Modelling power consumption femtocell
0 likes301 views
The document investigates the power consumption of femtocell base stations for different wireless technologies and bit rates. It proposes a power consumption model for femtocell base stations that accounts for components like the microprocessor, FPGA, transmitter, and power amplifier. The model is used in a deployment tool to design energy-efficient femtocell networks for a predefined area by activating base stations. Sleep modes are introduced, which reduce power consumption in the networks by 24%.
![Modelling the Power Consumption
in Femtocell Networks
Margot Deruyck, Dieter De Vulder, Wout Joseph, and Luc Martens
Ghent University/IBBT, Dept. of Information Technology
Gaston Crommenlaan 8 box 201, B-9050 Ghent, Belgium
Email: margot.deruyck@intec.ugent.be
Abstract—In this paper, the energy efficiency of a femtocell
base station is investigated and compared for various bit rates and
for three different wireless technologies namely, mobile WiMAX,
HSPA, and LTE. A power consumption model is proposed
and applied in a network deployment tool to develop energy-
efficient femtocell networks. Furthermore, it is investigated to
what extent the introduction of sleep modes can reduce the
power consumption in femtocell networks. Reductions of 24%
are obtained when sleep modes are introduced into the network.
I. INTRODUCTION
Today, a lot of people buy or already possess one or
more mobile and thus wireless devices. The Wireless World
Research Forum (WWRF) has a vision of 7 trillion wireless
devices serving 7 billion users by 2017 [1]. These mobile
devices are currently not only used outdoor, but are also
used in indoor environments. Therefore, the operators and the
manufacturers start searching for solutions to increase indoor
coverage. One possibility is the introduction of femtocell base
stations. A femtocell base station is rather cheap and can be
installed by the end user himself/herself. It communicates with
the cellular network through a broadband connection such as
DSL (Digital Subscriber Line) or cable modem, or through
a separate RF (Radio Frequency) backhaul channel, thereby
reducing the traffic on the cellular network.
The power consumption of one femtocell base station is
limited, however, when hundreds or thousands femtocell base
stations are used in a network, the power consumption can
become significant. Furthermore, Informa Telecoms & Media
concludes in its study on the occasion of the World Mobile
Congress in Barcelona 2011 that there will be 49 million
femtocell access points serving 114 million users in 2014 [2].
It is thus clear that it is interesting to investigate the power
consumption of these femtocell base stations.
The purpose of this paper is to investigate the energy
efficiency of femtocell base stations and compare it for various
bit rates and three different wireless technologies namely,
mobile WiMAX (Worldwide Interoperability for Microwave
Access), HSPA (High Speed Packet Access), and LTE (Long
Term Evolultion). Therefore, a power consumption model is
proposed. Based on this model, a deployment tool is designed
which develops energy-efficient femtocell networks (i.e., with
a minimal power consumption) for a predefined area. In a last
step, it is investigated to what extent the power consumption
can be reduced by introducing sleep modes in the developed
networks.
The outline of the paper is as follows: in Section II the
power consumption model is described. Section III gives
more information about the deployment tool we developed.
In Section IV, results obtained with the power consumption
model and the deployment tool are discussed and sleep modes
are introduced into the network. Finally, Section V summarizes
the most important conclusions.
II. ENERGY EFFICIENCY OF A FEMTOCELL
To allow the comparison of the performance of the different
technologies considered, the energy efficiency is used. Here,
the energy efficiency is defined as the power consumption
needed to cover a certain area (in W/m2
) [3], [4], [5]:
PCarea =
Pel
π · R2
(1)
with Pel the power consumption of the femtocell (in W) and
R the range of the femtocell (in m). The lower PCarea, the
more energy-efficient the femtocell is. How Pel and R are
determined, is described in Section II-A and II-B, respectively.
A. Power consumption model
The development of the power consumption model is based
on the hardware model for a femtocell base station of [6].
Three interacting blocks can be defined in a femtocell base
station. The first block consists of the microprocessor which
is responsible for the implementation and the management of
the standardised radio protocols and the associated base band
processing. Furthermore, it is also responsible of managing
the backhaul connection to the core network. The second
block is the FPGA (Field-Programmable Gate Array), along
with some other integrated circuitry to support a variety of
functions such as data-encryption, hardware authentication and
the Network Time Protocol. The third block contains the
RF (Radio Frequency) transmitter which allows to send and
receive signals and the power amplifier.
Based on this hardware model, a function is derived for the
power consumption Pel of the femtocell base station (in W):
Pel/femto = Pel/mp + Pel/F P GA + Pel/trans + Pel/amp (2)
with Pel/mp, Pel/F P GA, Pel/trans and Pel/amp the power
consumption (in W) of, respectively, the microprocessor, the
FPGA, the transmitter and the power amplifier.](https://image.slidesharecdn.com/modellingpowerconsumptionfemtocell-150422051734-conversion-gate02/85/Modelling-power-consumption-femtocell-1-320.jpg)
![The power consumption of each block is constant through-
out time and thus a fixed number is used for the power
consumption [6]. To determine the power consumption of the
power amplifier Pel/amp (in W) the following equation is
used [3], [4]:
Pel/amp =
PT x
η
(3)
wit PT x the input power of the antenna (in W) and η the
efficiency of the power amplifier which is the ratio of the RF
output power to the electrical input power. η can take values
between 5% and 40%. Here, a power amplifier is assumed
with a medium performance (i.e., 22.5%) for η. The values
for the other parameters can be found in Table I.
Component Value
Microprocessor Pel/mp 3.2 W
FPGA Pel/F P GA 4.7 W
Transmitter Pel/trans 1.7 W
Efficiency η 22.5%
TABLE I
POWER CONSUMPTION OF THE FEMTOCELL BASE STATION
COMPONENTS [6].
B. Determination of the range
To determine the range of a femtocell base station, a link
budget needs to be composed [3], [4]. A link budget takes
all of the gains and losses of the transmitter through the
medium to the receiver into account. Firstly, the maximum
allowable path loss PLmax a transmitted signal can experience
while still being detectable at the receiver is determined.
Table II gives an overview of the link budget parameters for
the considered technologies.
Parameter Mobile WiMAX HSPA LTE
Frequency [MHz] 2500 2100 2600
Maximum input power antenna 23 15 21
of base station [dBm]
Antenna gain of base station [dBi] 2 2 2
Antenna gain of mobile station [dBi] 2 0 0
Soft handover gain of mobile station [dB] 0 1.5 0
Feeder loss of base station [dB] 0.5 0 2
Feeder loss of mobile station [dB] 0 0 0
Fade margin [dB] 6 7.7 4
Interference margin [dB] 2 2 2
Bandwidth [MHz] 5 5 5
Receiver SNR [dB] [6, 8.5, 11.5, [-3.1, 0.1, 3.4, [-1.5, 3, 10.5
15 19 21]1 6, 7.1, 9.6 14, 19, 23
15.6]2 23, 29.4]3
Number of used subcarriers 360 — 301
Number of total subcarriers 512 — 512
Noise figure of mobile station [dB] 7 7 5
Implementation loss of mobile station [dB] 2 0 0
Shadowing margin [dB] 13.2 13.2 13.2
Duplexing TDD TDD TDD
(1) format: [1/2 QPSK, 3/4 QPSK, 1/2 16-QAM, 3/4 16-QAM, 2/3 64-QAM, 3/4 64-QAM]
(2) format: [1/4 QPSK, 1/2 QPSK, 3/4 QPSK, 3/4 8-QAM, 1/2 16-QAM, 3/4 16-QAM, 3/4 64-QAM]
(3) format: [1/3 QPSK, 1/2 QPSK, 2/3 QPSK, 1/2 16-QAM, 2/3 16-QAM, 4/5 16-QAM, 1/2 64-QAM, 2/3 64-QAM]
TABLE II
LINK BUDGET TABLE FOR THE TECHNOLOGIES CONSIDERED.
Once PLmax is known, the range can be calculated by using
a propagation model. As femtocell base stations are situated
indoor, the most appropriate propagation model is the ITU-
R P.1238 model [7]. This model predicts the path loss for
three possible scenarios: a residential building, a commercial
building or an office. Here, the office scenario is assumed. The
range R is then determined as follows:
R = 10
P Lmax−20·log10f−Lf (n)+28
N (4)
with f the frequency (in MHz), Lf (n) the floor penetration
loss factor (dB), n the number of floors between base station
and terminal and N the distance power loss coefficient. As
the office scenario is here considered, Lf (n) corresponds with
15 + 4 · (n − 1) and N equals 30.
III. DEPLOYMENT TOOL
The model described in Section II is used in a deployment
tool. This deployment tool allows to design energy-efficient
femtocell networks (i.e., with a minimal power consumption)
to provide indoor coverage in a predefined area, here noted
as the target area, and is based on the GRAND (Green
Radio Access Network Design) as defined in [8]. This area
is supplied as a shape file [8]. This shape file gives more
information about the buildings in the area and their position,
form and height.
Firstly, the tool determines a set of possible locations for the
femtocell base stations. Each building is a possible location
of a femtocell base station. This base station is placed at
the centre of the building. In this way, the size of the set is
somewhat limited to reduce the simulation duration. Initially,
no femtocell base stations are active.
Secondly, the tool computes an energy-efficient network based
on these locations. Therefore, a genetic search algorithm is
used. The algorithm will generate a first set of possible so-
lutions, the so-called population, by activating femtocell base
stations. The size of the population is specified by the end user.
Based on this first population, the algorithm will optimize the
solutions through a number of generations. A new generation
of solution is created by mutating (some of) the solutions
of the previous generation, by selecting the best solutions
of the previous solution and by the crossover operator. Four
mutations are defined: (i) activating an inactive femtocell
base station, (ii) deactivating an active femtocell base station,
(iii) adding 1 dBm to the input power of an antenna of an
active femtocell base station, and (iv) subtracting 1 dBm of
the input power of an antenna of an active femtocell base
station. Which femtocell base station and how it is adapted,
is randomly determined. To determine if a new solution is
better than the previous one, two different fitness functions
are defined. One tells how good a solution performs in terms
of coverage of the target area and the second in terms of power
consumption. The coverage fitness returns the percentage of
the indoor target area that is covered by the solution. The
maximum power consumption of the network (i.e., when each
possible location contains a femtocell base station) is used as
reference. The model of Section II is used to determine the
power consumption fitness. These two fitness functions are
combined into one global fitness function by using a weighted
average of the coverage fitness and the power consumption
fitness. The same functions as in [3], [8] are used. Furthermore,
these functions are also used to select the best solutions](https://image.slidesharecdn.com/modellingpowerconsumptionfemtocell-150422051734-conversion-gate02/85/Modelling-power-consumption-femtocell-2-320.jpg)
![in a generation. The crossover operator allows to create a
new hopefully better solution by using different parts of two
mediocre or bad solutions.
The algorithm will continue to create new solutions until a
stopping criterion is met. This stopping criterion is set by
the end user. The algorithm will stop if a certain amount of
generations is created or if the maximum allowable simulation
duration is exceeded. The longer the simulation duration, the
better the solution will be.
IV. RESULTS
A. Comparison of the energy efficiency
In this section, the energy efficiency of the considered
technologies, namely, mobile WiMAX, HSPA, and LTE is
compared for various bit rates in a 5 MHz channel at the fre-
quencies specified in Table II. The office scenario is assumed
with n equals to 1 (i.e., only one floor between base station
and receiver). Fig. 1 compares PCarea of the technologies
considered for different bit rates in a 5 MHz channel.
0 5 10 15 20
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
Bitrate [Mbps]
PCarea
[W/m
2
]
Mobile WiMAX
HSPA
LTE
Fig. 1. Comparison of P Carea for the technologies considered in a 5 MHz
channel.
The femtocell base station consumes 10.5 W, 10.0 W, and
9.7 W, respectively, for mobile WiMAX, LTE, and HSPA. The
differences in power consumption are due to the differences
in input power of the antenna (Table II). The higher the input
power of the antenna, the higher the power consumption of
the power amplifier (eq. (3)) and thus the higher the power
consumption of the base station.
For bit rates higher than 5 Mbps, mobile WiMAX is the
most energy-efficient technology (PCarea = 34.9 mW/m2
versus 4.2 mW/m2
for LTE and 2.3 mW/m2
for HSPA and
for a bit rate of approximately 11 Mbps). Although mobile
WiMAX has a higher power consumption, it obtains a higher
range (38.5 m versus 27.4 m and 9.4 m for LTE and HSPA
respectively) due to its lower receiver SNR (Signal-to-Noise
Ratio) and its higher input power of the antenna (Table II),
resulting thus in a higher energy efficiency.
For bit rates between 2.8 Mbps and 5 Mbps, LTE is the most
energy-efficient technology (PCarea = 0.4 mW/m2
versus
0.6 mW/m2
and 5.4 mW/m2
for, respectively, mobile WiMAX
and HSPA for a bit rate of approximately 4 Mbps). Again, a
higher range is obtained with LTE (93.5 m versus 76.8 m
and 24 m for mobile WiMAX and HSPA respectively and
for a bit rate of approximately 4 Mbps) then for the other
technologies due to its lower receiver SNR for the considered
bit rates (Table II).
Finally, HSPA is the most energy-efficient technology for bit
rates below the 2.8 Mbps as these bit rates are not supported
by the other technologies.
B. Application: a energy-efficient femtocell network for a part
of Ghent
In this section, it is investigated how much power is needed
to cover a certain part of Ghent by femtocells. The deployment
tool described in Section III is used for this purpose. Fig. 2
gives an overview of the target area and the possible locations
for femtocell base stations in this area. The surface of this
area is about 0.11 km2
from which 60.78% is covered by
buildings and where coverage needs to be foreseen. In this
area, 495 possible locations for the femtocell base stations are
found (i.e., in each building a femtocell base station can be
placed).
Fig. 2. Overview of the target area and the possible locations (squares) for
femtocell base stations in the target area (0.11 km2).
Fig. 3 gives an overview of the solution network for mobile
WiMAX (Fig. 3(a)), HSPA (Fig. 3(b)), and LTE (Fig. 3(c)).
A bit rate of 5 Mbps is considered in a 5 MHz channel, to
make a fair comparison between the different technologies.
Furthermore, for the genetic search algorithm, the population
size is chosen as 50, a maximum of 1000 generations can be
created, and the simulation duration is limited to 1h. However,
when the estimated value of the generation is not higher than
1% compared to the previous generation, the algorithm is also
stopped. The numerical results are listed in Table III for a
coverage requirement of 90%.
In agreement with the results from Section IV-A, mobile
WiMAX performs the best. For a coverage of approximately
98%, the WiMAX network only consumes 0.4 kW, resulting
in the lowest value for PCarea.
C. Introduction of sleep modes
The power consumption of the femtocell networks designed
in Section IV-B can be further minimized by introducing
sleep modes. When there is little or no activity in its cell, the](https://image.slidesharecdn.com/modellingpowerconsumptionfemtocell-150422051734-conversion-gate02/85/Modelling-power-consumption-femtocell-3-320.jpg)
![(a)
(b)
(c)
Fig. 3. Energy-efficient femtocell networks for mobile WiMAX (a), HSPA
(b), and LTE (c) for a bit rate of 5 Mbps in a 5 MHz channel resulting from
the deployment tool.
Parameter Mobile HSPA LTE
WiMAX
Number of femtocells used 38 272 63
Power consumption [kW] 0.4 2.6 0.6
Coverage percentage [%] 97.6 95.9 99.5
P Carea [mW/km2] 3.7 25.0 5.8
TABLE III
COMPARISON OF THE ENERGY-EFFICIENT FEMTOCELL NETWORK FOR THE
TECHNOLOGIES CONSIDERED AND A PHYSICAL BIT RATE OF 5 MBPS.
femtocell base station can enter the sleep mode i.e., a mode
wherein the femtocell base station consumes only a small
amount of power. When it is necessary, the femtocell base
station can be awaken to serve the users in its cell. In [6],
three sleep mode techniques are proposed: the sleep mode is
controlled by either the femtocell base station itself, or by the
core network, or by the UE (user equipment). In this paper,
the UE controlled approach will be used as this technique
does not need the overlaying macrocell network for signalling
the wake-up signals. In the UE controlled approach, the UE
can broadcast wake-up signals when it is necessary to wake
up a femtocell base station. Although the femtocell base
station is in sleep mode, it retains the capability to receive
wake-up signals from the UE and, if necessary, to switch
to an active state. In the sleep mode, the RF transmitter
(1.0 W [6]), the power amplifier (0.9 W, 0.03 W, and 0.4 W
for, respectively, mobile WiMAX, HSPA, and LTE) and
some other circuit elements (1.7 W [6]) can be switched off,
leading to a reduction in power consumption of 3.6 W, 2.8 W,
and 3.1 W for, respectively, mobile WiMAX, HSPA, and LTE.
1) Performance of the sleep mode technique:
In this section, it is investigated how much power is saved
by introducing the sleep mode for one femtocell base station.
In the next section, it is studied what the power consumption
reduction is in the networks developed in Section IV-B.
The same procedure as in [9] is used for this investigation.
The average power consumption reduction of the femtocell
base station (as a percentage) is defined as [9]:
Ω =
Psleep
Pactive
· (1 − η) · 100 (5)
with Psleep the power consumption of the femtocell base
station in sleep mode (in W), Pactive the power consumption
of the femtocell base station in active mode (in W) and
η (0 ≤ η ≤ 1) the average duty cycle (i.e., the time that
the femtocell base station spends in the active mode). To
determine η a traffic model is used [9]. This traffic model
indicates when there are periods of high and low traffic
during the day. Typically, it is modelled by a Poisson process
with exponentially distributed arrival time and exponentially
distributed service time (mean value is here chosen to be
3 min). The traffic model used is based on the traffic model
proposed in [9] but it is adapted to the office environment
here considered (Section II). Table IV gives an overview of
the traffic model used in this study.
Time Usage Arrival time λ Traffic/user µ
[min.] [Erlang]
0h-8h Low 240 0.0125
8h-18h High 30 0.1
18h-0h Low 240 0.0125
TABLE IV
TRAFFIC MODEL FOR AN OFFICE ENVIRONMENT.
Fig. 4 illustrates the results for Ω for the technologies
considered. The maximum power consumption reduction is
between 29% and 34% when no users are present. This is
a theoretical case as it is not realistic that the femtocell
base station sleeps all day. The highest reduction is obtained
with mobile WiMAX as the power consumption of the power
amplifier is the highest for this base station (as it has the
highest input power of the antenna, Table II) and thus more
power can be saved in sleep mode compared to HSPA and
LTE. Furthermore, the more users are present in the cell, the
lower the power consumption reduction is. This is logical
due to the fact that when there are more users in a cell,
the femtocell base station can sleep less, resulting in a lower
reduction. A femtocell base station has typically between 1 to
16 users. Based on the results from Fig. 4, it is concluded that](https://image.slidesharecdn.com/modellingpowerconsumptionfemtocell-150422051734-conversion-gate02/85/Modelling-power-consumption-femtocell-4-320.jpg)
![the sleep mode introduction is very useful until a boundary
of 8 users. For more than 8 users, there is still a power
consumption reduction, however, this reduction becomes very
small (2-3%).
Fig. 4. Power consumption reduction for the femtocell base station of the
technologies considered and for various numbers of users supported.
2) Power consumption reduction in femtocell networks:
Now, it is determined how much power can be saved in the
networks developed in Section IV-B. Therefore, a random
number of users between 1 and 8 is allocated to each cell
in the network. Based on this number and the results from
Section IV-C1, it is calculated how the dialy power consump-
tion is reduced for each femtocell in the network and thus
what the overall power consumption reduction for the whole
network is. Table V gives an overview of the results from this
study.
Parameter Mobile HSPA LTE
WiMAX
No. femtocells 38 273 64
No. users in network 198 1180 280
Average no. user 5.2 4.3 4.4
per femtocell
Dialy power consumption 9.6 63.3 15.12
without sleep modes [kWh]
Daily power consumption 7.1 49.0 11.5
with sleep modes [kWh]
Power consumption reduction 25.4% 22.7% 24.2%
TABLE V
POWER CONSUMPTION REDUCTION IN FEMTOCELL NETWORKS FOR THE
TECHNOLOGIES CONSIDERED.
Introducing sleep modes in the network results in a power
consumption reduction between 22% and 25%. The highest
reduction is achieved with mobile WiMAX which corresponds
with the results obtained in Section IV-C1. The number of
users served is the highest for HSPA as HSPA has the highest
number of femtocell base stations in its network. However,
the average number of users per femtocell is approximately
the same (i.e., 4.6 users per femtocell) for the three tech-
nologies considered. If the same number of users is assumed
for each network, the network with the highest number of
femtocell base stations will correspond with the highest power
consumption reduction as it will be possible to let more
femtocell base stations sleep. Therefore, the average number
of users per femtocell is kept (approximately) the same in this
investigation.
V. CONCLUSION
In this paper, the energy efficiency of femtocell base stations
is investigated and compared for three wireless technologies
namely, mobile WiMAX, HSPA, and LTE. Therefore, a power
consumption model is proposed. Based on this model, it was
found that a femtocell base station consumes, in general,
approximately 10 W for a range between 9 to 130 m (depend-
ing on the technology and the bit rate considered). For the
parameters assumed and a 5 MHz channel, mobile WiMAX
is the most energy-efficient technology for bit rates higher than
5 Mbps. For bit rates between 2.8 Mbps and 5 Mbps, LTE is
the most energy-efficient.
The power consumption model is then used in a deploy-
ment tool for the development of energy-efficient femtocell
networks. For the area considered and a bit rate of 5 Mbps,
the mobile WiMAX consumes 105.6 kWh per day, the HSPA
network 499.8 kWh, and the LTE network 146.4 kWh.
Furthermore, sleep modes are introduced in the network.
The sleep mode technique reduces the power consumption
significantly when it supports a maximum of 8 users. For a
higher number of users, there is still a power consumption re-
duction but this reduction is rather small (< 3% for the power
consumption of a single femtocell base station). Introducing
sleep modes in the network leads to a power consumption
reduction of approximately 24%.
ACKNOWLEDGMENT
The work described in this paper was carried out with
the support of the IBBT-project GreenICT. W. Joseph is a
Post-Doctoral Fellow of the FWO-V (Research Foundation
Flanders).
REFERENCES
[1] World Wireless Research Forum, WWRF, www.wireless-world-
research.org.
[2] D. Duffy, Femtocell market set for strong growth in 2011, Informa
Telecoms & Media, 2010.
[3] M. Deruyck, E. Tanghe, W. Joseph, L. Martens, Modelling
and optimization of power consumption in wireless access
networks, Elsevier Computer Communications, 2011, doi:
10.1016/j.comcom.2011.03.008.
[4] M. Deruyck, E. Tanghe, W. Joseph, W. Vereecken, M. Pickavet,
L. Martens, B. Dhoedt, Model for Power Consumption of Wireless
Access Networks, IET Science, Measurement & Technology, Vol. 5,
Issue 4, July 2011, pp. 155-161.
[5] L. M. Correia, D. Zeller, O. Blume, D. Ferling, Y. Jading, I. G´odor,
G. Auer, L. Van der Perre, Challenges and Enabling Technologies
for Energy Aware Mobile Radio Networks, IEEE Communications
Magazine, Vol. 48, Issue 11, November 2010, pp. 66-72.
[6] I. Ashraf, F. Boccardi, L. Ho, Power Savings in Small Cell Deployments
via Sleep Mode Techniques, 21st Annual IEEE Symposium on Personal,
Indoor and Mobile Radio Communications (PIMRC 2010): Workshop
W-Green, Istanbul, Turkey, September 2010, pp. 306-310.
[7] Recommendation ITU-R P.1238-2, Propagation data and prediction
methods for the planning of indoor radiocommunication systems and
radio local area networks in the frequency range 900 MHz to 100 GHz,
1997-1999-2001.](https://image.slidesharecdn.com/modellingpowerconsumptionfemtocell-150422051734-conversion-gate02/85/Modelling-power-consumption-femtocell-5-320.jpg)
![[8] M. Deruyck, E. Tanghe, W. Joseph, W. Vereecken, M. Pickavet,
B. Dhoedt, L. Martens, Towards a deployment tool for wireless
access networks with minimal power consumption, 21st Annual IEEE
Symposium on Personal, Indoor and Mobile Radio Communications
(PIMRC 2010): Workshop W-Green, Istanbul, Turkey, September 2010,
pp. 294-299.
[9] I. Ashraf, L. Ho, H. Claussen, Improving Energy Efficiency of Femtocell
Base Stations via User Activity Dection, Wireless Communications and
Networking Conference (WCNC 2010), Sydney, Australia, April 2010,
pp. 1-5.](https://image.slidesharecdn.com/modellingpowerconsumptionfemtocell-150422051734-conversion-gate02/85/Modelling-power-consumption-femtocell-6-320.jpg)
Modelling power consumption femtocell
- 1. Modelling the Power Consumption in Femtocell Networks Margot Deruyck, Dieter De Vulder, Wout Joseph, and Luc Martens Ghent University/IBBT, Dept. of Information Technology Gaston Crommenlaan 8 box 201, B-9050 Ghent, Belgium Email: margot.deruyck@intec.ugent.be Abstract—In this paper, the energy efficiency of a femtocell base station is investigated and compared for various bit rates and for three different wireless technologies namely, mobile WiMAX, HSPA, and LTE. A power consumption model is proposed and applied in a network deployment tool to develop energy- efficient femtocell networks. Furthermore, it is investigated to what extent the introduction of sleep modes can reduce the power consumption in femtocell networks. Reductions of 24% are obtained when sleep modes are introduced into the network. I. INTRODUCTION Today, a lot of people buy or already possess one or more mobile and thus wireless devices. The Wireless World Research Forum (WWRF) has a vision of 7 trillion wireless devices serving 7 billion users by 2017 [1]. These mobile devices are currently not only used outdoor, but are also used in indoor environments. Therefore, the operators and the manufacturers start searching for solutions to increase indoor coverage. One possibility is the introduction of femtocell base stations. A femtocell base station is rather cheap and can be installed by the end user himself/herself. It communicates with the cellular network through a broadband connection such as DSL (Digital Subscriber Line) or cable modem, or through a separate RF (Radio Frequency) backhaul channel, thereby reducing the traffic on the cellular network. The power consumption of one femtocell base station is limited, however, when hundreds or thousands femtocell base stations are used in a network, the power consumption can become significant. Furthermore, Informa Telecoms & Media concludes in its study on the occasion of the World Mobile Congress in Barcelona 2011 that there will be 49 million femtocell access points serving 114 million users in 2014 [2]. It is thus clear that it is interesting to investigate the power consumption of these femtocell base stations. The purpose of this paper is to investigate the energy efficiency of femtocell base stations and compare it for various bit rates and three different wireless technologies namely, mobile WiMAX (Worldwide Interoperability for Microwave Access), HSPA (High Speed Packet Access), and LTE (Long Term Evolultion). Therefore, a power consumption model is proposed. Based on this model, a deployment tool is designed which develops energy-efficient femtocell networks (i.e., with a minimal power consumption) for a predefined area. In a last step, it is investigated to what extent the power consumption can be reduced by introducing sleep modes in the developed networks. The outline of the paper is as follows: in Section II the power consumption model is described. Section III gives more information about the deployment tool we developed. In Section IV, results obtained with the power consumption model and the deployment tool are discussed and sleep modes are introduced into the network. Finally, Section V summarizes the most important conclusions. II. ENERGY EFFICIENCY OF A FEMTOCELL To allow the comparison of the performance of the different technologies considered, the energy efficiency is used. Here, the energy efficiency is defined as the power consumption needed to cover a certain area (in W/m2 ) [3], [4], [5]: PCarea = Pel π · R2 (1) with Pel the power consumption of the femtocell (in W) and R the range of the femtocell (in m). The lower PCarea, the more energy-efficient the femtocell is. How Pel and R are determined, is described in Section II-A and II-B, respectively. A. Power consumption model The development of the power consumption model is based on the hardware model for a femtocell base station of [6]. Three interacting blocks can be defined in a femtocell base station. The first block consists of the microprocessor which is responsible for the implementation and the management of the standardised radio protocols and the associated base band processing. Furthermore, it is also responsible of managing the backhaul connection to the core network. The second block is the FPGA (Field-Programmable Gate Array), along with some other integrated circuitry to support a variety of functions such as data-encryption, hardware authentication and the Network Time Protocol. The third block contains the RF (Radio Frequency) transmitter which allows to send and receive signals and the power amplifier. Based on this hardware model, a function is derived for the power consumption Pel of the femtocell base station (in W): Pel/femto = Pel/mp + Pel/F P GA + Pel/trans + Pel/amp (2) with Pel/mp, Pel/F P GA, Pel/trans and Pel/amp the power consumption (in W) of, respectively, the microprocessor, the FPGA, the transmitter and the power amplifier.
- 2. The power consumption of each block is constant through- out time and thus a fixed number is used for the power consumption [6]. To determine the power consumption of the power amplifier Pel/amp (in W) the following equation is used [3], [4]: Pel/amp = PT x η (3) wit PT x the input power of the antenna (in W) and η the efficiency of the power amplifier which is the ratio of the RF output power to the electrical input power. η can take values between 5% and 40%. Here, a power amplifier is assumed with a medium performance (i.e., 22.5%) for η. The values for the other parameters can be found in Table I. Component Value Microprocessor Pel/mp 3.2 W FPGA Pel/F P GA 4.7 W Transmitter Pel/trans 1.7 W Efficiency η 22.5% TABLE I POWER CONSUMPTION OF THE FEMTOCELL BASE STATION COMPONENTS [6]. B. Determination of the range To determine the range of a femtocell base station, a link budget needs to be composed [3], [4]. A link budget takes all of the gains and losses of the transmitter through the medium to the receiver into account. Firstly, the maximum allowable path loss PLmax a transmitted signal can experience while still being detectable at the receiver is determined. Table II gives an overview of the link budget parameters for the considered technologies. Parameter Mobile WiMAX HSPA LTE Frequency [MHz] 2500 2100 2600 Maximum input power antenna 23 15 21 of base station [dBm] Antenna gain of base station [dBi] 2 2 2 Antenna gain of mobile station [dBi] 2 0 0 Soft handover gain of mobile station [dB] 0 1.5 0 Feeder loss of base station [dB] 0.5 0 2 Feeder loss of mobile station [dB] 0 0 0 Fade margin [dB] 6 7.7 4 Interference margin [dB] 2 2 2 Bandwidth [MHz] 5 5 5 Receiver SNR [dB] [6, 8.5, 11.5, [-3.1, 0.1, 3.4, [-1.5, 3, 10.5 15 19 21]1 6, 7.1, 9.6 14, 19, 23 15.6]2 23, 29.4]3 Number of used subcarriers 360 — 301 Number of total subcarriers 512 — 512 Noise figure of mobile station [dB] 7 7 5 Implementation loss of mobile station [dB] 2 0 0 Shadowing margin [dB] 13.2 13.2 13.2 Duplexing TDD TDD TDD (1) format: [1/2 QPSK, 3/4 QPSK, 1/2 16-QAM, 3/4 16-QAM, 2/3 64-QAM, 3/4 64-QAM] (2) format: [1/4 QPSK, 1/2 QPSK, 3/4 QPSK, 3/4 8-QAM, 1/2 16-QAM, 3/4 16-QAM, 3/4 64-QAM] (3) format: [1/3 QPSK, 1/2 QPSK, 2/3 QPSK, 1/2 16-QAM, 2/3 16-QAM, 4/5 16-QAM, 1/2 64-QAM, 2/3 64-QAM] TABLE II LINK BUDGET TABLE FOR THE TECHNOLOGIES CONSIDERED. Once PLmax is known, the range can be calculated by using a propagation model. As femtocell base stations are situated indoor, the most appropriate propagation model is the ITU- R P.1238 model [7]. This model predicts the path loss for three possible scenarios: a residential building, a commercial building or an office. Here, the office scenario is assumed. The range R is then determined as follows: R = 10 P Lmax−20·log10f−Lf (n)+28 N (4) with f the frequency (in MHz), Lf (n) the floor penetration loss factor (dB), n the number of floors between base station and terminal and N the distance power loss coefficient. As the office scenario is here considered, Lf (n) corresponds with 15 + 4 · (n − 1) and N equals 30. III. DEPLOYMENT TOOL The model described in Section II is used in a deployment tool. This deployment tool allows to design energy-efficient femtocell networks (i.e., with a minimal power consumption) to provide indoor coverage in a predefined area, here noted as the target area, and is based on the GRAND (Green Radio Access Network Design) as defined in [8]. This area is supplied as a shape file [8]. This shape file gives more information about the buildings in the area and their position, form and height. Firstly, the tool determines a set of possible locations for the femtocell base stations. Each building is a possible location of a femtocell base station. This base station is placed at the centre of the building. In this way, the size of the set is somewhat limited to reduce the simulation duration. Initially, no femtocell base stations are active. Secondly, the tool computes an energy-efficient network based on these locations. Therefore, a genetic search algorithm is used. The algorithm will generate a first set of possible so- lutions, the so-called population, by activating femtocell base stations. The size of the population is specified by the end user. Based on this first population, the algorithm will optimize the solutions through a number of generations. A new generation of solution is created by mutating (some of) the solutions of the previous generation, by selecting the best solutions of the previous solution and by the crossover operator. Four mutations are defined: (i) activating an inactive femtocell base station, (ii) deactivating an active femtocell base station, (iii) adding 1 dBm to the input power of an antenna of an active femtocell base station, and (iv) subtracting 1 dBm of the input power of an antenna of an active femtocell base station. Which femtocell base station and how it is adapted, is randomly determined. To determine if a new solution is better than the previous one, two different fitness functions are defined. One tells how good a solution performs in terms of coverage of the target area and the second in terms of power consumption. The coverage fitness returns the percentage of the indoor target area that is covered by the solution. The maximum power consumption of the network (i.e., when each possible location contains a femtocell base station) is used as reference. The model of Section II is used to determine the power consumption fitness. These two fitness functions are combined into one global fitness function by using a weighted average of the coverage fitness and the power consumption fitness. The same functions as in [3], [8] are used. Furthermore, these functions are also used to select the best solutions
- 3. in a generation. The crossover operator allows to create a new hopefully better solution by using different parts of two mediocre or bad solutions. The algorithm will continue to create new solutions until a stopping criterion is met. This stopping criterion is set by the end user. The algorithm will stop if a certain amount of generations is created or if the maximum allowable simulation duration is exceeded. The longer the simulation duration, the better the solution will be. IV. RESULTS A. Comparison of the energy efficiency In this section, the energy efficiency of the considered technologies, namely, mobile WiMAX, HSPA, and LTE is compared for various bit rates in a 5 MHz channel at the fre- quencies specified in Table II. The office scenario is assumed with n equals to 1 (i.e., only one floor between base station and receiver). Fig. 1 compares PCarea of the technologies considered for different bit rates in a 5 MHz channel. 0 5 10 15 20 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 Bitrate [Mbps] PCarea [W/m 2 ] Mobile WiMAX HSPA LTE Fig. 1. Comparison of P Carea for the technologies considered in a 5 MHz channel. The femtocell base station consumes 10.5 W, 10.0 W, and 9.7 W, respectively, for mobile WiMAX, LTE, and HSPA. The differences in power consumption are due to the differences in input power of the antenna (Table II). The higher the input power of the antenna, the higher the power consumption of the power amplifier (eq. (3)) and thus the higher the power consumption of the base station. For bit rates higher than 5 Mbps, mobile WiMAX is the most energy-efficient technology (PCarea = 34.9 mW/m2 versus 4.2 mW/m2 for LTE and 2.3 mW/m2 for HSPA and for a bit rate of approximately 11 Mbps). Although mobile WiMAX has a higher power consumption, it obtains a higher range (38.5 m versus 27.4 m and 9.4 m for LTE and HSPA respectively) due to its lower receiver SNR (Signal-to-Noise Ratio) and its higher input power of the antenna (Table II), resulting thus in a higher energy efficiency. For bit rates between 2.8 Mbps and 5 Mbps, LTE is the most energy-efficient technology (PCarea = 0.4 mW/m2 versus 0.6 mW/m2 and 5.4 mW/m2 for, respectively, mobile WiMAX and HSPA for a bit rate of approximately 4 Mbps). Again, a higher range is obtained with LTE (93.5 m versus 76.8 m and 24 m for mobile WiMAX and HSPA respectively and for a bit rate of approximately 4 Mbps) then for the other technologies due to its lower receiver SNR for the considered bit rates (Table II). Finally, HSPA is the most energy-efficient technology for bit rates below the 2.8 Mbps as these bit rates are not supported by the other technologies. B. Application: a energy-efficient femtocell network for a part of Ghent In this section, it is investigated how much power is needed to cover a certain part of Ghent by femtocells. The deployment tool described in Section III is used for this purpose. Fig. 2 gives an overview of the target area and the possible locations for femtocell base stations in this area. The surface of this area is about 0.11 km2 from which 60.78% is covered by buildings and where coverage needs to be foreseen. In this area, 495 possible locations for the femtocell base stations are found (i.e., in each building a femtocell base station can be placed). Fig. 2. Overview of the target area and the possible locations (squares) for femtocell base stations in the target area (0.11 km2). Fig. 3 gives an overview of the solution network for mobile WiMAX (Fig. 3(a)), HSPA (Fig. 3(b)), and LTE (Fig. 3(c)). A bit rate of 5 Mbps is considered in a 5 MHz channel, to make a fair comparison between the different technologies. Furthermore, for the genetic search algorithm, the population size is chosen as 50, a maximum of 1000 generations can be created, and the simulation duration is limited to 1h. However, when the estimated value of the generation is not higher than 1% compared to the previous generation, the algorithm is also stopped. The numerical results are listed in Table III for a coverage requirement of 90%. In agreement with the results from Section IV-A, mobile WiMAX performs the best. For a coverage of approximately 98%, the WiMAX network only consumes 0.4 kW, resulting in the lowest value for PCarea. C. Introduction of sleep modes The power consumption of the femtocell networks designed in Section IV-B can be further minimized by introducing sleep modes. When there is little or no activity in its cell, the
- 4. (a) (b) (c) Fig. 3. Energy-efficient femtocell networks for mobile WiMAX (a), HSPA (b), and LTE (c) for a bit rate of 5 Mbps in a 5 MHz channel resulting from the deployment tool. Parameter Mobile HSPA LTE WiMAX Number of femtocells used 38 272 63 Power consumption [kW] 0.4 2.6 0.6 Coverage percentage [%] 97.6 95.9 99.5 P Carea [mW/km2] 3.7 25.0 5.8 TABLE III COMPARISON OF THE ENERGY-EFFICIENT FEMTOCELL NETWORK FOR THE TECHNOLOGIES CONSIDERED AND A PHYSICAL BIT RATE OF 5 MBPS. femtocell base station can enter the sleep mode i.e., a mode wherein the femtocell base station consumes only a small amount of power. When it is necessary, the femtocell base station can be awaken to serve the users in its cell. In [6], three sleep mode techniques are proposed: the sleep mode is controlled by either the femtocell base station itself, or by the core network, or by the UE (user equipment). In this paper, the UE controlled approach will be used as this technique does not need the overlaying macrocell network for signalling the wake-up signals. In the UE controlled approach, the UE can broadcast wake-up signals when it is necessary to wake up a femtocell base station. Although the femtocell base station is in sleep mode, it retains the capability to receive wake-up signals from the UE and, if necessary, to switch to an active state. In the sleep mode, the RF transmitter (1.0 W [6]), the power amplifier (0.9 W, 0.03 W, and 0.4 W for, respectively, mobile WiMAX, HSPA, and LTE) and some other circuit elements (1.7 W [6]) can be switched off, leading to a reduction in power consumption of 3.6 W, 2.8 W, and 3.1 W for, respectively, mobile WiMAX, HSPA, and LTE. 1) Performance of the sleep mode technique: In this section, it is investigated how much power is saved by introducing the sleep mode for one femtocell base station. In the next section, it is studied what the power consumption reduction is in the networks developed in Section IV-B. The same procedure as in [9] is used for this investigation. The average power consumption reduction of the femtocell base station (as a percentage) is defined as [9]: Ω = Psleep Pactive · (1 − η) · 100 (5) with Psleep the power consumption of the femtocell base station in sleep mode (in W), Pactive the power consumption of the femtocell base station in active mode (in W) and η (0 ≤ η ≤ 1) the average duty cycle (i.e., the time that the femtocell base station spends in the active mode). To determine η a traffic model is used [9]. This traffic model indicates when there are periods of high and low traffic during the day. Typically, it is modelled by a Poisson process with exponentially distributed arrival time and exponentially distributed service time (mean value is here chosen to be 3 min). The traffic model used is based on the traffic model proposed in [9] but it is adapted to the office environment here considered (Section II). Table IV gives an overview of the traffic model used in this study. Time Usage Arrival time λ Traffic/user µ [min.] [Erlang] 0h-8h Low 240 0.0125 8h-18h High 30 0.1 18h-0h Low 240 0.0125 TABLE IV TRAFFIC MODEL FOR AN OFFICE ENVIRONMENT. Fig. 4 illustrates the results for Ω for the technologies considered. The maximum power consumption reduction is between 29% and 34% when no users are present. This is a theoretical case as it is not realistic that the femtocell base station sleeps all day. The highest reduction is obtained with mobile WiMAX as the power consumption of the power amplifier is the highest for this base station (as it has the highest input power of the antenna, Table II) and thus more power can be saved in sleep mode compared to HSPA and LTE. Furthermore, the more users are present in the cell, the lower the power consumption reduction is. This is logical due to the fact that when there are more users in a cell, the femtocell base station can sleep less, resulting in a lower reduction. A femtocell base station has typically between 1 to 16 users. Based on the results from Fig. 4, it is concluded that
- 5. the sleep mode introduction is very useful until a boundary of 8 users. For more than 8 users, there is still a power consumption reduction, however, this reduction becomes very small (2-3%). Fig. 4. Power consumption reduction for the femtocell base station of the technologies considered and for various numbers of users supported. 2) Power consumption reduction in femtocell networks: Now, it is determined how much power can be saved in the networks developed in Section IV-B. Therefore, a random number of users between 1 and 8 is allocated to each cell in the network. Based on this number and the results from Section IV-C1, it is calculated how the dialy power consump- tion is reduced for each femtocell in the network and thus what the overall power consumption reduction for the whole network is. Table V gives an overview of the results from this study. Parameter Mobile HSPA LTE WiMAX No. femtocells 38 273 64 No. users in network 198 1180 280 Average no. user 5.2 4.3 4.4 per femtocell Dialy power consumption 9.6 63.3 15.12 without sleep modes [kWh] Daily power consumption 7.1 49.0 11.5 with sleep modes [kWh] Power consumption reduction 25.4% 22.7% 24.2% TABLE V POWER CONSUMPTION REDUCTION IN FEMTOCELL NETWORKS FOR THE TECHNOLOGIES CONSIDERED. Introducing sleep modes in the network results in a power consumption reduction between 22% and 25%. The highest reduction is achieved with mobile WiMAX which corresponds with the results obtained in Section IV-C1. The number of users served is the highest for HSPA as HSPA has the highest number of femtocell base stations in its network. However, the average number of users per femtocell is approximately the same (i.e., 4.6 users per femtocell) for the three tech- nologies considered. If the same number of users is assumed for each network, the network with the highest number of femtocell base stations will correspond with the highest power consumption reduction as it will be possible to let more femtocell base stations sleep. Therefore, the average number of users per femtocell is kept (approximately) the same in this investigation. V. CONCLUSION In this paper, the energy efficiency of femtocell base stations is investigated and compared for three wireless technologies namely, mobile WiMAX, HSPA, and LTE. Therefore, a power consumption model is proposed. Based on this model, it was found that a femtocell base station consumes, in general, approximately 10 W for a range between 9 to 130 m (depend- ing on the technology and the bit rate considered). For the parameters assumed and a 5 MHz channel, mobile WiMAX is the most energy-efficient technology for bit rates higher than 5 Mbps. For bit rates between 2.8 Mbps and 5 Mbps, LTE is the most energy-efficient. The power consumption model is then used in a deploy- ment tool for the development of energy-efficient femtocell networks. For the area considered and a bit rate of 5 Mbps, the mobile WiMAX consumes 105.6 kWh per day, the HSPA network 499.8 kWh, and the LTE network 146.4 kWh. Furthermore, sleep modes are introduced in the network. The sleep mode technique reduces the power consumption significantly when it supports a maximum of 8 users. For a higher number of users, there is still a power consumption re- duction but this reduction is rather small (< 3% for the power consumption of a single femtocell base station). Introducing sleep modes in the network leads to a power consumption reduction of approximately 24%. ACKNOWLEDGMENT The work described in this paper was carried out with the support of the IBBT-project GreenICT. W. Joseph is a Post-Doctoral Fellow of the FWO-V (Research Foundation Flanders). REFERENCES [1] World Wireless Research Forum, WWRF, www.wireless-world- research.org. [2] D. Duffy, Femtocell market set for strong growth in 2011, Informa Telecoms & Media, 2010. [3] M. Deruyck, E. Tanghe, W. Joseph, L. Martens, Modelling and optimization of power consumption in wireless access networks, Elsevier Computer Communications, 2011, doi: 10.1016/j.comcom.2011.03.008. [4] M. Deruyck, E. Tanghe, W. Joseph, W. Vereecken, M. Pickavet, L. Martens, B. Dhoedt, Model for Power Consumption of Wireless Access Networks, IET Science, Measurement & Technology, Vol. 5, Issue 4, July 2011, pp. 155-161. [5] L. M. Correia, D. Zeller, O. Blume, D. Ferling, Y. Jading, I. G´odor, G. Auer, L. Van der Perre, Challenges and Enabling Technologies for Energy Aware Mobile Radio Networks, IEEE Communications Magazine, Vol. 48, Issue 11, November 2010, pp. 66-72. [6] I. Ashraf, F. Boccardi, L. Ho, Power Savings in Small Cell Deployments via Sleep Mode Techniques, 21st Annual IEEE Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 2010): Workshop W-Green, Istanbul, Turkey, September 2010, pp. 306-310. [7] Recommendation ITU-R P.1238-2, Propagation data and prediction methods for the planning of indoor radiocommunication systems and radio local area networks in the frequency range 900 MHz to 100 GHz, 1997-1999-2001.
- 6. [8] M. Deruyck, E. Tanghe, W. Joseph, W. Vereecken, M. Pickavet, B. Dhoedt, L. Martens, Towards a deployment tool for wireless access networks with minimal power consumption, 21st Annual IEEE Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 2010): Workshop W-Green, Istanbul, Turkey, September 2010, pp. 294-299. [9] I. Ashraf, L. Ho, H. Claussen, Improving Energy Efficiency of Femtocell Base Stations via User Activity Dection, Wireless Communications and Networking Conference (WCNC 2010), Sydney, Australia, April 2010, pp. 1-5.