Transcript
Heterogeneou
Qualcomm CDMA Tec
Abstract 3GPP Long-term Evolution (LTEto use new and wider spectrum and complemwith higher data rates, lower latency andarchitecture. To further improve the broadbain an ubiquitous and cost-effective mannerworking on various aspects of LTE-Advancedperformance is quickly approaching theoretienhancements and LTE, the next performanfrom an evolved network topology. This paperfor an alternative deployment model anheterogeneous networks. The concept of LTEheterogeneous networks is about improving per unit area. Using a mix of macro, pico, femheterogeneous networks enable flexible and lowand provide a uniform broadband experiencperformance of these networks, advanceddescribed, which are needed to manage and cand deliver the full benefits of such networksinclude cell range expansion, adaptive intercoordination and interference cancellation rece
Keywords Heterogeneous Networks, LTE-A
I. INTRODUCTION
Data traffic demand in cellular networks toexponentially. To achieve further performanin LTE Advanced, 3GPP has been working oof LTE including higher order MIMO (mcarrier aggregation (multiple componenheterogeneous networks (picos, femtos, rsince improvements in spectral efficienapproaching its theoretical limits in 3G aenhancements are only possible by increasindensity of nodes.
Current wireless cellular networks are typihomogeneous networks using a macro-process. A homogeneous cellular system is stations in a planned layout, in which all the similar transmit power levels, antenna patterfloors and similar backhaul connectivity tonetwork. Moreover, all base stations offer uto user terminals in the network. The locatbase stations are carefully chosen through nand the base stations are properly configuredcoverage and control the interference betwethe traffic demand grows and the RF environhomogeneous network relies on cell splittcarriers to overcome capacity and link budg
us Networks in LTE-AdInvited Paper
Stefan Brueck
chnologies GmbH, Nordostpark 89, 90411 Nuremberg
sbrueck@qualcomm.com
E) allows operators ments 3G networks d a flat, IP-based and user experience r, 3GPP has been d. Since radio link ical limits with 3G nce leap will come r discusses the need d topology using E-Advanced based spectral efficiency
mto and relay cells, w-cost deployments ce. To enhance the d techniques are
control interference s. These techniques r cell interference eivers.
Advanced
oday is increasing nce improvements on various aspects
multiple antennas), nt carriers), and relays). However, cy per link are and LTE, further ng the deployment
ically deployed as -centric planning a network of base base stations have rns, receiver noise the (packet) data
unrestricted access tions of the macro network planning, d to maximize the
een each other. As nment changes, the ting or additional get limitations and
maintain uniform user experieprocess is complex and iteratfor macro base stations with tdifficult in dense urban aredeployment model is needebroadband user experience in way.
Wireless cellular systems haveisolated system (with just onoptimal performance, as deterlimits. Future gains of wirelemore from advanced netwonetworks, utilizing a diverse deployed to improve the specheterogeneous cellular systemcellular system consists of remacro base stations that typica(~ 5W 100W), overlaid withand relay cells, which transmlevels (~ 100mW 2W).
Figure 1: Heterogeneous netwofemto and
The low power cells can be dholes in the macro-only systemspots. Usually, three typesdistinguished:
• Pico nodes are reguldifference of havingtraditional macro cellwith omni-directionaindoors and outdoors i
• Femto nodes are (unplanned) network with a backhaul facildigital subscriber li
dvanced
g, Germany
ence. However, this deployment tive. Moreover, site acquisition owers becomes more and more eas. Hence, a more flexible ed for operators to improve a ubiquitous and cost effective
e evolved to the point where an ne base station) achieves near mined by information theoretic ess networks will be obtained ork topology. Heterogeneous
set of base stations, can be tral efficiency per unit area. A
m is depicted in Figure 1. This egular (planned) placement of
ally transmit at high power level h several pico cells, femto cells
mit at substantially lower power
ork utilizing a mix of macro, pico, d relay cells
deployed to eliminate coverage m and improve capacity in hot
s of low power nodes are
lar base stations with the only g lower transmit power than s. They are, typically, equipped
al antennas and are deployed in a planned manner. typically consumer deployed
nodes for indoor application itated by the consumer s home ne (DSL) or cable modem.
2011 8th International Symposium on Wireless Communication Systems, Aachen
978-1-61284-402-2/11/$26.00 ©2011 IEEE 171
Typically, omni-directional antenDepending on whether the femto baccess to all terminals or to a terminals only, femtos are classiclosed (CSG femto cells).
• Relay nodes are network nodes backhaul. The backhaul, whicconnection of the relay to the netand uses the air interface resourcesystem. Relays are typically directional antennas in the backdirectional antennas in the access lin
Due to their lower transmit power and smapico/femto/relay base stations can offacquisitions. Relay nodes offer additionbackhaul where a wireline backhaul is ueconomical.
Deployment options for heterogeneous netwin detail in [1] and the references thereiinclude multicarrier deployment, carrier agchannel deployments. In the following discon co-channel deployments. In such a scenarnodes are deployed in the same frequency deployments of heterogeneous networks issolution since it is applicable for any systenecessarily relying on the availability of a lait does not rely on the support of carrier aterminal. On the other side, severe interfarise in case of co-channel deployments as the next sections. To overcome these harseveral key design features of heterogenediscussed in the next sections.
II. KEY DESIGN FEATURES OF HETE
NETWORKS
A. Cell Range Expansion
A pico base station is characterized by a stransmit power as compared to a macro basof the large disparity between the transmimacro and pico base stations, the downlink cbase station is much smaller than that of a mThis is not the case for uplink, where the strereceived from a user terminal depends on thepower, which is the same for all uplinks frodifferent base stations. Hence, the uplink cbase stations is similar.
If serving cell selection is predominantly bsignal strength, as it is the case in LTE Rel-8pico base stations will be greatly diminscenario, the larger coverage of high powlimits the benefits of cell splitting by attrterminals towards macro base stations based without having enough base station resourserve these user terminals. Resources of stations may remain underutilized.
nnas are applied. base station allow
restricted set of ified as open or
without a wired ch provides the twork, is wireless es of the wireless
equipped with khaul and omni-nk.
aller physical size, fer flexible site nal flexibility in navailable or not
works are discussed in. These options gregation and co-cussions we focus rio, all the network layer. Co-channel a very attractive em bandwidth not arge spectrum and aggregation at the ference challenges
it will be seen in rsh RF conditions ous networks are
EROGENEOUS
ubstantially lower e station. Because t power levels of coverage of a pico
macro base station. ength of the signal e terminal transmit om the terminal to overage of all the
ased on downlink 8, the usefulness of ished. In such a wer base stations racting most user on signal strength
rces to efficiently low power base
The difference between the lobase stations can result in an and uneven user experiences anetwork. Therefore, from thcapacity, it is desirable to balanpico base stations by expandstations and subsequently increconcept is referred to as ceillustrated in Figure 2 (a) and (bof the pico cell is expandedassociate with the pico cell. macro cell and a more balancvarious nodes are achieved.
(a)
(b)
Figure 2: (a) Limited footprint signal, (b) Increased footprint o
A simple example of two catestations can be used to demonrange expansion. Figure 3 association with and without mixed macro and pico dedistribution (configuration 1 an
Figure 3: Pico cell user associatioexpa
oadings of high and low power unfair distribution of data rates among the user terminals in the he point of view of network nce the load between macro and ing the coverage of pico base
ease the cell splitting gains. This ell range expansion, which is b). In Figure 2 (b) the cell range d to allow more terminals to Traffic is offloaded from the
ced load distribution across the
of pico cells due to strong macro f pico cells with range expansion
egories of macro and pico base nstrate potential gains from cell
shows the statistics of user t cell range expansion for the eployment for uniform user nd 4b in [2]).
on statistics with and without range ansion
172
The range expansion here is achieved byassociation based on minimum path lossmaximum downlink signal strength. As it figure, cell range expansion allows manassociate with the pico cells and enabledistribution of air interface resources to eacis even more pronounced in hotspots layconfiguration 4b in [2], where users are clupico cells. Capacity gains can be achieved tthe resources allocated for low power bassufficient coverage is provided by high powethe resources that are allocated to them. illustrated in more detail in one of the follow
B. Interference Management
In a co-channel deployment of a heterosevere interference situations may arise. Typ
• Macro-Pico deployments with termcell range expansion. In this case, not necessarily the strongest andbecome much less than zero dB.macro cell is a strong interferer, thvictim. This situation is illustrated in
• Terminals in close proximity to a fbut barred from accessing them. femto cell is a strong interferer, thvictim. This situation is illustrated in
Figure 4: Barred terminal in proximity to a femmacro base station
In order to overcome such harsh interferennecessary to consider interference coordinatican solve these problems. To enable efficiechannel deployments of heterogeneousinterference management scheme should bedifferent traffic loads and different numbers stations at various geographical areas.
As opposed to homogeneous networknetworks necessitate more coordinatiopartitioning across base stations to minterference. Principally, the resource paperformed in time domain, frequency dodomain. A spatial domain solution basedscheduling is investigated for CSG femto cgeneral, time domain partitioning has the advbetter adapt to user distribution and traffic loa very attractive method for spectrum consttime domain solution that enables reso
y performing cell s rather than on is seen from the
ny more users to es more equitable ch user. The effect youts as given by ustered around the through sharing of se stations, while er vase stations on
These gains are wing sections.
ogeneous network pical scenarios are
minals operating in the serving cell is
d the SINR may In this case the he pico cell is the n Figure 2 (b). femto base station
In this case the he macro cell is a n Figure 4.
mto cell served by a
nce situations it is ion techniques that ent support of co-s networks, an e able to adapt to of low power base
s, heterogeneous on via resource
manage inter cell rtitioning can be omain, or spatial d on coordinated
cells e.g. in [3]. In vantage that it can oad changes and is trained markets. A ource partitioning
through X2 backhaul coordinaRel-10 (enhanced Inter Cell Int[4]. Performance analysis partitioning can be found e.g. CSG femto cells. In eICIC 3Gnegotiated time domain resourc
In order to enable time domaicalled almost blank subframes LTE Rel-10. In such a subfreference signals (RS), synchrmessages are transmitted compatibility to legacy terminaof a ABS subframe, where antenna ports are transmitted.elements are left idle that arecontrol channels PDCCH, PHIchannel PDSCH.
Figure 5: Almost Blank
Interference coordination betwand the victim cell is performeeach bit in the bitmap is mapindicates an ABS subframe. Thinferring that the interference pBased on the data traffic demoften as every 40ms. The compeer to peer, i.e. there is However, the cell creating controls which resources can serve terminals in harsh interfe
Figure 6 shows an exampleinterference management scen(strong interfering cell) and a p
Figure 6: X2 backhaul based interfor macro/pic
ation is included in 3GPP LTE terference Coordination, eICIC) for time domain resource
in [5] for pico cells and [6] for GPP LTE the granularity of the ces is one subframe, i.e. 1 ms.
in interference coordination, so (ABS) have been introduced in
frame only Rel-8 cell specific ronization signals and broadcast
to enable full backward als. Figure 5 shows an example only reference signals of two It is seen that many resource e usually used to transmit the ICH and PCFICH and the data
Subframe (Non-MBSFN)
ween the strong interfering cell ed by means of a bitmap, where pped to a single subframe and he size of the bitmap is 40 bits, pattern repeats itself after 40ms.
mand, the pattern can change as mmunication between the cells is
no master slave relationship. strong interference effectively be used by the victim cell to
erence conditions.
e of an X2 backhaul based nario in case of a macro cell
pico cell (victim cell).
rference management, illustrated on co cells (FDD)
173
In this scenario ABS subframes are statically and semi-statically allocated by the dominant macro interferer and indicated over the X2 backhaul to the victim pico cell. The pico cell serves its terminals in the expanded cell range on resources corresponding to a ABS subframe of the macro cell. This requires subframe synchronous operation of the two cells.
It is also obvious from Figure 6 that the interference caused by the macro cell can significantly vary from subframe to subframe depending on whether a subframe in the macro cell is a ABS subframe or a regular subframe. In order to ensure measurement accuracy for radio resource management purposes, it is necessary to restrict measurements to a desired set of subframes. Since such a measurement configuration is not intended to change frequently, certain ABS subframes should not change over time to provide static protection from the macro interference (statically assigned ABS). The remaining time resources can change more often (semi-statically assigned ABS) to adapt to the changes in the traffic load. The more resources have static interference protection, the more accurate the radio resource measurements. However, the more resources have static protection, the more constraint the adaptive resource partitioning algorithms. Gains of adaptive resource partitioning are shown in a following section. Adapting the resource partition over time requires communication between macro and pico cell over the X2 interface. The required messages of the X2 application protocol have been defined in LTE Rel-10 [7].
The considerations on adaptive resource partitioning hold for macro pico scenarios. The same principle can be applied in case of closed femtos. In this case the femto is the strong interferer and the macro the victim, i.e. the femto cell assigns almost blank subframes. However, for closed femto cells only a static OAM based solution is feasible due to the lack of an X2 interface to macro cells. Resource partitioning can not be adapted to the instantaneous characteristics of the actual data traffic in case of closed femto cells.
C. Future Trends: Advanced Terminal Receivers
As outlined in the previous section, resource partitioning can effectively mitigate interference from data and control channels. However, synchronization and reference signals are still present in ABS subframes to ensure backward compatibility. These channels and signals create interference to the terminals of the victim cell.
Interference mitigation for the synchronization signals in the victim cell is crucial for cell range expansion. It ensures that the terminals are able to detect and acquire a weak cell and measure and feedback the measurement reports to the network, which is a prerequisite for handover and cell range expansion. Interference mitigation for the reference signals has also an important role for system performance. Strong interference from reference signals, even though present only on a small fraction of resource elements, can significantly degrade Turbo code performance. Without such interference mitigation the potential gains of cell range expansion can be severely reduced.
Since there is no solution at network side to mitigate interference for synchronization and reference signals that works equally well in FDD and TDD systems, a solution at terminal side by utilizing advanced receivers provides a feasible and effective approach.
The main rationale for a terminal based solution for the synchronization and reference signals interference mitigation is that strong interference from another cell can reliably be estimated and subtracted. An interference cancellation receiver for both types of signals is feasible and applicable. As an example Figure 7 shows the achievable throughput in a ABS subframe w/ and w/o RS interference cancellation (RS IC) in presence of a strong interferer that is 16 dB above background noise. The link throughputs are obtained by running adaptive modulation and coding targeting 10% BLER of the first HARQ transmission and for 6 PRBs randomly selected per subframe. The channel models are ETU3 and EVA3 in the serving and interfering cell, respectively. The transmission mode is open loop spatial multiplexing (TM3) for 2 Tx antennas that allows single and dual layer transmission.
Figure 7: PDSCH performance with RS IC for non-colliding RS
It is seen that performance w/o CRS IC is degraded significantly although less than 10% of the resource elements in a ABS subframe are interfered by reference signals from the interfering cell (see Figure 5). Further simulation results for various scenarios applying interference cancellation for synchronization signals (PSS/SSS) and the broadcast channel (PBCH) are provided e.g. in [8].
III. SYSTEM PERFORMANCE FOR MACRO PICO
DEPLOYMENTS
In this section typical performance gains are illustrated that can be achieved with heterogeneous networks with pico cell deployments in case of cell range expansion/resource partitioning.
A hotspot scenario with two pico cells per macro cell randomly placed, and 2/3 of the terminals located within 40m radius of the two pico cells is considered. Interference cancellation for reference signals (RS) is assumed when
�
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 320
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000
Serv ing c e l l C/I (dB)
Th
rou
gh
pu
t (K
bp
s)
N onc o llid ingRS -Tx Mode3
CellID0-Tx Mode3
CellID0-Tx Mode3-Cel lID121-Tx Mode3-16dB
CellID0-Tx Mode3-Cel lID121-Tx Mode3-IC-16dB
174
resource partitioning is enabled. Without resource partitioning (i.e. without ABS subframes) RS interference cancellation is less effective since interference from data and control channels of the neighbor cell is more severe. User arrivals are modeled following a Poisson random process. Each user downloads a file of size 1 MByte. In case of resource partitioning a cell range expansion factor of 18 dB has been considered, i.e. a terminal can still be served by a pico cell when the dominant macro interferer is 18 dB stronger than the pico cell. This corresponds to a SIR of -18 dB. In [8] it is shown that operation at such low SIR is feasible if advanced receivers are applied. Without resource partitioning/cell range expansion 0 dB cell range expansion factor is considered. This means that corresponding to the Rel-8 cell selection criterion, a terminal is served by the macro cell instead of by the pico cell as soon as the received power from the macro cell becomes stronger. The additional simulation assumptions correspond to configuration 4b in [2].
Figure 8 compares the performance of the following cases:
1) Macro cells only, no RS IC receiver at terminal side 2) Two additional pico cells per macro cell, no resource
partitioning, no RS IC receiver at terminal side, no cell range expansion
3) Two additional pico cells per macro cell, resource partitioning, RS IC receiver at terminal side enabling 18 dB cell range expansion
It is seen that two additional pico cells without resource partitioning increase the cell throughput by about 50% due to cell splitting. However, the gains remain limited, since only a relatively low number of users is served by the pico cells.
With resource partitioning the gains increase further up to 130% for a resource utilization of about 75% of the macro cell if the terminal supports RS IC.
Figure 8: Served cell throughput gains with 2 picos/macro Hotspot
scenario
However, these gains are severely reduced, if the receiver does not support RS IC as can be anticipated from the link level results shown in Figure 7.
The system level results in Figure 8 show that in order to exploit the full potential gains of the deployment of pico cells, it is necessary to utilize resource partitioning. For a wireless
operator utilizing a byte-based charging policy, the gain in Figure 8 directly translates into increased revenue without degrading user experience.
It should be noted that resource partitioning techniques can increase delay and jitter. However, due to the short subframe duration of 1ms there is virtually no impact due to resource partitioning on user perception even for highly delay intolerant traffic classes, such as gaming and VoIP.
IV. SUMMARY
It has been shown that heterogeneous networks are a cost-effective approach to significantly enhance the capacity of LTE cellular networks. In the deployment strategy of such a network macro cells are used to ensure broad coverage, and low power cells can be deployed in hotspot areas to provide additional capacity. It has been outlined in the paper that three design features are crucial for heterogeneous networks:
• Interference management as severe interference limits the association of terminals to low power cells.
• Cell range expansion through adaptive resource partitioning as it enables traffic load balancing between high and low power cells.
• Interference cancellation receiver in the terminal as it ensures that weak cells can be detected and remaining interference is removed.
All three components together are needed to exploit the full potential of heterogeneous networks.
ACKNOWLEDGEMENT
The author acknowledges the valuable input of Tingfang Ji, Yongbin Wei, M. Awais Amin, Jochen Giese, Aleksandar Damnjanovic, Juan Montojo, Tao Luo, Madhavan Vajapeyam, Taesang Yoo, Osok Song, Alan Barbieri and Durga Malladi.
REFERENCES [1] A. Damnjanovic, J. Montojo, Y. Wei, T. Ji, T. Luo, M. Vajapeyam, T.
Yoo, O. Song, D. Malladi, A Survey on 3GPP Heterogeneous Networks, Invited Paper, IEEE Wireless Communications Magazine, June 2011
[2] 3GPP TR 36.814, Further advancements of E-UTRA physical layer aspects, v9.0.0 March 2010
[3] J. Giese, M. A. Amin, S. Brueck, Application of Coordinated Beam Selection in Heterogeneous LTE-Advanced Networks, IEEE ISWCS, Aachen, Germany, November 2011
[4] 3GPP TS 36.300, Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E-UTRAN); Overall Description, v10.4.0, June 2011
[5] M. Vajapeyam, A. Damnjanovic, J. Montojo, T. Ji, Y. Wei, D. Malladi, Downlink FTP Performance of Heterogeneous Networks for LTE-Advanced, IEEE ICC, Kyoto, Japan, June 2011
[6] A. Barbieri, A. Damnjanovic, T. Ji, J. Montojo, Y. Wei, D. Malladi, LTE Femtocells: System Design and Analysis, IEEE VTC, Budapest, Hungary, May 2011
[7] 3GPP TS 36.423, X2 application protocol (X2AP), v10.2.0, June 2011
[8] R1-103560, Enabling communication in harsh interference scenarios, Qualcomm Incorporated, TSG-RAN WG1 #61bis, June 2010
�
�
��
��
��
��
��
��
��
�� �� �� � � �� ��
������
�����
�����
� ��
��������������������������������
����������� ��������� ��������� ��������� ��������������������������������� ������������������������������
175
top related