Comparison of Relay and Pico eNB Deployments in
LTE-Advanced
Abdallah Bou Saleh, Simone Redana, Bernhard Raaf
Nokia Siemens Networks
St.-Martin-Stras 76, 81541, Munich, Germany abdallah.@, ,
bernhard.
Jyri Hmlinen
Helsinki Univer中午吃什么最好
sity of Technology P.O. Box 3000, FIN-02015 HUT, Finland jyri.hamalainen@tkk.fi
Abstract—Relaying is one of the most important novel elements
in 3GPP LTE-Advanced study item. It promis to offer significant gain for system capacity or coverage depending on the deployment prioritization. In this paper, we investigate the feasibility of relay node de
ployments in terms of system throughput and cell coverage area extension as compared to pico node deployments and traditional homogeneous single-hop macro cells. Relay backhaul link overhead is taken into consideration as
a limiting factor in a relay deployment; nevertheless, results show that its effect on coverage extension is in most cas negligible. While the effect of relay backhaul overhead is small in coverage limited scenarios, the limitations on throughput due to relaying are evident, in particular, for 500m ISD scenarios. This study also demonstrates that both relay and pico node deployments outperform clearly traditional macro cell deployment in terms of coverage and network capacity.
Keywords - LTE-Advanced; relay node; pico eNB; Iso-performance; system throughput
I.I NTRODUCTION
Multi-hop technologies have gained wide interest among the rearchers in the recent years, and now the technologies are finally finding their way to practical systems through standardization. Relaying is currently being considered as a novel enhancement to current radio technologies in the Third Generation Partnership Project (3GPP) Long Term Evolution (LTE) study item on LTE-Advanced. LTE-Advanced defines the framework for further advancement in LTE to fulfill the 4G req
uirements specified by the International Telecommunication Union – Radiocommunication ctor (ITU-R) [1]. Adhering to the requirements, LTE-Advanced should support significantly incread instantaneous peak data rates of 1Gbps and 500Mbps for the downlink and uplink respectively. Also lower C-plane and U-plane latency and c4月开什么花
onsiderably higher peak and average spectral efficiency are aimed, as well as more homogenous distribution of the ur experience over the coverage area, and hence, a high cell edge throughput.
Nevertheless, for high carrier frequencies, where LTE-Advanced will mostly operate, radio propagation loss are more vere, especially at the cell edge, resulting in significant capacity and coverage problems. Such problems could be tackled by increasing the density of enhanced Node B (eNB) or equivalently by decreasing the cell coverage area. Such a solution is, however, unappealing for network operators, as it implies high extra costs, inferred from the linear proportionality of the infrastructure costs of a homogeneous wireless system to the number of eNBs [2]. A promising solution is to implement relay nodes (RNs) near the cell edge to increa the capacity [3][4] and to better distribute resources in the cell, or alternatively, extend the cell coverage area [4][5][6]. RN deployment is also found to be a cost efficient solution [2][7][8][9]. Relay nodes are relatively small nodes with low power consumption, and hence, they offer deployment flexibility and
eliminate the need for high cost site lection, planning, acquisition and installation. As well, RN admits a wireless backhaul link, which eliminates the high costs of a fixed backhaul link.
In this paper, we asss the feasibility 孤单英文
of Decode-and-Forward (DF) Relay Nodes within the LTE-Advanced framework. Therefore, we consider a simple two-hop scenario, where the Ur Equipment (UE) can be rved by eNB or by RN. The terms direct link, backhaul link and access link are ud to refer to the eNB-UE, eNB-RN and RN-UE links, respectively. Our goal is two-folded. First, we consider the impact of the in-band wireless backhaul by making a comparison between RN and pico eNB deployments. Second, we make a comparison between heterogeneous (RN, pico eNB) deployments and conventional eNB-only deployment. We note that in our study the only difference between RN and pico eNB is that the former admit a wireless backhaul, which consumes part of the radio resources.
Performance evaluation focus on both coverage extension and network capacity. In the former ca, the results are given in terms of exchange ratio between small nodes (RNs, pico eNBs) and macro-cellular eNBs. That is, we examine how many small nodes are needed to replace a conventional eNB, while keeping the same system performance in terms of 10%-ile CDF throughput. This performance measure has been explained in more details in [6]. Throughput CDF plots assumin
g a fixed coverage area are also prented.
The comparisons between RN and pico eNB deployments have not attracted enough attention becau it is evident that the performance of the latter is better than the performance of RN deployment. Yet, our study shows that RN deployment is an appealing technology. It is also worth noting that pico nodes require a costly fixed backhaul link. Thus, RN deployments have the potential to save costs from wireless data tunneling, as well, benefit from more flexibility in deployment. Finally, it is
978-1-4244-2515-0/09/$25.00 2009 IEEE
worth stressing the fact that the study of relaying for LTE-Advanced is still in its initial pha, and hence more rearch on this subject within the framework defined by the LTE system is esntial.
The paper is organized as follows. Section II prents the channel model, system parameters and assumptions we consider in our study. In ction III, performance evaluation and analysis, in terms of coverage extension and network capacity, are carried out. We conclude the paper in ction IV.
II.SYSTEM MODEL
In what follows, the considered channel model is discusd, as well as the parameter choice and assumptions applied in the performance evaluation process.
A.Channel Model
We apply the latest 3GPP distance dependent path-loss model proposal for the three links (direct, backhaul and access) [10]. The access channel model is probability-bad, and compris line-of-sight (LOS) as well as non-LOS components. We note that the access channel model was, until very recently, an open issue in the 3GPP LTE-Advanced discussions. The preceding propod 3GPP model, given in [11], consists of only a non-line-of-sight (non-LOS) component and is bad on the non-LOS ITU-R Urban Micro model [12]. However, if UEs are clo to RNs, the probability that most of them are in LOS conditions increas significantly. Hence, the assumption of considering exclusively a non-LOS link as in [11] might be valid only to very denly populated cities and is, in general, pessimistic. In this study, indoor urs are considered and the channel model is applied where path-loss towards the building is determined before adding the penetration loss. Indeed, in many realistic scenarios, there is LOS or at least a channel that only compris few reflections on the propagation path between RN and the building where UE is. Therefore, the link suffers from smaller path-loss than the channel that considers propagation over rooftops as in [12].
Proper modeling of the access channel is esntial. It may have an undesirable bias on the system design and the complexity of radio resource management (RRM) schemes, and lead to a wrong conclusion about the benefits of deploying small nodes. Equation (1) below prents the considered access channel model, where PL is the path-loss and R is the distance, given in kilometers. According to 3GPP test scenarios our Ca 1 corresponds to an inter-site-distance (ISD) of 500m, whereas, ca 3 is related to ISD of 1732m. In a ca 1 scenario, a LOS is considered up to distances of 30m, whereas, in ca 3, it extends up to distances of 95m. NLOS is considered for distances above 156m and 300m for ca 1 and 3, respectively.
PL= Prob(R) PL LOS(R) + [1-Prob(R)]PL NLOS(R), (1) PL LOS(R) = 103.8+20.9log10(R)
PL NLOS(R) = 145.4+37.5log10(R)
Ca 1: Prob(R) = 0.5-min(0.5,5exp(-0.156/R))
+min(0.5, 5exp(-R/0.03))
Ca 3: Prob(R) = 0.5-min(0.5,3exp(-0.3/R))
+min(0.5, 3exp(-R/0.095)).
B.System tup
The simulated network is reprent面包用英语怎么说
ed by a regular
hexagonal cellular layout with 19 tri-ctored sites, variable
Figure 1. Small-node deployment.
number of small nodes (RNs or pico eNBs) per ctor and variable inter-site distance (ISD). The small nodes are deployed regularly at the ctor border. Considering the access channel
model in (1), the ctor border can be covered with 5 small nodes; 10 nodes form two tiers at the ctor border. Two extra
small nodes can be deployed nearby the edge with neighboring ctors. Fig. 1 prents the node deployments. Simulation parameters follow the current parameter ttings agreed in
3GPP [10], and summarized in Table 1.
TABLE I. S IMULATION P ARAMETERS
Parameter Value Carrier Frequency 2GHz
Bandwidth 10MHz
Highest Modulation Scheme 64-QAM
Extra Margins 30 dB on each link
Penetration Loss 20 dB on access link
Bandwidth Efficiency 0.88
SINR Efficiency 1.25
Thermal Noi PSD -174 dBm/Hz
eNB Parameters
eNB Transmit Power 46 dBm
eNB Elevation 25 m
eNB Elevation Gain 14 dBi
eNB Antenna Configuration Tx-2, Rx-2
eNB Antenna Pattern
A() = -min[12 (/ 3dB)2, A m]
3dB = 70o and A m = 25 dB
UE Parameters
UE Antenna Configuration Tx-1, Rx-2
UE Height 1.5m
UE Elevation + Antenna Gain 0dBi
UE Noi Figure 9 dB
RN / Pico eNB Parameters
Node Transmit Power 30 dBm
Node Elevation 5 m
Node Antenna Configuration Tx-2, Rx-2
Node-eNB Elevation Gain 7 dBi
Node-UE Elevation Gain 5 dBi
Node Antenna Pattern Omni-directional
Node Noi Figure 5 dB
Indoor urs are assumed to be distributed with equal probability over the ctor area and the full buffer traffic model is applied. Interference in the network is neglected, whereas shadowing and fading are implicitly accounted for by extra margins on each link. It is important to note that the relaying overhead in RN deployment is taken into account by using the parallel formula of [6]. There, optimum end-to-end throughput for the two-hop communication is given as follows:
TP opt = TP backhaul // TP access = (1/TP backhaul+ 1/TP access)−1, (2) where TP backhaul is the throughput on the backhaul link and TP accesss is the throughput on the access link. Link throughput calculation is bad on LTE hull curve approximation with LTE related bandwidth and SINR efficiency [13]. Shannon approximation with the implicit margins of Table I is applied. This approach simplifies the study and makes it transparent, since no sophisticated features like adaptive antennas or MIMO are introduced to the backhaul link. Nevertheless, such methods can be designed to take into account special properties of fixed backhaul link. On the other hand, pico eNBs
benefit from a perfect backhaul link and the end-to-end throughput in (2) is given only by the throughput on the access link.
III.P ERFORMANCE EVALUATION AND ANALYSIS In this ction, the performance of RN, pico eNB and macro eNB-only deployments are compared and analyzed in terms of coverage extension and network capacity. In the former ca, throughput level is fixed and the gain of heterogeneous deployments is highlighted with respect to ISD extension. In the latter ca, a fixed ISD is ud, and small nodes deployment gain is reprented in terms of throughput increa.
A.Coverage Extension
The propod evaluation methodology given in [6] is adopted. Different iso-performance scenarios, i.e. combinations of RNs or Pico eNBs and extended ISDs that provide the same performance in terms of coverage, are investigated. The 10%-ile of the throughput CDF, which reprents urs at the cell edge, is ud as a criterion for performance comparison. The gain on the 10%-ile throughput level, due to deployment of small nodes, is translated into coverage extension via extended ISDs.
Figs. 2 and 3 show, respectively, the iso-performance curves for reference ISDs of 500m and 1732m. An iso-performance curve is a t of iso-performance scenarios with different combinations
of RN/pico eNB and macro eNB while keeping the same coverage. It is worth noting that the overlapping curves in the coverage limited scenario, with ISD reference of 1732m, show that RN and Pico eNB deployments perform similarly in terms of coverage extension. In ca of reference ISD of 500m, the performance of RN deployments is clo to the performance of pico eNB deployments. Table II prents the ISD extensions achieved by the heterogeneous deployments in different scenarios, whereas Table III prents the exchange ratios between small nodes (RN or Pico eNB) and macro-cell eNB as well as the coverage area of the small nodes. The ratios are derived from the iso-performance curves, bad on the most cost efficient iso-performance scenario. Note that the gain in an iso-performance scenario is defined by the corresponding ISD extension, which is in turn bad on the coverage area of the small node. The exchange ratios give an indication on the cost savings when deploying RNs or Pico
eNBs in contrary to eNB-only deployment.
Figure 2.
Iso-performance curve - ISD 500m
Figure 3. Iso-performance curve - ISD 1732m
Results show that, for a 500m ISD, RN and Pico eNB deployments are justified, in terms of coverage extension, if their costs are less than 1/30 and 1/26 that of a macro eNB, respectively. The deployment costs are justified, in ca of a reference ISD = 1732m, if RN and pico eNB deployment costs
TABLE II. I NTER-SITE-DISTANCE EXTENSIONS
Number of
Small Nodes
per ctor
ISD Extensions [m]
with respect to ISD
500m Scenario
ISD Extensions [m] with
respect to ISD 1732m
Scenario
RN Pico eNB RN Pico eNB
5 nodes 112 12
6 413 418
10
nodes 190 219 667 675
12
nodes 235 276 846 855 TABLE III. COVERAGE EXTENSION EVALUATION
Inter-site-
distance (ISD)
[m]
Exchange Ratio between RN /
Pico eNB and macro eNB
Radius of RN /
Pico eNB
Coverage Area
[m]
RN Pico
eNB
500m 30 26 50 1732m 28 28 170
are less than 1/28 that of eNB. It should be noted that costs always include total costs of ownership (TCO), not only device costs [9].
Another important result of the coverage extension study is that the required number of pico eNBs is
not much smaller than that of RNs. This is mainly due the fact that RNs are located outdoors and thus they admit very good backhaul link towards the rving eNB. If network dimensioning is done assuming indoor urs, the r转正申请理由
adio resource usage by the backhaul of an outdoor relay is relatively small. RN deployments, as compared to pico eNB deployments, are hence justified in terms of coverage extension if their costs do not exceed 2.6/3 times that of pico eNB deployments in a 500m reference ISD. In ca of 1732m reference ISD, RN deployments are appealppt编辑
ing if their costs are simply less than tho of pico eNB deployments. B. Throughput Analysis
While Pico eNBs did not provide a significant advantage over RNs in terms of cell coverage area extension, the throughput distribution plots of Figs. 4 and 5 for pico eNB and RN deployments with ISD 500m show an advantage of Pico eNBs in improving the system throughput, especially in high throughput regimes. Whereas, urs rved by relay nodes are limited to half the maximum achievable throughput due to resource consumption on the backhaul link, pico eNBs rve urs up to throughputs which are only limited by the maximum spectral efficiency of the system.
According to Fig. 6, pico eNB deployment outperforms clearly RN deployment at both 10%ile and 50%ile throughput CDF levels, when ISD is 500m. The difference is larger at 50%ile level which relates to the average throughput in the cell. Numerical results are prented in Table IV. In contrary
to ISD=500m, the performance difference between pico eNB and RN deployments is small in ISD=1732m ca as en from Fig. 7. Actually, the cell edge performance related to 10%ile throughput CDF level is practically the same 做蛋糕的教程
for the deployments and the negative impact of backhaul resource consumption in RN deployment ca is visible only at 50%ile level. Again, numerical results can be found in Table IV. We note that the significant advantage of pico eNB deployment over RN deployment is, however, not noticeable in coverage extension gains, or equivalently in small node/macro eNB exchange ratios. We recall that according to the evaluation of Section IIIA, as the ISD extends, edge urs in
RN coverage area exhibit a low throughput on their access link,
Figure 4.
Throughput distribution [Kbps] - Pico eNB deployment
Figure 5. Throughput distribution [Kbps] – RN deployment
while the backhaul link is still good. Referring to (2), the end-to-end throughputs of such urs are, thus, dominated by the throughputs on the access link. Fig. 5 shows that such urs contribute to the 10%-ile throughput level, which is considered as a coverage extension criterion. Note that in pico eNB deployments, the throughput is fully defined by the throughput on the access link; hence, RNs perform similar to pico eNBs in terms of 10%-ile throughput or, equivalently, coverage extension.
An important obrvation in Fig. 6, which reflects the effect of resource consumption on the in-band backhaul link, is the step in the throughput CDF at 30Mbps for RN deployments. UEs connected to RNs are limited to half the maximum achieved throughput on the access link irrespective of possibly very high SNR on the link. Hence, the performance of RN deployments at such throughput level converge to that of the eNB-only deployment, as urs with higher throughputs are tho who are rved by the eNB only becau they are not limited by the resource consumption on the backhaul link. Another effect of the in-band backhaul link is that some urs at good throughput levels perfor
m wor in RN deployment networks than in traditional networks. This effect occurs when deploying a high number of relay nodes (2 tiers or more). UEs on the edge of the coverage area of the inner tier RNs already admit relatively high throughput in traditional single-hop networks, and are now rved by RN. With access point lection being bad on downlink signal strength, the throughput gain of the slightly higher signal strength from the rving RN is overtaken by the throughput loss due to the resource consumption on the backhaul link.
TABLE IV.
T HROUGHPUT G AIN O F H ETEROGENEOUS D EPLOYMENTS Throughput Level
Number of Small Nodes Throughput gain [%] in ISD 500m
Scenario
Throughput gain [%] in ISD 1732m
Scenario
RN
Pico eNB
RN
Pico eNB
10%-ile
5 nodes
48 74 163 164 10 nodes 69 137 672 672 12 nodes 73 164 1023 1046
50%-ile
5 nodes
19 58 386 400 10 nodes 25 87 1086 1194 12 nodes
25
97
1285
1490
Figure 6.
Throughput CDF - ISD 500m
Figure 7. Thoughput CDF - ISD 1732m
We further note that in coverage limited scenario of ISD=1732m, the RN coverage area is large, with
radius up to 170m. Hence, urs at the edges experience relatively low throughput on the access link, and along with the eNB coverage area edge urs contribute to the lno的音标
ow throughput regime. Considering a relatively high SNR on the backhaul link in RN deployments, the access channel throughput, as depicted in (2), defines, thus, the end-to-end throughput, which is the ca also in pico eNB deployments. Finally, we state that it is worth noting the significant gains at the 50%-ile as well as the 10%-ile throughput levels in heterogeneous deployments as compared to homogeneous deployments.
From the before-hand obrvations on ISD 1732m scenario, and recalling that RN deployments achieved also similar to pico eNB deployments in the coverage extension evaluation, RN deployment is en as an appealing solution for cell edge capacity and coverage problems in coverage limited scenarios.
IV. C ONCLUSION AND F UTURE W ORK
We have compared the performance and feasibility of relay nodes and Pico eNBs within the LTE-Advanced framework in terms of system throughput and cell coverage area extension. Relaying overhead is considered as a limiting factor in RN deployments. Results show that, in most cas, the
performance difference between RN and Pico eNB deployments is small making the RN deployment attractive, especially in the ca where network dimensioning assumes indoor urs and outdoor relays. In a coverage limited scenario, RN deployments perform almost equally well with pico eNB deployments in terms of tackling both capacity and coverage gaps on the cell edge. However, they lag behind pico eNB deployments in throughput performance in ISD 500m scenario and in hig孔子见南子
h throughput regimes. Our study also highlights the significant gain of heterogeneous deployments over traditional networks. Future work will focus on extending the study to include more sophisticated features such as shadowing models, ur drops, scheduling, and evaluate different resource partitioning schemes ta新郎父亲婚礼致辞
king into consideration the interference, especially in a 500m ISD scenario. Future work on the relaying concept will be considered within the LTE-Advanced framework.
R EFERENCES
[1] TR 36.913 v8.0.1, “Requirements for Further Advancements for
Evolved Terrestrial Radio Access (E-UTRA)”, March 2009.
[2] K. Johansson, A. Furuskaer, P. Karlsson, and J. Zander, “Relation
Between Ba Station Characteristics and Cost Structure in Cellular Systems”, IEEE Personal, Indoor and Mobile Radio Communications (PIMRC) 2004, 5-8 Sept. 2004, Vol. 4, pp. 2627-2631.
[3] A. So, B. Liang, “Effect of Relaying on Capacity Improvement in
Wireless Local Area Networks”, WCNC 2005, 13-17 March 2005, Vol. 3, pp. 1539-1544
[4] R. Schoenen, W. Zirwas, and B. H. Walke, “Capacity and Coverage
Analysis of a 3GPP-LTE Multihop Deployment Scenario”, IEEE International Conference on Communications Workshops (ICC) 2008, 19-23 May 2008, pp. 31-36.
[5] R. Schoenen, R. Halfmann, B. H. Walke, “An FDD Multihop
Cellular Network for 3GPP-LTE”, IEEE Vehicular Technology Conference (VTC) 2008, 11-14 May 2008, pp. 1990-1994
[6] T. Beniero, S. Redana, J. Hmlinen, B. Raaf, “Effect of Relaying
on Coverage in 3GPP LTE-Advanced”, IEEE Vehicular Technology Conference (VTC) Spring 2009, April 2009.
[7] B. Timus, “Cost Analysis Issues in a Wireless Multihop
Architecture with Fixed Relays”, IEEE Vehicular Technology Conference (VTC) Spring, 2005, vol. 5, pp. 3178- 3182. [8] P. Moberg, P. Skillermark, N. Johansson, A. Furuskar,
“Performance and Cost Evaluation of Fixed Relay Nodes in Future Wide Area Cellular Networks”, IEEE Personal, Indoor and Mobile Radio Communications (PIMRC), 2007, pp. 1-5.
[9] E. Lang, S. Redana, B. Raaf, “Business Impact of Relay Deploy-ment for Coverage Extension in 3GPP LTE-Advanced”, LTE Evo-lution Workshop, IEEE International Conference on Communications Workshops (ICC) 2009, 14-18 June 2009.
[10] 3GPP TR 36.814 v1.0.1, “Further Advancements for E-UTRA
Physical Layer Aspects”, March 2009.
[11] 3GPP TR 36.814 V1.0.0, “Further Advancements for E-UTRAN
Physical Layer Aspects”, February 2009.
[12] “Guidelines for evaluation of radio interface technologies for IMT-Advanced”, Document 5D/TEMP/99(Rev.1)
[13] 3GPP 36.942 V8.0.1 (2008-09), Evolved Universal Terrestrial
Radio Access (E-UTRA); Radio Frequency (RF) system scenarios
