. According o [2], BLER is compued afer Radio Link (RL) combinaion in he Rake Receiver of he UE using he formulae given in (). number of received ranspor blocks incrc error BLER () number of oal received ranspor blocks The CDF analysis of downlink ranspor channel BLER in boh Case and 2 is presened in Figure 2. According o Figure 2, a much beer BLER in Case is obained. In Case, for 8% of he ime no error has been measured on BLER, while only BLER values of -2% are obained in measuremens. The wors BLER value is approximaely 2% based on average of wo es runs in Case. In Case 2, he BLER value of % corresponds o a CDF of approximaely 5. Therefore he poor link qualiy siuaion (i.e. BLER>%) have been obained for approximaely 45% of he session ime and maximum BLER value is as high as 8% case s run case 2nd run 5 case2 s run case2 2nd run Downlink ranspor channel BLER [%] Figure 2. CDF of downlink ranspor channel BLER in each measuremens when sho is enabled and disabled, respecively 3) Primary Common Pilo Channel (CPICH) Ec/No The direc indicaor of RL qualiy is he received CPICH Ec/No by he UE. By definiion in [2], CPICH Ec/No is he received energy per chip divided by power densiy in he band. Generally, sronger he received CPICH Ec/No a UE, lower he inerference and noise level in he RL, hence beer signal qualiy can be achieved, or less downlink ransmission power from he BS is required for receiving similar qualiy signal. case s run case 2nd run. case2 s run case2 2nd run Bes received CPICH Ec/No [db] Figure 3. CDF of Bes received CPICH Ec/No in Case, and received CPICH Ec/No from Scr. Case 2, respecively Figure 3 compares he bes received CPICH Ec/No a UE in Case and 2, respecively. Higher CPICH Ec/No values are obained in Case. Considering he CDF value of, in Case, i corresponds approximaely -db in CPICH Ec/No value, which indicaes ha half of he possibiliy for CPICH Ec/No values are obained above -db; however, in Case 2, half of he possibiliy for CPICH Ec/No values are obained approximaely higher han -4dB. Tha is, 3dB gain in Case. Furhermore, 5% of he CPICH Ec/No values in Case 2 are lower han he performance requiremen of -2dB [2]. 4) Primary Common Pilo Channel (CPICH) RSCP For availabiliy of service for QoS [3], i is essenial for UE o say wihin he service coverage. If a UE wih an in-used VT moved ino he hole [8] in 3G service coverage, he UE

5 "power up" % "power up" % CDF may suffer from e.g. low hroughpu, long buffering ime, or call drop. Technically, CPICH RSCP received a UE indicaes he coverage probabiliy of WCDMA nework. The coverage of a cell in WCDMA nework is usually indicaed by cell radius, i.e. maximum disance away from he cenre of cell where he BS of ha cell is able o provide mobile service o UE. However, in WCDMA, cell coverage is dynamic as balance o is capaciy [2]; hence, coverage probabiliy is used. Higher CPICH RSCP measured a UE, higher probabiliy he UE is locaing in wihin nework coverage. The relaion of CPICH RSCP value and disance are shown in (2) and (3). RSCP CPICH P, PahLoss dbm (2) Tx CPICH A PahLoss (3) r Where, PTx, CPICH is he ransmission power as he oupu of BS anenna; A is a consan and he value of A depends on he properies of he BS anenna; r denoes he disance beween UE and he BS anenna; α is pahloss exponen, which would be differen in value as depending on environmen. In map analysis (Figure 4), he values of on UE along he es pah are marked in disinc colors based on he applied color se in Figure 4. High CPICH RSCP values (i.e. >= - 6dBm) are obained more ofen in he measuremens in Case (upper picure in Figure 4) han hose in Case 2 (lower picure in Figure 4). Whereas, in Case 2, some bes CPICH RSCP value of lower han -7dBm have been measured in he UE. case s run case 2nd run. case2 s run case2 2nd run Bes received CPICH RSCP [dbm] Figure 5. CDF of bes received CPICH RSCP on UE from each measuremens in Case and Case 2, respecively 5) Downlink power up command and. uplink ransmission power of UE Anoher pair of RL qualiy indicaors is from he power conrol procedures in WCDMA, i.e. he downlink power up command from he nework and he uplink ransmission power of UE. In downlink, he radio nework conroller (RNC) performs ouer loop power conrol o adjus he received signalo-noise arge (SIR arge) from he served UE. The adjusmen is based on he received signal qualiy of he UE repored by he serving BS(s) o RNC. The adjused SIR arge is hen sen o he BS(s) from RNC. The SIR arge is used as reference for BS(s) o evaluae he UE should ransmi in higher or lower power, and hence sends a power up or power down command in downlink. According o he received power up or power down command, UE adjus is uplink ransmission power. In SHO, UE receives power conrol commands from muliple BS s. In he UE used in his work, he combining algorihm follows any reliable power down command as received [4]. Figure 6 shows he downlink power up command received by UE in each measured case, and Figure 7 presens he CDF and PDF analysis of he UE ransmission power in each measuremens. Case s run Figure 4. Map analysis of bes received CPICH RSCP a UE in Case (upper) and Case 2 (lower), respecively. The color se is shown on righ-hand side. Given ha he UE in SHO measures CPICH RSCP of muliple BS s, he bes received CPICH RSCP is used in daa analysis (Figure 5). According o Figure 5, he CDF curves of measurmens in Case shows approximaely 4dB shif owards higher (lef) value of bes received CPICH RSCP from he CDF curves of Case 2 measuremens. Tha is, an approximaely 4dB gain has been achieved Case 2 s run Figure 6. Downlink Power up command in Case s run and Case 2 s run, respecively.

6 CPICH Ec/No of bes Acive Se [db] Downlink ranspor channel BLER CPICH Ec/No of bes Acive Se [db] Downlink ranspor channel BLER i, E c N o polluioni, E c N o bes ASET, [ db] (5) i, i, [ linear ] (6) Thus, we furher define he oal level of insananeous polluion as he sum of effec from all polluing pilos, i.e. N Toal, Ec N o Ec N o [ db] (7) i polluioni, bes ASET, Figure 7. PDF (bars) and CDF (lines) of UE ransmission power from each measuremens in Case and Case 2, respecively In Figure 6, he downlink power up command in Case 2 (lower figure in Figure 6) occured more frequenly han ha in Case (upper figure in Figure 6). Tha indicaes ha he UE is 3 suffering a lo by poor RL condiion, seen by he BS. In oher words, he ransmission of UE has been beer received on 2.8 he nework in Case. According o Figure 7, 5% (CDF=) of he UE 2.4 ransmission power values are measured lower han 2.2 approximaely -3dBm in Case and -24dBm in Case 2, based 2 on average of es runs. Tha is, a reducion of approximaely 5.5dB in UE ransmission power is achieved in SHO scenario..8 Since he service availabiliy is also limied by he ransmission.6 power of UE, he reducion of required ransmission power is,.4 in urn, benefi for he UE o be kep service availabiliy from he nework when i moving furher away from he originally.2 serving BS. Thus, service coninuiy has been improved. B. Pilo polluion Regarding o Secion, Pilo polluion is ofen obained in cell overlapping areas. However, since he execuion of SHO requires cell overlapping (i.e. he area where more han one BS is capable o provide service o UE), pilo polluion is 3 commonly exising in WCDMA neworks. In his work, he crieria of idenifying pilo polluion are 2.6 presened in Table II. TABLE II. PILOT POLLUTION CRITERIA Crierions Symbol Values CPICH Ec/No (of oher cells) - CPICH Ec/No (of bes serving cell) ζ i, * >= -6dB CPICH Ec/No (of bes serving cell) <= -8dB * Noe ζi, denoes he polluion from pilo i a ime insance.2. According o he descripion of pilo polluion siuaion, he insananeous polluion level from each polluing pilo, ζi,, is defined in his paper as he closeness of oher cell CPICH Ec/No measuremens oward he bes CPICH Ec/No in Acive Se of he UE (i.e. he bes of curren serving cells) a ime insan. The definiion of ζi, is also shown in (5) and (6). 2.6 Toal, N i i, [ linear ]. (8) By analysing he measuremens in his paper, pilo polluion has significan influences on he degradaion of RL qualiy, which are shown in Figure 8 and Number of pilos Figure 8. CPICH Ec/No of bes received in Acive Se decrease when more polluing pilos have been deeced; while BLER DL increases only afer more han 3 pilos are obained as polluing pilos CPICH Ec/No case s run CPICH Ec/No case 2nd run BLER DL on avg. in case s&2nd runs CPICH Ec/No case s run CPICH Ec/No case 2nd run BLER DL on avg. in case s&2nd runs Toal pilo polluing level (in linear) Figure 9. CPICH Ec/No of bes received in Acive Se decrease when oal level of pilo polluion increase. Meanwhile, BLER DL increases significanly along wih he rise of oal pilo polluion level..

7 According o Figure 8, he bes CPICH Ec/No in Acive Se of he UE decreases significanly when he number of polluing pilos increases. Moreover, noe ha he Acive Se size is se o be maximum 3 cells. I is reasonable o express a paricular ineres o he poin ha number of polluing pilos equals o 2. The CPICH Ec/No has dropped approximaely 3.5dB when 2 polluing pilos are deeced in comparison wih he condiion ha no polluing pilo has been deeced. However, he increase of has become significan only afer number of polluing pilos is more han 3. In Figure 9, he BLER DL higher sensiiviy o he increase of oal level of pilo polluion. Ye, he deducion of bes CPICH Ec/No in Acive Se is rapid bu slighly slower when oal level of pilo polluion goes up. Thus, he negaive impac of pilo polluion on RL qualiy is imporan. The possible consequences of he impac include: degradaion of performance of SHO algorihm [6]; increase of call drop rae; capaciy loss [9], ec. IV. CONCLUSION From live measuremens, he presence of SHO in WCDMA sysem brings improvemens o RL qualiy for he UE in using VT, hence improve user mobiliy and suppor he nework beer mainaining QoS. However, on he oher hand, pilo polluion causes he reducion of he RL qualiy, significanly. Since pilo polluion and SHO usually appear in he same area, i.e. cell coverage overlapping area, i is necessary o miigae he impac of pilo polluion for geing he full benefi o he overall service qualiy in WCDMA. Las bu no he leas, live measuremen provides a pracical approach o evaluae performance of commercial nework. So i can be compared wih heoreical analyses. REFERENCES [] ETSI, "Universal Mobile Telecommunicaions Sysem (UMTS); Radio Resource Managemen sraegies," ETSI TR25.922, 27 [2] H. Holma and A. Toskala, WCDMA for UMTS: HSPA Evoluion and LTE, Fourh ed.: John Wiley & Sons Ld., 27 [3] 3GPP, " RRC Proocol Specificaion," 3GPP TG25.33, 29. [4] R. M. Joyce, T. Griparis, L. J. Osborne, B. Graves, and T. M. Lee, "Sof Handover Gain Measuremens and Opimisaion of A WCDMA Nework," 3G Mobile Communicaion Technologies, 24. 3G 24. 5h IEE Inernaional Conference, pp , 24. [5] T. Isoalo, J. Niemela, J. Borkowski and J.Lempiainen,"Impac of Pilo Polluion on SHO Performance," in Proc. 8h IEEE Inernaional Symposium on Wireless Personal Mulimedia Communicaions (WPMC'5), Aalborg, Denmark, Sep. 25. [6] M. Malkowski and D. ClaBen, "Performance of Video Telephony Services in UMTS using Live Measuremens and Nework Emulaion," [online] in Wireless Personal Communicaions:Springer Neherlands, Neherlands: Vol.46, No. : 9-32, Jul. 28. [7] R. Weber, " Mulimedia Performance Assessmens in Deployed UMTS Neworks," in Proc. 25 Sysems Communicaions (ICW 5), pp. 66-7, Aug.25 [8] M. Rahnema, UMTS Nework Planning, Opimisaion and Iner- Operaion wih GSM, Singapore: John Wiley & Sons (Asia) Pe Ld., 28. [9] 3GPP, " Feasibiliy sudy on he miigaion of he effec of he Common Pilo Channel (CPICH) inerference a he User Equipmen," 3GPP TR25.99, 22. [] R. Kreher, UMTS Performance Measurmen.: John Wiley & Sons Ld., 26. [] 3GPP, "End-o-end mulimedia services performace merics," 3GPP TR26.944, 29. [2] 3GPP, "Physical Layer Measuremens (FDD)," 3GPP TS25.25, 2. [3] ITU-T, "Definiions of erms relaed o qualiy of service," ITU-T E.8, 28. [4] 3GPP, "Physical layer procedures (FDD)," 3GPP TS25.24, 2