femtocell networks a survey - chandrasekhar - ieee comm sept 2008

IEEE Communications Magazine • September 2008 590163-6804/08/$25.00 © 2008 IEEE

INTRODUCTION

The demand for higher data rates in wireless net-works is unrelenting, and has triggered the designand development of new data-minded cellularstandards such as WiMAX (802.16e), the ThirdGeneration Partnership Project’s (3GPP’s) HighSpeed Packet Access (HSPA) and LTE stan-dards, and 3GPP2’s EVDO and UMB standards.In parallel, Wi-Fi mesh networks are also beingdeveloped to provide nomadic high-rate data ser-vices in a more distributed fashion [1]. Althoughthe Wi-Fi networks will not be able to supportthe same level of mobility and coverage as thecellular standards, to be competitive for homeand office use, cellular data systems will need toprovide service roughly comparable to thatoffered by Wi-Fi networks.

The growth in wireless capacity is exemplifiedby this observation from Martin Cooper ofArraycomm: “The wireless capacity has doubledevery 30 months over the last 104 years.” Thistranslates into an approximately millionfoldcapacity increase since 1957. Breaking downthese gains shows a 25× improvement from widerspectrum, a 5× improvement by dividing thespectrum into smaller slices, a 5× improvementby designing better modulation schemes, and awhopping 1600× gain through reduced cell sizesand transmit distance. The enormous gainsreaped from smaller cell sizes arise from effi-cient spatial reuse of spectrum or, alternatively,higher area spectral efficiency [2].

The main problem in this continued micro-ization of cellular networks is that the networkinfrastructure for doing so is expensive. A recentdevelopment is femtocells, also called home basestations (BSs), which are short-range low-costlow-power BSs installed by the consumer forbetter indoor voice and data reception. Theuser-installed device communicates with the cel-lular network over a broadband connection suchas digital subscriber line (DSL), cable modem,or a separate radio frequency (RF) backhaulchannel. While conventional approaches requiredual-mode handsets to deliver both in-home andmobile services, an in-home femtocell deploy-ment promises fixed mobile convergence withexisting handsets. Compared to other techniquesfor increasing system capacity, such as distribut-ed antenna systems [3] and microcells [4], thekey advantage of femtocells is that there is verylittle upfront cost to the service provider. Table1 provides a detailed comparison of the keytraits of these three approaches.

Studies on wireless usage show that morethan 50 percent of all voice calls [Airvana, 5]and more than 70 percent of data traffic origi-nate indoors [PicoChip, 5]. Voice networks areengineered to tolerate low signal quality, sincethe required data rate for voice signals is verylow, on the order of 10 kb/s or less. Data net-works, on the other hand, require much highersignal quality in order to provide the multi-megabit per second data rates users have cometo expect. For indoor devices, particularly at thehigher carrier frequencies likely to be deployedin many wireless broadband systems, attenuationlosses will make high signal quality and hencehigh data rates very difficult to achieve. Thisraises the obvious question: why not encouragethe end user to install a short-range low-powerlink in these locations? This is the essence of thewin-win of the femtocell approach. The sub-scriber is happy with the higher data rates andreliability; the operator reduces the amount oftraffic on their expensive macrocell network, andcan focus its resources on truly mobile users. Tosummarize, the key arguments in favor of femto-cells are the following.

Better coverage and capacity. Due to theirshort transmit-receive distance, femtocells cangreatly lower transmit power, prolong handsetbattery life, and achieve a higher signal-to-inter-ference-plus-noise ratio (SINR). These translateinto improved reception — so-called five-barcoverage — and higher capacity. Because of the

ABSTRACT

The surest way to increase the system capaci-ty of a wireless link is by getting the transmitterand receiver closer to each other, which createsthe dual benefits of higher-quality links andmore spatial reuse. In a network with nomadicusers, this inevitably involves deploying moreinfrastructure, typically in the form of microcells,hot spots, distributed antennas, or relays. A lessexpensive alternative is the recent concept offemtocells — also called home base stations —which are data access points installed by homeusers to get better indoor voice and data cover-age. In this article we overview the technical andbusiness arguments for femtocells and describethe state of the art on each front. We alsodescribe the technical challenges facing femto-cell networks and give some preliminary ideasfor how to overcome them.

TOPICS IN RADIO COMMUNICATIONS

Vikram Chandrasekhar and Jeffrey G. Andrews, The University of Texas at Austin

Alan Gatherer, Texas Instruments

Femtocell Networks: A Survey

CHANDRASEKHAR LAYOUT 8/20/08 5:09 PM Page 59

IEEE Communications Magazine • September 200860

reduced interference, more users can be packedinto a given area in the same region of spectrum,thus increasing the area spectral efficiency [2] or,equivalently, the total number of active users perHertz per unit area.

Improved macrocell reliability. If the trafficoriginating indoors can be absorbed into thefemtocell networks over the IP backbone, themacrocell BS can redirect its resources towardproviding better reception for mobile users.

n Table 1. Femtocells, distributed antennas, and microcells.

Infrastructure Expenses Features

Femtocell: Consumer installed wireless data accesspoint inside homes, which backhauls data througha broadband gateway (DSL/cable/Ethernet/WiMAX)over the Internet to the cellular operator network.

Capital expenditure. Subsidized fem-tocell hardware.

Operating expenditure. a) Providing ascalable architecture to transport dataover IP; b) upgrading femtocells tonewer standards.

Benefits. a) Lower cost, better cover-age and prolonged handset batterylife from shrinking cell-size; b) capaci-ty gain from higher SINR and dedicat-ed BS to home subscribers ; c)reduced subscriber churn

Shortcomings. a) Interference fromnearby macrocell and femtocell trans-missions limits capacity; b) increasedstrain on backhaul from data trafficmay affect throughput.

Distributed antennas: Operator installed spatiallyseparated antenna elements (AEs) connected to amacro BS via a dedicated fiber/microwave backhaullink.

Backhaul

Capital expenditure. AE and backhaulinstallation.

Operating expenditure. AE mainte-nance and backhaul connection.

Benefits. a) Better coverage sinceuser talks to nearby AE; b) capacitygain by exploiting both macro- andmicro-diversity (using multiple AEsper macrocell user).

Shortcomings. a) Does not solve theindoor coverage problem; b) RF inter-ference in the same bandwidth fromnearby AEs will diminish capacity; c)backhaul deployment costs may beconsiderable.

Microcells: Operator installed cell towers, whichimprove coverage in urban areas with poorreception.

Capital expenditure. Installing newcell towers.

Operating expenditure. Electricity,site lease, and backhaul.

Benefits. a) System capacity gainfrom smaller cell size; b) completeoperator control.

Shortcomings. a) Installation andmaintenance of cell towers is pro-hibitively expensive; b) does not com-pletely solve indoor coverageproblem.


IEEE Communications Magazine • September 2008 61

Cost benefits. Femtocell deployments willreduce the operating and capital expenditurecosts for operators. A typical urban macrocellcosts upwards of $1000/month in site lease, andadditional costs for electricity and backhaul. Themacrocell network will be stressed by the operat-ing expenses, especially when subscriber growthdoes not match the increased demand for datatraffic [Airvana, 5]. The deployment of femto-cells will reduce the need for adding macro-BStowers. A recent study [6] shows that the operat-ing expenses scale from $60,000/year/macrocellto just $200/year/femtocell.

Reduced subscriber turnover. Poor in-build-ing coverage causes customer dissatisfaction,encouraging them to either switch operators ormaintain a separate wired line whenever indoors.The enhanced home coverage provided by fem-tocells will reduce motivation for home users toswitch carriers.

The goal of this article is to provide anoverview for these end-user-deployed infra-structure enhancements and describe in moredetail how the above improvements come about.We also describe the business and technicalchallenges that femtocells present, and providesome ideas about how to address them.

TECHNICAL ASPECTS OFFEMTOCELLS

The capacity potential of femtocells can be veri-fied rapidly from Shannon’s law, which relatesthe wireless link capacity (in bits per second) ina bandwidth to the SINR. The SINR is a func-tion of the transmission powers of the desiredand interfering transmitters, path losses, andshadowing during terrestrial propagation. Pathlosses cause the transmitted signal to decay asAd–α, where A is a fixed loss, d is the distancebetween the transmitter and receiver, and α isthe path loss exponent. The key to increasingcapacity is to enhance reception between intend-ed transmitter-receiver pairs by minimizing dand α. Simultaneously, additional benefits in

network-wide spatial reuse can be obtained by,but not restricted to, exploiting diversity, andemploying interference cancellation, interferencesuppression, and interference avoidance tech-niques.

Femtocells enable reduced transmit powerwhile maintaining good indoor coverage. Pene-tration losses insulate the femtocell from sur-rounding femtocell transmissions. Assuming afixed receive power target with a path loss prop-agation model (no fading), and denoting α (β)as the outdoor (indoor) path loss exponent,overlaying an area L2 with N femtocells resultsin a transmit power reduction on the order of[10(α–β) log10 L + 5 β log10 N] dB. For exam-ple, choosing a cell dimension of L = 1000 mand N = 50 femtocells, with equal path lossexponents α = β = 4, femtocells give a transmitpower saving of nearly 34 dB. When the indoorpath loss exponent is smaller, say β = 2, thetransmit power savings increase to nearly 77 dB.

To summarize, the capacity benefits of femto-cells are attributed to:• Reduced distance between the femtocell

and the user, which leads to higher receivedsignal strength.

• Lowered transmit power, and mitigation ofinterference from neighboring macrocelland femtocell users due to outdoor propa-gation and penetration losses.

• As femtocells serve only around one to fourusers, they can devote a larger portion oftheir resources (transmit power and band-width) to each subscriber. A macrocell, onthe other hand, has a larger coverage area(500 m–1 km radius) and a larger numberof users; providing quality of service (QoS)for data users is more difficult.The first two points illustrate the dual

improvements in capacity through increased sig-nal strength and reduced interference. The thirdpoint shows that deploying femtocells will enablemore efficient usage of precious power and fre-quency resources. The assumption here is thatthe wired broadband operator provides sufficientQoS over the backhaul. Otherwise, backhaul

n Figure 1. Femtocell vs. macrocell throughput.

Normalized throughput per user (b/s/Hz)10

0

0.1

Cum

ulat

ive

dist

ribu

tion

fun

ctio

n

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

2 3 4

Select 20strongest users

All femtocells transmitsimultaneously

Number of femtocells: 50Users per femtocell: 2Median throughput gain: 1.8 b/s/Hz/user

5 6 7 8 9 10

Femtocell (voice)Macrocell (voice)Femtocell (data)Macrocell (data)

Normalized sum throughput (b/s/Hz)500

0

0.1

Cum

ulat

ive

dist

ribu

tion

fun

ctio

n

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

100 150 200

Number of femtocells: 50Users per femtocell: 2Median sum throughput gain: 250 b/s/Hz

Simultaneous low-power femtocelltransmissions enable enormous gainsin sum throughput for a femtocell networkrelative to a macrocell network.

250 300 350

Femtocell (data)Macrocell (data)

Macrocell radius: 400 m Femtocell radius: 40 mTransmit/receive antennas: 1Lognormal std. dev. (macro): 8 dBMacrocell path loss (dB): 34 + 40log(d)Transmit power (femto): 23 dBmLognormal std. dev. (neighboring femto): 12 dBLognormal std. dev. (indoor): 4 dBIndoor path loss (dB): 37 + 30log(d)Penetration loss: 10 dB



capacity limitations could reduce the indoorcapacity gains provided by femtocells.

Example. Consider a cellular orthogonal fre-quency-division multiple access (OFDMA) sys-tem with 100 active users. One scenario consistsof a single macrocell serving all 100 users simul-taneously; another scenario consists of 50 femto-cells, with two active users in each femtocell.Figure 1 illustrates the cumulative distributionfunction of the normalized per user and sumthroughput (normalized by the overall band-width), considering voice and data traffic. Simu-lations show a nearly 0.6 b/s/Hz gain innormalized median user throughput in femtocelldeployments for voice only networks.

With data traffic, it is infeasible for a macro-cell to provide data services to all 100 userssimultaneously because of the limited transmis-sion power and spectrum availability per user.We therefore consider a somewhat pathologicalcase, in which the macrocell always schedulesthe 20 strongest users for transmission. On theother hand, femtocells can transmit simultane-ously over the entire bandwidth. Compared to amacrocell, a femtocell deployment shows a nor-malized user throughput gain equaling 1.8 b/s/Hzand a huge system-wide median sum throughput

gain of nearly 250 b/s/Hz. This shows that thebiggest benefits of femtocells are massiveimprovement in the system spatial spectral effi-ciency.

BUSINESS ASPECTS OF FEMTOCELLSEven though femtocells offer savings in sitelease, backhaul, and electricity costs for theoperator, they incur strategic investments. Oper-ators will need to aggressively price femtocells,despite tight budgets and high manufacturingcosts, to compete with ubiquitous Wi-Fi. Forexample, the North American operator Sprintcharges a subsidized price of $99.99 per Airavefemtocell for subscribing to a $20/month familyplan. At the same time, the features femtocellshave to provide are in many ways more sophisti-cated than what is in a consumer grade Wi-Fiaccess point. The nascent femto vendors are fac-ing cost targets set by the mature high-volumeWi-Fi market and the demands of operators forminimal subsidy to reduce return-on-investment(ROI) time. Consequently, cost issues are in mostcases the central factor driving the selection ofsolutions to each technical challenge.

Table 2 shows a predictive cost breakdown ina femtocell network deployment, conducted by

n Table 2. Predictive return on investment from femtocell deployments

Gartner

Airvana

New core network infrastructureCapital expenditure = $35 per consumer

Subscriber addition

a) Femtocell price = $200b) Price to consumer = $50c) Marketing expense = $50Net expense per additional subscriber = $200

Monthly revenues

a) Customer revenue = $25b) Network operating expenditure = $8c) Customer support = $3

Monthly revenue

1) Electricity savingsfrom reduced macro BS powerconsumption = $0.50 per consumer

2) Monthly savings from reductionin number of macrocells= $1.50 per consumer

3) Monthly revenue from indoor datadownloads = $5 per consumer

4) Bundled voice services, broadbandservices over LTE/WiMAX/UMB= $2.50 per consumer

Monthly expenses

1) Free voice calls insidefemtocell = $2.50 per consumer

2) Operating expenditurefor core network upgrades= $2 per consumer

Expenseper consumer= $4.50 per month

Revenueper consumer= $9.50 per month

Implied ROI with$100 subsidy perconsumer equals20 months

Fixed expense per consumer= $235

Revenue per consumer= $14 per month

Implied ROI equals 16 months

Even though

femtocells offer

savings in site lease,

backhaul, and

electricity costs for

the operator, they

incur strategic

investments.

Operators will need

to aggressively price

femtocells, despite

tight budgets and

high manufacturing

costs, to compete

with ubiquitous

Wi-Fi.



Airvana and Gartner [5]. On balance, it can beassumed that after 1.5 years operator investmentwill be recovered, allowing future profits.

CURRENT STANDARDIZATION ANDDEPLOYMENTS

Given the aggressive cost challenges, standard-ization of requirements across customers isimportant to accomplish a low-cost femtocellsolution. Toward this end, a collaborative orga-nization called the Femto-Forum comprisingoperators and femtocell vendors was formed in2007 with the objective of developing open stan-dards for product interoperability.

Table 3 shows the current state of femtocelldeployments. The operator Sprint currently pro-vides code-division multiple access (CDMA) 1xEVDO femtocell services in the United States;concurrently, a number of operators — Verizonand AT&T (United States), O2 Telefonica andMotorola (Europe), and Softbank (Japan) — areconducting femtocell trials prior to market. ABIResearch [5] predicts 102 million users world-wide on more than 32 million femtocells by 2012.

TECHNICAL CHALLENGESThis section overviews the key technical chal-lenges facing femtocell networks:• Broadband femtocells: Resource allocation,

timing/synchronization, and backhaul• Voice femtocells: Interference management

in femtocells, allowing access to femtocells,handoffs, mobility, and providing Emergen-cy-911 services

• Network infrastructure: Securely bridgingthe femtocell with the operator networkover IP

PHYSICAL AND MEDIUM ACCESS LAYERS:BROADBAND FEMTOCELLS

Confronting operators will be the dual problemsof mitigating RF interference and efficiently

allocating spectrum in femtocell networks. Inter-ference mitigation will require innovative solu-tions since the low-cost target potentiallynecessitates scaled down signal processing capa-bilities inside femtocells. The RF interferencewill arise from:a Macrocell-to-femtocell interferenceb Femtocell-to-femtocell interferencec Femtocell-to-macrocell interferenceThe near-far effect due to uneven distribution ofreceive power is the main contributor for a andc, while femtocell-to-femtocell interference isrelatively smaller due to low transmit power andpenetration losses.

Challenge 1: How Will a Femtocell Adaptto Its Surrounding Environment and Allo-cate Spectrum in the Presence of Intra-and Cross-Tier Interference? — The 3GPPLTE and WiMAX standards ensure intracellorthogonality among macrocellular users andmitigate intercell interference through fractionalfrequency reuse. Since femtocells will be placedby end consumers, the ad hoc locations of fem-tocells will render centralized frequency plan-ning difficult.

Due to the absence of coordination betweenthe macrocell and femtocells, and between fem-tocells, decentralized spectrum allocationbetween macrocell and femtocell users is anopen research problem that can provide answersto the following questions:• Should macrocell and femtocell users be

orthogonal through bandwidth splitting? Isthere an “optimal” splitting policy? Howdoes this vary with the femtocell density?

• Alternatively, with shared bandwidth (i.e.,universal frequency reuse), what fraction ofthe spectrum should the macrocell andfemtocells assign their users?

• Which of these two schemes is “better” invarious configurations?

Challenge 2: How Will Femtocells ProvideTiming and Synchronization? — Femtocells

n Table 3. Existing femtocell offerings.

Manufacturer Partner/operator Region Technology

Samsung (Ubicell) Sprint (Airave) North America a) IS-95, CDMA2000 1xEV-DOb) WCDMA

AirWalk Communications Tatara Systems andTango Networks North America CDMA 1x RTT & 1x-EVDO

Ericsson Europe GSM/3GPP UMTS

Airvana Nokia-Siemens 3GPP UMTS

Alcatel-Lucent North America 3GPP UMTS

Axiom Wireless PicoChip United Kingdom a) 3GPP UMTSb) WiMAX

IP Access (Oyster) PicoChip United Kingdom 3GPP UMTS

Ubiquisys (Zonegate) Kineto Wireless, Google United Kingdom 3GPP UMTS/HSPA

Given the aggressive

cost challenges,

standardization of

requirements across

customers is

important to

accomplish a

low-cost femtocell

solution. Toward this

end, a collaborative

organization called

the Femto-Forum

comprising operators

and femtocell

vendors was

formed in 2007.



will require synchronization to align received sig-nals to minimize multi-access interference andensure a tolerable carrier offset. Synchronizationis also required so that macrocell users can handoff to a femtocell or vice versa, which is mademore difficult due to absence of centralizedcoordination between them. With an IP back-haul, femtocells will experience difficulty inobtaining a time base that is immune to packetjitter. For fourth-generation (4G) OFDMA airinterfaces, ranging procedures to achieve timing(~ 1 µs) and frequency accuracy (~250 ppb) [7,8] are needed for two reasons:• The intercarrier interference arising from a

carrier offset causes loss of subcarrierorthogonality. Additionally, femtocells willhave to compensate for frequency errorsarising from the handsets, which typicallyhave poor oscillators.

• In time-division duplex (TDD) systems,femtocells will require an accurate refer-ence for coordinating the absolute phasesfor forward and reverse link transmissions,and bounding the timing drift.Although both points apply to the macrocell

BS as well, the low cost burden and difficulty ofsynchronizing over backhaul will make efficientsynchronization especially important for femto-cells. Network solutions such as the IEEE-1588Precision Timing Protocol over IP, with poten-tial timing accuracy of 100 ns, and self-adaptivetiming recovery protocols (e.g., the G.8261 stan-dard) are promising. Another possibility is equip-ping femtocells with GPS for synchronizing withthe macrocell, which relies on maintaining stableindoor satellite reception and keeping low costsin a price-sensitive unit. Finally, high-precisionoven-controlled crystal oscillators may be usedinside femtocells, incurring additional cost andperiodic calibration.

Challenge 3: How Will Backhaul ProvideAcceptable QoS? — IP backhaul needs QoSfor delay-sensitive traffic and providing serviceparity with macrocells. Additionally, it should

provide sufficient capacity to avoid creating atraffic bottleneck. While existing macrocell net-works provide latency guarantees within 15 ms,current backhaul networks are not equipped toprovide delay resiliency. Lack of net neutralityposes a serious concern, except in cases wherethe wireline backhaul provider is the same com-pany or in a tight strategic relationship with thecellular operator.

Another issue arises when femtocell usageoccurs when the backhaul is already being usedto deliver Wi-Fi traffic. Trials by Telefonica [5]reveal that when users employed Wi-Fi, femto-cells experienced difficulty transferring data andeven low-bandwidth services like voice. This isespecially important considering that improvedvoice coverage is expected to be a main driverfor femtocells.

PHYSICAL AND MEDIUM ACCESS LAYERS:VOICE FEMTOCELLS

For voice users, an operator faces two choices:either allocate different frequency bands tomacrocell and femtocell users to eliminate cross-tier interference, or serve both macrocell andfemtocell users in the same region of bandwidthto maximize area spectral efficiency. Consideringthe scarce availability of radio resources andease of deployment, using the same region ofbandwidth is preferable, if at all possible.

Challenge 4: How Will Femtocells HandleCross-Tier Interference? — CDMA networks(without femtocells) employ fast power control tocompensate for path loss, shadowing, and fading,and to provide uniform coverage. When femto-cells are added, power control creates dead zones(Fig. 2), leading to nonuniform coverage. On thereverse link, a cell edge macrocell user transmit-ting at maximum power causes unacceptableinterference to nearby femtocells. Consequently,cell edge femtocells experience significantly high-er interference than interior femtocells. On theforward link, at the cell edge — where femtocellsare most needed — macrocell users are disrupt-ed by nearby femtocell transmissions since theysuffer higher path loss than cell interior users.

Challenge 5: Should Femtocells ProvideOpen or Closed Access? — A closed accessfemtocell has a fixed set of subscribed homeusers that, for privacy and security, are licensedto use the femtocell. Open access femtocells, onthe other hand, provide service to macrocellusers if they pass nearby. Radio interference ismanaged by allowing strong macrocell interfer-ers to communicate with nearby femtocells.

Although open access reduces the macrocellload, the higher numbers of users communicat-ing with each femtocell will strain the backhaulto provide sufficient capacity (related to Chal-lenge 3) and raise privacy concerns for homeusers. Open access will need to avoid “starving”the paying home user so that they never see “allcircuits busy.” Since femtocells are typically mar-keted as offering flat rate calling, open accesswill need to differentiate the zero tariff homeusers from the pay-per-minute visitor. For bothreasons, operators are looking at hybrid models

n Figure 2. Dead zones in CDMA femtocell networks.

Scenario A: forward link(BS to mobile)

Scenario A: reverse link(mobile to BS)

Nearby macrocell usertransmits with high power to macro BS and

creates dead zone for femtocell user

Femtocell transmits tohome user and creates

dead zone for macro user

Macrocell transmits to cell edgeuser in femtocell dead zone

CHANDRASEKHAR LAYOUT 8/20/08 5:09 PM 4


where some of the femto’s resources are reservedfor registered family members while others areopen for roamers.

Challenge 6: How Will Handoff Be Per-formed in Open Access? — In general, hand-off from a femtocell to the macrocell network issignificantly easier (as there is only one macroBS) than handoffs from the macrocell to thefemtocell. Current second-generation (2G) and3G systems broadcast a neighbor list used by amobile attached to the current cell to learn whereto search for potential handover cells. Such ahandoff protocol does not scale to the large num-bers of femtocells that “neighbor” (actuallyunderlay) the macrocell, and the underlying net-work equipment is not designed to rapidly changethe lists as femtocells come and go. This moti-vates 4G handover procedures to take the pres-ence of femtocells into consideration.

In open access, channel fluctuations maycause a passing macrocell user to perform multi-ple handovers. In co-channel deploymentsClaussen et al. [10] have proposed auto configu-ration that periodically reduces the pilot powerinside a femtocell when no active calls are inprogress, thereby minimizing handoffs frompassing macrocell users. An open research areais developing low-complexity algorithms to pre-dict the dwell time before handing off a macro-cell user to a nearby femtocell. Yeung andNanda [11] proposed controlling handoff eventsby choosing velocity thresholds based on usermobility and sojourn times when a macrocelluser travels in the vicinity of a femtocell.

Challenge 7: Can Subscribers Carry TheirFemtocells for Use Outside the Home Area?— The portability of the femtocell presents aconundrum. Unlike Wi-Fi networks that operatein unlicensed spectrum in which radio interfer-ence is not actively managed, femtocell networkswill operate in licensed spectrum. Femtocellmobility can cause problems when a subscriberwith operator A carries their femtocell to anoth-er location where the only service provider isrival operator B. In such a scenario, should thefemtocell be allowed to transmit on operator B’sspectrum? Viable options are providing GPSinside femtocells for location tracking and lock-ing the femtocell within a geographical area.Alternatively, interoperator agreements facilitatecharging the home subscriber on roaming.

Challenge 8: How Will Femtocells ProvideLocation Tracking for Emergency-911, andShould They Service Nearby MacrocellUsers with Poor Coverage? — Governmentmandated Emergency-911 services require oper-ators to provision femtocells for transmittinglocation information during emergency calls.Femtocell location may be obtained by eitherusing GPS inside femtocells (added cost withpossibly poor indoor coverage), querying the ser-vice provider for location over the backhaul,gathering information from the macrocell pro-viding the femtocell falls within the macrocellradio range, or even inferring the location fromthe mobile position (estimated by the macro net-work) at handoff to the femtocell.

Ethical/legal dilemmas can arise on whether afemtocell should service macrocell users withpoor outdoor coverage for making emergencycalls if they are located within its radio range. Inopen access networks this problem can be solvedby handoff. Closed access femtocells should beprovisioned to allow communication with unsub-scribed users in the event of emergencies.

NETWORK INFRASTRUCTUREIn a femtocell environment the operator willneed to provide a secure and scalable interfaceover the Internet at a reasonable cost. Tradition-al radio network controllers (RNCs) areequipped to handle tens to hundreds of macro-cells. How will they provide equal parity service tofemtocells over the Internet?

Three network interfaces have been pro-posed, of which the IP multimedia subsystem(IMS)/Session Initiation Protocol (SIP) and unli-censed mobile access (UMA)-based interfacesappear to be the architectures of choice.

Iu-b over IP: Existing RNCs connect to fem-tocells through standard Iu-CS (circuit-switched)and Iu-PS (packet-switched) interfaces presentin macrocell networks. The advantage is that thecapital expenditure is comparatively low insofaras the operator can leverage existing RNCs. Theshortcomings are the lack of scalability and thatthe interface is not yet standardized.

IMS/SIP: The IMS/SIP interface provides acore network residing between the femtocell andthe operator. The IMS interface converts sub-scriber traffic into IP packets and employs voiceover IP (VoIP) using SIP, and coexists with themacrocell network. The main advantages arescalability and rapid standardization. Disadvan-tages include the capital expenditure forupgrade, and the operating expenditure in main-taining two separate core networks for themacrocell and femtocell respectively.

RAN-gateway-based UMA: A radio accessnetwork (RAN) gateway exists between the IPand operator networks, aggregating traffic fromfemtocells. This gateway is connected to theoperator network using a standard Iu-PS/CSinterface. Between the femtocell and the RANgateway, the UMA protocol makes use of secureIP tunneling for transporting the femtocell sig-nals over the Internet. Current UMA-enabledservices such as T-Mobile’s Hotspot@Homerequire dual-mode handsets for switchingbetween in-home Wi-Fi and outdoor cellularaccess. Integrating the UMA client inside femto-cells rather than the mobile would enable futuredeployments support use of legacy handsets.

RESEARCH DIRECTIONSWe now consider the key areas and tools forconducting further femtocell research, with thegoal of designing efficient femtocell architec-tures.

INTERFERENCE MANAGEMENTDue to the ad hoc topology of femtocell loca-tions, interference suppression techniques alonewill prove ineffective in femtocell networks. Suc-cessive interference cancellation, in which eachuser subtracts out the strongest neighboring

In a femtocell

environment the

operator will need to

provide a secure and

scalable interface

over the Internet at a

reasonable cost.

Traditional radio

network controllers

(RNCs) are equipped

to handle tens to

hundreds of macro-

cells. How will they

provide equal parity

service to femtocells

over the Internet?



interferers from their received signal, appearspromising initially, but cancellation errors quick-ly degrade its usefulness [12]. Consequently, aninterference avoidance approach wherein usersavoid rather than suppress mutual interference ismore likely to work well in geography-dependentfemtocell networks. The low cost requirement islikely to influence the design of low-complexityfemtocell BS receivers (e.g., simple matched fil-ter processing) and low-complexity transmissionschemes for sensing nearby available frequencychannels to avoid collisions.

In CDMA femtocell networks with universalfrequency reuse, for example, interference avoid-ance through time hopping and directionalantennas provides a 7× improvement in systemcapacity [9] when macrocell and femtocell usersshare a common bandwidth.

Frequency and time hopping. In Global Sys-tem for Mobile Communications (GSM) net-works, the interference avoidance offeredthrough slow frequency hopping enables femto-cell users and nearby transmitting macrocellusers to avoid consistent mutual interference.Similarly, frequency-hopped OFDMA networkscan use random subchannel assignments in orderto decrease the probability of persistent collisionwith neighboring femtocells.

In time-hopped CDMA, the CDMA durationG Tc (G is the processing gain and Tc is the chipperiod) is divided into Nhop hopping slots, whereeach user randomly selects a hopping slot fortransmission and remains silent during theremaining slots. Random time hopping reducesthe average number of interfering users by a fac-tor of Nhop while trading off the processing gain.The trade-off is that femtocells are accommodat-ed by changing the way an existing CDMAmacrocell network operates.

Directional antennas inside femtocells offerinterference avoidance, with zero protocol over-

head, by restricting radio interference within anantenna sector. Providing a reasonable unit costand easy end-user deployment are the key chal-lenges confronting this approach.

Adaptive power control strategies vary thefemtocell receive power target depending on itslocation. Commercial femtocells such as Sprint’sAirave femtocell tackle cross-tier interferenceusing an “automatic adaptation” protocol thatadjusts the femtocell transmit power. Over theforward link in closed access, Ericsson [13] hasproposed reducing interference to macrocellusers by reducing the femtocell transmit powerwith increasing distance from the macro BS. Thetrade-off is decreased home coverage at far-offfemtocells. Over the reverse link in closed access,we suggest providing a higher receive power tar-get to femtocell users relative to macrocell users[9], which will vary based on femtocell location.

Figure 3 shows the reductions in the outageprobability for a femtocell user (the femto-macro receive power ratios are 1 and 10), inconjunction with interference avoidance usingCDMA time hopping and antenna sectoring. Foropen access, Kishore et al. [14] propose over-coming the near-far effect by extending femto-cell coverage-allow a user to communicate to amacrocell only if their channel gain is apprecia-bly higher — at the expense of increased inter-ference between neighboring femtocells.Multimode phones [7] switch protocols in closed-access systems for mitigating interference, forexample, transmit in WCDMA mode inside afemtocell, and revert to GSM mode otherwise.

MIMO FEMTOCELLSMultiple antennas at the transmitter and/or thereceiver (multiple input multiple output, MIMO)exploit the spatial diversity of the wireless chan-nel. Femtocells can perform temporal link adap-tation through adaptive modulation and coding;additionally, MIMO spatial link adaptation willenable a femtocell to switch between providinghigh data rates and robust transmission. Highdata rates are obtainable by transmitting multi-ple spatial streams (spatial multiplexing) overhigh SINR links. Over low SINR links, MIMOprovides robustness through open and closedloop diversity schemes such as space-time codesand beamforming. Interesting areas for futureresearch are:• Link adaptive mode switching for femtocells

between diversity and spatial multiplexing[15]

• Analyzing the effect of channel state infor-mation errors induced by co-channel inter-ference on MIMO femtocell performance

• The complexity limitations of MIMO femto-cell receivers, which may be significant vs.macrocell receivers due to cost considera-tions

• Channel models for MIMO femtocells, sincethe diversity characteristics may be very dif-ferent from macrocells

CONCLUSIONSFemtocells have the potential to provide high-quality network access to indoor users at lowcost, while simultaneously reducing the burden

n Figure 3. Outage probability in a closed-access femtocell.

Mean macrocellular users50

10-3

10-4

Fem

toce

ll ou

tage

pro

babi

lity

10-2

10-1

100

10 15 20 25

Time hopping slots = 128Femtocell-macro power ratio = 1

Femto antenna sectors = 1

Time hopping slots = 4Femto-macro power ratio = 10


No interferenceavoidance and equalfemto/macroreceivepowers

CDMA hopping slots = 4Femto-macro power ratio = 10


Maximum allowable outage

30 35 40 45 50

Macrocell radius = 500 mFemtocell radius = 20 mDistance from femtocell to

macrocell = 250 mCDMA processing gain = 128Outdoor lognormal std. dev. = 8 dBIndoor path loss (dB) = 37+20log(d)Outdoor to indoor path loss (dB) = 44.24+40log(d)



on the whole system. This article has identifiedthe key benefits of femtocells, the technologicaland business challenges, and research opportuni-ties. From a technical standpoint, operators facechallenges in providing a low-cost solution whilemitigating RF interference, providing QoS overthe IP backhaul, and maintaining scalability.From a business perspective, generating long-term revenue growth and overcoming initial end-user subsidies are key challenges.

ACKNOWLEDGMENTSThe authors would like to thank the reviewersfor providing excellent feedback, which greatlyimproved the quality and readability of this arti-cle. The authors also thank Andrew Hunter(University of Texas) and Shalinee Kishore(Lehigh University) for their comments and sug-gestions on the article, and Vijay Sundararajan(Texas Instruments) for ongoing feedback onour work.

REFERENCES[1] Tropos Networks, “Picocell Mesh: Bringing Low-Cost

Coverage, Capacity and Symmetry to Mobile WiMAX,”White Paper.

[2] M.-S Alouini and A. J. Goldsmith, “Area Spectral Effi-ciency of Cellular Mobile Radio Systems,” IEEE Trans.Vehic. Tech., vol. 48, no. 4, July 1999, pp. 1047–66.

[3] A. Saleh, A. Rustako, and R. Roman, “Distributed Antennasfor Indoor Radio Communications,” IEEE Trans. Commun.,vol. 35, no. 12, Dec. 1987, pp. 1245–51.

[4] C.-L. I, L. J. Greenstein, and R. D. Gitlin, “A Microcell/Macrocell Cellular Architecture for Low- and High-Mobility Wireless Users,” IEEE JSAC, vol. 11, no. 6, Aug.1993, pp. 885–91.

[5] Presentations by ABI Research, Picochip, Airvana, IP.access,Gartner, Telefonica Espana, 2nd Int’l. Conf. Home AccessPoints and Femtocells; http://www.avrenevents.com/dallas-femto2007/purchase_presentations.htm

[6] Analysys, “Picocells and Femtocells: Will Indoor Base-Stations Transform the Telecoms Industry?” http://research.analysys.com

[7] Picochip, “The Case For Home Base Stations,” WhitePaper, Apr. 2007.

[8] J. G. Andrews, A. Ghosh, and R. Muhamed, Fundamen-tals of WiMAX, Prentice Hall, 2007.

[9] V. Chandrasekhar and J. Andrews, “Uplink Capacity andInterference Avoidance in Two-Tier Femtocell Net-works,” to appear, IEEE Trans. Wireless Commun.,http://arxiv.org/abs/cs.NI/0702132

[10] L. T. W. Ho and H. Claussen, “Effects of User-Deployed,Co-Channel Femtocells on the Call Drop Probability in aResidential Scenario,” 18th Annual IEEE Int’l. Symp. Person-al, Indoor and Mobile Commun., Sept. 2007, pp. 1–5.

[11] K. L. Yeung and S. Nanda, “Channel Management inMicrocell/Macrocell Cellular Radio Systems,” IEEE Trans.Vehic. Tech., vol. 45, no. 4, Nov 2006, pp. 601–12.

[12] S. Weber et al., “Transmission Capacity of Ad Hoc Networkswith Successive Interference Cancellation,” IEEE Trans. Info.Theory, vol. 53, no. 8, Aug. 2007, pp. 2799–2814.

[13] Ericsson, “Home NodeB Output Power,” 3GPP TSGWorking Group 4 meeting; http://www.3gpp.org/ftp/tsg_ran/WG4_Radio/TSGR4_43bis/Docs/

[14] S. Kishore et al, “Uplink User Capacity in a Multicell CDMASystem with Hotspot Microcells,” IEEE Trans. Wireless Com-mun., vol. 5, no. 6, June 2006, pp. 1333–42.

[15] A. Forenza et al, “Adaptive MIMO Transmission forExploiting the Capacity of Spatially Correlated Chan-nels,” IEEE Trans. Vehic. Tech., vol. 56, no. 2, Mar.2007, pp. 619–30.

BIOGRAPHIESVIKRAM CHANDRASEKHAR ([email protected]) is a Ph.D.candidate at the University of Texas (UT) Austin. He com-pleted his B.S. at IIT Kharagpur and his M.S. at Rice Univer-sity. His current research focuses on fundamental limitsand algorithms for microcellular and hotspot-aided broad-band cellular networks. He has held industry positions atNational Instruments, Freescale, and Texas Instruments.

JEFFREY G. ANDREWS ([email protected]) received aB.S. from Harvey Mudd College, and M.S. and Ph.D.degrees in electrical engineering from Stanford University.He is an associate professor in the Electrical and ComputerEngineering Department at UT Austin, where he is directorof the Wireless Networking and Communications Group.He is author of Fundamentals of WiMAX (Prentice Hall,2007) and recipient of the 2007 NSF CAREER award. Hehas held industry positions at Intel and Qualcomm.

ALAN GATHERER received a B.Eng. from Strathclyde Universi-ty, and M.S. and Ph.D. degrees in electrical engineeringfrom Stanford University. He is the CTO for the Communi-cations Infrastructure and Voice Business and a Fellow atTexas Instruments. He is responsible for the strategic devel-opment of TI’s digital baseband modems for 3G and 4Gwireless infrastructure as well as high-performance medicalproducts..

From a technical

standpoint, operators

face challenges in

providing a low-cost

solution while

mitigating RF

interference, provid-

ing QoS over the IP

backhaul, and

maintaining

scalability. From a

business perspective,

generating long-term

revenue growth and

overcoming initial

end-user subsidies

are key challenges.


femtocell networks a survey - chandrasekhar - ieee comm sept 2008

Documents