OFDMA Femtocells: A Roadmap on Interference Avoidancexanthippi.ceid.upatras.gr/courses/mobile/2009_10/ofdmafemtocells.pdf · so-called femtocell access points (FAPs) or home base

IEEE Communications Magazine • September 2009 410163-6804/09/$25.00 © 2009 IEEE

INTRODUCTION

According to recent surveys [1], 50 percent ofphone calls and 70 percent of data services willtake place indoors in the next years. Therefore,indoor coverage providing high data rates andquality of service (QoS) will soon be needed byoperators. Since macrocell coverage becomesexpensive to serve indoor customers with largeservice demands, new solutions for the indoorcoverage/capacity problem are required.

One solution to enhance indoor coverage areso-called femtocell access points (FAPs) orhome base stations [2]. These are low-powerbase stations designed for indoor usage thatallow cellular network providers to extendindoor coverage where it is limited or unavail-able. On the air interface, FAPs provide radiocoverage of a given cellular standard (e.g.,GSM, UMTS, WiMAX, LTE), while the back-haul connection makes use of a broadband con-nection such as optical fiber or digital subscriberline (DSL). In this article FAP is used to denotethe device itself, while femtocell refers to thecoverage area.

The use of femtocells will benefit both usersand operators. Users will enjoy better signalquality due to the proximity between transmitterand receiver, the result being communicationswith larger reliabilities and throughputs. Fur-

thermore, this will also provide power savings,reducing electromagnetic interference and ener-gy consumption. This way, more users will accessthe same pool of radio resources or use largermodulation and coding schemes, while operatorswill benefit from greater network capacity andspectral efficiency. In addition, since indoor traf-fic will be transmitted over the Internet Protocol(IP) backhaul, femtocells will help the operatorto manage the exponential growth of traffic andincrease the reliability of macrocell networks.Moreover, given that they are paid for and main-tained by the owers, femtocells will also reducethe overall network cost.

It is estimated that by 2012, there could bearound 70 million FAPs installed in homes oroffices around the world, serving more than 150million customers [3]. Consequently, the co-channel deployment of such a large femtocelllayer will impact existing macrocell networks,affecting their capacity and performance [4].Therefore, to mitigate this impact, severalaspects of this new technology such as the accessmethods, frequency band allocation, timing andsynchronization, and self-organization need fur-ther investigation before FAPs become widelydeployed. Since the number and position of theFAPs will be unknown, interference manage-ment cannot be further handled by the operatorusing traditional network planning and optimiza-tion techniques. Therefore, special attentionmust be paid to the mitigation of interferencebetween the macro- and femtocell layers, as wellas between femtocells.

In order to handle these issues, orthogonalfrequency-division multiple access (OFDMA)femtocells are a more promising solution thancode-division multiple access (CDMA) ones,mainly due to its intracell interference avoidanceproperties and robustness to multipath. InOFDMA systems the available spectrum is divid-ed into orthogonal subcarriers, which are thengrouped into subchannels. OFDMA works as amulti-access technique by allocating differentusers to different groups of orthogonal subchan-nels. Moreover, OFDMA femtocells can exploitchannel variations in both frequency and timedomains for the avoidance of interference, whileCDMA can only exploit the time domain.

The aim of this article is to provide anoverview of the technical challenges faced byoperators when deploying an OFDMA femtocell

ABSTRACT

OFDMA femtocells have been pointed out bythe industry as a good solution not only to over-come the indoor coverage problem but also todeal with the growth of traffic within macrocells.However, the deployment of a new femtocelllayer may have an undesired impact on the per-formance of the macrocell layer. The allocationof spectrum resources and the avoidance of elec-tromagnetic interference are some of the moreurgent challenges that operators face beforefemtocells become widely deployed. In this arti-cle a coverage and interference analysis basedon a realistic OFDMA macro/femtocell scenariois provided, as well as some guidelines on howthe spectrum allocation and interference mitiga-tion problems can be approached in these net-works. Special attention is paid to the use ofself-configuration and self-optimization tech-niques for the avoidance of interference.

TOPICS IN RADIO COMMUNICATIONS

David López-Pérez, Alvaro Valcarce, Guillaume de la Roche, and Jie Zhang, University of Bedfordshire

OFDMA Femtocells: A Roadmap onInterference Avoidance

LOPEZ LAYOUT 8/20/09 3:54 PM Page 41

Authorized licensed use limited to: Christos Maurokefalidis. Downloaded on December 28, 2009 at 05:10 from IEEE Xplore. Restrictions apply.

IEEE Communications Magazine • September 200942

network. Since interference is the main limita-tion in the deployment of two-layer networks,special attention will be given to those problemsin relation to spectrum allocation and interfer-ence avoidance.

CHALLENGES FOROFDMA FEMTOCELLS

Femtocells will face several issues that still needto be addressed in order to guarantee interoper-ability with existing macrocells. These are ana-lyzed in the following.

THE ACCESS METHODFemtocells can be configured in three ways to

allow or restrict their usage by certain users:• Open access: All users are allowed to con-

nect.• Closed access: The femtocell allows only

subscribed users to establish connections.• Hybrid access: Nonsubscribers use only a

limited amount of the femtocell resources.It has been shown that open access improves

the overall capacity of the network [5], mainlybecause macrocell users can connect to nearbyfemtocells in locations where the macrocell cov-erage is deficient. From an interference view-point, this avoids femtocells behaving asinterferers since outdoor users can also connectto indoor femtocells. As a drawback, open accesswill increase the number of handoffs and signal-ing. Furthermore, with this type of access, secu-rity issues apply. Moreover, recent customersurveys show that open access is commerciallychallenging for operators. This is because femto-cells are paid for by subscribers, who are notkeen to accept nonsubscribers as users of theirown femtocells unless they obtain some kind ofbenefit/revenue.

Closed access femtocells are thus more likelyto be deployed in the home environment. How-ever, this implies that power leaks through win-dows and doors (Fig. 1) will be sensed asinterference by passing macrocell users, thusdecreasing their signal quality. As will be shown,

algorithms for the allocation of OFDMA sub-channels can help to solve this problem.

Hybrid approaches allow the connectivity ofnonsubscribers while restricting the amount ofOFDMA subchannels that can be shared. In thisway, most of the interference problems of closedaccess are eliminated while controlling theimpact on the femtocell owner.

TIME SYNCHRONIZATIONSince femtocells are deployed by users, there isno centralized management of their radioresources. However, network time synchroniza-tion is necessary between macrocells and femto-cells in order to minimize multi-accessinterference, as well as for the proper perfor-mance of handoffs. Without timing, transmissioninstants would vary between different cells. Thiscould lead to the uplink period of some cellsoverlapping with the downlink of others, thusincreasing intercell interference in the network.

Since FAPs are aimed at the consumer elec-tronics market, they are intended to attain lowprices. The manufacture of low-cost femtocellsequipped with high precision oscillators is nottrivial, so other approaches need to be consid-ered in order to achieve reliable time synchro-nization.

The use of GPS receivers, which provideaccurate timing over satellite links, has been pro-posed as a possible solution. However, their per-formance depends on the availability of GPScoverage inside user premises. Another solutionis the use of the IEEE-1588 Precision TimingProtocol as a feasible method to achieve syn-chronization [6]. However, some modificationsare necessary in order for it to perform efficient-ly over asymmetric backhaul links such as ADSL.

PHYSICAL CELL IDENTITYPhysical cell identity (PCI) is normally used toidentify a cell for radio purposes; for example,camping/handoff procedures are simplified byexplicitly providing the list of PCIs that mobileterminals have to monitor. Note that this list isusually known as the neighboring cell list (seebelow). The PCI of a cell does not need to beunique across the entire network; however, itmust be unique on a local scale to avoid confu-sion with neighboring cells. This represents achallenge in femtocell networks, since they mustselect their PCIs dynamically after booting orchanging their position in order to avoid collisionwith other macro/femtocells. Furthermore, inextensive femtocell deployments and due to thelimited number of PCIs (e.g., 504 in LTE), thereuse of PCIs among femtocells in a given areamay be unavoidable, thus causing PCI confusion.

NEIGHBORING CELL LISTSince femtocells can be switched on/off or movedat any time, their neighbors vary often. Femto-cells must thus be able to set up their neighbor-ing cell list in a dynamic manner. Due to this,the relationships between femtocells must behandled differently than those between macro-and femtocells. In addition, it is expected thatthe number of neighboring femtocells within amacrocell will grow beyond the 32 that are cur-rently considered in macrocells. Therefore, new

�� Figure 1. Coverage of a femtocell emitting an EIRP = 10 dBm at f = 3.5GHz.

Distance (km)

0.02 0

0.1

Dis

tanc

e (k

m)

0.08

0.06

0.04

0.02

0

0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 Power(dBm)

-110

-120

-90

-100

-80

-70

-50

-40

-60



IEEE Communications Magazine • September 2009 43

techniques must be developed to allow macro-and femtocells to support a larger number ofneighboring cells and handle them rapidly.

MOBILITY MANAGEMENTIn macrocell networks, cell handoffs are trig-gered when users enter the coverage area ofother cells. However, given the coverage size ofopen/hybrid access femtocells, this occurs moreoften than in the macrocell case, hence increas-ing network signaling. Different handoff man-agement procedures are thus needed to allownonsubscribers to camp for longer periods onnearby femtocells. Furthermore, a hierarchicalcell structure (HCS) can also be used to distin-guish between macro- and femtocells. In thisway, the signaling across layers can be minimizedas well as the neighboring cell list that users scanwhen performing a handoff.

INTERFERENCE ANALYSISIn Fig. 1 it can be seen that a femtocell not onlyprovides coverage at the customer premises, butalso radiates toward neighboring houses as wellas outdoors, introducing interference. Due tothis and given that femtocells are deployed with-in the coverage area of existing macrocells, theycan cause strong degradation of the macrocells’performance. Furthermore, the deployment ofnew femtocells could also disturb the normalfunctioning of already existing femtocells. There-fore, in order to reduce the appearance of deadzones within the macrocells and successfullydeploy a femtocell network, interference avoid-ance, randomization, or cancellation techniquesmust be applied.

In the following, it is assumed that the clocksof the femtocells are synchronized with that ofthe umbrella macrocell. Therefore, and consider-ing that these networks define two separate lay-ers (the femtocell and macrocell layers),interference can be classified as follows:

• Cross-layer: This refers to situations inwhich the aggressor (e.g., an FAP) and thevictim (e.g., a macrocell user) of interfer-ence belong to different network layers.

• Co-layer: In this case the aggressor (e.g., anFAP) and the victim (e.g., a neighboringfemtocell user) belong to the same networklayer.To overcome the effects of interference,

cancellation techniques have been proposedbut often disregarded due to errors in the can-cellation process [7]. The use of sectorialantennas at the FAP has also been suggestedin [8] as a means of reducing interference bydecreasing the number of interferers. Similarly,a dynamic selection of predefined antenna pat-terns has been used in [9] to reduce the powerleakage outdoors. However, hardware-basedapproaches usually imply an increase in FAPcost. On the other hand, strategies based oninterference avoidance also represent efficientalternatives, (e.g., power and subchannel man-agement).

Power control algorithms and radio resourcemanagement are tools often used in cellularsystems to mitigate interference. If they are notapplied, users located far from a base stationwill be jammed by users in much closer posi-tions. These techniques are also necessary inFAPs for the same reasons plus the addedproblem of cross-layer interference. For exam-ple, in closed access femtocells, users locatedfar from the FAP and being asked to raisetheir power level might produce high levels ofinterference to neighboring femtocells or evento the macrocell. This is illustrated in the fol-lowing example.

From the point of view of the frequency allo-cation in OFDMA systems, Fig. 2 represents adownlink scenario where a loaded macrocelluses all the available subchannels to transmitinformation to its users. However, when a macro-cell user is located close to a femtocell (e.g., user

�� Figure 2. Downlink allocation of OFDMA subchannels in a macro/femtocells network with co-channelassignment.

1

L1

1 2 3 4OFDMA subchannels

5 6 7 8

3

L3

4

5

2

L2 L1

L6

L5

L7

67

L4

L2

L7

L3

L6

L4

L5

Since femtocells can

be switched on/off

or moved at any

time, their neighbors

vary often.

Femtocells must thus

be able to set up

their neighboring cell

list in a dynamic

manner. Due to this,

the relationships

between femtocells

must be handled

differently than those

between macro-

and femtocells.




6), the femtocell may also transmit in some orall of the subchannels required by that user. Thisis illustrated by the spectrum occupation of linksL2 and L6, which use subchannels 3 and 4 simul-taneously, resulting in heavy downlink interfer-ence for macrocell user 6. Therefore, moreefficient management of the OFDMA subchan-nels is necessary. Such an approach would helpto enhance frequency reuse in the femtocellslayer and maximize the overall cell throughput.For instance, the spectrum occupation of users 1and 5 illustrate how interference can be avoidedin OFDMA femtocells while providing sufficientfrequency resources for a satisfactory user expe-rience.

SPECTRUM ALLOCATION INTWO-LAYER NETWORKS

The different approaches that can be adoptedto manage the OFDMA subchannels areschematized in Fig. 3. An approach that com-pletely eliminates cross-layer interference is todivide the licensed spectrum into two parts(orthogonal channel assignment). This way, afraction of the subchannels would be used bythe macrocell layer while another fractionwould be used by the femtocells. This approachis supported by companies such as Comcast,which have acquired spectrum to be used exclu-sively by their WiMAX femtocells. Althoughoptimal from a cross-layer interference stand-point, this approach is inefficient in terms ofspectrum reuse. Therefore, co-channel assign-ment of the macrocell and femtocell layersseems more efficient and profitable for opera-tors, although far more intricate from the tech-nical point of view.

Ideally to mitigate cross- and co-layer inter-ference, there would be a central entity in chargeof intelligently telling each cell which subchan-nels to use. This entity would need to collectinformation from the femtocells and their users,

and use it to find an optimal or a good solutionwithin a short period of time. However, since thenumber and position of the femtocells are ini-tially unknown due to the individualistic natureof the FAPs, this approach poses some hardproblems. The presence of hundreds of femto-cells makes the optimization problem too com-plex, and latency issues arise when trying tofacilitate the femtocells’ communication with thecentral subchannels broker throughout the back-haul.

A distributed approach to mitigate cross- andco-layer interference, where each cell managesits own subchannels, is thus more suitable in thiscase (i.e., self-organization). In a non-coopera-tive solution, each femtocell would plan its sub-channels so as to maximize the throughput andQoS for its users. Furthermore, this would bedone independent of the effects its allocationmight cause to neighboring femtocells, even if itsupposes larger interference. The access to thesubchannels then becomes opportunistic, and itis possible that the method decays to greedy. Onthe other hand, in a cooperative approach, eachFAP gathers information about its neighboringfemtocells and may perform its allocation takinginto account the effect it would cause to itsneighbors. In this way, the average femtocells’throughput and QoS, as well as their global per-formance can be locally optimized.

SELF-CONFIGURATION ANDSELF-OPTIMIZATION

As discussed above and due to the uncertainty inthe femtocells’ number and positions, FAPsmust be self-configurable and self-optimizingunits [10], capable of integrating themselves intothe existing radio access network causing theleast interference to existing systems. In this casethe FAPs must be capable of sensing the airinterface and tuning their own parametersaccording to changes in the network or channel.

�� Figure 3. Classification and examples of subchannel allocation techniques for OFDMA femtocells.

Orthogonal channelassignment Co-channel assignment

Example Problem

Solution

Static Dynamic

Dependent on the geographic area

Spectrum allocation

Centralized

Cooperative Non-cooperative

Distributed

Subchannels sharing

Subchannels allocation in OFDMA Femto/Macrocells networks

Dependent on traffic demand

Random access

Fragmented spectrum access

Distributed frequency planning

Automatic frequency planning

Dependent on user mobility

The presence of

hundreds of

femtocells makes the

optimization

problem too

complex and latency

issues arise when

trying to facilitate

the femtocells

communication with

the central

sub-channels broker

throughout the

back-haul.




SENSING PHASE

FAPs must be aware of the presence of neigh-boring cells (i.e., macros or femtos) as well astheir respective spectrum allocations. Differentstrategies can be used to achieve this cognitiveradio stage, in which the femtocell is able tolearn about the state of the network (e.g., archi-tecture and load) and channel conditions (e.g.,interference and fading). Note that the followingapproaches can be used not only to obtain infor-mation about the resource allocation of otherfemtocells, but also of nearby macrocells.

A first approach consists of building the sens-ing capability into the FAP itself. As representedin Fig. 4a (left), the FAPs will sense the airinterface and identify which cells and subchan-nels are active within their range. Then the fem-tocells use this information to configure itselfand perform its resource allocation. This can bedone using a network listening mode, similar toa user terminal operating in idle mode.

A second approach is based on the exchangeof information between femtocells. As represent-ed in Fig. 4a (center), the FAPs could directlyexchange information about their subchannelusage, spectrum needs, and so on. This way,femtocells can be aware of the present actions

and future intentions of their neighboring cells,and act accordingly. These messages can beexchanged through the femtocell gateway, usinga direct link between cells (similar to the X2interface in LTE), but also broadcasting mes-sages or using mobile terminals to relay informa-tion.

However, none of these would work in situa-tions where the FAPs are not within the cover-age area of other femtocells. For instance, inFig. 4b, user 3 is located at the cell edge of twooverlapping femtocells whose FAPs are not visi-ble to each other. In this situation, such a usermight suffer from interference because the fem-tocells are not able to coordinate their resourceallocation as a result of the lack of informationobtained during the sensing phase.

A third approach that can avoid this problemis the use of measurement reports, which are peri-odically performed by users and then sent backto their FAPs. They contain information such asthe received signal strength and active subchan-nels of the serving and strongest neighboringcells (both macros and femtos). They can thusbe post-processed and used to establish interfer-ence relationships between neighboring cells aswell as to identify forbidden subchannels. Withthis technique, users can forward information

�� Figure 4. Different approaches for the sensing of OFDMA sub-channels: a) self-sensing and relay-sensing;b) sensing problem in non-overlapping femtocells; and c) measurement reports.

1 2 3 4 5 OFDMA

subchannels

6 7 8

(a)

(b)

(c)

L2

L1

1

2

L2 L1

L1

1

2

L2 L3

L1

1

3

2

L1

L2

1

2

L2

1 2 3 4 5 OFDMA

subchannels

6 7 8

L2

L1

L1

L2

L3

1 2 3 4 5 OFDMA subchannels

6 7 8

The combination of

these sensing

strategies together

with other statistics

such as the number

of mobility events or

packet drop ratio

can be used to

enhance the

reliability of the

sensing phase and

therefore the

quality of the

femtocell tuning.




about their immediate radio environment (userlocation) to their FAPs and aid them to mitigateinterference (Fig. 4c).

The combination of these sensing strategiestogether with other statistics such as the numberof mobility events or packet drop ratio can beused to enhance the reliability of the sensingphase and therefore the quality of the femtocelltuning.

TUNING PHASEFAPs should be able to select and modify theirparameters in different situations. In this articlea distinction is made between self-configurationand self-optimization. Self-configuration pro-vides the initial settings of the FAP when it isturned on, while self-optimization updates theconfiguration of the FAP in order to adapt itsparameters to the environment.

Self-Configuration — After powering on, thefemtocell is registered into the network of theoperator, and its radio parameters are set to adefault configuration. This is done over thebackhaul, which provides fundamental informa-tion such as the carrier frequency and other net-work parameters (e.g., location, routing andservice area, initial PCI, and neighboring list).Afterward, the sensing of the environment isdone throughout the network listening mode. Bydecoding the control channels, the femtocell isable to synchronize with the external network.Moreover, the femtocell also uses this informa-tion to reselect its PCI, and optimize its neigh-bor list and handoff parameters. Finally, theFAP adapts its power and selects its subchannelsaccording to the obtained sensing information inorder to ensure that it provides the dominantsignal in the desired coverage area.

Self-Optimization — It is well known that theradio channel conditions can change rapidly dueto shadowing and multipath fading. In addition,the network load and user traffic vary quicklydepending on the location and time, and becauseof the packet nature of the services. Due to thisfluctuating behavior, OFDMA femtocells must

dynamically adapt their radio resources to theirenvironments in order to optimize the perfor-mance of the network. Such changes can bedetected during the sensing phase (e.g., increasein traffic or interference, decrease in mobilityevents or dropped calls). Taking into accountthis information, the FAPs tune their parameters(power or subchannel assignment) online inorder to mitigate interference across cells.

EXPERIMENTAL EVALUATIONIn this section system-level simulations are used toaid network designers and operators to under-stand the advantages and drawbacks of the orthog-onal and co-channel assignment. Furthermore, thedifferent features of centralized vs. distributed,and cooperative vs. non-cooperative approaches tofrequency planning are also illustrated.

The simulations presented here were carriedout based on the simulation framework present-ed in [5], and they are based on a realisticWiMAX (IEEE 802.16e) network comprisingtwo macrocells and a large number of femtocells(Fig. 5). It operates in the 3.5 GHz band with a5 MHz bandwidth. Moreover, the number offemtocells has been selected according to recentforecasts that indicate a femtocell penetration of10 percent.

Adaptive modulation and coding (AMC) hasbeen selected as a permutation scheme, wherethe subchannels comprise contiguous subcarri-ers. In this case the number of OFDMA sub-channels in the DL subframe is set to 8, whilethe number of OFDM data symbols is 20. More-over, closed access has been selected as theaccess method to the femtocells and a largenumber of end-users has been spread over thescenario. The number of outdoor users has beenselected according to a density of 40 users/km2,while a random number of indoor users between2 and 4 has been placed per femtocell/house.Outdoor users have a minimum throughputrequirement of 64 kb/s (video), whereas theindoor users’ minimum throughput requirementis 350 kb/s (data).

Interference will occur only if several usersare allocated to the same subchannel and OFDMsymbol. Furthermore, an indoor-to-outdoorpropagation model calibrated with femtocellmeasurements [11] has also been built into thesimulation tool. The key performance indicatorsused to analyze the behavior of the network inthis experimental evaluation is the networkcapacity in terms of user outage and throughput.

The details of the subchannel allocationstrategies evaluated in this section are:• Orthogonal assignment: The spectrum is

divided into two independent fragments:one used by the macrocells and another bythe femtocells. For the first setup, SM = 5subchannels have been assigned to themacrocell layer and SF = 3 to the femto-cells. For the second setup, SM = 6 and SF= 2.

• Co-channel assignment FRS1: In thisscheme the spectrum is not fragmented, soall the macrocells and femtocells haveaccess to all available subchannels. There-fore, interference coordination is neglected.

�� Figure 5. Coverage plot of two WiMAX macrocells with an EIRP of 30 dBm,in an scenario with a femtocells penetration of 10.25 percent of the householdsand femtocell EIRP of 10 dBm. The work frequency is that of WiMAX inEurope, f = 3.5 GHz.

Distance (km)

0.1 0

0.45

0.5

Dis

tanc

e (k

m)

0.4

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

-90

-80

-70

-60

-50

-40

-30

-20

Power(dBm)




• Co-channel assignment FRSx: The spectrumis divided into x fragments. Macrocells canuse all the spectrum, but each femtocellonly uses one randomly selected fragment.This way, the collision probability betweenneighboring femtocells is reduced by a fac-tor x. In this evaluation x = 3 and x = 4have been tested.

• Co-channel assignment and distributed plan-ning: In the distributed-dynamic frequencyplanning (D-DFP) approach, measurementreports are used to sense the environment.Using this information, each femtocell inde-pendently configures its own subchannels pri-ority list, which is sorted by priority(subchannels suffering the least interferencego first) and is periodically updated. Toachieve a stable solution and avoid differentfemtocells making changes on their list at thesame time, a random time counter is used towake up the femtocell and take action.

• Co-channel assignment and centralizedplanning: In the centralized DFP (C-DFP)approach, measurements reports are alsoused to sense the environment. Then theinterference information is sent from thefemtocells to a centralized subchannel bro-ker, which plans the frequency usage andpasses it on to the femtocells.Orthogonal assignments remove cross-layer

interference and perform well when the load ofthe network is low or the subchannel demand isless than the availability. In Table 1 it can beseen that when switching from a (5, 3) to a (6, 2)scheme, the macrocell layer throughput increas-es and the femtocell layer throughput decreases.This behavior is because the resources of themacrocell layer have increased, while theresources of the femtocells layer have decreased.As a result, some users cannot access their cellsor carry their services due to the lack ofresources. Therefore, when using an orthogonalassignment approach, the size of the fragmentsmust be carefully planned according to the traf-fic forecast for each layer.

On the other hand, when using a co-channeldeployment, the capacity problems of the macro-cell and femtocell layers disappear due to thelarger spectrum availability. However, cross-layerinterference comes up, and the macrocell userssuffer from interference due to the presence ofnearby femtocells.

It can be seen that when interference coordi-nation is neglected (FRS1), the macrocell net-work performance is strongly degraded. Sincethere is a heavy traffic load and no interferencecoordination, the probability of subchannel colli-sion between macrocells and femtocells is large.The result is thus poor signal quality (reducedthroughput) and an increase in the outage prob-ability for some users.

When using FRS3 the throughput in themacrocell layer increases, since the probability ofcollision has been reduced by a factor of 3. Onthe other hand, when using FRS4, problems dueto the lack of resources appear in the femtocellsbecause the fragments are too small. Therefore,the trade-off between capacity and probability ofsubchannel collision should be taken intoaccount.

Finally, DFP results in the best network per-formance. This is logical since a dynamic fre-quency planning strategy has been used, able toadapt the system resources to the changing con-ditions of the traffic and channel. Note that thecentralized approach performs better than thedistributed one, since the optimization processgathers information not only from neighboringcells, but using a more global viewpoint of thenetwork.

CONCLUSIONThe multi-subcarrier nature of OFDMA is morethan appropriate to cope with the interferenceproblems that arise between femtocells andmacrocells. While femtocells configure them-selves to reduce interference with neighboringcells, this technology can help to minimize theimpact of the deployment of femtocells in amacrocell network.

It has been shown that co-channel assign-ment of the spectrum can lead to higher cellthroughputs thanks to the greater availability ofsubchannels. However, this will be in exchangefor a higher computational load in the networknodes (femtocells and macrocells). WithOFDMA, and since the number and positionsof the femtocells are unknown, traditional net-work planning needs to be replaced with effi-cient spectrum allocation algorithms for thepurpose of interference avoidance, since novelalgorithms are still needed to perform such allo-cation in realistic scenarios.

■ Table 1. Performance of various resource allocation algorithms.

Orthogonal channel assignment Co-channel assignment

(SM, SF) = (5, 3) (SM, SF) = (6, 2) FRS1 FRS3 FRS4 D-DFP C-DFP

Successful users (%) 93.14 88.44 94.10 95.43 87.49 96.32 97.92

Users in outage (%) 4.89 4.45 5.40 3.94 11.24 2.73 2.35

Macrocell throughput (Mb/s) 4.07 4.57 3.93 4.32 4.06 4.54 4.81

Femtocells throughput (Mb/s) 122.02 112.50 123.47 123.32 113.67 123.60 123.64

Total throughput (Mb/s) 126.09 117.07 127.4 127.64 117.73 128.14 128.45



ACKNOWLEDGMENT

This work is supported by the EU FP6 RAN-PLAN-HEC project on 3G/4G Radio AccessNetwork Design under grant number MEST-CT-2005-020958.

REFERENCES[1] G. Mansfield, “Femtocells in the US Market -Business

Drivers and Consumer Propositions,” FemtoCellsEurope, ATT, London, U.K., June 2008.

[2] V. Chandrasekhar and J. G. Andrews, “Femtocell Net-works: A Survey,” IEEE Commun. Mag., vol. 46, no. 9,Sept. 2008, pp. 59–67.

[3] S. Carlaw, “IPR and the Potential Effect on FemtocellMarkets,” FemtoCells Europe, ABIresearch, London, UK,June 2008.

[4] H. Claussen, “Performance of Macro- and Co-ChannelFemtocells in a Hierarchical Cell Structure,” IEEE 18thPIMRC 2007, Athens, Greece, Sept. 2007, pp. 1–5.

[5] D. López-Pérez et al., “Access Methods to WiMAX Fem-tocells: A Downlink System-Level Case Study,” 11th IEEEICCS, Guangzhou, China, Nov. 2008, pp. 1657–62.

[6] S. Lee, “An Enhanced IEEE 1588 Time SynchronizationAlgorithm for Asymmetric Communication Link usingBlock Burst Transmission,” IEEE Commun. Lett., vol. 12,no. 9, Sept. 2008, pp. 687–89.

[7] S. P. Weber et al., “Transmission Capacity of WirelessAd Hoc Networks With Successive Interference Cancel-lation,” IEEE Trans. Info. Theory, vol. 53, no. 8, Aug.2007, pp. 2799–2814.

[8] V. Chandrasekhar and J. G. Andrews, “Uplink Capacityand Interference Avoidance for Two-Tier Femtocell Net-works,” IEEE Trans. Wireless Commun., Feb. 2008.

[9] H. Claussen, “Femtocell Coverage Optimization usingSwitched Multi element Antennas,” IEEE ICC, Dresden,Germany, June 2009.

[10] H. Claussen, L. Ho, and L. Samuel, “Self-Optimizationof Coverage for Femtocell Deployments,” WTS, Apr.2008, pp. 278–85.

[11] A. Valcarce et al., “Applying FDTD to the CoveragePrediction of WiMAX Femtocells,” EURASIP J. WirelessCommun. Net., vol. 2009, Feb. 2009, article ID: 308606.

BIOGRAPHIESDAVID LOPEZ-PÉREZ ([email protected]) received hisBachelor and Master degrees in telecommunication from

Miguel Hernandez University, Elche, Spain, in 2003 and2006, respectively. He joined Vodafone Spain in 2005,working in the Radio Frequency Department in the area ofnetwork planning and optimization. He took up a researchscholarship at the Cork Institute of Technology, Ireland, in2006, where he researched indoor positioning systems andsensor networks. Nowadays, he is a Ph.D. Marie-Curie Fel-low at the Centre for Wireless Network Design (CWiND) atthe University of Bedfordshire, United Kingdom. He isinvolved in different FP6 and FP7 European projects, aswell as EPSRC. His research interests are in two-tier net-works, OFDMA, self-organization, and interference avoid-ance.

ALVARO VALCARCE obtained his M.Eng. in telecommunica-tions engineering from the University of Vigo, Spain, in2006. Then he was with the WiSAAR consortium in Saar-brücken, Germany, performing WiMAX trials and mea-surements data analysis. He joined CWiND at theUniversity of Bedfordshire in 2007 with the support of aMarie Curie Fellowship. During 2008 he worked on “thefeasibility study of WiMAX-based femtocells for indoorcoverage.” His Ph.D. work is on the FP6 RANPLAN-HECproject, which studies the indoor-to-outdoor wirelesschannel and its applicability to network planning andoptimization.

GUILLAUME DE LA ROCHE has been working as a research fel-low at CWiND since 2007. From 2001 to 2002, he was aresearch engineer by Infineon in Munich (Germany). From2003 to 2004 he worked in a small French company wherehe deployed WiFi networks. From 2004 to 2007 he waswith the CITI Laboratory at the National Institute ofApplied Sciences (INSA), France. He holds a Dipl-Ing fromCPE Lyon, France, and an M.Sc. (2003) and a Ph.D. (2007)in wireless communications from INSA Lyon.

JIE ZHANG is a professor in the Department of ComputerScience and Technology, University of Bedfordshire. Hereceived his Ph.D. degree from East China University of Sci-ence and Technology in 1995. From 1997 to 2001 he wasa research fellow with University College London, ImperialCollege London, and Oxford University. He is the foundingdirector of CWiND, one of the largest leading researchgroups in 3G/4G radio access network planning and opti-mization in Europe. Since 2003, as principal investigator,he has been awarded several projects worth over US$5 mil-lion (his total project funding is much larger) by the EPSRC,the European Commission (FP6/FP7), and the NuffieldFoundation. He is also a lead author of the book Femto-cells: Technologies and Deployment.

48 IEEE Communications Magazine • September 2009



Documents

OFDMA Femtocells: A Roadmap on Interference Avoidancexanthippi.ceid.upatras.gr/courses/mobile/2009_10/ofdmafemtocells.pdf · so-called femtocell access points (FAPs) or home base