Content uploaded by Wenchi Cheng
Author content
All content in this area was uploaded by Wenchi Cheng on May 14, 2014
Content may be subject to copyright.
Energy Efficient Spectrum Allocation for Green
Radio in Two-tier Cellular Networks
Wenchi Cheng, Hailin Zhang, Liqiang Zhao and Yongzhao Li
State Key Laboratory of Integrated Services Networks
Xidian University
Xi’an, Shaanxi, China
Email: wccheng@mail.xidian.edu.cn
Abstract—Recently, Green Radio, which aims to reduce energy
consumption of information and communication technologies
(ICTs), has been concerned by telecommunication operators
and researchers. Cellular networks account for a rather large
share of energy consumption in ICTs. Macro-femto networks,
as a typical two-tier cellular network topology, have obvious
advantages in lowering cellular energy consumption. In this
paper, we firstly introduce a spatial two-tier model and two
new metrics for analyzing the cellular network. Secondly, we
analyze the downlink energy consumption. Thirdly, we provide
a novel energy efficient spectrum allocation strategy in marco-
femto cellular networks to approach the minimal downlink
energy consumption. Finally, simulation results show that this
new spectrum allocation strategy can improve energy efficiency
of cellular networks.
I. INTRODUCTION
Past ten years have witnessed the explosive growth in
the number of subscribers for mobile telephony. This un-
precedented growth has made information and communication
technologies (ICTs) to be a major contributor to overall
green house gas emissions [1]. For reducing the impact of
ICTs on environment, efforts to increase energy efficiency
have received more and more attention recently appearing
a new concept - Green Radio, which means reduing the
CO2emissions in ICTs. Over 80% of the power in mobile
communications is consumed in the radio access network,
especially cost in base stations [2]. So an energy efficient
cellular access network will make mobile communications
easy to approach ”green” ICTs.
Actually, there are already a lot of energy efficient strategies
in cellular access networks. However, these strategies mainly
focus on smart ends with uplink energy constrains, limited
battery life or just transmission energy consumption [3-4].
Obviously, these strategies do not fit on Green Radio because
the energy consumption of ends is not the most in the cell,
even not the second and the third. In a cell, the BS costs the
most energy and all operating energy consumption of the BS
should be considered.
For getting an energy efficient cellular access network, we
must not sacrifice the already obtained system capacity or at
least satisfy users’ requiring for spectrum efficiency. For a
wireless link, the surest way to increase the system capacity
and reduce the system radiated energy consumption is by
getting the transmitter and receiver closer to each other, which
creates higher-quality links, more spectrum reuse and most
important - higher energy efficiency. Moreover, we must take
the whole operating energy consumption into consideration.
Two-tier cellular networks, as effective access networks
for high spectrum efficiency, have been developed for many
years. Previous works mainly aim at only spectrum efficiency
and ignore the energy consumption of the access network.
However, due to the shorter coverage provided by the second
tier cell such as microcells, picocells, femtocells, the two-tier
network is potential to be energy efficient.
Femtocell networks [5], as low-cost and low-power two-
tier cellular access networks, have been proposed as a so-
lution to the poor indoor coverage problem. But until now,
operators and researches just focus on the spectrum efficiency
improvement by adding femtocell access points (FAPs) into a
macrocell, and do not give enough attention to the potential
energy saving coming with FAPs. It is very dangerous because
it will push femtocell networks to be power hungry networks.
For avoiding this, we must consider the energy consumption
in femtocell networks.
In this paper, section II gives a spatial macro-femto two-tier
cellular network model with spectrum and energy efficiency
metrics. Section III shows the explanation of the problem
– why energy efficiency designing backs to downlink, and
discusses the energy consumption of the macro-femto cellular
networks, resulting in a convenient model to calculate the
energy consumption. In section IV, a novel spectrum allocation
strategy for energy efficiency two-tier cellular networks is
discussed. Section V gives simulation results.
II. SYSTEM MODEL
In a two-tier cellular network, the first tier and second
tier are macrocells and femtocells respectively. The central
macrocell is a hexagonal region Am=3
√3R2
c/2with a
central BS, which height is H. Two rings of interfering
macrocells are around the central macrocell. Every macrocell
has its theoretic spatial coverage S=AmH/3=√3HR2
c/2.
The central macrocell is overlaid with femtocell access
points (FAPs) of radius Rfand height hf, which are randomly
distributed on R3according to a homogeneous Spatial Poisson
Point Process (SPPP) Ωfwith intensity λf. The mean number
of femtocells per cell site is readily obtained as Nf=λfS.
Users are assumed to be uniformly distributed inside each cell
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
site. Fig.1 shows the two-tier macro-femto cell (MU: macrocell
user; FU: femtocell user).
1
MU
femtocell
BS
FAP
FU
f
h
f
R
c
R
Fig. 1. The two-tier macro-femto cell
A. Access Method
When deploying femtocell networks, the access method,
which refers to the rights of the users when making use of the
femtocells, needs to be defined. Two main different strategies
for femtocell access have been proposed so far [6]:
•Public Access where all the users can access all the
femtocells of a given operator.
•Private Access where only the subscriber of the femtocell
and a list of invited users can access a given femtocell.
It has been shown that public access will provide a better
network performance than private access. However, those who
pay for a femtocell are generally not keen to share their
resources with any other users. Operators are inclined to use
femtocells with private access.
In this paper, femtocells operate with private access. Let
U=Um+NfUfdenote the average number of users per cell
with Umand Ufreferring to the number of outdoor users of
per macrocell and indoor users of per femtocell respectively.
B. Channel Model and Scheduling Strategy
Generally, in propagation, three factors have bad effects
on signal quality: path loss, shadowing (slow fading), multi
path fading (fast fading). The lower of signal quality in
transmission, the more energy should be needed to guarantee
the spectrum efficiency.
In our model, the downlink channel between each BS and its
users (each FAP and its users) is composed of a fixed distance
dependent path loss, a slowly varying component modeled
by lognormal shadowing and Rayleigh fast fading with unit
average power. The thermal noise is ignored for simplicity.
Available spectrum consists of Ffrequency subchannels
each with bandwidth WHz. All subchannels are assigned equal
transmit power. Each user is assumed to track their Signal-to-
Interference Ratio (SIR) in each subchannel. Each BS assigns
rate adaptively based on the received SIR per user, i.e. assigns
ibps/Hz when SIR lies in [τi,τ
i+1),1≤i≤Iin per
subchannel.
So, the throughput per subchannel is:
T=
I−1
i=1
i·Pr[τi≤SIR < τi+1]+I·Pr[SIR ≥τI](1)
Besides, we consider Round Robin scheduling strategy for
femtocell and macrocell users.
C. Metrics
Normally designing of wireless networks always try to ob-
tain the maximal spectrum efficiency under Quality of Service
(QoS) constraints such as minimum required bandwidth, maxi-
mum tolerable delays and jitter, and packet loss rate. However,
it will make wireless networks easy to be energy hungry.
For energy efficient two-tier cellular networks, the designing
criterion should be maximizing the energy efficiency under
QoS constraints. And in this paper we maximize the energy
efficiency under specific spectrum efficiency. Therefore, the
general metrics both for spectrum and energy efficiency should
be provided.
Some existing spectrum efficiency metrics for wireless
networks such as b/s/Hz spectral efficiency, b/s/Hz/m2
area spectral efficiency (ASE) [7], b/s/Hz/antenna,etc
have been used for many years. For cellular access networks,
previous researches prone to use b/s/Hz/m2ASE to evaluate
the spectrum efficiency. However, the ASE has its limitation
in measuring femtocell networks. In the vertical direction of
femtocell networks, it may exist more than one femtocells. For
example, one house is a femtocell with another femtocell on
its top. So the spatial spectral efficiency is urgently needed.
There are also already some energy efficiency metrics for
wireless networks, such as b/J energy efficiency, b/T E N U
power efficiency, where TNEU refers to the amount of signal
energy identical to the variance of the complex-valued AWGN
samples recorded at the receiver [8]. b/J energy efficiency is
good applying for most cases in wireless communications. So
we use it with a little adjustment.
Two novel metrics to evaluate the spectrum efficiency and
energy efficiency of cellular networks are below:
•spatial spectral efficiency: the data rate of each user per
unit bandwidth per unit space supported by a cell
•green factor: the b/J power efficiency per subchannel
III. MODEL FOR ENERGY CONSUMPTION IN
DOWNLINK
A. Energy Efficiency Designing Backs to Downlink
Because of the powerful energy supply for BSs, researchers
always ignore the energy saving of downlink and give more
focus on uplink. However, as the appearance of Green Radio
concept, it deserves to give more attention on downlink energy
saving because BSs cost the most energy in cellular access
networks. Moreover, from operators’ perspective, reducing the
downlink energy consumption not only minimizes the environ-
mental impact of the industry, but also benefits for economical
reasons. For example, reduced energy consumption translates
directly to lower operating expenditure (OPEX) [9].
For reducing the CO2emission and improving the interests
of operators (lower OPEX), we consider cellular downlink
energy consumption. Comparing to BSs and FAPs, which cost
the most energy in the cell, mobile users and femtocell users
cost little energy to receive the signal from BSs and FAPs. So
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
ignoring the receiving energy consumption of mobile users
and femtocell users is reasonable.
B. BS Power Consumption
For getting the radiated power of BS Ptx, the average
consumed power Pmis needed. Until now, no exact equation
shows the relation between Ptx and Pm. It is difficult to get
the exact relation between Ptx and Pmbecause the power
consumption of BS corresponds to a lot of factors, such
as amplifier and feeder losses, cooling, signal processing,
battery backup, etc. The power consumption of amplifier and
feeder losses as well as cooling is linear with Ptx.The
power consumption of signal processing and battery backup
is constant with Ptx. All of these constitute PM. Set PLm
and PCm representing the linear part and the constant part
respectively.
We can employ the general model for power consumption:
PM=PLm +PCm =aPtx +PCm (2)
where ais constant. The typical values for aand PCm can be
found in [10].
C. FAP Power Consuption
Same as BSs, FAPs can also employ the model:
PF=PLf +PCf =bPtxf +PCf (3)
where PLf and PCf mean the linear and constant part power
consumption of FAPs, bis constant, Ptxf is the transmitting
power of FAPs. However, the components of PLf and PCf
are different from PLm and PCm. For example, there is no
cooling device for FAPs.
D. Downlink Energy Consumption
According to the above analyzing of BSs and FAPs power
consumption, we can obtain the whole downlink power con-
sumption Psystem.
Psystem =PM+NfPF(4)
IV. SPECTRUM ALLOCATION FOR GREEN RADIO
Considering the participating of FAPs in the cell, the
spectrum should be repartitioned for energy saving. Existing
spectrum allocation strategies in femtocell networks should
be departed into two different ones. In a splitting spectrum
network, femtocells use different frequency bands from those
employed by macrocells. This avoids interference between
macrocells and femtocells but additional frequency bands
are needed, decreasing the spectrum efficiency of cellular
networks. In a shared spectrum network, femtocells use the
same frequency band as macrocells. This is beneficial for
increasing the spectrum efficiency but coming with serious
interference between macrocells and femtocells.
All of these strategies aim to improve spectrum efficiency
but do not consider the energy saving. For energy efficient
networks, we propose a new split spectrum allocation strategy,
which maximize the green factor of the cell under spectrum
efficiency constraint.
A. Spectrum Allocation
For a cell with frequency band WF, once be participated
by FAPs, some MUs, which originally belong to the BS are
handed off to FAPs, i.e. becoming FUs. Due to this change,
excessive frequency subchannels will be allocated for MUs
now because the required spectrum efficiency of FUs can
be provided by FAPs. So it is rational to allocate some
subchannels to femtocells, which belong to macrocell before.
Fig. 2 shows the spectrum partition in a cell with WF
mand
WF
fpertaining to macrocell and femtocell respectively.
WFmWFf
WF
Fig. 2. The spectrum allocation in a cell
According to recent surveys, over 50 percent of calls and 70
percent of data services will take place indoors in the future
[11]. More subchannels should be allocated to femtocells for
getting required spectrum efficiency as shown in Fig. 2.
However, due to interference among femtocells, it is not
sensible to use all frequency band WF
fin each femtocell.
So we consider part frequency access of femtocells. The
frequency access ratio is rf=Fa/Ff, where Fasubchannels
are active for each FAP.
B. System Throughput
Define Tmand Tfas the throughput (in b/s/Hz) in each
subchannel provided by BSs and FAPs respectively. So
Tm=
I−1
i=1
i·[Zm(τi+1)−Zm(τi)] + I·[1 −Zm(τI)]
Tf=
I−1
i=1
i·[Zf(τi+1)−Zf(τi)] + I·[1 −Zf(τI)]
(5)
where Zm(τi)=Pr (SIRm≤τi) and Zf(τi)=Pr (SIRf≤τi)
denote spatially cellular SIR distribution for macrocell and
femtocell users respectively. There are some convenient strate-
gies to calculate Tmand Tf[12].
Macrocell BS transmits over WF
mand FAPs transmit over
WF
f, where WF
m=rWF and WF
f=(1−r)WF.Sothe
two-tier throughput (b/s) per subchannel can be calculated as
below:
Tc=W[rTm+(1−r)NfrfTf(rf,N
f)] (6)
And the cellular throughput can be obtained by multiply (10)
by F.
C. Maximize Green Factor under Required Spectrum Effi-
ciency
Some existing energy efficiency strategies use b/J energy
efficiency as the only target ignoring the spectrum efficiency
[13]. This is bad because it will obtain not only low energy
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
TAB L E I
SIMULATION PARAMETERS
Parameters Values
Macrocell/Femtocell Radius 288m,40m
Total users in a cell 300
Users per femtocell 2
BS transmitting power 20W
FAP transmitting power 0.1W
Wall partition loss 2dB
Energy model coefficient a, b 21.54, 7.84
Energy model constant PCm,PCf 354.44, 71.50
The ratio of throughput η=0.01;0.1;0.5
Femtocell numbers in a cell Nf= 10; 30; 50; 70; 90; 110
consumption but also low spectrum efficiency wireless sys-
tems.
Different from available works which are improving system
energy efficiency only, we consider green factor under required
spectrum efficiency, expecting to find the maximum value of
green factor for a specific spectrum efficiency. With frequency
band WF, the spatial spectrum efficiency can be converted to
throughput. So we have
max green factor = max
W[rTm+(1−r)NfrfTf(rf,N
f)]
Psystem (7)
under throughput constraint:
TmrF
Um
·(1 −η)= Tf(rf,N
f)(1 −r)Fr
f
Uf
·η(8)
where ηis the ratio of throughput per user in the first tier to
throughput per user in the second tier.
To calculate the green factor, we firstly obtain the equation for r:
r(rf,N
f)= 1
1+ UfTm(1 −η)
UcTf(rf,N
f)η
(9)
Then, combining equation (7) and (9), we can obtain the optimal
spectrum allocation which maximizes the green factor. The maximal
green factor relates to rfand Nf. In the next section, we will show
the simulation.
V. SIMULATIONS AND RESULTS
We consider macro-femto cellular networks, which consists of a
central macrocell BS with two rings of 18 interfering macrocells
and several FAPs. FAPs are assumed to be scattered according to
SPPP. Users are placed randomly in the cell area following a uniform
distribution on R2(Here, we propose that all users are on the ground
for simplicity). Table 1 shows the simulation parameters.
Fig. 3 shows the minimum required spectrum WF, satisfying
a target data rate 0.1Mbps for MUs, the corresponding data rate
for FUs are 10Mbs, 1Mbps, 0.1Mbps for η=0.01,η=0.1,
η=0.5respectively. So the required spatial spectrum efficiency
are 0.1M bps/W F for MUs, and 10M bps/W F ,1Mbps/W F ,
0.1M bps/W F for FUs.
Fig. 4 to Fig.6 show the relation between green factor and
the frequency access ratio rfunder the spatial spectral efficiency
0.1M bps/W F for MUs, and 10M bps/W F (Fig. 4), 1M bps/W F
(Fig. 5), 0.1M bps/W F (Fig. 6) for FUs. From Fig. 4 to Fig. 6 we
can see that
•When Nfis less than 30, the green factor increases with the
increasing of rf.WhenNfis large than 50, the green factor
firstly increases, then decreases with the increasing of rf.
•The maximal green factor is increasing with the increasing of
Nf. However, due to the different rfto achieve the maximal
green factor of different Nf, adjustment of rfis needed to
improve energy efficiency.
•When ηis small(η= 0.01 or 0.1), the maximal green factor for
different Nfwill be obtained with different rf. For example,
in Fig. 4 for Nf=10 and 110, the green factor is maximal when
rfis 1 and 0.2, respectively. While ηis relatively large(η=0.5),
the maximal green factor can be obtained at almost rf=0.2 for
different Nf.
10 20 30 40 50 60 70 80 90 100 110
0
1
2
3
4
5
6
7
8
9x 107
The number of femtocells in a cell
Required spectrum WF(Hz)
Eta=0.01
Eta=0.1
Eta=0.5
Fig. 3. Required spectrum WF meeting a target average data rate of 0.1Mbps
for macro cellular user
00.2 0.4 0.6 0.8 1
0
0.5
1
1.5
2
2.5
3x 105
the frequency access ratio of femtocells
green factor(b/J)
Nf=10
Nf=30
Nf=50
Nf=70
Nf=90
Nf=110
Fig. 4. Green factor versus frequency access ratio of femtocells(η=0.01)
VI. CONCLUSION AND FURTHER WORK
Based on the spatial model and two new metrics, this paper
provides a novel energy efficient spectrum allocation strategy for
two-tier cellular networks. Two new metrics are spatial spectral
efficiency and green factor, which are the general spectrum and
energy efficiency metrics for cellular networks. Maximizing the green
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
00.2 0.4 0.6 0.8 1
0
0.5
1
1.5
2
2.5 x 104
the frequency access ratio of femtocells
green factor(b/J)
Nf=10
Nf=30
Nf=50
Nf=70
Nf=90
Nf=110
Fig. 5. Green factor versus frequency access ratio of femtocells(η=0.1)
00.2 0.4 0.6 0.8 1
1000
1200
1400
1600
1800
2000
2200
2400
2600
the frequency access ratio of femtocells
green factor(b/J)
Nf=10
Nf=30
Nf=50
Nf=70
Nf=90
Nf=110
Fig. 6. Green factor versus frequency access ratio of femtocells(η=0.5)
factor under spatial spectral efficiency is an effective way to improve
energy efficiency. Through properly adjusting frequency access ratio
of FAPs, the maximal green factor under specific spatial spectral
efficiency constraint can be achieved.
After the spatial spectral efficiency has been defined, the multi-
floor model (users are on different floors) is better than one-floor
model (users are on the ground) for analyzing the energy efficiency
of macro-femto two-tier cellular networks. We will maximize green
factor under spatial spectral efficiency constraint in the multi-floor
model in our future work.
ACKNOWLEDGMENT
This work is supported by the 111 Project (B08038), State Key
Laboratory of Integrated Services Networks (ISN090105), Program
for New Century Excellent Talents in University (NCET-08-0810),
National Natural Science Foundation of China (No. 61072069 &
60772137), and the Fundamental Research Funds for the Central
Universities (No.72101855 & 72105242 & 72105377), and UK-China
Science Bridges: R&D on (B)4G Wireless Mobile Communications.
REFERENCES
[1] Smart 2020: Enabling the low carbon economy in the information age.
The Climate Group, Globe e-Sustainability Initiative (GeSI), 2008
[2] G. P. Fettweis and E. Zimmermann, ”ICT energy consumption - trends
and challenges,” in Proceedings of the 11th International Symposium
on Wireless Personal Multimedia Communications, Lapland, Finland,
September 2008.
[3] Y. Xiao, ”Energy saving mechanism in the IEEE 802.16e wireless
MAN,” IEEE Commun. Letters, vol. 9, no. 7, pp.595-597, July 2005.
[4] F. Meshkati, H. V. Poor, S. C. Schwartz, and N. B. Mandayam,
”An energy-efficient approach to power control and receiver design in
wireless networks,” IEEE Trans, Commun., vol. 5, no. 1, pp. 3306-3315,
Nov. 2006.
[5] V. Chandrasekhar, J. G. Andrews, and A. Gatherer, ”Femtocell networks:
a survey,” IEEE Comm. Mag., vol. 46, pp.59-67,2008.
[6] D. L. Perez, A. Valcarce, G. D. L. Roche, E. Liu, and J. Zhang, ”Access
methods to WiMAX femtocells: a downlink system-level case study,” in
11th IEEE ICCS, Guangzhou, China, Nov. 2008, pp 1657-1662
[7] M. S. Alouini and A. J. Goldsmith, ”Area spectral efficiency of cellular
mobile radio systems,” IEEE Trans. Veh. Technol., vol. 48, no. 4, pp.
1047-1066, July 1999.
[8] J. Akhtman and L. Hanzo, ”Power versus bandwidth efficiency in
wireless communications: the economic perspective,” IEEE VTC Fall,
Anchorage, Alaska, USA, 2009.
[9] S. Fletcher, ”Green Radio@ Sustainable Wire-
less Networks,” VCE Core5 Programme. Available:
http://kn.theiet.org/magazine/rateit/communications/green-radio-
article.cfm
[10] F. Richter, A. J. Fehske and G. P. Fettweis, ”Energy efficiency aspects
of base station deployment strategies for celluar networks,” IEEE
GlobeCom, Hawaii, the USA, Nov, 2009.
[11] G. Mansfield, ”Femtocells in the US Market-Business drivers and
consumer propositions,” FemtoCells Europe, ATT, London, U.K., June
2008.
[12] V. Chandrasekhar, and J. G. Andrews, ”Spectrum allocation in tiered
cellular networks,” IEEE Trans. Commun., vol. 57, no. 10, Oct. 2009,
pp. 3059-3068.
[13] G. Miao, N. Himayat, Y. Li, and D. Ormann, ”Energy-efficient design
in wireless OFDMA,” IEEE ICC, Beijing, China, May, 2008.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
