A Novel Architecture for LTE-B

Transcription

1 GC'12 Workshop: International Workshop on Emerging Technologies for LTE-Advanced and Beyond-4G A Novel Architecture for LTE-B C-plane/U-plane Split and Phantom Cell Concept Hiroyuki Ishii DOCOMO Innovations, Inc 3240 Hillview Avenue Palo Alto, CA USA ishii@docomoinnovations.com Yoshihisa Kishiyama and Hideaki Takahashi Radio Access Network Development Department NTT DOCOMO, INC. 3-5 Hikari-no-oka, Yokosuka-shi, Kanagawa-ken Japan kishiyama@nttdocomo.com, h-takahashi@nttdocomo.co.jp Abstract This paper introduces a novel approach in increasing the capacity of LTE cellular networks. The solution is based on massive deployment of small cells by leveraging high frequency reuse at high frequency bands in conjunction with a Macrocell. The presence, discovery and usage of the small cells are controlled dynamically by a Macrocell in a master-slave configuration hence they are called Phantom Cells. To realize this concept, a new method of managing the connections between mobile terminals and small cell nodes is introduced. It is achieved by splitting the Control and User (C/U) planes of the radio link. The combination of C/U-plane split and Phantom Cells can achieve high capacity enhancement using small cells at the same time taking into consideration mobility, scalability and flexibility requirements for massive deployment. The advantages of this approach as well as the implementation aspects are described in the paper. Simulations were also conducted to verify the concept and the results show some promising capacity enhancements. The rest of the paper describes the Phantom Cell concept as well as the challenges of deploying small cells in LTE networks. Keywords-LTE; C-plane/U-plane Split; Phantom Cell, HetNet I. INTRODUCTION There has been a significant growth in cellular traffic over the past few years with the introduction of smart phones, tablet devices and other new mobile devices supporting a wide range of applications. If significant traffic growth continues, current cellular capacity will not be able to support future demand. According to recent forecasts, mobile data traffic is expected to grow beyond 500 times in ten years ( ). Last year alone, mobile data traffic grew 2.3 fold and the average smartphone usage nearly tripled [1]. To meet the increasing demand, studies on the further development of Long Term Evolution (LTE/LTE-A) technologies referred to as LTE-B, are being started in the 3GPP standardization [2]. The capacity enhancement approaches consist of three main technology components: spectrum efficiency, spectrum extension, and network density as shown in Fig. 1. Spectrum efficiency solutions include beam-forming, Massive MIMO, Coordinated Multipoint Transmission and Reception (CoMP), and advanced receiver techniques while spectrum extension includes Carrier Aggregation (CA) [3]-[7]. Significant capacity enhancement can be achieved using Heterogeneous Networks (HetNet) [8], where various types of small cells are densely deployed in addition to conventional Macrocells. In the current discussions, there is also a strong emphasis for improving celledge throughput including data rate fairness, a major issue in the current LTE networks [9]. In fact LTE capacity decreases as more terminals are located near the cell edges [10]. For network density enhancements, co-channel deployments of Macrocells and small cells were intensively discussed and studied in 3GPP Release 10/11 [8]. It includes small cell solutions such as pico/femto nodes, remote radio heads and relay nodes sharing the same frequency with the Macrocells. However dense deployment of small cells in cochannel deployments can result to interference between Macrocells and small cells. As the Macrocells provide the underlying network coverage, operators will hesitate to deploy small cell solutions that would impact the Macrocell layer s key performance indicators. A more straight-forward solution for enhancing network density in cellular networks is introducing additional layers at different frequencies (previously known as Hierarchical Cell Structure). For LTE this means deploying smaller cells at a different frequency band. In the network evolution cube in Fig. 1 this corresponds to both spectrum extension and network density enhancements. Enhanced MIMO, Advanced Receiver, CoMP, Controller Spectrum Efficiency Small Cells Low Figure 1. Capacity Evolution Cube Spectrum Extension High (Wider BW) F /12/$ IEEE 624

2 Small cells can provide higher capacity in traffic hotspots and using a different frequency band or carrier avoids interference, re-optimization and re-planning of the existing Macrocells. However, there are still two fundamental problems when reducing the cell sizes to very small areas mobility and connectivity. Smaller cells can increase handovers and cell changes in high mobility areas [11]. Moreover, small cells in dense environments require careful cell planning and are logistically demanding to operate in a large scale. Therefore capacity enhancement proposals using small cells must be simple and capable of addressing these two problems. In this paper, a higher frequency band solution called Phantom Cells with C-plane/U-plane Split is discussed. In Section II, the high frequency deployment scenario is described followed by a discussion of C-plane/U-plane Split in section III. In Section IV the Phantom Cell implementation is discussed in more detail together with its advantages. Performance evaluation is shown in Section V followed by conclusions. II. HIGHER FREQUENCY BANDS FOR LTE-B Spectrum below 2.5 GHz is currently fully utilized for cellular systems. Hence future capacity expansion for LTE is envisaged at 3.5 GHz and higher. However, higher frequency bands suffer from higher propagation loss compared to lower bands. Conversely, higher frequency bands have more available spectrum making them suitable for considerable capacity expansion. With pathloss frequency factor in the order of 20dB/decade, higher frequency cells will never match the coverage of the existing Macrocells if using co-located sites [12]. Therefore a practical approach is to deploy small cells at different locations from the Macrocell. In the current deployments, there are a number of capacity solutions for indoor environment such as WiFi, Femto cells, and in-building cells with Distributed Antenna Systems (DAS). However capacity solutions for the outdoor environment that support good mobility and connectivity are still missing. It is therefore necessary for small cells solution at high frequency bands to support mobility to also address the capacity demands in outdoor environments. There are two scenarios where small cells can be deployed together with Macrocells (see Fig. 2). In normal high traffic spots small cells are sparsely deployed (such as railway stations in sub-urban areas). In this scenario, good inter-frequency mobility is required for both connected mode and idle mode (or idle-like mode if Always-ON operation is assumed). That is, the small cells need to be detected quickly with low user equipment (UE) power consumption. However, the current inter-frequency handover procedures are based on signal strength and signal quality primarily for coverage reasons. They are not optimized for boosting data rate or increasing cell capacity. Therefore more efficient inter-frequency cell identification and measurements procedures may need to be introduced to meet the mobility requirements in sparsely deployed scenarios. In a super high traffic area (i.e. downtown Tokyo) small cells need to be densely deployed in a continuous or partial coverage. In this scenario, good intra-frequency mobility is required between small cells. Furthermore, small outdoor cells are expected to support medium-to-high speed mobility (up to 50km/h) while maintaining good quality of service without high control signaling overhead due to the so-called ping-pong effects. Performance impairment to small cells at high mobility is mainly due to high failure rates and packet loss rates during handover. III. C-PLANE/ U-PLANE SPLIT To provide an efficient, flexible and low-cost solution at higher frequency bands, a C-plane/U-plane Split configuration is proposed. In this configuration, the C-plane must be supported by a continuous more reliable coverage layer at lower frequency band while the U-plane can be provided by high capacity smaller cells. This concept is shown in Fig. 3. The C-plane is provided by a Macrocell at low frequency band to maintain good connectivity and mobility. The Macrocell also works as a normal cell supporting both C-plane and U- plane signaling. On the other hand, the U-plane is provided by a small cell at higher frequency band to boost user data rate. However, the small cells are not conventional cells because they are not configured with cell-specific signals/channels, i.e. primary/secondary synchronization signals (PSS/SSS), cellspecific reference signals (CRS), master information block/ system information blocks (MIB/SIB), etc. Hence they are called Phantom Cells which are only intended to carry user traffic. The Radio Resource Control (RRC) connection procedures between the UE and a Phantom Cell such as channel establishment and release are managed by the Macrocell. Effectively, a Phantom cell provides one-to-one connection for UE without any cell specific signals/channels, resulting to a flexible and efficient operation. With this concept, network operators can maintain basic mobility/connectivity performance even when they deploy only a small number of higher frequency band cells. Super high traffic area Normal high traffic spot Macro cell (low frequency) C-plane SPLIT!!! U-plane Phantom cell (high frequency) Figure 2. Outdoor Deployment Scenarios for Higher Frequency Bands Figure 3. C-plane/U-plane Split and Phantom Cell 625

3 In terms of scalability, it enables network operators to gradually add capacity. Since the Phantom cell coverage does not have to be continuous, its deployment can start with a few small cells in one higher frequency carrier (i.e. hotspot areas). Further capacity requirements can be met by gradually increasing the number of cells forming a continuous Phantom Cell layer. Then additional layers at second and third carrier frequencies can be added at a later stage. This is one of its attractive features in terms of deployment. Similar concept to this configuration is adopted in the shared cell concept [8] and the IEEE j standard by providing low-cost range expansions for base stations supporting small cells or relay nodes. In contrast, the Phantom Cell concept s main motivation is achieving high capacity using small cells at the same time taking into consideration the mobility, scalability and flexibility requirements for massive deployment. Furthermore, when the C and U-planes for each connection are established separately at different frequency bands, new technical challenges are expected as opposed to cochannel solutions. A. Single Node C-Plane/U-plane Split Configuration A single node C-plane/U-plane split configuration requires the baseband processing to be performed in one node. This is possible even if the C-plane and the U-plane are having different transmission points (i.e. different locations as in Fig. 3). This configuration can be realized by the conventional Carrier Aggregation (CA) utilizing Remote Radio Heads (RRH) described in Annex J of TS Scenario 4 in [13]. Hence a RRH assigned with U-plane only and higher frequency carriers is one type of Phantom Cell implementation. In RRH implementation of Phantom Cells the baseband processing are all performed at the Macrocell. Thus the Macrocell and the Phantom Cell is logically a single node. One requirement for this implementation is the mandatory optical fiber link between the Macrocell and Phantom Cell. Additionally, the coverage area of the Phantom Cells must be inside the coverage of the Macrocells as they operate as a single node. B. C-plane/U-plane with Separate Nodes In addition to the conventional CA-based implementation, the C/U-plane Split can be provided by separate nodes, as illustrated in Fig. 4. Control signaling for managing the Phantom Node is transferred by Macro Node via New Interface. In this configuration, the relationship between the Macro Node and the Phantom Node is master-slave rather than peer-to-peer. This new configuration has a number of advantages. Firstly the throughput/latency requirements for backhaul between Phantom Cells and Macrocells can be relaxed, because the Macro Node and Phantom Node are different transmission points with independent baseband processing. More specifically, optical fiber backhaul is not required in this configuration, i.e., non-ideal backhaul is applicable to this configuration. Secondly, the independent U-plane processing Figure 4. C-Plane and U-Plane with Separate Nodes and Independent U- Plane Processing allows a Phantom Cell to cooperate with more than one Macro Node, i.e. the Phantom cell can be located among multiple Macro Nodes at coverage boundaries. As a result, the Phantom cell can enhance the user throughput at Macrocell s edge regions. Thirdly, system scalability can be supported. The main limitation of the conventional CA-based configuration is the centralized U-plane processing at the Macro node which can potentially limit the number of Phantom Cells it can support. If more Phantom Cells are required per Macrocell, there is a need to assign some of the U-plane processing to the Phantom nodes. This configuration allows for further addition of new Phantom Cells or new carriers for even more capacity. It also offers flexibility in allocating U-plane resources. IV. PHANTOM CELL CONCEPT AND ARCHITECTURE An efficient realization of the Phantom Cell is by adapting the separate-node configuration. To support mobility and high capacity using this configuration the following requirements need to be satisfied: Macro Node New Interface (Non-Ideal Backhaul) Phantom Node The Phantom Cell radio link requires a new carrier type with no (or much less) common signals/channels, new discovery signals and L1/L2 mobility support between Phantom cell and UE The C-plane of the Phantom Cell is managed by the Macro Node in a master-slave mode The Phantom Cell baseband processing does not have to be located in the Macro Node The U-plane of the Phantom Cell requires a new data path (backhaul) to the core network. A new RAN architecture is proposed to satisfy the above assumptions in the next subsection. A. New RAN Architecture Supporting Phantom Cells To fully benefit from the C/U-plane split configuration, the Phantom Cell implementation requires a novel Radio Access Network (RAN) architecture (see Fig. 5). The master-slave relationship between the Macro Node and the Phantom Node requires a new interface (X3). However, the Phantom Node does not require a C-plane interface towards the core network (S1-C). The data path for user traffic is provided by S1-U interface between the Phantom Node and the core network. 626

4 UE Phantom cell Macro cell Server Macro Node C-plane UE X3 - interface U-plane S1-C Phantom Node S1-U user data path Figure 5. LTE RAN Architecture for Phantom Node From the above configuration, a new channel establishment procedure is shown in Fig. 6 detailing some of the control procedures. The figure shows the signaling exchange between the UE, Macrocell and the Phantom Cell related to measurements, preamble and connection setup for the U-plane while the Macrocell maintains the C-plane. It also shows how a normal data transfer in the LTE U-plane can be reconfigured via the Phantom Cell. The physical layer procedures for Phantom Cells are still under study. However, there is a general trend in 3GPP Release 10/11 that most procedures, such as demodulation, channel estimation, and measurements, are independent from Cellspecific Reference Signals (CRS). For example data and control channels can be transmitted by UE specific reference signals (Transmission mode 9/e-PDCCH) and measurements can be conducted by using Channel State Information Reference Signals (CSI-RS) [14][15]. With these features already considered in the specifications, it implies that the oneto-one connection between Phantom Cell and UE can be realized with marginal efforts. B. Implementation Aspects of Phantom Cells As discussed in Section II, the future capacity solutions using small cells (Phantom Cells) needs to support both indoor and outdoor environments, and therefore needs to offer good mobility support. This involves efficient cell discovery and L1/L2 mobility. Since Phantom Cells are deployed in a carrier frequency different than the Macrocell carrier, the UEs need to perform inter-frequency measurements to discover them. To implement a fast discovery procedure, a new discovery signal is proposed. For fast and power-efficient detection, it is necessary for the discovery signals to be transmitted synchronized with the Macrocell in a periodic manner. Such time synchronization with the Macrocell can significantly reduce the detection effort of the UE for a different carrier, because it does not have to search the discovery signals all the time. Furthermore, these signals also need to be orthogonal for easy Phantom Cell identification in densely deployed scenarios. Thus fast and power-efficient detection can be achieved through rough initial synchronization and efficient decoding of orthogonal signals. For the densely deployed scenarios, efficient L1/L2 mobility among Phantom Cells must be introduced in order to support medium to high UE speeds (e.g. up to 50 km/h) and minimize control signaling caused by many small cell handovers. That is, a notion of large phantom coverage area consisting of multiple Phantom Cells must be supported Data transfer in normal LTE link (U-plane) Measurement control for Phantom to UE (RRC) Measurement for Phantom-to-UE Measurement report for Phantom to UE (RRC) Preparation Preparation confirm Connection setup request for Phantom to UE (RRC) RA preamble for Phantom to UE (L1/ L2) RA Response (L1/ L2) Connection setup confirm for Phantom to UE (RRC) Data transfer in Phantom-to-UE link Data transfer in normal LTE link (U-plane) Figure 6. An Example of Connection Establishment Procedure coupled with efficient point association (Phantom Cell association), which is so-called L1/L2 mobility. This can be realized by an architecture where multiple Phantom Cells are controlled by a center-node Phantom Cell with ideal backhaul (e.g. optical fibers). In the execution of handover, the hysteresis or time-to-trigger can be removed in the L1/L2 mobility. In this way medium to high UE speeds can be supported. Furthermore, since L1/L2 handover can be ignored by higher layer nodes, control signaling increase can be avoided even in dense small cell deployments. There are other implementation aspects that need to be addressed such as interference coordination, energy saving and load balancing. These are discussed as follows: For maintaining connection quality, interference between Phantom Cells needs some coordination depending on the actual network configuration. For example, the inter-cell interference coordination (ICIC) depends on the duplexing scheme. As traffic is expected to be asymmetric, Dynamic- TDD may be adopted. Depending on the number of frequency bands available, interference coordination would be necessary. For Dynamic TDD, ICIC comes in a variety of techniques such as Macrocell-controlled slot allocations, frequency domain scheduling, and time-domain slot partitioning. These are optimized for user data rates and block error rates. Such ICIC techniques can be realized easily by the characteristics of Phantom, i.e. the absence of cell-specific channels/signals. Energy saving is also essential in a massive deployment of Phantom Cells. Depending on the traffic levels, some Phantom cells can be turned ON/OFF (in conjunction with power adjustments) without affecting the coverage. Such dynamic ON/OFF behaviors are well fit to Phantom Cells. Different network configurations are currently under study to determine the most flexible option when implementing this feature. Load balancing is another aspect to be addressed in the Phantom Cell deployments. In small cells, load fluctuations are expected to be even higher than in conventional Macrocell network due to lower average users per cell. If the cells have sufficient coverage overlaps, these can be exploited for load balancing whereby one cell can potentially transfer overflow 627

5 traffic to the next cell resulting to a better handling of large traffic variances. The absence of cell-specific channels makes coverage adjustments easier. In contrast, cell-specific channels/signals in conventional cells need to be set according to the cell s planned coverage. Phantom Cells can perform such load balancing in a simple and efficient manner by adjusting the downlink powers as well as handling of inter-cell interference or frequency planning. This feature is currently investigated for various network configurations. V. PERFORMANCE EVALUATION System-level simulations were performed to evaluate the performance. The aim was to obtain some capacity improvement figures as well as to assess the mobility performance and compare them against a conventional Macrocell deployment. The performance indicators used were (i) total system throughput, (ii) distribution of the Geometry Factor (i.e. ratio of total serving cell power and interference from the other cells including thermal noise), and (iii) distribution of user data rate with (i.e. user throughput vs. distance from the Macrocell). In the capacity simulations a number of Phantom cells were placed inside the Macrocell areas to simulate a very dense deployment. We observed the Geometry Factor in order to see any changes in the behavior of inter-cell interference. For the mobility performance the handover (HO) rates and their corresponding failure rates were studied comparing a conventional HO procedure against L1/L2 mobility procedure in a continuous Phantom Cell layout. A. Simulation Models and Assumptions Models and assumptions are basically aligned with 3GPP simulation case 1 [16]. Macro base stations were simulated uniformly in a 19-Macrocell area (wraparound) with inter-site distance of 500 m. Phantom Cells assigned with a different carrier from the Macrocell. The number of Phantom Cells per Macro sector was set from 4-20 to evaluate the capacity increase due to high frequency reuse. A series of snapshot simulations were performed. In each simulation run, UEs and Phantom Cells were randomly positioned with the minimum distance of 40 m among the Phantom cells in the system area. The radio channel between each Macrocell and UE or between each Phantom Cell and UE was calculated based on the propagation and fading models. We assumed a log-normal shadow fading with zero-mean and standard deviation of 8dB and 10dB for Macro and Phantom respectively together with six-ray typical urban model with 3 km/h speed. A proportional fairness scheduling was used both for Macrocells and Phantom Cells. LTE-based adaptive modulation and coding were used, i.e. transport formats specified in [15] were selected based on radio link quality for each sub-frame (1 msec). Chase combining with the round trip delay of 8 msec was also implemented as hybrid automatic repeat request (HARQ) in the simulation. For user data rate calculations, it was assumed that 30 users with full buffer are located in each macro sector. The average number of UEs per Phantom Cell is 7.5, 3.8, 2.5, 1.9, and 1.5 in the 4, 8, 12, 16, and 20 Phantom Cell cases, respectively. The main parameters are listed in Table I. For the Phantom Cell propagation, the pathloss model is described in Eq. (1) [17]. L = log W log <H> { (<H>/h b ) 2 }log h b + ( log h b ) log (d/1000) + 20 log f c (3.2 (log (11.75h m )) ) (1) where: h b : BS antenna height [m] (= 20) <H>: Average building height [m] (= 20) W: Street width [m] (= 20) f c : Carrier frequency [MHz] (=3500) h m : UE antenna height [m] (= 1.5) Additional parameters for the mobility simulations are shown in Table II based on 3GPP mobility simulation [11]. The downlink SINR was calculated using exponential effective SINR mapping (EESM) at each measurement period, and HO failures were triggered if the downlink SINR is lower than -10 db. UE speed was set to 3 km/ h or 50 km/ h. Two sets of measurement configurations were evaluated in the simulations. One configuration is a typical parameter setting for conventional HO, in which Time-To-Trigger (TTT) and HO hysteresis (Hyst) are set to 320 msec and 3 db, respectively. In the other configuration, TTT and HO Hyst are set to 0 msec and 0 db, respectively, which can be regarded as a parameter setting for the L1/ L2 mobility. In the L1/ L2 mobility operations, conventional HOs (i.e. higher layer HOs), are triggered when the UE crosses the boundary of a large phantom coverage area discussed in Section IV.B. TABLE I. MAIN SIMULATION PARAMETERS Parameters Values Traffic model Full buffer model Transmission power 43 dbm (Macrocell), 30 dbm (Phantom Cell) Path loss (Macro) L = log (d), d = distance in meters Log-normal, 8 db standard deviation Shadow Fading (Macrocell) and 10 db standard deviation (Phantom Cell) Multipath fading Six-ray Typical Urban model, 3 km/ h Cell layout Hexagonal grid, 3-sector sites Inter-site distance 500 m 5 MHz channel bandwidth at 2 GHz (Macrocell) Spectrum allocation 5 MHz channel bandwidth at 3.5 GHz (Phantom Cell)^ BS antennas 2 (Macrocell), 2 (Phantom Cell) UE antennas 2 ^ For comparison purposes, 5 MHz is utilized for both Macrocell carrier and Phantom Cell carrier, but wider bandwidth can be available in 3.5 GHz) TABLE II. SIMULATION PARAMETERS FOR MOBILITY Parameters Values Measurement metric Reference Signal Received Power Measurement bandwidth 6 resource blocks Measurement internal 40 msec Measurement period 5 measurement samples L3 filter coefficient k = 4 SINR threshold for HO Failure -10 db UE speed 3 km/ h, 50 km/ h 628

6 B. Capacity and Throughput Results The combined capacity of Macrocell together with the Phantom Cells under its control (C-plane) is shown in Fig. 7. The total or combined capacity is almost linear as the number of Phantom Cells increase. This finding show that spatial reuse of frequency is possible even for small cells. This result is consistent with the theoretical prediction in [18]. The relatively linear capacity increase can be explained in Figure 8 which presents the distribution of geometry as the number of Phantom Cells increases to very dense values (4-20 cells). The results indicate that the geometry distribution hardly change across all geometry values. This means the inter-cell interference is not a major issue as long as each terminal is connected to the best Phantom Cell. This also provides support for the same theoretical conclusion in the analysis found in [19]. It is expected that the capacity gains higher than what is shown in Fig. 7 can be achieved if deployed in traffic hotspots areas as Phantom Cells can provide better SINR and better handling of traffic overflow from the Macrocell. Figure 9 shows the resulting user data rate vs. user distance from the Macrocell site. As expected, the user data rate from a Macrocell carrier degrades as a user is located farther away from the site. Also there is data rate degradation in the vicinity of the Macrocell site is due to vertical antenna pattern. On the other hand, the average user data rate in the Phantom Cells carrier is constant irrespective of the user s distance from the Macrocell site. One of the reasons for this behavior is the avoidance of Macrocell inter-cell interference near its cell edges because the Phantom Cells use their own carrier with higher SINRs (i.e. effective frequency reuse). To that effect, Phantom Cells can improve cell-edge throughput as well as cell capacity providing a very effective solution to combat the near-far problem of user throughput in LTE networks. C. Mobility Performance The handover performance is shown in Table III in terms of the number of HOs and HO failure ratio. They were simulated based on 4 Phantom Cells per macro sector. The number of HOs was calculated for Phantom-Cell-to-Phantom-Cell HO in the Phantom Cell carrier. From the results, the typical parameter setting (TTT: 320 msec, Hyst: 3 db) cannot support medium to high UE speeds due to high HO failure ratio, which is caused by HO delay. The other parameter setting (TTT: 0 msec, Hyst: 0 db) can suppress the HO failure ratio, but can significantly increases the number of HOs. The 0 msec TTT and 0 db hysteresis does not provide enough margins to handover ping-pong phenomenon. Expectedly, the HO failure problems could get worse as the density of Phantom Cells further increases. The results show that conventional HO procedures are not suitable for small cells. This is in agreement with the finding in [20]. In the L1/ L2 mobility operations, the number of higherlayer HOs can be reduced based on the large phantom coverage area size. For example, if the large phantom coverage area is the same as the Macrocell sector, the higher-layer handover rate caused by the Phantom Cells is equivalent to the conventional handover rate of the Macrocell. This means L1/ L2 mobility can support medium-to-high speed UEs without high control signaling overhead. Thus L1/L2 mobility support is essential for Phantom Cell implementation. Figure 7. Capacity Increase with Phantom Cells CDF Phantom Cells 8 Phantom Cells 12 Phantom Cells 16 Phantom Cells 20 Phantom Cells Geometry (db) Figure 8. Geometry Distribution Figure 9. User Data Rate vs. Distance from Macrocell Site 629

7 TABLE III. TTT: 320 msec, Hyst: 3 db (Macrocell carrier) TTT: 320 msec, Hyst: 3 db (Phantom Cell carrier) TTT: 0 msec, Hyst: 0 db (Phantom Cell carrier) UE velocity (km/ h) HANDOVER PERFORMANCE Number of HOs per minute (^) HO Failure Ratio (%) (x4.3) (x3.2) (x35.8) (x6.8) 9.7 ^ The ratio to the Macrocell carrier is shown for 3 km/ h and 50 km/ h, respectively in (.) VI. CONCLUSIONS This paper proposed an LTE/LTE-A evolution path for supporting future traffic demands using small cells at higher frequency bands. The C-plane/U-plane Split together with Phantom Cell concept is described as a solution for a costeffective and efficient deployment of small cells to address the problem of capacity and cell-edge user data rates. Simulations were conducted to verify the performance for both capacity and mobility. The results show that this approach of utilizing small cells is even applicable in dense area deployments. The capacity figures obtained provide further motivation in developing this concept for future LTE standardization. The paper also described some of the implementation aspects and architecture which this technology requires such as the need for a new carrier type, signaling interface, mobility support and new cell discovery procedure. These techniques are considered as fundamental for small cells and are currently discussed in 3GPP/LTE-B standardization. REFERENCES [1] CISCO Whitepaper, CISCO Visual Networking Index: Global Mobile Data Traffic Forecast Update, , 15 Feb 2012 [2] 3GPP RWS , Requirements, Candidate Solutions & Technology Roadmap for LTE Rel-12 Onward [3] O. Yilmaz, S. Hamalainen, and J. Hamalainen, System level analysis of vertical sectorization for 3GPP LTE, Proc. ISWCS 2009, Sept [4] T.L. Marzetta, Non-cooperative cellular wireless with unlimited numbers of base station antennas, IEEE Trans. Wireless Commun., vol. 9, no. 11, Nov [5] Y. Ohwatari, N. Miki, T. Asai, T. Abe, and H. Taoka, Performance of interference rejection combining receiver to suppress inter-cell interference in LTE-Advanced downlink, IEICE Trans. Commun. vol. E94-B, no. 12, pp , Dec [6] R. Cadambe and S. A. Jafar, Interference alignment and degrees of freedom of the K-user interference channel, IEEE Trans. Inf. Theory, vol. 54, no. 8, pp , Aug [7] R Ratasuk, D Tolli and E Ghosh, Carrier Aggregation in LTE-Advanced in Proc. Vehicular Technology Conference (VTC-Spring) May, [8] S Parkvall et al, Heterogenous Network Deployments in LTE Ericsson Review [9] Future Networks 5th FP7 Concertation Meeting, ARTIST4G, 2010 [10] O Osterbo, Scheduling and Capacity Estimation in LTE, The 23rd International Teletraffic Congress, September 6-8, 2011 in San Francisco, USA. [11] K Yagyu et al, Investigation on Mobility Management for Carrier Aggregation in LTE-Advanced in Proc. Vehicular Technology Conference (VTC-Fall) [12] K Kitao and S Ichitsubo, Pathloss Prediction Formula in Urban Area for the Fourth-Generation Mobile Communications Systems, IEICE Trans. Commun, Vol E91 No.6 June [13] 3GPP TS , V10.7.0, Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E- UTRAN); Overall description. [14] 3GPP TS , V10.5.0, Physical Channels and Modulation [15] 3GPP TS , V10.5.0, Physical layer procedures [16] 3GPP TR , V9.0.0, Further advancements for E-UTRA physical layer aspects [17] Report ITU-R M.2135, Guidelines for evaluation of radio interface technologies for IMA-Advanced [18] H.S. Dhillon, R.K. Ganti, F. Baccelli, and J.G. Andrews, Modeling and Analysis of K-Tier Downlink Heterogeneous Cellular Networks, IEEE Journal on Selected Areas in Communications, vol. 30, no. 3, pages , April [19] B. Blaszczyszyn, M.K. Karray, F.X. Klepper, Impact of the Geometry, Path-Loss Exponent, and Random Shadowing on the Mean Interference Factor in Wireless Cellular Networks, Proceedings of the IEEE Wireless and Mobile Networking Conference (WMNC) [20] 3GPP TR , V0.5.0, Mobility Enhancements in Heterogeneous Networks 630