You are on page 1of 5

On Coverage and Capacity of Relaying in LTEAdvanced in Example Deployments

Ralf Irmer, Member, IEEE, and Fabian Diehm1 Vodafone Group R&D, The Connection, Newbury, RG142FN, United Kingdom. Email ralf.irmer@vodafone.com
The coverage and network capacity could be increased by increasing the Macro Site density, regardless what air interface is deployed. However, the grid of Macro Sites is already very dense in urban areas, and it is almost prohibitive from a cost perspective to increase it further. Micro Sites, which could be deployed on street level (e.g. walls, lamp-posts) have the potential to relieve some of the costs associated with site deployment. However, backhauling, one of the main cost elements, would still be required. Installing backhaul such as fiber or microwave to lamp-posts is also very challenging. The licensed frequency bands used for radio access are very attractive for backhaul as well access and backhaul could use the same technology, antennas etc, and synergies could be achieved in a flexible and modular way. Another expression for these in-band backhauling concepts is Relaying, which has been discussed in academia for a long time [8]. Relaying has found its way into the IEEE 802.16j standard [6],[10], and a WiMAX profile for relaying is currently discussed, as well as relaying concepts for IEEE 802.16m. Relaying is an integral part of the WINNER air interface, a beyond 3G system concept. The benefits of relaying were shown in [9]. Relaying involving mobile terminals is also studied in the literature, but beyond the scope of this paper. This paper assumed decode-and-forward relays because they seem to be the most promising approach. Amplify-and-forward relays (repeaters) are not studied, since they suffer from some issues like transmitreceive attenuation, and noise/interference enhancement. The key advantage of relaying is that coverage and capacity can be traded in some way the huge capacity LTE has to offer in theory would not only be available to few users underneath the base station, but also to shadowed and indoor users. From that perspective, it seems to be reasonable to even take a small loss in capacity into account, if fairness of the system can be improved, leading to a more homogenous user experience. However, relaying also has several drawbacks. Access capacity is wasted for the backhaul link, and complexity and delay are added. It is therefore important to assess the relaying performance very carefully to study all trade-offs, before standardization and deployment are envisaged. Evaluating the capacity and coverage of a mobile communications system which includes relaying is very challenging, since relay deployment and radio channels have to be modeled realistically at the one hand and simply at the other. Also, results on the performance of relaying vary

Abstract Relaying as a means of in-band backhaul has the potential to extend the coverage of Beyond 3G networks, enabling the expected high data rates of these networks to be delivered without increasing the density of traditional macro base stations. Assessing the performance of relaying is not simple, since traditional metrics fail and the performance depends strongly on the actual deployment. The current literature considers usually very artificial deployment environments and propagation models. This paper shows the coverage and achievable peak data rates for an urban area in central London using three-dimensional building data and a ray-tracing simulator. The number of relays per sector is determined for different scenarios, and compared to a Macro Deployment. Index TermsLTE, LTE-Advanced, Relaying, Deployment, Coverage, Capacity

I. INTRODUCTION Third generation mobile networks such as HSPA (High Speed Packed Access) currently experience a huge uptake in data services such as mobile internet access mainly due to the availability of simple, affordable, small and plug-andplay modem devices, attractive flat-rate tariffs, and good network coverage, at least from leading network operators. However, the capacity of current networks is limited; and indoor coverage for high-speed data access poses a challenge. NGNM (Next Generation Mobile Networks Alliance) have addressed the requirements on next generation mobile networks [1], which include increased spectral efficiency and re-use of existing infrastructure including cell sites. 3GPP Long Term Evolution (LTE), IEEE 802.16m and Ultra Mobile Broadband (UMB) all address these requirements with MIMO-OFDMA system concepts. The most important benchmark for further innovation will be LTE Release 8 [4][5]. It comes very close to fundamental limits in terms of link-level performance by using advanced spatial processing techniques [3]. However, fundamental limits are not yet reached, when the system level is considered and deployment topologies are taken into account. Within 3GPP [5], discussions on further development of LTE have started in April 2008 in the context of LTE-Advanced, which addresses operator requirements and envisages to fulfill requirements set out by the ITU for newly identified frequency bands.

1 Fabian Diehm is now with the Vodafone Chair, Dresden University of Technology, Germany

978-1-4244-2644-7/08/$25.00 2008 IEEE

significantly depending on the scenario and the evaluation methodology. There are attempts to describe relay deployments in a semi-analytic way, e.g. in IEEE 802.16j [7] or WINNER [9], but it is very hard for an operator to draw general conclusions. The limitations of published relaying performance assessments were the starting-point of this paper. A real world example area in a dense urban environment (central London) is selected instead. Topology, buildings and the radio channel are modeled using the three-dimensional raytracing tool Pace3D. Relays are placed in an intelligent way by filling coverage holes. II. LTE RELAYING Currently LTE has not developed a concept for relaying, but the majority of 3GPP partners expressed at the LTE Advanced workshop in Shenzen (China) in April 2008 that relaying is an important technology element to address the LTE-Advanced requirements. To enable an analysis of potential relaying concepts anyway, some assumptions are made in this paper on what a relay actually is: Standardized decode-and-forward radio relay (inband backhaul) using the same frequency band as the access link, and the same air interface technology (e.g. LTE). Low equipment cost: The price of relays as assumed in this study is very low compared to conventional Macro Sites, and hence allows deployments with many units to be considered. Low cost is targeted for both the equipment and the actual deployment. Small form factor and low weight: Relays are expected to be small and lightweight, enabling easy installation and support of units on lampposts or building walls. The form factor should be similar to municipal WiFi radio nodes, which are nowadays commercially deployed in some cities. For simulation it is assumed that relays are deployed at a height of 5m.. Low transmission power: The relays used in this study have a maximum transmission power of 30dBm (1 Watt). Omnidirectional Antennas: For deployment simplicity and to address the low-cost requirement, omni-directional relays are assumed in this paper. However, deployed WiFi mesh networks have shown that even 8-fold sectorization is possible in small form-factor. Low operational costs: Using in-band backhaul and low power transmission (30dBm) relays are expected to have low operational costs. Air conditioning and other active cooling are not required. Avoiding directional antennas, personnel need no more qualification for deployment than for any street-lamps. Conforms to radio emission health and safety regulations: This is particularly important as relays are deployed at a low height where it cannot be guaranteed that people do not get close to the devices

To integrate relaying into a standard, a couple of issues have to be addressed and choices have to be made: The relaying concept should be applicable to both
TABLE I MAXIMUM PATHLOSS FOR UPLINK BASE CASE CorresponAssumed ReMaximum ding data User/Relay quired allowable rate (Mbps) Bandwidth SNR pathloss (dB) (MHz): (dB): 1.08 3 103.5 0.4 1.08 3 96.5 0.4 1.08 17 120.2 2 1.08 17 113.2 2

Links UE -> BS UE -> RS RS -> BS RS -> RS

TDD (Time Division Duplex) and FDD (Frequency Division Duplex) schemes. Frame structure to accommodate efficient and low delay relaying. Transparent or non-transparent mode has to be selected this affects preamble design Protocol architecture has to be defined Radio resource management that allows for spectrally efficient relaying HARQ (Hybrid Automatic Repeat Request) concept hop-by-hop or end-to-end Control channels and measurement procedures have to be designed Power control has to be adopted to enable a wellperforming overall system Existing security mechanisms such as authentication and encryption have to be extended to relaying, and to be enhanced if necessary

It is beyond the scope of this paper to address these issues, but the coverage and capacity analysis of this study is expected to give guidance for the development of the relaying concept. III. DEPLOYMENT
LONDON IN

DENSE

URBAN

AREA CENTRAL

A. Existing Macro-Cellular 3G Network For this study, an area with an existing 3G/HSPA MacroCellular deployment by Vodafone was chosen as a reference case. The area shown in Figure 1 in central London has a size of approximately 1km2, but interfering cells outside this area were considered in this study as well to avoid edge effects. Most sites in the examined area have three sectors, and the downtlilt of the antennas has been optimized. The site density in this particular area is about 4.1 sites/ km2. B. LTE Link Budget For the coverage prediction, it is important to base it on accurate link budgets, which are influenced by both, environment and system specific parameters. The environment specific parameters are based on a threedimensional ray-tracing tool, which takes the topology data and all buildings into account. The used tool was calibrated by drive-tests. It takes account of antenna patterns, pathloss and shadowing. For indoor penetration loss, 20 dB is assumed as in the NGMN evaluation methodology [2]. Indoor penetration loss is applied to buildings according to

the topography data base. For coverage analysis, only the uplink (UL) direction is considered since it is more challenging, whereas the average peak rate is calculated for the downlink direction.
529,800 530,000 530,200 530,400 530,600

1542A

1852D

Relays fill coverage holes whilst it is ensured that they have good connectivity to one of the Macro sites directly or via another Relay Node. This deployment procedure is done manually with the assistance of a network planning tool. All simulations are carried out at a carrier frequency of 2.6 GHz. The network planning strategy in this paper is purely coverage oriented. The data traffic is assumed to be equally distributed within the cell. Only the strongest signal at each location is taken into account. It is assumed that inter-cell interference can be treated with other concepts, such as advanced scheduling, fractional frequency reuse or radio resource management (RRM). Interference is discussed later in this paper.
Define Link Budgets
007'40"W

181,600

181,600

181,400

181,400

181,200

786A

358C

23009G

Apply BS to UE budgets to existing macro sites and observe coverage at 1.5m height
181,000

5131'00"N

181,200

5131'00"N

1852D

181,000

007'40"W 007'40"W

Figure 1: Central London current 3G/HSPA Macro Sites The system specific parameters can be described by SINRto-data rate mappings. LTE SINR-to-data-rate mappings are used, which include HARQ, adaptive modulation and coding (AMC) and MIMO schemes up to 2x2 spatial multiplexing in the downlink and 1x2 SIMO in the uplink. A simple Pedestrian B channel model with 3 km/h is used in the SINR-to-data rate mapping to take into account frequency-selectivity and fading. This approach is wellsuited for the analysis in this paper, but could be extended in the future by using more elaborate channel models such as SCM-E (see e.g. [2]) and using link-level measurements from LTE equipment. LTE downlink is based on OFDMA, whereas the uplink is based on SC-FDMA (Single Carrier Frequency Division Multiple Access). For the UL coverage link budgets we assume an SNR of 3 dB on both the relay and base station access links to be achieved for users with an assignment of 6 simultaneously used LTE resource blocks. This equals to a bandwidth of 1.08 MHz per user if a 10 MHz FDD system with 50 available resource blocks is considered. These assumptions translate into an UL target user data rate of about 400 kbps. User data rates are calculated by assuming a total overhead of 30% for both, up- and downlink. On the relay links (relays to base stations and relays to other relays) an SNR of 17 dB is assumed. Together with the assumption of 6 simultaneously used resource blocks this translates into a data rate of about 2 Mbps. These assumptions and the resulting allowable pathlosses which make up the base case link budgets are given in Table I. The presented pathlosses do not account for antenna gains (which are taken into account by the ray-tracing tool) and include an indoor penetration margin of 20dB. C. Relay Deployment In this study, relays are deployed by the procedure illustrated in Figure 2. Basically, the target area coverage of 90% (including indoors) is tried to be met with the existing Macro Sites (BS) and a minimum amount of additional Relays (RS) deployed at 5m height on street level. The

Apply BS to RS budgets to existing macro sites and observe coverage at 5m height (single hop coverage)

5131'00"N

529,800

530,000

530,200

530,400

530,600

5131'00"N

1852D

007'40"W 007'40"W

5131'00"N

5131'00"N

Add relays and place them in coverage holes at 1.5m height. Make sure they are not placed in coverage holes at 5m height.

Site71 1852D Site5 Site69

007'40"W 007'40"W

5131'00"N

If target coverage is not met and relays can still be placed at locations with coverage at 5m height.

5131'00"N

Site71 1852D Site5 Site69

Simulate combined Relay/ Macro Coverage at 1.5m height

007'40"W

If target coverage is not met and relays can not be placed at locations with coverage at 5m Height.

Optimze placement and simulate combined Relay/Macro coverage at 1.5m height (apply BS to UE and RS to UE budgets)

If target coverage is met

Done

Simulate combined Relay and Macro coverage at 5m height (apply BS to RS and RS to RS budgets) Multi hop case

Add relays and place them in coverage holes at 1.5m height. Make sure they are not placed in coverage holes at 5m height.

Figure 2 Relay Deployment Procedure IV. DEPLOYMENT DENSITIES TO ACHIEVE COVERAGE FOR TARGET DATA RATE WITH DIFFERENT DEPLOYMENT STRATEGIES Figure 3 shows the resulting relay deployment if the above described relay deployment strategy is applied to achieve the 90% uplink coverage target of 3dB. Additionally to the 4.1 Macro Sites per km2, 44 Relays/ km2 have to be deployed i.e. 3.6 relays per sector. This represents the base case result. Because the results are very sensitive to link budgets, the assumed link budgets were varied by plus and minus 5dB to study the impact of link variations. Table II shows the number of relays per sector which are necessary to achieve the uplink coverage target for different access link variations. Depending on the assumptions, between 10 and 88 relays are necessary for the investigated dense urban area which equals 0.8 to 7.2 relay stations per sector. The presented numbers underline the strong influence of link budget assumptions. Simulations show also that variations on the macro access link have a greater influence on the number of deployed relay nodes.

Figure 3: Relay placement for 90% Coverage. The colors indicate current Macro Site only coverage (only red area meets uplink signal strength condition). Macro sites with three-fold sectorization are depicted by three arrows, whereas relays are depicted by one arrow. To have a comparison to conventional macro deployments, we also created Greenfield Macro (hexagonal grid) deployments in the same area for the same UL indoor data rate target coverage. In the case of Macro Sites three possible access link variations (-5/0/+5 dB) result in site densities of 17.8/8.9/4.4 sites/km2, respectively. Hence, in order to achieve the target coverage, we expect to have the options of doubling the Macro Site density or adding 3-4 relays per sector, if we assume our base case link budgets. Using link budget variations is a very powerful approach,
TABLE II NUMBER OF RELAYS PER SECTOR FOR DIFFERENT LINK BUDGET ASSUMPTIONS UE BS -5 dB 0 dB +5 dB -5 dB 7.2 4.4 3.2 0 dB 4.1 3.6 2.3 +5 dB 1.4 1.1 0.8

resource split penalty of relaying is taken into account. In the case of orthogonal resource reuse of resources within one cell, none of the resource elements can be reused by any network element within one cell. In the case of full resource reuse, relays can reuse all resource except the ones used the BS-RS feeder links. In this paper, the resource split is optimized for both cases to uniformly distribute the capacity over the coverage area, i.e. fairness is prioritized opposed to maximization of overall capacity. The relaying overhead is modeled by 5% additional overhead on the feeder link. This value for additional overhead is motivated measurements in commercial WLAN-based mesh networks. It is impossible to have an accurate modeling of interference in a frequency reuse one based OFDMA system such as LTE with Relaying, as the LTE Relaying concept does not yet exist. Thus assumptions on the interference and relaying concepts were made and performance bounds are calculated. Three different cases are considered: Interference is fully neglected noise limited environment (optimistic case) Interference is taken into account by an interference margin (realistic case) Interference is taken into account by full frequency reuse one between all Micro BSs and all Relays, considering the SINR values (pessimistic case) It is up to the LTE relaying concept to find a good trade-off between interference impact and efficient resource utilization.
Existing Macro: BS -> UE -5 dB TP (Mbps) 12.6 Greenfield Macro: BS -> UE -5 dB TP (Mbps) 21.9 Relay Deployments: BS -> UE TP (Mbps) -5 dB RS -> UE 0 dB +5 dB 18.8 20.1 22.5 -5 dB 13.5 14.7 16.9 11.0 10.3 9.9 20.8 22.7 23.9 0 dB 15.2 16.8 18.1 10.9 11.1 10.9 25.3 25.7 26.5 +5 dB 19.3 19.7 20.5 10.3 10.4 10.1 15.6 6.4 25.6 0 dB 19.0 10.2 25.6 +5 dB 19.0 10.3 9.5 8.9 17.9 0 dB 13.6 8.9 24.3 +5 dB 19.1 8.9

since it covers a variety of cases: Dependency on carrier frequencies LTE deployments are considered from 700 MHz to 2.6 GHz in different parts of the world Use of more elaborative antennas at base station, UE or relay, i.e. 4x4 antennas in the downlink or 1x4 antennas in the uplink. The effect could be modeled either in the SNIR-to-data rate mapping, or in link budgets. Different assumptions on target coverage data rate Variation of antenna deployment (i.e. use of remote radio heads, antenna gain, cable loss) V. ACHIEVABLE DOWNLINK DATA RATES AND CAPACITY An LTE FDD system with 10 MHz bandwidth on the upand downlink is assumed. Considering a guard band of 1 MHz, 50 available LTE resource blocks are assumed. To calculate user data rates we take an overhead of 30% into account. Resources used for the BS-RS link are not reused on the RS-UE link or on the direct BS-UE link, i.e. the

UE R S

Figure 4: Downlink Average Peak Throughput in Mbps for different deployments, link budgets and interference assumptions. The left blue bar represents the noise limited case, the middle green bar the interferencemargin-based realistic case, and the right brown bar the interference limited pessimistic case Figure 4 shows the average DL peak throughput for all deployments and all interference assumptions. We calculated the average peak throughput as the average of the peak throughputs (all resources assigned to one user) for all tiles in the selected area.

Downlink Peak Data Rates vs Area Coverage Max Interference 100 90 % area (Peak Data Rate >= abscissa) 80 70 60 50 40 30 20 10 0 0 10 20 30 Peak Data Rate (Mbps) 40 50 60 Interference Margin No Interference

Figure 5: Downlink Peak Data Rate for the Base Case Relay Deployment Figure 5 shows how the peak data rates are distributed over the selected area for the base case relay deployment for different interference assumptions. The graphs for no interference and maximum interference are the upper and lower bounds, respectively. We expect a real LTE system to be somewhere in between these bounds. Table III shows the area downlink capacity for different deployments. The existing macro sites (2G/3G), which can not meet the 90% uplink indoor coverage target (3 dB / 400 kbps), has a capacity of 167 Mbps/sqkm is an interference margin is considered for inter-sector interference. The coverage target can be achieved by doubling the site density (Greenfield Macro Deployment), and the area capacity can be increased then to 507 Mbps/sqkm. If instead relays are added to the existing macro sites, the 90% UL coverage target can also be met and the overall capacity is 180-283 Mbps/sqkm, depending on the resource reuse strategy of the relaying concept. The usage of resources for the BS-RS feeder link is taken into account in this analysis. The
TABLE III AREA D OWNLINK CAPACITY IN MBPS FOR ONE SQUARE KILOMETER, BASE CASE LINK BUDGET (0 DB) No Interference interference margin Existing Macro Sites Only (4.1 220 167 sites/sqkm) Greeefield Macro Deployment to 685 507 achieve target coverage (8.9 sites/sqkm) Existing Macro Sites + Relays (Full Resource Reuse) Existing Macro Sites + Relays (Orthogonal Resource Reuse) 359 234 283 180

deployment areas. This paper compares macro site and relay deployments in an example urban area (central London) using ray-tracing simulations and three-dimensional building data. The uplink data coverage for indoor and outdoor users can be significantly improved by adding 3-4 relays per sector, or alternatively doubling the macro site density. The average peak throughput within the considered area is 17.8 Mbps in a 10 MHz FDD LTE system with relay deployments, but this figure depends strongly on how interference is handled in the relaying concept. This average peak throughput is substantially higher than in an LTE deployment considering only existing macro sites. The area capacity can also be increased by adding relays if a coverage target is taken into account in the deployment strategy. The resource reuse strategy between base stations and relays within one cell has a critical impact on the area capacity. Coverage and capacity can be increased by just deploying more macro sites. But this is very often just not feasible from economic and technical perspectives relays offer an attractive alternative solution as shown in this paper. VII. OUTLOOK Many assumptions had to be made to take into account the relaying concept, especially on the resource allocation and interference. Once a relaying concept is developed for the LTE standard, the established bounds can be narrowed down. This investigation, which is based on well-calibrated ray-tracing simulations and three-dimensional building data is valid for a particular area. Simulations with statistical performance evaluation models could be compared to the approach taken in this paper. This study assumed very simple, cheap relays with omni-directional antennas. Sectorized relays might increase capacity and offer flexibility for resource reuse and interference handling, but trade-offs to cost, size and power consumption increase. REFERENCES
Next Generation Mobile Networks Beyond HSPA & EVDO, NGMN Alliance White Paper 3.0. December 2006, www.ngmn.org [2] Next Generation Mobile Networks Radio Access Performance Evaluation Methodology, NGMN Alliance White Paper, June 2007, www.ngmn.org [3] P. Mogensen et.al.: LTE Capacity Compared to the Shannon Bound, In . Proc. IEEE VTC Spring 2007, pp 1234-1238 [4] E. Dahlman, S. Parkvall, S. Soeld, P. Beming:3G Evolution. HSPA and LTE for Mobile Broadband, Academic Press 2007 [5] www.3gpp.org [6] D. Soldani and S. Dixit: Wireless Relays for Broadband Access. IEEE Communications Magazine Volume 46, No. 3, pp 58-66, March 2008. [7] G. Senarath, W. Tong et.al: Multi-hop Relay System Evaluation Methodology (Channel Model and Performance Metric), IEEE 802.16j-06/013r3, 19.2.2007, www.ieee802.org/16/relay [8] R. Pabst, et.al.: Relay-based Deployment Concept for Wireless and Mobile Broadband Radio, IEEE Communications Magazine, Volume 42, No. 9, pp 80-89, Sept. 2004 [9] S. Redana, K. Doppler et.al.: Final assessment of relaying concepts for all CG scenarios under consideration of related WINNER L1 and L2 protocol functions, IST WINNER II deliverable D3.5.3, September 2007, www.ist-winner.org [10] www.ieee802.org/16/relay [1]

sensitivity to link budget variations is not shown here, but relays improve the area capacity in all cases compared to the existing macro-only deployment despite the feeder link resource use penalty. The reason is that macro sites can use their resources in a much more efficient way when relays are deployed, since more efficient coding, modulation and spatial processing schemes can be utilized by both the BS and the relay. The capacity difference between orthogonal and full resource reuse is very substantial any relaying concept should try to come close to full resource reuse if possible but provide means to deal with interference. VI. CONCLUSION The performance of relaying is debated very intensely in the literature but most studies use statistical models of

You might also like