A Survey on 3GPP Hetrogeneous Networks

IEEE Wireless Communications • June 201110 1536-1284/11/$25.00 © 2011 IEEE

OpeHeN

MacroeNB

ALEKSANDAR DAMNJANOVIC, JUAN MONTOJO, YONGBIN WEI, TINGFANG JI, TAO LUO,MADHAVAN VAJAPEYAM, TAESANG YOO, OSOK SONG, AND DURGA MALLADI,

QUALCOMM INC.

A SURVEY ON 3GPP HETEROGENEOUS NETWORKS

IN V I T E D ART I C L E

INTRODUCTIONData traffic demand in cellular networks today isincreasing at an exponential rate. As the linkefficiency is approaching its fundamental limits,further improvements in system spectral efficien-cy are only possible by increasing the nodedeployment density. In a relatively sparse deploy-ment of macro base stations, adding anotherbase station does not severely increase intercellinterference, and solid cell splitting gains areeasy to achieve. However, in already densedeployments today, cell splitting gains are signifi-cantly reduced due to already severe intercell

interference. Moreover, site acquisition costs ina capacity limited dense urban area can get pro-hibitively expensive.

Challenges associated with the deploymentof traditional macro base stations can be over-come by the utilization of base stations withlower transmit power. As we will further discussbelow, we classify low power nodes as pico,femto and relay nodes. If the low power nodesare intended for outdoor deployments, theirtransmit power ranges from 250 mW to approx-imately 2 W. They do not require an air condi-tioning unit for the power amplifier, and aremuch lower in cost than traditional macro basestations, which transmit power typically variesbetween 5 and 40 W. Femto base stations aremeant for indoor use, and their transmit poweris typically 100 mW or less. Unlike pico, femtobase stations may be configured with a restrict-ed association, allowing access only to its closedsubscriber group (CSG) members. Such femtobase stations are commonly referred to asclosed femtos. A network that consists of a mixof macrocells and low-power nodes, wheresome may be configured with restricted accessand some may lack wired backhaul, is referredto as a heterogeneous network and is illustratedin Fig. 1.

Femto cells have recently attracted signifi-cant attention. There are a number of technicalstudies associated with various aspects of femtocells deployments based on cellular technologysuch as third-generation (3G) High Speed Pack-et Access (HSPA) or 1xEV-DO. These studiesconsider operations, administration, and man-agement (OAM) and self organizing network(SON) protocols, network architecture, local IPaccess (LIPA), access (open, closed, andhybrid), and interference management [1–3, ref-erences therein]. In addition, heterogeneousnetworks based on different access technologies,where macro network is based on a cellulartechnology and low power access points arebased on WLAN have also been studied in liter-ature [4, 5]. Reduced cost is one of the maindrivers for the adoption of femto cells. It wasshown in [6] that in urban areas a combinationof publicly accessible home base stations orfemto cells (randomly deployed by the end

ABSTRACTAs the spectral efficiency of a point-to-point

link in cellular networks approaches its theoreti-cal limits, with the forecasted explosion of datatraffic, there is a need for an increase in thenode density to further improve network capaci-ty. However, in already dense deployments intoday’s networks, cell splitting gains can beseverely limited by high inter-cell interference.Moreover, high capital expenditure cost associat-ed with high power macro nodes further limitsviability of such an approach. This article dis-cusses the need for an alternative strategy, wherelow power nodes are overlaid within a macronetwork, creating what is referred to as a hetero-geneous network. We survey current state of theart in heterogeneous deployments and focus on3GPP LTE air interface to describe futuretrends. A high-level overview of the 3GPP LTEair interface, network nodes, and spectrum allo-cation options is provided, along with theenabling mechanisms for heterogeneous deploy-ments. Interference management techniques thatare critical for LTE heterogeneous deploymentsare discussed in greater detail. Cell range expan-sion, enabled through cell biasing and adaptiveresource partitioning, is seen as an effectivemethod to balance the load among the nodes inthe network and improve overall trunking effi-ciency. An interference cancellation receiverplays a crucial role in ensuring acquisition ofweak cells and reliability of control and datareception in the presence of legacy signals.

As the spectral efficiency of a point-to-point link incellular networksapproaches its theoretical limits,with the forecastedexplosion of datatraffic, there is aneed for an increasein the node densityto further improvenetwork capacity.

DAMNJANOVIC LAYOUT 6/6/11 10:56 AM 0

IEEE Wireless Communications • June 2011 11

user), and macrocells deployed by an operatorfor area coverage in a planned manner, canresult in significant reductions (up to 70 percentin the investigated scenario) of the total annualnetwork costs compared to a pure macro-cellu-lar network deployment. If a wired backhaul isnot available, relay nodes can be deployedwhere the air interface spectrum is used forbackhaul connectivity and to provide access toterminals [7]. In this case, the relay nodeappears as user equipment (UE) to the macrobase station and as a regular base station to theUE it serves.

Straightforward co-channel deployment oflow-power nodes has its own challenges. Theintroduction of low-power nodes in a macro net-work creates imbalance between uplink anddownlink coverage. Due to larger transmit powerof the macro base station, the handover bound-ary is shifted closer to the low-power node,which can lead to severe uplink interferenceproblems as UE units served by macro base sta-tions create strong interference to the low-powernodes. Given the relatively small footprint oflow-power nodes, even in the case of the mostoptimized placement, low-power nodes maybecome underutilized due to geographic changesin data traffic demand. The performance of amixed deployment of macro, pico, and openfemto cells was evaluated in [8, 9], which showedthat the limited coverage of low-power nodes isthe main reason for limited performance gain inheterogeneous networks.

Furthermore, some of the deployed femtocells may have enforced restricted association,which effectively creates a coverage hole andexacerbates the interference problem. In orderto cope with the interference, it is necessary tointroduce techniques that can adequately addressthese issues. A well-known baseline interferencemanagement and self-configuration solution for3G femtocells (i.e., HSPA or 1xEVDO femto-cells) is downlink transmitter power control. Asmentioned above, due to restricted association, aclosed femtocell creates a downlink coveragehole at the vicinity of the femtocell for non-member UE [8]. In order to reduce the coveragehole, a femtocell could adapt the transmittedpower intelligently such that interference is notleaked outside the intended coverage area suchas a residential home. This technique is oftenassisted by a network listening module (NLM),where a femtocell monitors macrocell signalstrength occasionally on the same channel. Sincethe femto coverage and femto interference tothe macro UE are directly proportional to thefemto to macro signal strength ratio, a femtocellcould set the proper transmit power by estimat-ing the additional interference that could be tol-erated by non-member UE near the femtocell.As a result, the femtocell maintains a similarcoverage area regardless of its location withinthe embedded macrocell. The network capacityof heterogeneous networks with mixed macro-and closed femtocell deployments were investi-gated in [10–12], and significant gain was shownwhen dynamic femtocell transmit power controlwas utilized.

Further enhancement to this technique isUE-enhanced femtocell transmit power setting.

For example, when non-member UE attempts toaccess a femtocell, the femtocell rejects theattempt due to restricted access. At the sametime, the femtocell could also obtain informationabout the macro UE during the access hand-shake. By intelligently processing the femto UEand macro UE information, a proper femtocelltransmitter power could be chosen.

On the uplink, femto UE and macro UEcould create high interference, which can lead tohigh interference variation. Due to the downlinkpower mismatch, UE that receives similar signalstrength from the macro- and femtocells is muchcloser to the femtocell than the macrocell. In thecase of nominal transmit power of the macrobase station at 43 dBm and femto at 20 dBm,the received power difference on uplink wouldbe as high as 23 dB (i.e., the received power atthe femtocell is 23 dB higher than the receivedpower at the macrocell). If this UE is served bythe macrocell with a targeted received signal-to-noise ratio of 5 dB, the interference caused bythis UE would be 28 dB above the thermal atthe femtocell when this UE starts transmitting.In order to cope with such large interferencevariation, in 3G systems a femtocell could useadaptive noise padding at the receiver to dampdown the total interference plus noise level vari-ation [13].

In 3G networks, pico deployments are alsopopular for hotspot coverage. Maximum gain isachieved by adjusting the handover boundarybetween the macro- and picocells. For example,when a macro experiences a high load, it couldhand over UE to a picocell early. However, dueto co-channel interference, adjustment of hand-over boundary is quite limited,; handover to avery weak cell could cause radio link failure.

Given the interference scenarios deploymentof low-power nodes can create, it is clear that anintroduction of a mechanism that allows flexibleassociation between a serving base station and aUE, and that mitigates downlink and uplinkinterference would be beneficial. In 4G net-works, new physical layer design allows for flexi-ble time and frequency resource partitioning.This added flexibility enables macro- andfemto/picocells to assign different time-frequen-

Figure 1. Heterogeneous network topology utilizing a mix of high-power(macro) and low-power base stations.

ClosedHeNB

OpenHeNB

PicoeNB

MacroeNB

RelayeNB

DAMNJANOVIC LAYOUT 6/6/11 10:56 AM Page 11

IEEE Wireless Communications • June 201112

cy resource blocks within a carrier or differentcarriers (if available) to their respective UE.This is one of the intercell interference coordi-nation (ICIC) techniques that can be used onthe downlink to mitigate data interference [14,15]. With additional complexity, joint processingof serving and interfering base station signalscould further improve the performance of het-erogeneous networks [16, 17], but these tech-niques require further study for the scenarioscommonly seen in practice.

For those reasons, in the context of LongTerm Evolution (LTE) standardization, theThird Generation Partnership Project (3GPP) isdiscussing the concept of cell range expansionthrough handover biasing and resource partition-ing among different node power classes. Thebiasing mechanism allows for load balancing,where depending on the bias value, the networkcan control the number of UE units associatedwith the low-power nodes and therefore controltraffic demand at those nodes. Resource parti-tioning, which can also be adaptive, allows con-figuration and adjustment of interferenceprotected resources, enabling UE in a cellexpanded area to receive data. Moreover, thesame resource partitioning technique provides amechanism to mitigate uplink interference andallows non-CSG member UE to receive servicewhen in proximity of closed femtocells.

The main goal of this article is to provide anoverview of the topology and the deploymentoptions for heterogeneous networks. We focuson LTE downlink, where we review the currentstate of the art, describe technical challenges,and give some thoughts on future research direc-tions. Uplink data transmissions are slaved tothe downlink control channel, and interferencemanagement techniques enabled to partitiondownlink resources apply to the uplink transmis-sions too.

LTE OVERVIEWThe first release of LTE was published in March2009 and is referred to as LTE Rel-8 [18]. 3GPPhas developed the LTE standard for fourth-gen-eration (4G) cellular networks based on orthog-onal frequency-division multiplexing (OFDM)waveform for downlink (DL) and single-carrierFDM (SC-FDM) waveform for uplink (UL)communications mainly to improve the userexperience for broadband data communications.Compared to 3G technologies, such as 3GPP’sHSPA,1 LTE Rel-8 offers higher peak data ratesdue to larger system bandwidth (up to 20 MHzwas allowed) and higher-order multiple-inputmultiple-output (MIMO) spatial processing tech-niques (up to 4 Tx × 4 Rx open and closed loopMIMO schemes are supported in the DL ofLTE Rel-8).

Figure 2 illustrates the LTE network [18]nodes and the interfaces among them. The basestations are denoted eNode-B (eNB) and themobile stations or terminals as UE. The low-power nodes include picocells, femtocells, homeeNBs (HeNBs), and relay nodes (RNs). TheeNB serving the RN (i.e., scheduling RN back-haul traffic) is denoted donor eNB (DeNB). Thesame eNB can be the DeNB for one RN and theregular serving cell for UE, as shown in Fig. 2.The mobility management entity (MME) andserving gateway (S-GW) serve as local mobilityanchor points for the control and data planes,respectively.

The X2 interface defined as a direct eNB-to-eNB interface allows for inter-cell interferencecoordination (ICIC). The Rel-8 ICIC techniquescan be summarized as:• Proactive: Techniques that facilitate fractional

frequency reuse (FFR) or “soft reuse” opera-tion in the DL and UL with the goal of reduc-ing interference experienced in certain

Figure 2. LTE heterogeneous network nodes and their interfaces.

RN

UE

UE

UE

Pico Pico

HeNBHeNB HeNB

HeNB GW

E-UTRAN

UE

UE

eNB

DeNB

MME / S-GWMME / S-GW

X2

X2

S1 S1

S1X2

S1

S1S1

S1 S1

S1

S1

S1

S1X2Un

S11

1 HSPA Rel-10 supportscarrier aggregation mode,where four 5 MHz can beaggregated offering broad-band wireless communi-cation to a single UE over20 MHz bandwidth.

Compared to 3Gtechnologies, such as3GPP’s HSPA, LTERel-8 offers higherpeak data rates dueto larger systembandwidth (up to 20MHz was allowed)and higher-orderMIMO spatial processing techniques.



frequency subbands in order to increase thecell edge user throughput

• Reactive: Techniques that respond to high-interference conditions and enable tight con-trol of the interference-over-thermal (IoT)level in the UL

These ICIC techniques are expanded in Rel-10to enable efficient support of co-channel hetero-geneous network deployments, discussed later.S1 and S11 interfaces support transfer or userand data traffic between the correspondingnodes and are not utilized for ICIC. The Uninterface refers to an air interface betweenDeNB and RN. Un is based on a modified inter-face between the eNB and UE in order to allowhalf duplex operation for the RN.

As mentioned above, the LTE air interface isbased on OFDM in the DL and SC-FDM in theUL. The basic time and frequency unit in theDL (UL) is one OFDM (SC-FDM) symbol andone subcarrier (virtual subcarrier), respectively.The subcarrier spacing is 15 kHz and therefore,the OFDM symbol duration is 66.67 us. EachOFDM/SC-FDM symbol is pre-appended with acyclic prefix (CP) to suppress the inter-symbolinterference and mitigate multi-path. Two CPdurations are defined; the normal CP has aduration of 4.7 us and the extended CP has aduration of 16 us. One resource element corre-sponds to one subcarrier (virtual subcarrier) inone OFDM (SC-FDM) symbol. OFDM (SC-FDM) symbols are grouped in subframes of 1ms duration. Each subframe is composed of two0.5 ms slots. In order to limit the signaling over-head of data allocations, the minimum schedul-ing unit for the DL and UL of LTE is referredto as a resource block (RB). One RB pair con-sists of 12 subcarriers in the frequency domain(i.e., 180 kHz) and one subframe in the timedomain (i.e., 1 ms). In the DL, all the controlinformation is time-division multiplexed (TDM)with the data transmission. The DL controlinformation is concentrated in the first slot ofthe first subframe, and dynamically spans thefirst one, two, or three OFDM symbols of thesubframe.

Subframes are further grouped in 10 ms radioframes. Figure 3 illustrates the physical layerframe structure for FDD. Each radio frame hastwo 5 ms halves containing the signals necessaryto obtain the physical identity of the cell. Thesesignals are what we call the acquisition channels,which are the primary and secondary synchro-nization signals, providing the physical cell iden-tity (PCI) of the cell, and the physical broadcastchannel (PBCH), which provides some criticalsystem information such as the DL transmissionbandwidth and the number of DL antenna ports.The acquisition channels share the property ofspanning the middle six RBs of the system band-width. This enables having the same acquisitionchannels irrespective of the actual system band-width (up to 20 MHz is supported for Rel-8).

LTE defines a reference or pilot signal inRel-8, referred to as a common reference signal(CRS), which is used for mobility measurementsas well as for demodulation of the DL controland data channels. The CRS transmission is dis-tributed in time and frequency, as shown in Fig.3, to enable adequate time and frequency inter-

polation of the channel estimates for the pur-pose of coherent reception of the transmittedsignals in time- and frequency-selective channels.

As discussed later, co-channel deployments ofheterogeneous networks rely on the coordinationof almost blank subframes. Almost blank sub-frames are intended to reduce the interferencecreated by the transmitting node while providingfull legacy support. For that reason, on almostblank subframes, eNB does not schedule unicasttraffic while transmitting acquisition channelsand CRS to provide legacy support.

3GPP has been working on further improvingthe spectral efficiency of LTE as part of its Rel-10 version. LTE Rel-10 is being developed tomeet ITU requirements for IMT-Advanced tech-nology. LTE Rel-10 [19], also termed LTE-Advanced (LTE-A), supports improved MIMOoperation as DL MIMO support is enhanced (8Tx × 8 Rx is supported), and UL MIMO (4 Tx ×4 Rx) is introduced to improve link spectral effi-ciency. Signaling mechanisms enabling aggrega-tion of multiple carriers are also introduced inLTE Rel-10, offering improvements in peak userdata throughput. Up to five 20-MHz componentcarriers can be aggregated, offering a peak datarate of more than 1 Gb/s. However, theseimprovements, while significant from the linkperspective and for users in good coverage, or interms of the peak data rates for lightly loadedsystems, do not translate into significantimprovements in terms of system spectral effi-ciency in bits per second per Hertz. System gainsare only achievable through increased node den-sity and deployment of low-power nodes, such aspico, femto, and relay base stations.

PICOCELLSPicocells are regular eNBs with the only differ-ence of having lower transmit power than tradi-tional macro cells. They are, typically, equippedwith omni-directional antennas, i.e., not sector-ized, and are deployed indoors or outdoors oftenin a planned (hot-spot) manner.

Their transmit power ranges from 250 mW toapproximately 2 W for outdoor deployments,while it is typically 100 mW or less for indoordeployments. Since picocells are regular eNBsfrom the architecture perspective, as can be seenin Fig. 2, they can benefit from X2-based inter-cell interference coordination (ICIC).

FEMTOCELLSFemtocells or HeNBs are typically consumerdeployed (unplanned) network nodes for indoorapplication with a network backhaul facilitatedby the consumer’s home digital subscriber line(DSL) or cable modem. Femtocells are typicallyequipped with omnidirectional antennas, andtheir transmit power is 100 mW or less.

Depending on whether the femto cells allowaccess and hence usage of the consumer’s homeDSL or cable modem to all terminals, or to arestricted set of terminals only, femto cells areclassified as open or closed. Closed femtosrestrict the access to a closed subscriber group(CSG), while open femtos are similar to pico-cells but with the network backhaul provided bythe home DSL or cable modem. A femtocell canalso be hybrid, whereby all terminals can access

Picocells are regulareNBs, with the onlydifference of having

lower transmit powerthan traditional

macrocells. They are,typically, equipped

with omnidirectionalantennas (i.e., notsectorized) and aredeployed indoors oroutdoors, often in a

planned (hotspot)manner.



but with lower priority for the terminals that donot belong to the femto’s subscriber group.

Since closed femtos do not allow access to allterminals, they become a source of interferenceto those terminals. Co-channel deployments ofclosed femtos therefore cause coverage holesand hence outage of a size proportional to thetransmit power of the femtocell.

Figure 2 shows the architecture of LTE fem-tocells (HeNBs), and, as can be seen, no X2interface is defined for HeNBs in Rel-8/9. Theabsence of an X2 interface for closed femtosdoes not make ICIC possible for this type of

node. Instead, OAM-based techniques in con-junction with possibly autonomous power con-trol techniques are the only viable interferencecontrol techniques for Rel-10. These techniquesseek to minimize the outage these network nodescause around them by enabling reception of thesignal from the closest macrocell in close prox-imity to the closed femto.

RELAY NODEAn RN is a network node without a wired back-haul. The backhaul, which provides the attach-ment of the RN to the rest of the network, is

Figure 3. LTE DL physical layer structure for FDD.

SubcarrierRB

Subcarrier

S1

R0 CRS for antenna port 0




Slot 0

Time

Freq

OFDMsymbol

6 RBs 6 RBs

Normal CP

Normal CP

P B B B B

S1

Extended CP

P B B B B

10 ms

5 ms Subframe

Slot

0

0 1 2 3 4 5 6

0 1 2 3 4 5 6 0 1 2 3 4 5 6

Extended CP

0 1 2 3 4 5 0 1 2 3 4 5

R0 R2 R1 R0 R3 R1 R0 R2 R1 R0 R3 R10 0

1 1

2 2

R1 R3 R0 R1 R2 R0 R1 R3 R0 R1 R2 R03 3

4 4

5 5

R0 R2 R1 R0 R3 R1 R0 R2 R1 R0 R3 R16 6

7 7

8 8

R1 R3 R0 R1 R2 R0 R1 R3 R0 R1 R2 R09 9

10 10

11

OFDMsymbol

Slot 0

11

7 8 9 10 11 12 13 14 15 16 17 18 19

1 2 3 4 5 6 7 8 9

P Primary synchronization signal

S1 Secondary synchronization signal 1

S2 Secondary synchronization signal 2

B PBCH

Time

Freq Slot 0

OFDMsymbol

S2

Slot 10OFDMsymbol

Normal CP

P

S2

Extended CP

P 6 RBs

Slot 10

OFDMsymbol

Slot 1

Subframe

OFDMsymbol

Slot 0 Slot 1

Subframe



wireless and uses the air interface resources ofthe wireless system in question. In case the back-haul communication takes place in the same fre-quency as the communication to/from UE onDL/UL, respectively, the relays are denoted asin-band. If the backhaul communication takesplace at a frequency different to that used byUE in DL and UL, the relay node is classified asout-of-band. Note that unlike in-band relays,out-of-band relays do not pose many physicallayer design challenges. However, out-of-bandrelays require additional dedicated spectrum forbackhaul operation and therefore are far lessattractive than their in-band counterparts. As aresult, in-band relaying has been the focus ofstudy in 3GPP.

The backhaul link denotes the DeNB to relaylink; the access link denotes the RN to UE link;and the direct link denotes the eNB to UE link(without relay participation). An RN can behalf-duplex (HD) or full-duplex (FD). This cate-gorization is relevant for in-band relays wherefor the DL, the eNB-to-relay transmission takesplace in the same frequency as the relay-to-UEtransmission (i.e., backhaul and access links areat the same frequency). Similarly, for the UL,the UE-to-relay transmission takes place in thesame frequency as the relay-to-eNB transmis-sion. HD operation is motivated to avoid desens-ing the RN receiver if it had to transmit andreceive at the same frequency and at the sametime. Therefore, HD relays will, at a given time,either transmit to the UE on the access link orlisten to its DeNB on the backhaul link for DLcommunications, and transmit to the eNB on thebackhaul link or listen to the UE on the accesslink for UL communications. FD relays are ableto transmit and receive at the same time on dif-ferent links, relying on spatial separationbetween access and backhaul links (e.g., byantenna directivity or antenna location for bothlinks) and possibly on interference cancellationcapabilities of such relay nodes. Relay nodes arefull-fledged eNBs without a wired backhaul.There are independent HARQ entities at thebackhaul link, i.e., between the donor eNB andthe relay node, and at the access link, i.e.,between the relay node and the UE.

Relay nodes are typically equipped with direc-tional antennas in the backhaul link (pointing tothe DeNB) and omnidirectional antennas in theaccess link. RNs are deployed indoors or out-doors. Their transmit power ranges from 250mW to approximately 2 W for outdoor deploy-ments, while it is typically 100 mW or less forindoor deployments. Figure 2 shows the archi-tecture of the RNs. Since there is an X2 inter-face to the DeNB, RNs can benefit fromX2-based ICIC.

DEPLOYMENT SCENARIOS ANDTECHNICAL CHALLENGES

Heterogeneous networks are networks deployedwith a mix of traditional (high-power) macro and(low-power) pico, femto, and/or relay nodes. Inaddition to possibly having a mix of open andclosed subscriber access, this type of networks ischaracterized by large disparities in the transmit

power used by different types of network nodes.Such power disparities, in general, put the low-power nodes (pico, femto, and relay) at a disad-vantage relative to the high-power nodes(macrocells). In addition, the deployment ofCSG cells poses the challenge of how to sharephysical resources (time/frequency) with themacrocell to avoid creating coverage holes in themacro network. Figure 4 illustrates deploymentoptions that are further discussed below.

MULTICARRIER DEPLOYMENTA simple solution to avoid macro coverage beingaffected by the deployment of closed femtocellsis, as illustrated in Fig. 4a, to deploy the closedfemtocells on a carrier frequency different thanthat of the macro network. While this solution iseffective in avoiding the interference caused bythe closed femtocells, it is highly inefficient as itrequires at least two carrier frequencies and cre-ates undesirable bandwidth segmentation.

The same solution can be used for the deploy-ment of picocells, open femtos, and RNs, whereone carrier frequency is used by the high-powernodes (macrocells) and another carrier frequen-cy is used by the low-power nodes (picocells,open femtos, and RNs). Since the transmitpower characteristics of the nodes transmittingon each of the component carriers are the same,there is no power disparity and hence no inter-ference problem: each frequency layer is homo-geneous. While for the closed femto case themotivation to deploy closed femto on a carrierfrequency different than that of the macro net-work is to avoid the creation of macro coverageoutages from the deployment of closed femtos,the same cannot be said for the other (openaccess) low-power nodes. For this case, themacro network and the low-power nodes couldboth use the two carrier frequencies; and there-fore, terminals that can aggregate the receptionof more than one carrier could benefit from theavailability of the entire available spectrum.

CARRIER AGGREGATIONConsider the scenario where, as shown in Fig.4b, one carrier frequency is used for macro cov-erage and another is shared by macrocells andclosed femtocells. Interference from the closedfemtocell to the macro coverage is avoided sinceone carrier frequency is used only by macrocells.The other carrier frequency is shared betweenmacrocells and closed femtocells.

Carrier-aggregation-capable UE can benefitfrom ubiquitous coverage of the macro networkin one of the carrier frequencies and partial cov-erage (outside of the closed femto coverage) ona second carrier frequency. At the same time,closed femtocells will continue providing cover-age in their respective carrier frequency. Notethat this example can be generalized to the casewhere more than one carrier frequency is dedi-cated to closed femtocells to alleviate the closed-femto-to-closed-femto interference problemarising from having a single carrier frequency forthe entire closed femto deployment. Therefore,carrier aggregation enables the full use of (fre-quency) resources by the macro network withoutcompromising the closed femto coverage.

While the closed femto case is a very obvious

Relay nodes are full-fledged eNBs without a wired

backhaul. There areindependent HARQ

entities at the backhaul link (i.e.,between the donoreNB and the relaynode( and at theaccess link (i.e.,

between the relaynode and the UE).



example showcasing the benefits of carrier aggre-gation in heterogeneous networks, a similar rea-soning can follow for other types of (low-power)nodes:, picocells, open femtos, and relay nodes.These types of nodes, as discussed above, areintrinsically disadvantaged with respect to high-power macrocells. It is important to note that inthe case of deploying low-power nodes withopen access, the coverage of the macro networkis not jeopardized by the deployment of thesenodes. However, targeting the regular macrocoverage in one of the carrier frequencies by set-ting the power of the macrocells to their nomi-nal value while deploying low-power nodes onanother carrier frequency would maximize thecapacity gains.

The macrocells can also transmit at the carri-er frequency of the low-power nodes but with

reduced power to avoid a large power disparitythat would overwhelm the coverage of the low-power nodes. Conversely, the (harmless) low-power nodes can be deployed in the carrierfrequency originally used by the macro network.As a result, for this example, one carrier fre-quency would be used by the macro network toprovide nominal macro coverage (macrocelltransmission at full power). The low-powernodes would also use this carrier frequency forcapacity enhancement at locations close to thesenodes. Another carrier frequency would be usedby low-power nodes and by macrocells withreduced transmit power. In this case, clearly,carrier-aggregation-capable UE would benefitfrom the fact that both types of nodes use twocarrier frequencies.

Therefore, carrier aggregation can be a pow-

Figure 4. Multicarrier, carrier aggregation, and co-channel deployments.

Pico/femto eNBMacro eNBDL datachannel

DL controlchannel

(b)

(c)

(a)

Subframe

F1

F2

F1

F2

F1

Subframe

Subframe

Subframe

Subframe

Macro eNB primary component carrierPico/femto eNB secondary carrier

Pico/femto eNB primary component carrierMacro eNB secondary carrier

Primary carrier is typically utilized for transmission of DL control channel.Secondary carrier is mainly used for receiving data. Fractional frequencyreuse and/or reduced transmit power are commonly utilized.

Subframe

Datascheduled

Datascheduled

Data not scheduled -increased coverage allowed

for pico/femto eNB

Data not scheduled -increased coverage allowed

for macro eNB

While the closedfemto case is a veryobvious exampleshowcasing the benefits of carrieraggregation in heterogeneous networks, a similarreasoning can followfor other types of(low-power) nodes:pico cells, open femtos, and relay nodes.



erful tool for efficient deployment of heteroge-neous networks without causing a loss of band-width. However, carrier-aggregation-basedsolutions require the availability of a goodamount of spectrum and do not really scaledown to small system bandwidths because of theinefficiency associated with such transmissionbandwidths, including peak rate limitations forthe UE that is not carrier-aggregation-capable.The cost of carrier-aggregation-capable UE andlow-power nodes could become an issue as well.

CO-CHANNEL DEPLOYMENTA third option to deploy heterogeneous net-works is illustrated in Fig. 4c, and we refer to itas co-channel deployment. In this scenario, allthe network nodes are deployed in the same fre-quency layer to avoid bandwidth segmentation,to be applicable for any system bandwidth notnecessarily relying on high spectrum availability,and to avoid relying on carrier aggregation sup-port at the terminals.

As discussed above, deploying low-powernodes at the same frequency layer as the (high-power) macrocells presents severe interferenceproblems in the case of closed femtos. Also, forthe case of open access nodes (open femtos,picocells, and RNs), the coverage of the low-power nodes is overshadowed by the transmis-sions of the high-power nodes; therefore, thebenefits of the introduction of low -power nodesfor the overall system capacity are somewhatlimited.

It is therefore necessary to consider interfer-ence coordination techniques that can solvethese problems. In summary, the following isdesired:• Reduction of closed femto interference to the

macro layer• Maximizing the performance benefits from the

introduction of open access low-power nodesTo enable efficient support of co-channel deploy-ment of heterogeneous networks, an interfer-ence management scheme should be able toadapt to different traffic loads and differentnumbers of low-power nodes at various geo-graphical areas. In this article, we focus on co-channel deployments, and the next section goesinto the details of interference managementtechniques that enable efficient co-channeldeployment of heterogeneous networks.

Co-channel deployment of heterogeneousnetworks is attractive for many reasons. It is theonly feasible solution when spectrum is limited(20 MHz or less), and the peak data rate perfor-mance of legacy Rel-8 UE does not need to beimpacted. It is suitable for deployment of low-cost single-carrier low-power base stations, anddoes not require higher-cost carrier-aggregation-capable UE.

INTERFERENCE MANAGEMENT FORCO-CHANNEL DEPLOYMENTS

The role of resource partitioning for interfer-ence management is critical in co-channeldeployments. But, as seen below, to reach its fullpotential, resource partitioning needs to bepaired with interference-cancellation-capable

UE. Even though the focus of the article is onthe DL, we should note that the same partition-ing scheme applies to uplink as well since uplinkdata transmissions are slaved to the downlinkcontrol channel.

PICO/RELAY DEPLOYMENTSIn this section, we focus on the case with openaccess low-power nodes. Throughout the text werefer to low-power nodes as pico base stations. Itshould be noted, however, that the same mecha-nism can be applied in the case of relay base sta-tions, where the only caveat is that the X2backhaul link between the macro and relay basestations is over the LTE air interface. Similartechniques can be applied when closed femto-cells are deployed, but as discussed earlier, dueto the lack of an X2 interface, only a staticOAM-based solution is feasible.

Adaptive Resource Partitioning — Adaptive interfer-ence management in LTE Rel-10 is enabledthrough X2 backhaul coordination of resourcesused for scheduling data traffic. The granularityof the negotiated resource is a single subframe.The main motivation behind resource partition-ing is to enable cell range expansion through cellbiasing. In a typical case, cell range expansion isenabled to improve system capacity, and a cellbias is applied to low-power nodes. The biasvalue refers to a threshold that triggers handoverbetween two cells. A positive bias means thatUE will be handed over to a picocell as soon asthe difference in the signal strength from themacro- and picocells drops below a bias value.The high-power node (macro base station)informs the low-power node (pico base station)of which resources will be utilized for schedulingmacro UE and which subframes would remainunutilized (almost blank subframes). This is howlow-power nodes are made aware of the interfer-ence pattern from the high-power base stationand can therefore schedule UE in the cellextended areas on subframes protected fromhigh-power interference, that is, subframes cor-responding to almost blank subframes at thehigh-power base station.

Interference coordination is performed bymeans of a bitmap, where each bit in the bitmapis mapped to a single subframe. The size of thebitmap is 40 bits, inferring that the interferencepattern repeats itself after 40 ms. Based on thedata traffic demand, the pattern can change asoften as every 40 ms. The communicationbetween nodes is peer to peer; that is, there isno formal master-slave relationship. However,the node creating dominant interference condi-tions effectively controls which resources can beused to serve UE in the cell range extensionarea. In a typical scenario, a macro base stationwould effectively be a master, and a pico basestation is a slave since cell range expansionwould most frequently be established for picobase stations. The scenario is illustrated in Fig.5. Note that the cell range expansion may bedesirable for a macro base station as well. Acommon use case for cell range expansion at themacro base station is to reduce the number ofhandovers. In a scenario with a large number ofhigh-mobility users, the number of handovers in

The role of resourcepartitioning for

interference management is

critical in co-channel

deployments. But toreach its full

potential, resourcepartitioning needs to

be paired with interferencecancellation- capable UE.



the network can become a problem, particularlywhen in the deployment of pico base stations thecells become small. The same resource partition-ing schemes can then be utilized to allow high-mobility macro UE to remain attached to amacrocell while deep inside coverage of a pico-cell. In this case, the pico base station needs torestrict scheduling data traffic on someresources, allowing coverage for high-mobilitymacro UE.

Resource Restricted Measurements — Subframeresource partitioning creates an interference pat-tern that requires a new radio resource measure-ment paradigm. As dominant interference couldpotentially significantly vary from one subframeto the other, in order to ensure radio resourcemanagement measurement accuracy, it is neces-sary to restrict measurements to a desired set ofsubframes. Since the measurement configurationin LTE is not intended to change frequently, fre-quent changes in the resource partitioning con-figuration can negatively impact themeasurement procedure.

For that reason, it is desirable to defineanchor interference-protected resources that canbe utilized for radio resource measurement, andthe 40-bit bitmap includes indications of whichsubframes have static protection from interfer-ence and therefore are suitable for measure-ments. As discussed above, the remainingresources can change more often. The moreresources have static interference protection, themore accurate the radio resource managementmeasurements. However, the more resourceshave static interference protection, the moreconstrained the adaptive resource partitioningalgorithm.

CLOSED FEMTO DEPLOYMENTS

Due to the lack of an X2 interface, only a staticOAM-based solution is feasible when closedfemto nodes are deployed on the same frequen-cy as the macro network. The same principle ofextending coverage of one cell into the area cov-ered by the other can be achieved, as shown inFig. 6. In this case, however, it is the macrocellservice that needs to be extended into an areacovered by the closed femtocell. Due to the stat-ic nature of the solution, however, resource par-titioning cannot be adapted to the instantaneouscharacteristics of actual data traffic. Neverthe-less, “time-of-day” type adaptation is feasiblethrough the use of OAM.

FUTURE TRENDS: ADVANCED UE RECEIVERCo-channel deployments are of interest due tocost effectiveness and high spectral efficiency.The resource partitioning can effectively miti-gate interference from the data channel. Howev-er, interference from the acquisition channelsand CRS signal remains, since these signals arenecessary to be transmitted for the backwardcompatibility reason. Subframe time shifting canbe utilized in FDD systems to avoid collision ofthe acquisition channels between eNBs of differ-ent power classes that require partitioning, but itcannot be used in TDD systems. There is no net-work solution for CRS interference mitigationfor either FDD or TDD systems. Without addi-tional interference mitigation of the acquisitionchannels and CRS, only limited medium biasvalues (roughly 6–10 dB) can be supported,which limits the potential for cell range expan-sion.

Interference mitigation for the acquisitionchannels is crucial for cell range expansion. Itensures that UEs is able to detect and acquire aweek cell and then measure and feedback themeasurement report to the network, which is aprerequisite for handover and cell range expan-sion.

CRS interference mitigation has an importantrole in system performance. Strong CRS inter-ference, even though present only on a relativelysmall fraction of resource elements, can signifi-cantly degrade turbo code performance and theoverall signal-to-interference-plus-noise radio(SINR); therefore, the potential gains of cellrange expansion can be severely reduced. Giventhat acquisition channels and CRS are all broad-cast at full power targeting UE at the cell edge,a robust UE solution is feasible.

INTERFERENCE CANCELLATION RECEIVERThe main rationale for the UE solution for theacquisition signals and CRS interference mitiga-tion is that strong interference can reliably beestimated and subtracted so that in the end itdoes not represent significant interference at all.Acquisition channels are transmitted at thesame location in all cells, which means that theacquisition channels interference consists onlyof the acquisition channels from the neighbor-ing interfering cells. This structure lends itself toa design of an interference canceller, illustratedin Fig. 7a. A UE receiver first decodes the

Figure 5. Backhaul-based interference management, illustrated on amacro/pico example for FDD systems. The same mechanism can be appliedfor the macro/relay scenario.

X2

X2 backhaul link

Semi-statically assigned regular subframe or almost blanksubframe

0

Statically assigned almost blank subframe

Semi-statically assigned almost blank subframe

1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9

Interference bitmap transmitted over X2 backhaul link

Pico eNBMacro eNB

Cell rangeexpansion

UEs served on resources corresponding to almostblank subframes at macro eNBs.



strongest signal, performs channel gain estima-tion toward the interfering cell, cancels theinterfering signal, and continues the procedureuntil acquisition channels of the serving cell areacquired.

A similar procedure can be performed toremove CRS interference. The difference isthat CRS tones may also interfere with datatones since interference between CRS tonesbetween high- and low-power nodes can beavoided by CRS tone shifting. As illustrated inFig. 7b, the procedure is similar. If strong CRSinterference is detected, and after the channelgain toward the interfering cell is estimated, theCRS signal can be cancelled. The procedure isrepeated until all interfering signals are sub-tracted.

SYSTEM PERFORMANCE(MACRO/PICO DEPLOYMENTS)

An illustration of a typical performance gainthat can be achieved with the heterogeneousnetworks with pico eNB deployments is shownin two scenarios in Figs. 8 and 9. The first sce-nario considers four picos randomly placed inthe coverage of a macrocell and random loca-tion of users. The second scenario is a hotspotcase with two picos per macrocell, randomlyplaced, and 2/3 of users located within 40 mradius of the two picos. CRS interference can-cellation is assumed for both scenarios whenresource partitioning is considered. Userarrivals are modeled as a Poisson random pro-cess, and each user downloads a 1 MByte file.Biasing of 18 dB toward picocells for theresource partitioning case and 0 dB biasingwhen resource partitioning is not utilized areconsidered. Note that the results are not appli-cable for half duplex inband relays, since insuch a scenario some resources would need tobe reserved for backhaul relay communicationswith the DeNB. As shown in Fig. 8, for the firstscenario and typical macrocell loading of 50percent, random deployment of 4 low-powernodes can offer only a relatively modest 15 per-cent gain in terms of the served cell through-put. For the higher loads, close to 30 percentgain is achievable, and resource partitioningcan provide an additional 44–50 percent

improvement, bringing the total gain to above90 percent for the loaded networks. Theseresults show that in order to exploit the fullpotential of the deployment of picocells, it isnecessary to utilize resource partitioning tech-niques. For a wireless network operator utiliz-ing a byte-based charging policy, this gaindirectly translates into increased revenue with-out degrading user experience. The results forthe hotspot scenario are illustrated in Fig. 9showing even larger gains than the random userlocations scenario. The total gain for the loadednetworks is over 150 percent, even though onlytwo picos are deployed per macrocell. Thesegains, however, can severely be reduced if CRSinterference is not removed. The simulationresults shown in Table 1 illustrate the impor-

Figure 6. Illustration of the static interference management technique for FDDsystems, applicable to the macro/closed femto scenario.

Macro eNB serviced enabled for macro UEs on protected subframes

Coverage hole for macro UEs on unprotected subframes

Macro eNB

Closed femto eNB subframe configuration example

Macro eNB

Closed HeNB

Closed HeNB

Subframes utilized by open access macro eNB only

Subframes shared by open access macro eNBs and closed femto eNBs

Statically assigned regular subframe or almost blank subframe

0

Statically assigned almost blank subframe

123456789012345678901234567890123456789

Figure 7. Interference cancellation of acquisition signals and CRS.

(a)

Desired celldetected

Detection processcompleted

Yes

No

Remove acquisitionsignals interference

(b)

Strong CRSinterference

detected

Decode datachannel

Yes

No

Remove CRSinterference


tance of CRS interference cancellation. For theconsidered scenario, it is estimated that evenwhen biasing optimization is taken into account,roughly 3/4 of the gains disappear if a CRSinterference canceller is not utilized.

It should be noted that the resource parti-tioning techniques can increase delay and jitter.However, the impact is practically negligibledue to short subframe duration of 1 ms. Forexample, in a scenario where 50 percent ofresources are unavailable due to resource parti-tioning, the partition can be realized with analternating pattern where even subframes areavailable for scheduling and odd ones are not.

This partitioning pattern does not add morethan 1 ms of delay and jitter; hence, there isvirtually no impact on user perception even forhighly delay-intolerant traffic, such as gamingor VoIP.

CONCLUSIONS

Heterogeneous deployment is seen as a prag-matic and cost-effective way to significantlyenhance the capacity of LTE cellular networks.Macro nodes are used to ensure broad coverage,and low-power nodes may be located close toplaces with increased demand for data. Interfer-ence management represents the first crucialcomponent of this strategy as severe interferencefrom the macro nodes significantly limits theoffloading potential of low-power nodes. Cellrange expansion enabled through resource parti-tioning is necessary as it creates the potential fortraffic load balancing, which improves the trunk-ing efficiency of the network. The interferencecancellation receiver at the UE ensures that cellacquisition channels of weak cells can be detect-ed and CRS interference removed, fully exploit-ing the potential of heterogeneous networkdeployments. All three components are neces-sary to achieve the full potential of heteroge-neous deployments.

REFERENCES[1] M. Yavuz et al., “Interference Management and Perfor-

mance Analysis of UMTS/HSPA+ Femtocells,” IEEE Com-mun. Mag., 2009.

[2] V. Chandrasekhar, J. G. Andrews, and A. Gatherer, “Femto-cell Networks: A Survey,” IEEE Commun. Mag., 2008.

[3] C. Patel, M. Yavuz, and S. Nanda, “Femtocells [IndustryPerspectives],” IEEE Wireless Commun. Mag., Oct. 2010.

[4] M. Coupechoux, J-M. Kelif, and P. Godlewski, “NetworkControlled Joint Radio Resource Management for Het-erogeneous Networks,” IEEE VTC Spring 2008.

[5] W. Song, H. Jiang, and W. Zhuang, “Performance Anal-ysis of the WLAN-First Scheme in Cellular/WLAN Inter-working,” IEEE Trans. Wireless Commun., May 2007.

[6] H. Claussen, L.T.W. Ho, and L. G. Samuel, “FinancialAnalysis of a Pico-Cellular Home Network Deployment,”IEEE ICC ‘07.

[7] A. Bou Saleh et al., “Comparison of Relay and Pico eNBDeployments in LTE-Advanced,” IEEE VTC 2009.

[8] T. Nihtila and V. Haikola, “HSDPA Performance withDual Stream MIMO in a Combined Macro-Femto CellNetwork,” IEEE VTC 2010.

[9] H. R. Karimi et al., “Evolution Towards Dynamic Spec-trum Sharing in Mobile Communications,” IEEE PIMRC2006.

[10] J. Gora and T. E. Kolding, “Deployment Aspects of 3GFemtocells,” IEEE PIMRC 2009.

[11] H. A. Mahmoud and I. Guvenc, “A Comparative Studyof Different Deployment Modes for Femtocell Net-works,” IEEE PIMRC 2009.

[12] Han et al., “Optimization of Femtocell Network Configura-tion Under Interference Constraints,” WiOPT 2009.

[13] Y. Tokgoz et al., “Uplink Interference Management forHSPA+ and 1xEVDO Femtocells,” IEEE GLOBECOM 2009.

[14] G. Boudreau et al., “Interference Coordination and Cancel-lation for 4G Networks,” IEEE Commun. Mag., 2009.

[15] A. Khandekar et al., “LTE-Advanced: HeterogeneousNetworks,” European Wireless Conf. 2010.

[16] O. Simeone, E. Erkip, and S. Shamai , “Robust Trans-mission and Interference Management For Femtocellswith Unreliable Network Access,” IEEE JSAC, Dec. 2010.

[17] S. Annapureddy et al., “http://www.ieee-ctw.org/2010/mon/Gorokhov.pdf,” 2010 IEEE Commun. TheoryWksp., Cancun, Mexico, 2010.

[18] 3GPP Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio AccessNetwork (E-UTRAN);Overall Description; Stage 2(Release 10).

[19] 3GPP TR 36.814, “E-UTRA; Further Advancements forE-UTRA Physical Layer Aspects.”

Figure 8. Served cell throughput gains with 4 picos/macrocell — scenario 1:random UE distribution and 18 dB biasing toward picocell for resource parti-tioning.


Macro cell resource utilization [%]30

5

Serv

ed t

hrou

ghpu

t [M

bps]

0

10

15

20

25

30

35

40

20 40

Performance summary

Served cell throughput(gain vs. macro only/gain ofresource partitioning)

Macro only 4 picos, RP4 picos, no RP

50% Macro utilization 12.3 Mbps 20.5 Mbps(1.67x/1.44x)

14.2 Mbps(1.15x/1x)


17.7 Mbps(1.27x/1x)

50

Served throughput for random UE distribution

60 70 80 90

Macro only4 picos, no resource partitioning4 picos, resource partitioning

Figure 9. Served cell throughput gains with 2 picos/macrocell — scenario 2:hotspot and 18 dB biasing toward picocell for resource partitioning.

Macro cell resource utilization [%]

5

Serv

ed t

hrou

ghpu

t [M

bps]

0

10

15

20

25

30

35

40

30 40

Performance summary

Served cell throughput(gain vs. macro only/gain ofresource partitioning)

Macro only 2 picos, RP2 picos, no RP


16.9 Mbps(1.37x/1x)


21.7 Mbps(1.56x/1x)

50

Served throughput for the hotspot scenario

60 70 80 90

Macro only2 picos, no resource partitioning2 picos, resource partitioning



BIOGRAPHIESALEKSANDAR D. DAMNJANOVIC ([email protected])received his Diploma in electrical engineering from the Uni-versity of Nis, Serbia, in 1994 and his D.Sc. degree in elec-trical engineering from George Washington University in2000. He joined Ericsson Wireless Communications Inc.,San Diego, California, in 2000, where he worked oncdma2000 base station controller development. In 2003 hejoined Qualcomm Inc., San Diego, where he worked on 3Gcellular standards and, since 2005, standardization andprototyping of 3GPP LTE and LTE-Advanced systems lead-ing MAC design efforts. He is a co-author of a book, Thecdma2000 System for Mobile Communications.

JUAN I. MONTOJO has been with Qualcomm Inc. since 1997where he has worked as a systems engineer on differentwireless systems. He has been heavily involved in thedesign and specification of the physical layer of LTE and isthe editor of one of the PHY layer specifications (3GPP TS36.212). He holds engineering degrees from the UniversitatPolitècnica de Catalunya (UPC), Barcelona, Spain, and Insti-tut Eurecom, Sophia-Antipolis, France. He also holds anM.S. from the University of Southern California (USC) and aPh.D. from the University of California San Diego (UCSD),both in electrical engineering.

YONGBIN WEI received B.E. and M.S. degrees from the Uni-versity of Science and Technology of China, and a Ph.D.degree from Purdue University, all in electrical engineering.Since he joined Qualcomm in 2000, he has worked on sys-tem design, evaluation, and standardization for cdma2000and 1xEV-DO as well as base station ASIC chip develop-ment. Since 2006 he has been leading the system effortsfor system design, standardization, and prototype of LTEand LTE-Advanced. He has 52 U.S. issued patents andmany more pending.

TINGFANG JI received his B.S. degree in electronic engineer-ing from Tsinghua University, China, in 1995, his M.S andPh.D degrees, both in electrical engineering, from the Uni-versity of Toledo, Ohio, and the University of Michigan,Ann Arbor in 1997 and 2001, respectively. From 2001 to2003 he was a member of technical staff in the Bell Labo-ratories Advanced Technology Department involved in cel-lular technologies and ultra wideband research. Since 2003he has been with the Corporate R&D division of QualcommInc., where he is currently a senior staff engineer and man-ager. From 2003 to 2007 he was one of the key technicalcontributors to the development of wideband OFDMAtechnologies that evolved into IEEE 802.20 and 3GPP2UMB standards. Since 2007 he has been actively contribut-ing to the development and standardization of 3GPP LTE-Advanced technology. His research interests include thePHY/MAC/RF aspects of heterogeneous networks, carrieraggregation, network MIMO, and relay networks.

MADHAVAN VAJAPEYAM received his B.S. degree in electricalengineering from the Universidade Federal da Paraiba,Campina Grande, Brazil, in 2000, and M.S. and Ph.D.

degrees in electrical engineering from USC, Los Angeles, in2002 and 2007, respectively. Since 2007 he has been withthe Corporate R&D division of Qualcomm Inc., where he isactively involved in the development and standardizationof next-generation 3GPP LTE and LTE-Advanced technolo-gies. His research interests include algorithms for physicaland MAC layer design of wideband OFDMA systems, inter-ference management in heterogeneous networks, andMIMO communications.

TAO LUO received his Ph.D. in EECS from the University ofToronto, Canada, in June 2001. From July 2001 to October2004, he was with Wireless Design Center of Texas Instru-ments, San Diego, and worked on a WCDMA related pro-ject and was a key contributor to the WCDMA basebandand RF chipset development. Since November 2004 he hasbeen with the Standards group and later Corporate R&D atQualcomm. He has actively contributed to the standardiza-tion of HSUPA, LTE, and, more recently, LTE-Advanced.Since 2008 he has been leading the LTE Rel-8 and thenLTE-A UE prototyping for HetNet. He has over 100 issuedor pending U.S. patents in the area of wireless communica-tions.

TAESANG YOO received his B.S. degree (Hons.) in electricalengineering from Seoul National University, Korea, in 1998,and M.S. and Ph.D. degrees in electrical engineering fromStanford University in 2003 and 2007. Since 2006 he hasbeen with Corporate R&D, Qualcomm, Inc., where he iscurrently working on 3GPP LTE and LTE-Advanced wirelesssystems. His research interests include multiple antennasystems, communications and information theory, hetero-geneous cellular networks, and various design and perfor-mance aspects of LTE/LTE-A systems and standards.

OSOK SONG received his B.S. and Ph.D. in electrical engi-neering from Seoul National University. He joined SamsungElectronics in 2003 and has been involved in 3GPP stan-dardization. He served as vice chairman of 3GPP SA2, theworking group responsible for 3GPP network architecture,and rapporteur (editor) of several work items. In April 2007he joined Qualcomm Inc., and worked on LTE and LTE-Advanced network protocol research and standardization.He is an active participant in 3GPP RAN3, the workinggroup responsible for 3GPP radio network architecture andprotocols.

DURGA MALLADI received his B.Tech (1993) degree fromthe Indian Institute of Technology, Madras, and M.S.(1995) and Ph.D. (1998) degrees from the University ofCalifornia, Los Angeles. He joined Qualcomm in 1998and worked on system design and implementation ofGlobalstar from 1998 to 2001. From 2001 to 2004 heled the system design and standardization of 3GPPHSDPA and HSUPA. From 2004 onward he has led theresearch and development activities related to LTE andLTE Advanced, including system design, standardization,imp lementation, and testing. His research interestsinclude MIMO, signal processing techniques, communica-tion theory, and cognitive radio.

Table 1. Impact of CRS interference on system performance (random user locations scenario).

Impact of CRS interference

Served cell throughput at 75%loading

4 picos, RP w/o CRS IC (10 dBbias) 4 picos, RP w/ CRS IC (18 dB bias)

Gain compared to no RP 1.1x 1.44x

It should be notedthat the resource

partitioning techniques can

increase delay andjitter. However, theimpact is practically

negligible due toshort subframe

duration of 1 ms.


A Survey on 3GPP Hetrogeneous Networks

Technology

Transcript of A Survey on 3GPP Hetrogeneous Networks