THROUGHPUT ENHANCEMENT THROUGH

FEMTO-CELL DEPLOYMENT

Zubin Bharucha and Harald HaasInstitute for Digital Communications, School of Engineering and Electronics

The University of Edinburgh, Edinburgh EH9 3JL, UK

{z.bharucha, h.haas}@ed.ac.uk

Ivan Cosovic and Gunther AuerDOCOMO Euro-Labs

Landsberger Strasse 312, 80687 Munich, Germany

auer@docomolab-euro.com

Abstract This paper studies the impact of femto-cell underlay deployment thatshare radio frequency resources with urban macro-cells. Due to the ran-dom and uncoordinated deployment, femto-cells potentially cause de-structive interference to macro-cells and vice versa. On the other hand,femto-cells promise to substantially enhance the spectral efficiency dueto an increased reuse of radio resources. The performance of systemswith femto base station (FBS) deployment is compared to a systemwhere all users, including indoor users, are served by the macro basestation (MBS). In addition, the impact of closed-access and open-accessfemto-cell operation is examined. It is demonstrated that significantthroughput gains can be achieved through such FBS deployment, re-gardless of whether closed-access or open-access is considered. Resultsclearly indicate that the benefits of FBS deployment by far outweightheir impact on the macro-cell capacity.

1. Introduction

There is a growing demand for increased user and system throughputin wireless systems. Naturally, such rapidly increasing demand is servedby higher bandwidth allocation, but since bandwidth is scarce and ex-pensive, a key to substantial throughput enhancement is to improve thereuse of radio frequency resources. One of the most powerful methodsof boosting wireless capacity is by shrinking the cell size. The reason forthis is that smaller cell sizes enable the efficient spatial reuse of spec-

2 Zubin Bharucha, Harald Haas, Ivan Cosovic and Gunther Auer

trum [1]. Furthermore, the shorter transmission distances allow for theuse of higher order modulation schemes. While decreasing the cell sizeboosts system capacity, the cost involved is becoming increasingly pro-hibitive, due to the required installation of new network infrastructure.

Studies indicate that a significant proportion of data traffic originatesindoors [2]. Poor signal reception caused by penetration losses throughwalls severely hampers the operation of indoor data services. Therefore,the concept of 3rd generation (3G) femto-cells has recently attractedconsiderable interest. FBSs are low-cost, low-power, short-range, plug-and-play base stations, which aim to extend and improve macro-cellcoverage in indoor areas. FBSs are directly connected to the backbonenetwork and user equipment (UE) located indoors communicates directlywith FBSs. FBSs therefore offload indoor users from the macro-cell, thuspotentially enhancing the capacity both, indoors as well as outdoors.

In [3, 4], the authors propose the TDD underlay concept. Owing tothe asymmetric nature of traffic, one of the frequency division duplex(FDD) bands (the underloaded one) can be split in time such that theFBS transmits and receives information from its associated femto userequipment (FUE) in a time division duplex (TDD) fashion. In this work,both macro and femto cells operate in the same radio frequency spectrumin frequency division duplex (FDD) mode, compliant with the specifi-cations for beyond 3G mobile communication systems [5]. Like in theoriginal TDD underlay concept [3], the FBS backhauls data through adedicated broadband gateway (DSL/cable/Ethernet/etc.) to the cellularoperator network. There are several obvious advantages from femto-celldeployment, the most important of which is that the operator is able toconcentrate on providing better service to the outdoor macro-cell UEs(MUEs). Another selling point for operators is that high throughputcoverage is extended to the indoor environment.

However, it must be kept in mind that FBSs are deployed withoutnetwork planning. Since FBSs operate on the same bands as MBSs,their deployment introduces additional interference. Furthermore, ques-tions regarding security are raised with the deployment of FBSs, suchas “should the FBS allow any UE in its vicinity to connect to it?” or“should the FBS maintain a list of UEs that are exclusively allowedto use its resources?” and “who should control access to the FBS?”.The discussion is therefore taken in the direction of open-access versusclosed-access systems. Open-access allows any UE to connect to theFBS whereas closed-access only allows a specified set of UEs to connectto the FBS. Closed-access systems are therefore susceptible to higherinterference from indoor UEs that lie geographically in a femto-cell butare not connected to its FBS.

Throughput Enhancement through Femto-Cell Deployment 3

Figure 1. The different access methods and the benchmark system. Interferencescenarios are highlighted with dashed lines.

Recent studies on femto-cells shows the geographic distribution ofthroughput for a deployment of open-access femto-cells in a cellularnetwork [6] and the impact of FBS deployment on macro-cell perfor-mance [7]. In this work, system-level simulations are carried out for threedifferent scenarios. In all systems, indoor and outdoor users exist andare randomly distributed. A system without FBS deployment, where allusers (indoors or outdoors) are exclusively served by macro-cells servesas benchmark system. Systems with FBS deployment are distinguishedbetween open and closed access (see Figure 1). We demonstrate throughsystem level simulations that FBS deployment, no matter what the ac-cess method is, has a clear advantage over the benchmark system interms of user and overall system throughput.

The remainder of this paper is organised as follows. Section 2 de-scribes the system model and simulator setup. Simulation results arepresented in Section 3 and Section 4 highlights the key findings.

2. System Model and Simulation Setup

2.1 User Distribution

The simulation area comprises a two-tier, tessellated hexagonal celldistribution. In order to eliminate edge effects, an additional two tiersare simulated, however statistics are taken only from the first two tiers.Femto-cells have a circular area, each with a fixed radius and they areuniformly distributed over the four-tier structure with a given density.A random number of users and one FBS are uniformly distributed ineach femto-cell. In the macro-cell, the MUEs are uniformly distributedover the entire region. This ensures that there is a certain probability(dependent on femto-cell and MUE density) that MUEs lie within femto-cells. In the case of open-access systems, such MUEs are assimilated

into the femto-cell and are served by the associated FBS. In the caseof closed-access systems, these MUEs are still served by the MBS. For

4 Zubin Bharucha, Harald Haas, Ivan Cosovic and Gunther Auer

the benchmark system, all UEs are MUEs (and are located indoors oroutdoors).

2.2 Path Loss Models

Three path loss models (along with their respective delay profiles)are used – the urban micro (UMi) model for the macro-cell channel,the indoor hotspot (InH) model [8] for the femto-cell and the indoor-to-outdoor model to simulate the channel between entities lying indoorsand outdoors. For each link in the system, the probability of line of sight(LoS) is calculated as specified in the appropriate path loss model andbased on the LoS condition, the associated path loss model is applied tothat link. Furthermore, a wall penetration loss of 20 dB is used to modeloutdoor-to-indoor links (and vice-versa). The delay profiles associatedwith these path loss models used to generate frequency-selective fadingare provided in [8].

2.3 Interference and SINR Calculation

Once the users are distributed, log-normal shadowing maps containingcorrelated shadowing values in space are generated. Using these and thepath loss models, MUEs are associated with the MBS to which theyhave the least path loss. Depending on whether open or closed accessis considered, MUEs that lie within femto-cells are either assimilated ornot.

A resource block (RB) represents one basic time-frequency unit. TheRBs are distributed equally among the MUEs of a macro-cell and thesame resources are reused within each femto-cell. In order to allow fora fair comparison, the transmit power per RB is computed a priori andis kept constant for all systems.

The signal-to-interference-plus-noise ratio (SINR) for user i on RB j

(γji ) is calculated as

γji = P

ji

/ (

Iji + N

)

, (1)

where Pji is the useful received power for user i on RB j, I

ji is the inter-

ference seen by that user on that RB and N is the thermal noise in thesystem. There are 8 interference scenarios: MBS↔MUE, FBS↔FUE,MBS↔FUE, FBS↔MUE, where “↔” represents bi-directional interfer-ence. The first four represent “inter-site” interference scenarios, i.e., theinterference originates from neighboring cells (macro or femto). The lastfour interference scenarios can be caused within the same cell. Since allthe resources are reused within the femto-cell, FUEs are affected by theMBS of the macro-cell within which they lie (and vice-versa) and FBSsare interfered with by the MUEs which lie in the same macro-cell. The

Throughput Enhancement through Femto-Cell Deployment 5

Table 1. Simulation Parameters

Parameter Value

Femto-cells per macro-cell 20

Macro-cell major radius 200 m

Femto-cell radius 10 m

Tot. Number of available RBs 100

RB Bandwidth 100 kHz

Thermal Noise per RB −144 dBm

MBS Tx power per RB 24 dBm

FBS Tx power per RB −50, −30 & 0 dBm

MUE Tx power per RB 18.8 dBm

FUE Tx power per RB −46, −26 & 4 dBm

last two interference scenarios are potentially the most detrimental inclosed-access systems. This is because for closed-access, an MUE maylie within a femto-cell while still being connected to the MBS, therebycausing severe MUE→FBS interference in the uplink (UL). Only thefirst two interference scenarios exist for the benchmark system, since allusers are served by macro-cells.

3. Results

Simulations are run for a full-buffer traffic model, i.e., all users inthe system are active simultaneously. Furthermore, a quasi-static chan-nel model is assumed, where channel variations due to mobile veloci-ties of mobile users are neglected. Perfect synchronisation in time andfrequency is assumed, so that interference between neighboring RBs isavoided. Based on the achieved SINR per RB, the aggregate throughputfor user j is calculated as

T j =∑

i∈R j

WRB log2

(

1 + γji

)

, (2)

where R j is the set of RBs allocated to user j, WRB is the bandwidth ofone RB and γ

ji is the achieved SINR on the ith RB of user j. The sum sys-

tem throughput is generated as the sum of the individual user through-puts. The number of FUEs per femto cell is uniformly distributed withthe minimum being one and the maximum being four. The rest of theparameters used in the simulation are taken from [5, 9–11] and are shownin Table 1.

6 Zubin Bharucha, Harald Haas, Ivan Cosovic and Gunther Auer

−30 −20 −10 0 10 20 10

2

103

104

105

FBS/FUE tx. pow. [dBm]

Ove

rall

Sys

tem

Th

rou

gh

pu

t [M

bp

s]

Closed; ULClosed; DLOpen; ULOpen; DLBenchmark; ULBenchmark; DL

Figure 2. Sum system throughput for all 3 systems.

The overall sum system throughput for both uplink (UL) and down-link (DL) is depicted in Figure 2. For systems with FBS deployment theaggregate femto and macro user throughputs is compared to the bench-mark system where all users are served by the macro-cell. These resultshighlight the achieved gains through FBS deployment. It is seen that theUL system throughput for systems with FBS deployment are approxi-mately two orders of magnitude higher than for the benchmark system.While the corresponding DL gains through FBS deployment are stillimpressive, the increase in DL system throughput is significantly lowercompared to the UL. The reason for this is twofold: firstly, in the DLthe MBS transmit power per RB is higher than in the UL, which resultsin an increased DL macro-cell throughput, which in turn cause moreDL interference to femto-cells. Hence the DL throughput of femto-cellsis lower compared to the UL. Secondly, in the UL, even if there is anMUE very close to a femto-cell, it typically does not transmit on all RBs.Thus, fewer RBs are affected.

In Figure 2 we also observe that the system throughput for closedand open-access systems are nearly identical in DL and UL despite thefact that MUEs lying indoors suffer in terms of throughput. This isbecause in these simulations, the number of MUEs lying indoors is smallin comparison to the total number of UEs in the cell and therefore thisdoes not significantly affect the system throughput of closed-access. Forthe parameters used in this simulation, the probability of an MUE lyingindoors is less than 5%, which is reflected in the results.

Figure 3(a) shows the throughput for only femto users in the ULfor varying FUE transmit powers. It can clearly be seen that as theFUE transmit power is increased, the user throughput is boosted. The

Throughput Enhancement through Femto-Cell Deployment 7

0 50 100 150 200 2500

0.2

0.4

0.6

0.8

1

User Throughput [Mbps]

CD

F

Open UL; P=−30 dBmClosed UL; P=−30 dBmOpen UL; P=−10 dBmClosed UL; P=−10 dBmOpen UL; P=20 dBmClosed UL; P=20 dBm

(a) FUE uplink

0 50 100 150 2000

0.2

0.4

0.6

0.8

1

User Throughput [Mbps]

CD

F

Open DL; P=−30 dBmClosed DL; P=−30 dBmOpen DL; P=−10 dBmClosed DL;P=−10 dBmOpen DL;P=20 dBmClosed DL; P=20 dBm

10 20 30 40

0.88

0.9

0.92

0.94

(b) FUE downlink

Figure 3. Femto user throughput for open and closed-access systems. The CDF ofthe user throughput for FUEs is plotted for UL (a) and for DL (b).

close-up also shows that closed-access outperforms open-access by a slimmargin. This is because more FUEs are served by the FBS for open-access than for closed-access, as a result of the fact that some userslying inside femto-cells for closed-access are served by the MBS. Due tothis, for open-access, each user, on average, is allocated fewer RBS, thusmarginally bringing down the user-throughput for open-access, despitethe increased interference caused by MUEs lying inside femto-cells forclosed-access. The two different slopes in the CDF arise from the factthat there can be FUEs having LoS and non LoS conditions with theassociated FBS.

The throughput for only femto users in the DL for varying FBS trans-mit powers is shown in Figure 3(b). Here, the trends are the same as inthe previous case. However, it must be noted that the user throughputis reduced in comparison to those seen in Figure 3(a). The reason isthat the DL interference is, on average, higher than in the UL, due tothe higher available transmit power at the MBS.

Figure 4(a) shows the cumulative distribution function (CDF) of macro-user throughput in the UL for varying FUE transmit powers. For thebenchmark system, all users are included in the statistics because theyare all served by the MBS. Interestingly, FBS deployment achieves anapproximately 5-fold gain at the 90th percentile compared to the bench-mark system. In case of FBS deployment, the FBS offloads the MBSfrom serving femto-cell users, thus freeing resources that can be givento macro users. Looking closer (inset), it is seen that open-access out-performs closed-access. This is expected because for open-access, thereare no MUEs lying inside femto-cells, avoiding destructive interference.

8 Zubin Bharucha, Harald Haas, Ivan Cosovic and Gunther Auer

0 2 4 6 80

0.2

0.4

0.6

0.8

1

User Throughput [Mbps]

CD

F

Open UL; FUE tx. pow. −30dBmClosed UL; FUE tx. pow. −30dBmMacro−cell ULOpen UL; FUE tx. pow. −10dBmClosed UL; FUE tx. pow. −10dBmOpen UL; FUE tx. pow. 20dBmClosed UL; FUE tx. pow. 20dBm

1.75 1.8 1.85

0.81

0.815

0.82

(a) MUE uplink

0 2 4 6 8 100

0.2

0.4

0.6

0.8

1

User Throughput [Mbps]

CD

F

Open DL; FBS tx. pow. −30 dBmClosed DL; FBS tx. pow. −30 dBmMacro−cell DLOpen DL; FBS tx. pow. −10 dBmClosed DL; FBS tx. pow. −10 dBmOpen DL; FBS tx. pow. 20 dBmClosed DL; FBS tx. pow. 20 dBm

1.4 1.5 1.6

0.775

0.78

0.785

(b) MUE downlink

Figure 4. Macro user throughput for the 3 systems. The CDF of the user through-put for MUEs is plotted in parts (a) for UL and (b) for DL.

On the other hand, the signal reaching the MBS from indoor MUEs forclosed-access is highly attenuated (due to wall penetration losses, etc.),thus significantly bringing down the UL throughput for such users. Fi-nally, it is seen that varying the FUE transmit powers does not sig-nificantly affect the macro-cell performance for systems with femto-celldeployment.

The same trends continue in Figure 4(b) which depicts the CDF ofuser throughput served by the macro-cell in the DL for varying FUEand FBS transmit powers. As in the previous case, for the benchmarksystem the statistics of all users in the system are gathered, while onlymacro users are measured for systems with femto-cell deployment. Here,all systems exhibit slightly higher user throughput owing to the highertransmit power per RB in the DL. Hence, in Figure 4(b) the differencebetween open and closed-access is slightly higher than in Figure 4(a)(see inset).

In order to assess the difference in performance between open andclosed-access systems, those MUEs lying indoors (for closed-access) areidentified and their throughput performance is compared against open-access. Figures 5(a) and 5(b) show the throughputs of such MUEs in ULand DL (notice the difference in scale of the x-axes). As expected for allFBS/FUE transmit powers, open-access significantly outperform closed-access, as for open-access, UEs are assimilated by the FBS. Therefore,for closed-access systems, wall penetration losses and longer transmitdistances drastically reduce throughput. Furthermore, for open-accessthe UL performance is superior to the DL. This is attributed to the factthat interference is lower in UL than in DL. The reverse holds true for

Throughput Enhancement through Femto-Cell Deployment 9

0 20 40 60 80 1000

0.2

0.4

0.6

0.8

1

User Throughput [Mbps]

CD

F

DL; −30 dBmUL; −30 dBmDL; −10 dBmUL; −10 dBmDL; 20 dBmUL; 20 dBm

(a) Indoor MUE throughput, open-access

0 1 2 3 40

0.2

0.4

0.6

0.8

1

User Throughput [Mbps]

CD

F

DL; −30 dBmUL; −30 dBmDL; −10 dBmUL; −10 dBmDL; 20 dBmUL; 20 dBm

0 0.1 0.2 0.3 0.4

0.9

0.95

1

(b) Indoor MUE throughput, closed-access

Figure 5. Throughput for MUEs located indoors in open and closed-access systems.

closed-access systems. Only for low FBS transmit powers of −30 dBm,a median throughput of ≈ 200 kbps is achieved. At the other powers,the interference from the femto-cell is too high to support reasonablethroughputs.

4. Conclusion

Femto-cell deployment poses a viable complement to cellular net-works. Operators need to bear low cost in their deployment, since theyare installed directly by the users themselves. Furthermore, since theyshare both the radio access scheme and the frequency band with MBSs,they are compatible with legacy UEs. Aside from these benefits, a cel-lular network stands to significantly gain in overall system throughputthrough the widespread deployment of FBSs. Not only do FBSs im-prove indoor coverage, bringing broadband-like experience directly tothe handset, but they also offload resources from the MBS which can beutilised to improve coverage to outdoor users.

In terms of overall throughput, the difference between closed-access

and open-access systems is almost negligible, because of the relativelylow probability that a MUE is inside a femto-cell. However, even if aMUE is within a femto-cell, it might be able to receive from its MBSif the FBS transmit power is sufficiently low — which indeed is possi-ble given the short distances in femto-cells. This means power controland intelligent resource scheduling in femto-cells are envisaged to fur-ther reduce the interference that femto-cell entities cause to macro-cellentities.

10 Zubin Bharucha, Harald Haas, Ivan Cosovic and Gunther Auer

Acknowledgement

Initial parts of this work were supported by DFG grant HA 3570/2-1as part of program SPP-1163 (adaptability in heterogeneous communi-cation networks with wireless access – AKOM) while some latter partsof this work have been performed within the framework of the CELTICproject CP5-026 WINNER+.

References

[1] M.-S. Alouini and A. Goldsmith, “Area Spectral Efficiency of Cellular MobileRadio Systems,” IEEE Transactions on Vehicular Technology, vol. 48, no. 4, pp.1047–1066, 1999.

[2] V. Chandrasekhar, J. Andrews, and A. Gatherer, “Femtocell Networks: A Sur-vey,” IEEE Communcations Magazine, vol. 46, no. 9, pp. 59–67, 2008.

[3] H. Haas and G. J. R. Povey, “Capacity Analysis of a TDD Underlay Applicablefor UMTS,” in Proc. of the 10th IEEE International Symposium on Personal,Indoor and Mobile Radio Communications (PIMRC), Osaka, Japan, 12-15 Sep.1999, pp. A6–4.

[4] Z. Bharucha and H. Haas, “Application of the TDD Underlay Concept to HomeNodeB Scenario,” in Proc. of the Vehicular Technology Conference (VTC). Ma-rina Bay, Singapore: IEEE, May 11–14, 2008, pp. 56–60.

[5] 3rd Generation Partnership Project (3GPP), Technical Specification Group Ra-dio Access Network, Physical Channels and Modulation (Release 8), 3GPP TS36.211 V 8.2.0 (2008-03), 3GPP Std., Mar. 2008.

[6] H. Claussen, “Performance of Macro- and Co-Channel Femtocells in a Hierar-chical Cell Structure,” in Proc. of the 18th IEEE International Symposium onPersonal, Indoor and Mobile Radio Communications (PIMRC), Athens, Greece,Sep. 3–7 2007, pp. 1–5.

[7] L. Ho and H. Claussen, “Effects of User-Deployed, Co-Channel Femtocells onthe Call Drop Probability in a Residential Scenario,” in Proc. of the 18th IEEEInternational Symposium on Personal, Indoor and Mobile Radio Communica-tions (PIMRC), Athens, Greece, Sep. 3–7 2007, pp. 1–5.

[8] ITU-R Working Party 5D (WP5D) - IMT Systems, “Report 124, Report ofcorrespondence group for IMT.EVAL,” May 2008, United Arab Emirates.

[9] 3rd Generation Partnership Project (3GPP), Technical Specification GroupRadio Access Network, Base Station (BS) Radio Transmission and Reception(FDD), 3GPP TS 25.104 V 8.3.0 (2008-05), 3GPP Std., May 2008.

[10] 3rd Generation Partnership Project (3GPP), Technical Specification Group Ra-dio Access Network, User Equipment (UE) Radio Transmission and Reception(FDD), 3GPP TS 25.101 V 8.3.0 (2008-05), 3GPP Std., May 2008.

[11] 3rd Generation Partnership Project (3GPP), Technical Specification Group Ra-dio Access Network, 3G Home NodeB Study Item Technical Report, 3GPP TR25.820 V 8.1.0 (2008-05), 3GPP Std., May 2008.