Extended Synchronization Signals for Eliminating PCI Confusion in Heterogeneous LTE

Transcription

1 1 Extended Syncronization Signals for Eliminating PCI Confusion in Heterogeneous LTE Amed H. Zaran Department of Electronics and Electrical Communications Cairo University Egypt. Abstract Heterogeneous long term evolution (LTE) networks evolve as a possible solution to accommodate te exponential growt of mobile data. However te eterogeneity introduces several design callenges suc as increasing interference and mobility overead. In tis paper we propose Extended Syncronization Signals (ESS) to eliminate te unavoidable pysical cell identity (PCI) confusion problem in dense eterogeneous LTE deployments. Te ESS is also designed to reduce signaling overead and andover delay. Our analysis sows a andover delay reduction of 65% can be realized. Additionally we sow tat PCI confusion probability can be practically eliminated in case of centralized planning and a reduction of five order of magnitude in confusion probability can be attained in case of distributed planning. I. INTRODUCTION Te deployment of eterogeneous long term evolution (LTE) cellular systems represents a very promising approac to accommodate te exponential growt in mobile data [1]. In eterogeneous deployment a macrocell tier typically provides full coverage using base stations (BS) wit large footprints wile oter BSs wit smaller footprints suc as relay BSs and femto-bss are implemented to improve or extend te coverage in teir locality. Tis eterogeneous deployment especially femtocells not only boost te system spectral efficiency and enables traffic offloading from te congested macro-tier but also increases te average revenue per user extends te mobile terminal (MT) battery life and enances te customer loyalty. Te introduction of femtocells introduces several design issues including but not limited to interference mitigation mobility overead and self-organization aspects. Typically femtocells increase te rate of cell crossing in te system and consequently increase signaling overead for location update and andover. Additionally LTE standards define tree access modes for femtocells [2] including closed access open access and ybrid access. Closed access cells are restricted to a set of users. Hence additional signaling overead may be incurred if a MT tried to camp on a cell tat it as no rigt to access. On performing andovers te MT provides te serving BS wit a measurement report including te signal strengt and a PCI for eac detected cell. If te serving BS as two or more neigbors aving te same PCI it would not be able to determine te MT target BS. Tis problem is known as PCI confusion [3]. In te eterogeneous deployment te PCI confusion problem seems to be unavoidable due to te Tis work is part of te 4G++ project (4gpp-project.net) supported by te National Telecom Regulatory Autority of Egypt ( expected large number of femtocells and te limited range of PCI in LTE standards. Typically te PCI is derived from two pysical layer signals known as te primary syncronization signal (PSS) and te secondary syncronization signal (SSS). Te PSS and SSS ave a sort transmission time interval of 0.5ms to speed te cell searc process. Additionally tey are used to derive te cell scrambling codes. Hence every BS sould ave a unique PCI in its neigborood. In tis paper we present a framework for eliminating te PCI confusion problem in eterogeneous LTE using our novel concept of extended syncronization signals (ESS). Te ESS extends te basic syncronization signals (BSS) by embedding additional information about te femtocell. Based on ESS we propose a centralized PCI planning algoritm tat practically eliminates te PCI confusion possibility. Additionally we sow tat ESS reduces bot signaling load and andover delay wen compared to existing design alternatives. Furtermore te proposed design enables te te MT to identify te femtocell access mode leading to a furter reduction in te overall system signaling overead as well as te MT battery consumption. Te rest of tis paper is organized as follows. In Section II we present an overview for background and related work. In Section III we present our framework tat tackles PCI confusion design issues in eterogeneous deployments followed by our performance evaluation in Section IV. Finally we conclude and present future work in Section V. II. BACKGROUND AND RELATED WORK Tis section first briefly introduce LTE access procedure after wic a brief overview for femtocell access modes is ten provided. PCI relevant topics are te presented presented in te last subsection. A. LTE Access Procedure In order to communicate wit LTE system a MT sould first acquire syncronization wit te cell and ten receive and decode cell system information. Te syncronization process is known as cell searc and is performed on powering up te MT and is repeated wenever te MT is interested in using a new BS. During te cell searc te MT searces for te PSS and SSS witin LTE frames. Eac LTE frame spans 10ms and is divided into 1ms sub-frames wic are furter subdivided into two slots containing six or seven OFDM symbols. In frequency division duplexing (FDD) LTE te PSS

2 2 is transmitted witin te last symbol of te first slot of subframes 0 and 5 wile te SSS is transmitted witin te second last symbol of te same slot. On locking te PSS and SSS te MT sould ave identified te frame timing and te cell PCI. Te PCI is obtained by direct mapping from te PSS (wic represents one of tree possible cell identities) and te SSS (wic represents one of 168 cell-group-identities). Once MT syncronizes wit a new cell te MT can acquire te reference signal wic enables te MT to determine te received signal strengt for mobility purposes and decode te broadcast transport cannel for obtaining relevant system information. Te system information includes a Master information block (MIB) and System information blocks (SIBs). Te MIB contains limited information and is transmitted over te broadcast cannel (BCH) wile te SIBs contain more detailed information and are transmitted over te downlink sared cannel (DL-SCH). It is wort noting tat for reading DL-SCH te MT as to first acquire MIB wose transmission time interval (TTI) is 40ms. SIB-1 contains information about te access rigts of te BS and as a typical TTI of 80ms. Oter SIB (SIB2-SIB13) contains oter relevant information and are typically transmitted at longer TTIs [4]. B. Femtocell Access Modes Te LTE standards define tree different access modes [2] including closed open and ybrid. Closed access femtocells are restricted to a specific set of users; open access femtocells can be used by any MT; and ybrid access femtocells can be accessed by any MT wit a prespecified set of users olding iger access priorities. In 3GPP standards te concept of closed subscriber group (CSG) is introduced to define te group of users wit specific access rigts to a closed or ybrid access femtocells. Te identification of te femtocell access mode is required to avoid unnecessary signaling sent by MTs requesting camping to femtocells tat tey do not ave te rigt to access. For example a MT tat as no subscription wit any CSG would not be interested in accessing closed access femtocells. Note tat femtocell access requests are forwarded by te serving BS to te network core were suc requests are processed. Hence suc unnecessary requests cause significant signaling and processing overead. In order to avoid suc overead a combination of a CSG indicator (one bit flag) and a CSG identifier are broadcasted in te SIBs to identify te access mode. However tis approac is expensive in terms of te associated delay wic is critical for active users aving ongoing real-time applications. Anoter alternative to avoid tis expensive delay is to split PCI range into groups corresponding to open access 1 ybrid access and closed access BSs. Hence te MT can identify te type of te femtocell by te end of te cell searc pase. However te limited range of PCI in eac cluster would magnify te PCI confusion problem. 1 Open access BSs include macro and open access femtocells. (a) PCI Collision C. PCI (b) PCI Confusion Figure 1: PCI Problems Typically te PCI sould represent a local unique identifier in te serving cell neigborood as it is directly mapped from te syncronization signals and is used in te derivation of te cell scrambling codes. Hence tere exist two typical PCI planning requirements including collision-free medium and confusion-free medium. Te former requirement mandates tat any two neigbor cells i.e. cells wit coverage overlap sould not ave te same PCI. If a PCI collision occurs te MT located in te common coverage of te two cells may not be able to decode te cannels of te serving base station as illustrated in Figure 1a. Te second requirement mandates tat any two cells aving te same PCI sould not sare a neigbor. In case of PCI confusion te serving cell would not be able to identify te target cell on receiving a measurement report for andover purposes as illustrated in Figure 1b. Two approaces are proposed in te literature to avoid PCI confusion for femtocells. Te first solution is based on including an additional identifier wit te PCI in te measurement report. Te global cell identifier (GCI) is a typical identifying parameter. Te GCI in LTE consists of two parts [5]: PLMN Identity: Te identity of te Public Land Mobile Network. CIPL: Unique Cell Identity for a cell witin a PLMN. Tis approac is a delay critical solution especially for active MTs because GCI would not be available until te MT reads te SIBs. Te second approac is to send conditional andover messages to all cells aving te same PCI in te serving cell neigborood. Te latter approac leads to a significant signaling and processing 2 load increase especially in dense femtocell deployment. PCI planning plays a role in reducing te PCI confusion problem. Two approaces are proposed for PCI allocation in LTE including centralized and distributed mecanisms. In te former a central entity typically located in te evolved packet core would collect te initiating femtocell information and assign it a new PCI based on a wider overview for te system wile in te latter eac femtocell allocates its PCI based on te local collected information. Note tat in macro-tier te base 2 Note tat all femtocell signaling as to to go troug te network core due to te independent nature of its implementation.

3 3 stations can communicate to eac oter using X2 interface to collect information about neigbor cells. On te contrary femtocells only depends on te sensed information due to its deployment nature. In [6] and [7] te autors present PCI allocation under centralized and distributed settings for macrotier. In [8] te autors propose a PCI allocation algoritm for incremental network growt wit minimal overead using grap coloring as a typical tool for suc type of planning problems. In [9] te autors propose a centralized PCI assignment function entity tat collect information form initiating femtocells and ten assign te femtocell a suitable PCI. Tis information may include a list of sensed neigbors te configured access mode information (including te CSG ID) and previously assigned PCI if applicable. III. ESS FRAMEWORK In tis section we first present te concept of extended syncronization signal (ESS). Second we present te ESSbased PCI planning strategy in centralized and distributed deployments. A. Extended Syncronization Signals Bot PSS and SSS are based on Zadoff Cu (ZC) sequences [4]. Te PSS and SSS are respectively transmitted in te last and second-last symbols of sub-frame 0 and 5. Te PSS uses a 63-bit ZC sequence (only 62 bits are used after dropping te middle bit) tat is transmitted in bot PSS symbols. Te SSS uses two 31-bit ZC sequence tat are interleaved before transmission in te assigned slots; i.e two versions of SSS are transmitted in te defined OFDM symbols. Tis sift will allow te MT to acquire complete frame timing by detecting te PSS and one of te SSSs. Te syncronization ZC sequences are typically padded wit five zeros before and after te syncronization signals resulting in a 72-bit sequence tat is passed to te OFDM modulator. In te proposed extended syncronization signal design te padding zeros are replaced by additional information bits tat can be used for different purposes. Typically eigt 5-bit groups (totaling forty padding bits) can be used before and after te syncronization signals witin eac LTE frame. However te syncronization signal is extended wit only twenty bits since te cell searc process is concluded by locking one PSS and one SSS witin tis frame. Tese twenty bits are split into four five-bit groups denoted G1-G4. Tese groups are transmitted as sown in Figure 2. As sown in te figure G1 and G2 are respectively transmitted before and after PSS Zadoff- Cu codes. Similarly G3 and G4 are respectively transmitted before and after SSS Zadoff-Cu sequences. Note tat only femtocells would advertise suc additional information bits. Hence once a MT decodes te symbols of te syncronization signals it can identify it as a macrocell if all te information bits are set to zero. Additionally tese bits can be defined in different ways to obtain additional information about te femtocell. One possible design for suc additional information bits is Te most significant bit is used to determine te access mode of te femtocell. On decoding one in tis bit te Figure 2: Extended Syncronization Signal Structure (FDD LET) MT would identify te femtocell as a closed access cell; wile on decoding it as zero te MT would identify te femtocell as an open or ybrid access. Hence if a MT is not interested in CSG cells it would not proceed wit detecting te reference signal and can dismiss tis cell. Additionally it may start looking for anoter cell instantaneously. Consequently tis bit is considered as an early stage filtering rule tat reduces te MT battery consumption speeds mobility function and reduces overall signaling. Te remaining information bits are taken from one of te femtocell identifiers. For example te least significant bits of te GCI. Tis identifier would be sent witin te measurement report togeter wit te PCI to te serving BS. Tis additional identifier would not only significantly reduce te PCI confusion probability but will also improve te andover performance especially for real-time applications by reducing te time required for measurement gaps as presented in Section IV. To tis end it is wort noting tat ESS is compatible wit legacy devices tat can only send PCI in teir measurement report or employ measurement gaps to obtain te GCI. However tese legacy devices would not enjoy te benefits of ESS due to te iger probability of PCI confusion. B. PCI Planning 1) Centralized PCI Planning : By employing centralized PCI planning and ESS concept te PCI confusion can be practically eliminated for eterogeneous LTE systems. In tis section we assume tat te macro-tier is PCI confusion free and we focus on eliminating PCI confusion and collision problems at te femto-tier as well. In order to acieve tis te following PCI sets are defined - Ω s : a set of sensed PCIs at te download link (DL) of te initiating femtocell. - Ω n : a set of PCIs of te sensed neigbor BSs. Tese cells are determined by te PA-function according to te information provided in te Ω s.

4 4 - Ω nn : a set of PCIs of te neigbors of neigbors. Tis set is determined troug te information read from te SIBs and information stored in te PA-Function. - Ω a : te set of all possible PCIs. Tis set typically includes te 504 PCI defined in te standard. If te PCI range is split to define different ranges for various femtocell access modes tis set would include te corresponding PCI range for te type of te initiated femtocell. - Ω id : te set of PCIs of femtocells tat ave bot a common overlaying macrocell and identical ESS information bits to te initiating femtocell. For eliminating te PCI confusion te initiating femtocell sould not be assigned a PCI from Ω id. - Ω c : te set of te candidate PCIs. Tis set includes all te possible PCIs after excluding all te PCIs tat can cause a PCI collision or confusion problems i.e. Ω c = Ω a /[Ω n Ωnn Ωid ]. Once te Ω c is determined a PCI is selected and assigned to te initiating femtocell according to te adopted PCI selection policy. Tis policy can be selecting a random PCI from te list or coosing te least used PCI or any oter policy. 2) Distributed PCI Planning: In distributed PCI planning te process is mainly dependent on sensed information in te initiating femtocell DL. In enterprise implementations te femtocell may be able to communicate wit oter femtocells peers connected to te same communication backbone to collect more information about femtocells in te local neigborood. Hence Ω s Ω n Ω nn and Ω id are all based on locally sensed or collected information. Te allocated PCI is cosen randomly from te calculated Ω c. Hence te probability of PCI confusion is a non-zero probability. However by including te ESS information in te measurement report tis probability is far less tan te probability or PCI confusion in systems using basic syncronization signals (BSS). IV. PERFORMANCE EVALUATION A. Handover Delay Analysis Te andover delay denoted as d o in LTE as several components including - te cell searc delay denoted as d s required for locking PSS and SSS. Since te TTI of syncronization signals is 5ms te time required to lock te PSS is 2.5ms on average. Te following SSS would be transmitted after 5ms. Hence te cell searc delay is expected to be 7.5ms. - measurement delay denoted as d m required for reading te reference signal wic is transmitted in every resource block witing any sub-frame. Hence d m is estimated as 0.25ms. - reading system information delay denoted as d i required for obtaining GCI of femtocells from system information blocks. Tis delay include te delay required for reading MIB (20ms on average) and SIB-1 (40ms on average) totaling 60ms on average 3. - total andover messages propagation delay denoted as d l required for message propagation in between te serving 3 It is wort noting tat te MIB and SIB are transmitted several times witin teir TTIs. Depending on te te signal to interference and noise ratio users would be able to decode te information blocks after different number of block reception. and target BSs in te access part and te involved entities in te network core suc as te mobility management entity (MME) and te femto-gateway (F-GW). In any andover involving a femtocell two round trips are required. One way delay is required for submitting te measurement report to te evolved packet core network to ceck access rigts and decision taking. Assuming te MT as te rigt to access te target cell a round trip delay is required to forward te andover request to target BS and receiving te target BS confirmation. Finally anoter one way delay is required to send te andover acknowledgment to te MT. For more details about te andover procedure and signaling te reader is referred to [3]. In [10] one-way delay is estimated as 5ms. Hence te expected d l equals 20ms. - processing delay denoted as d p wic includes processing at MT macro-bs femto-bs and involved core network entities. Tis delay is common between bot BSS and ESS and is approximated to 0.5ms per involved entity. Hence te total processing delay at different involved entities is approximately assumed to be 5ms i.e. d p = 5ms. Te summation of tese delay components provides an estimate for te required andover delay. Yet te main difference between BSS- and ESS-based andover is eliminating te time required for reading te system information. Hence te andover delay for BSS-based system d BSS can be estimated as d BSS = d s + d m + d i + d l + d p 92.75ms. Te andover delay for ESS-bases system d ESS estimated as d ESS = d s + d m + d l + d p 32.75ms. can be Hence te proposed concept of ESS leads to a reduction of 65% in andover delay in comparison to BSS-based system tat depends on reading system information to avoid PCI confusion. Additionally adopting ESS would also prevent introducing any transmission gaps required for reading system information. Hence using ESS would improve te andover performance of real-time applications suc as voice and video communications. B. Access Mode Identification Identifying te femtocell access mode witout reading te system information is a required design feature. Tis feature not only reduces te battery usage of te MT [2] but also enables avoiding transmission gaps required for reading te system information for active MTs. ESS realizes tis goal witout compromising oter system configurations by including an access mode indicator in te most significant bit of te ESS. Hence by adopting ESS if a MT is only interested in open access or ybrid access femtocells te MT will instantaneously ignore closed access femtocells. Note tat te oter design alternative is to adopt PCI slicing into ranges dedicated for different access modes. Clearly te latter approac will magnify te PCI confusion problem especially if BSS is used. It is wort noting tat bot PCI range slicing and ESS can be used togeter and in tis case te entire 20 bits can be used for te additional identifier.

5 BSS ESS To tis end it is wort noting tat P c in systems wit centralized allocation function is deterministic and can be expressed as { } 0 nf n P c centralized = l 1 n f > n l Confusion Probability e-05 1e-06 were n l respectively equals n c and n c n id for BSS- and ESSbased systems. Hence using ESS would practically provide a PCI confusion free system as te number of femtocells witin an LTE macrocell will not practically approac n c n id even in 3d realizations. 1e-07 1e Number of Femtocells per Macrocell Figure 3: PCI confusion probability for BSS and ESS C. PCI Confusion Probability In tis section we compare te probability of PCI confusion denoted as P c of BSS and ESS for distributed PCI allocation. On using BSS te PCI confusion would occur if two or more femtocells use te same PCI under te same macrocell coverage. Note tat we assume no confusion would take place in te same tier since tis type of confusion can be avoided using self-ealing mecanisms wic are based on femtocell reconfiguration according to user measurement reports. We also assume tat te PCI is selected randomly from te estimated candidate set Ω c. Let n f denotes te number of femtocells installed in a specific macrocell and n c denotes te number of candidate PCI. Note tat n c equals te cardinality of Ω c. Hence P c for BSS can be expressed as P c = 1 P c were P c represents te probability of no PCI confusion. On using BSS Pc BSS can be derived as } P BSS c = { nc P nf n n f c n f n c 0 n f > n c were nc P nf represents te n f permutations. On using ESS a PCI confusion would occurs if two or more femtocells ave te same PCI te same access mode bit and identical 19 least significant bits of GCI. Hence te Pc ESS can be derived as { } P ESS c = (α 2 + (1 α) 2 ) (ncn id ) P nf (n id n c) nf n f n c n id 0 n f > n c n id were α represents te probability tat a femtocell access mode is closed and n id represents te number of possible combinations of te 19 least significant bits of te GCI; i.e. n id = Figure 3 plots te confusion probability for BSS- and SSSbased systems versus te number of femtocells per macrocell wit α = 0.5. Te figure clearly sows tat te confusion probability in BSS-based systems is very ig. On using ESS te PCI confusion probability significantly drops to about five order of magnitude less in comparison to BSS-based systems. V. CONCLUSION Te dense deployment of femtocell in LTE cellular networks introduces several design issues to te system. PCI confusion is one of te most critical mobility management callenges in eterogeneous LTE networks. Te existing solutions would eiter lead to an increase in andover delay or te signaling and processing overead in te network core and backaul. Te proposed ESS represents a novel practical framework to eliminate PCI confusion in eterogeneous deployment. Additionally te presented ESS design provides a means tat enables te MT to identify te femtocell access mode witout reading te system information. Hence ESS not only improves te andover performance for active MTs but also reduces te battery drainage and signaling load required for some essential mobility functions for bot active and idle devices. REFERENCES [1] A. Damnjanovic J. Montojo Y. Wei T. Ji T. Luo M. Vajapeyam T. Yoo O. Song and D. Malladi A survey on 3GPP eterogeneous networks IEEE Wireless Communications vol. 18 no. 3 pp [2] A. Golaup M. Mustapa and L. B. Patanapongpibul Femtocell access control strategy in UMTS and LTE IEEE Communication Magazine vol. 47 no. 9 pp [3] J. Zang and G. de la Roce Femtocells: Tecnologies and Deployment. Wiley Dec [4] E. Dalman S. Parkvall and J. Skold 4G: LTE/LTE-Advanced for Mobile Broadband. Academic Press [5] M. Amirijoo P. Frenger F. Gunnarsson H. Kallin J. Moe and K. Zetterberg Neigbor Cell Relation List and Pysical Cell Identity Self-Organization in LTE in Proc. IEEE Int. Conf. Communications Worksops ICC Worksops 08 pp [6] Y. Liu W. Li H. Zang and W. Lu Grap based automatic centralized PCI assignment in LTE in Proc. IEEE Symp. Computers and Communications (ISCC) pp [7] Y. Liu W. Li H. Zang and L. Yu Distributed PCI Assignment in LTE Based on Consultation Mecanism in Proc. 6t Int Wireless Communications Networking and Mobile Computing (WiCOM) Conf pp [8] T. Band G. Carle and H. Sanneck Grap coloring based pysicalcell-id assignment for LTE networks in Proc. of te 2009 Intl Conf. on Wireless Communications and Mobile Computing IWCMC 09 pp New York NY USA: ACM [9] Y. Wu H. Jiang Y. Wu and D. Zang Pysical Cell Identity Self- Organization for Home enodeb Deployment in LTE in Proc. 6t Int Wireless Communications Networking and Mobile Computing (WiCOM) Conf pp [10] D. Singal M. Kunapareddy and V. Cetlapalli LTE-Advanced: Latency Analysis for IMT-A Evaluation. Wite paper 2010.