Power consumption model for macrocell and microcell base stations

TRANSACTIONS ON EMERGING TELECOMMUNICATIONS TECHNOLOGIESTrans. Emerging Tel. Tech. (2012)

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/ett.2565

RESEARCH ARTICLE

Power consumption model for macrocell andmicrocell base stationsMargot Deruyck*, Wout Joseph and Luc Martens

Department of Information Technology (INTEC), Ghent University/IBBT, Gaston Crommenlaan 8, 9050 Ghent, Belgium

ABSTRACT

In this paper, a power consumption model for both macrocell and microcell base stations is proposed. This model isvalidated by temporal power measurements on actual base stations, and an excellent agreement is obtained. Furthermore,the power consumption’s evolution during the day is investigated by means of these measurements. The energy efficiencyof three different wireless technologies is compared namely mobile Worldwide Interoperability for Microwave Access,Long Term Evolution and High Speed Packet Access. With the model proposed, the deployment tool Green Radio AccessNetwork Design is implemented, which allows to design an energy-efficient access network for a predefined area.

In general and with the assumptions made, a macrocell base station consumes about 4.4 times more than a microcellbase station. However, a microcell base station is less energy efficient than a macrocell base station because of its lowercoverage range. Despite this, it is still useful to introduce them in the network as the same coverage can be obtainedwith a lower total power consumption than with a network where only macrocell base stations are used. Copyright © 2012John Wiley & Sons, Ltd.

*Correspondence

Margot Deruyck, Department of Information Technology (INTEC), Ghent University/IBBT, Gaston Crommenlaan 8, 9050 Ghent,Belgium.E-mail: [email protected]

Received 17 February 2012; Revised 11 April 2012; Accepted 5 July 2012

1. INTRODUCTION

An operator’s wireless access network has typically a hier-archical structure with different base station types such asthe macrocell, the microcell and the picocell base station.As 9% of the Information and Communication Technologypower consumption is caused by radio access networks [1],it is important to determine the power consumption of thewhole wireless access network and thus to model the powerconsumption of each part of this network.

In this study, a power consumption model is proposedfor both the macrocell and the microcell base stations.With this model, the energy efficiency can be comparedfor various bit rates and for various wireless technologies,such as mobile Worldwide Interoperability for MicrowaveAccess (WiMAX) [2], Long Term Evolution (LTE) [3]and High Speed Packet Access (HSPA) [4]. Further-more, the evolution of the power consumption duringthe day is investigated experimentally using temporalpower measurements of both an actual macrocell andmicrocell base stations. The predicted power consump-tion will be compared with these measured values. Thispower consumption model is then used in the deploymenttool Green Radio Access Network Design (GRAND).

GRAND allows to design energy-efficient wireless accessnetworks (i.e. with minimal power consumption) for apredefined area.

The remainder of this paper is organized as follows. InSection 2, a power consumption model is proposed forboth the macrocell and the microcell base stations, andthe temporal behaviour of the power consumption is char-acterized on the basis of temporal power measurements.Section 3 describes the deployment tool GRAND intodetails. In Section 4, the power consumption of differentwireless technologies (obtained with the model fromSection 2) is compared, and the influence of the intro-duction of microcell base stations in the wireless accessnetworks of the technologies considered is investigated(on the basis of the GRAND tool described in Section 3).Section 6 gives the conclusion.

2. METHOD

2.1. Energy efficiency of a base station

An important parameter to investigate is the energy effi-ciency of a base station. The energy efficiency is here

Copyright © 2012 John Wiley & Sons, Ltd.

M. Deruyck, W. Joseph and L. Martens

defined as the power consumption needed to cover a cer-tain area (in W/m2). The power consumption PCarea percovered area is then defined as [5]

PCarea DPel

� �R2(1)

with Pel the power consumption of the base station andR the range of the base station. The lower the PCarea,the more energy efficient the base station is. Note that thepower consumption over the number of information bits isalso commonly used in literature [6]. Here, it is assumedthat the users are uniformly distributed and all requiringthe same bit rate. In this case, this metric is equivalent toPCarea. These assumptions are proposed to make a faircomparison between the energy efficiency of the individ-ual macrocell and microcell base stations and between themacrocell and the hybrid macrocell microcell networks.

2.2. Power consumption of a macrocellbase station

2.2.1. Theory.

A base station is here defined as the equipment neededto communicate with the mobile stations and with thebackhaul network [5]. The area covered by a base stationis called a cell, which is further divided into a numberof sectors. Each sector is covered by a sector antenna,which is a directional antenna with a sector-shapedradiation pattern.

A macrocell base station typically consists of differ-ent power-consuming components. These components areshown in Figure 1(a) and can be divided into two groups.The first group is the equipment that is used per sector suchas the digital signal processing (responsible for system pro-cessing and coding), the power amplifier, the transceiver

(responsible for sending and receiving of the signals to themobile stations) and the rectifier. The power consumptionof these components has to be multiplied by the numberof sectors nsector supported by the base station. A macro-cell base station has typically three sectors. The secondgroup is the equipment that is common for all sectors andis thus used once, such as the microwave link and the airconditioning. The microwave link is used for the commu-nication with the backhaul network, and nowadays, it isoften replaced by a fibre link. In this study, the worst-casescenario for a base station is investigated, and thus, it isassumed that a microwave link is present.

Considering the power consumption of each component,the components can also be divided in the followingcategories. The first category consists of the componentswhose power consumption is not load dependent such asthe rectifier, the microwave link and the air conditioning.The load on a base station is determined by the number ofusers and the services they use in the base station’s cell.The higher the number of users and the higher the require-ments for the services, the higher the load is. The powerconsumption of these load-independent components is thusconstant throughout time. However, a nuance should bemade for the air conditioning. The power consumption ofthis component depends on the internal and ambient tem-perature of the base station cabinet (on the basis of infor-mation retrieved from data sheets of a manufacturer [7]).An internal and ambient temperature of 25ıC is assumed,which results also in a constant power consumption forthe air conditioning. The second category is the equipmentthat has a load-dependent power consumption (denoted asPel/load) such as the power amplifier, the transceiver andthe digital signal processing. The power consumption ofthese components should be scaled with a factor Fi , whichis here called the load factor, to determine the power con-sumption for a specific hour at weekdays or weekends.

(a) (b)

Figure 1. Block diagram of the macrocell (a) and the microcell (b) base station.

Trans. Emerging Tel. Tech. (2012) © 2012 John Wiley & Sons, Ltd.DOI: 10.1002/ett


The power consumptionPel/amp of the power amplifier alsodepends on the input power PTx of the antenna and iscalculated as follows (in Watt):

Pel/amp DPTx

�(2)

with � the efficiency of the power amplifier, which isthe ratio of the RF output power to the electrical inputpower [8].

Once the power consumption of each component isknown, the power consumption Pel/macro of the macrocellbase station can be determined as follows (in Watt):

Pel/macro D Pel/const CPel/load �Fi (3)

with

Pel/const D nsector �Pel/rect CPel/link CPel/airco

Pel/load D nsector � .nTx � .Pel/ampCPel/trans/

CPel/proc/

nsector is the number of sectors supported by the macrocellbase station, Fi (i = 0.23, 0 6 Fi 6 1) the load factor,nTx the number of transmitting antennas and Pel/rect,Pel/link, Pel/airco, Pel/amp, Pel/trans and Pel/proc the powerconsumption of, respectively, the rectifier, the microwavelink (if present), the air conditioning, the power amplifier,the transceiver and the digital signal processing (in Watt).

Table I summarizes the typical power consumption ofthe different components for the technologies considered.These values are retrieved from data sheets of variousnetwork equipment manufacturers and private interviewswith the operator [9–14].

2.2.2. Measurements.

To determine the load factor Fi , measurements are per-formed for an actual HSPA macrocell base station in thesuburban area of Ghent, Belgium. During 6 days (5 week-days and 1 weekend day), the power consumption of thebase station is measured. The group of load-independentcomponents, that is, the rectifier, the air conditioning and

the microwave link, was not included in these measure-ments. For the equipment considered, the voltage is con-stant (i.e. approximately 54 V), and thus, only the currentwas measured. The power consumption P .t/ at a certaintime t is then determined as follows (in Watt):

P .t/D V � I .t/ (4)

with V the voltage (in volt) and I the current at time t(in ampere).

The current was measured with an AC/DC current clamp(Fluke i410). Every second, the value of the current wassaved, which results in 466 319 samples for the measure-ment period of 6 days. Figure 2 shows the power consump-tion (in Watt) as a function of the time. In Figure 2, days1–5 are weekdays, and day 6 is a weekend day. For theweekdays, it is noticed that the power consumption at night(i.e. from midnight till 8 AM) is lower than during the day(i.e. from 8 AM till midnight) because during the day morepeople are active than at night. Also during weekdays,more people are coming to town, for example, to work. Forthe weekend day, this difference in power consumption issmaller than for weekdays.

With the data shown in Figure 2, the average power con-sumption per hour for weekdays and for the weekend iscalculated. These averages are presented in Table II. Themeasured equipment consumes between 1016 and 1087 W.

Determining the power consumption with our model(Equation (3), Fi D 1, PTx D43 dBm, including load-independent equipment), and the data from Table I resultin a power consumption of 1673 W. Excluding the powerconsumption of the rectifiers, the microwave link and theair conditioning result in a power consumption of 1068 W,which agrees very well (standard deviation of 2.0%) withthe range of power consumptions in Table II. The load fac-tor F10 for the time between 10 and 11 AM on a weekdayis chosen to be 1 as this time interval corresponds withthe highest power consumption (which corresponds withan average of 165 voice calls and 766 data calls duringthis hour). The minimum, maximum and average num-ber of voice and data calls for different time periods areshown in Table III. As expected, the highest number of

Table I. Power consumption of the macrocell and microcell base stationcomponents for the technologies considered (mobile WiMAX, HSPA and LTE).

Macrocell Microcell

Equipment base station base station

Number of sectors nsector 3 1Number of transmitting nTx 1 1antennas per sectorDigital signal processing Pel/proc 100 W 100 WPower amplifier � 12.8 % 12 %Transceiver Pel/trans 100 W 100 WRectifier Pel/rect 100 W 100 WAir conditioning Pel/airco 225 W 60 WMicrowave link Pel/link 80 W —



Figure 2. Evolution of the power consumption in time (days 1–5 weekdays; day 6 weekend day).

Table II. Average power consumption per hour during weekdays and weekends for a macrocell and a microcell base stations(only load-dependent equipment).

Macrocell base station Microcell base station

Hour Weekday (W) Weekday (W) Factor week Weekday (W) Weekday (W) Factor week

0 AM–1 AM 1018 1029 F0 = 0.94 220 227 F0 = 0.96

1 AM–2 AM 1017 1028 F1 = 0.94 214 224 F1 = 0.93

2 AM–3 AM 1017 1027 F2 = 0.94 221 220 F2 = 0.97

3 AM–4 AM 1016 1026 F3 = 0.93 220 221 F3 = 0.96

4 AM–5 AM 1016 1024 F4 = 0.93 221 227 F4 = 0.97

5 AM–6 AM 1017 1024 F5 = 0.94 217 226 F5 = 0.95

6 AM–7 AM 1020 1025 F6 = 0.94 224 218 F6 = 0.98

7 AM–8 AM 1026 1025 F7 = 0.94 221 224 F7 = 0.97

8 AM–9 AM 1044 1027 F8 = 0.96 229 232 F8 = 1.00

9 AM–10 AM 1057 1032 F9 = 0.97 219 223 F9 = 0.96

10 AM–11 AM 1087 1031 F10 = 1.00 225 225 F10 = 0.98

11 AM–12 PM 1073 1033 F11 = 0.99 222 229 F11 = 0.97

12 PM–1 PM 1055 1036 F12 = 0.97 219 229 F12 = 0.96

1 PM–2 PM 1060 1035 F13 = 0.98 226 223 F13 = 0.99

2 PM–3 PM 1070 1037 F14 = 0.98 219 222 F14 = 0.96

3 PM–4 PM 1065 1035 F15 = 0.98 221 224 F15 = 0.97

4 PM–5 PM 1049 1038 F16 = 0.97 220 227 F16 = 0.96

5 PM–6 PM 1044 1036 F17 = 0.96 221 226 F17 = 0.97

6 PM–7 PM 1042 1034 F18 = 0.96 222 223 F18 = 0.97

7 PM–8 PM 1034 1039 F19 = 0.95 220 221 F19 = 0.96

8 PM–9 PM 1034 1033 F20 = 0.95 219 225 F20 = 0.96

9 PM–10 PM 1033 1033 F21 = 0.95 221 228 F21 = 0.97

10 PM–11 PM 1033 1030 F22 = 0.95 219 228 F22 = 0.96

11 PM–0 AM 1024 1030 F23 = 0.94 221 223 F23 = 0.97



Table III. Overview of the maximum, minimum and average number of voice and data calls, and Erlang, respectively, for a macrocelland microcell base stations for different periods.


Minimum Maximum Average Minimum Maximum Average

No. voice No. data No. voice No. data No. voice No. data Erlang Erlang Erlang

Weekday 0 1163 227 28 86 375 0 3.093 0.861Weekend day 0 44 79 207 23 113 0 1.088 0.328Daytime 7 28 227 311 105 109 0.015 3.093 0.834Night 0 44 59 1163 7 422 0 1.078 0.092

voice and data calls is found during weekdays and at day-time. The load factor Fi (i D 0::23) for each hour of aweekday can then be determined. These load factors canbe found in Table II. One might argue that Fi representsthe impact of the time on the power consumption of thebase station rather than the impact of the load. However,a different point in time will result in a different load asshown in [15] and thus reflects Fi implicitly the impact ofthe load. Furthermore, it is also important that the Fi modelis representative for other macrocell base stations in thesuburban city environment as well. For other environmentssuch as residential areas, rural terrain and industrial envi-ronment, a new model for Fi should be determined follow-ing the previously described procedure. This is a similarapproach as in [15] where for each environment, one site isselected to perform measurements. Here, the suburban cityenvironment is chosen as this is the most interesting caseto investigate.

For the weekend, the fluctuations in power consump-tion are small as shown in Table II because the differencebetween the number of active users at night and duringthe day is smaller than for weekdays. For example, dur-ing the weekend, offices are closed and less people arecoming to town. Therefore, only one average load fac-tor F is determined for the weekend. The average powerconsumption for the weekend is 1031 W per hour (exclud-ing load-independent equipment, average of 22 voice and113 data calls during the weekend), which correspondswith F equalling 0.95. Table III shows that there are lessusers active during a weekend day, which results thus in alower F .

Note that the off-peak load for a base station is verylow compared with the peak value (Table III), althoughthis cannot be seen in the evolution of the power consump-tion (Figure 2) because variations in traffic are not alwaysfollowed by similar variations in energy consumption asalready concluded by [16, 17].

2.3. Power consumption of a microcellbase station

2.3.1. Theory.

Analogously as for the macrocell base station, a micro-cell base station consists of several power-consuming com-ponents, which are shown in Figure 1(b). The following

components are present: the transceiver, the digital signalprocessing, the power amplifier, the rectifier and the airconditioning. However, the air conditioning is not alwaysrequired. In this paper, the worst-case scenario is inves-tigated as mentioned before, and therefore, the air condi-tioning is included. In contrast to a macrocell base station,a microcell base station supports only one sector (omni-directional antenna), and each component is thus usedonce. These assumptions are made on the basis of theconfidential information retrieved from an operator.

The power consumption Pel/micro of a microcell basestation is then determined as follows (in Watt):

Pel/micro D Pel/const CPel/load �Fi (5)

with

Pel/const D Pel/rect CPel/airco

Pel/load D Pel/amp CPel/trans CPel/proc

with Fi (i = 0..23, 0 6 Fi 6 1) the load factor and Pel/rect,Pel/airco, Pel/amp, Pel/trans and Pel/proc the power consump-tion of, respectively, the rectifier, the air conditioning,the power amplifier, the transceiver and the digital signalprocessing (in Watt). Pel/amp is again determined withEquation (2). Table I summarizes the typical power con-sumption of the different components. These values aresimilar to those of the macrocell base station, except forthe air conditioning. The cooling power of the air condi-tioning is significantly lower for a microcell base station,resulting in a lower power consumption (on the basis ofprivate interviews with the operator). The load factor Fi isagain determined on the basis of measurements describedin the next section.

2.3.2. Measurements.

As mentioned before, the load factor Fi for the microcellbase station is also determined on the basis of measure-ments in a suburban environment. The same measurementprocedure as for the macrocell base station is used(Section 2.2.2). However, for the microcell base station, thevoltage amounts to 48 V instead of 54 V as for the macro-cell base station. During these measurements, 469 919samples were obtained. Figure 3 illustrates the power con-sumption (in Watt) as a function of time. Although the



Figure 3. Evolution of the power consumption in time (days 1–5 weekdays; day 6 weekend day).

power consumption fluctuates again over time, the differ-ence in power consumption during day and night, as noti-cied by the macrocell base station, is not present for themicrocell base station.

With the data shown in Figure 3, the average power con-sumption per hour for weekdays and for the weekend iscalculated. These averages are presented in Table II. Themeasured equipment consumes between 214 and 232 W.

Determining the power consumption with our model(Equation (5), Fi D 1, PTx D33 dBm, including load-independent equipment) and the data from Table I resultsin a power consumption of 377 W. Excluding the powerconsumption of the rectifiers, the microwave link and theair conditioning results in a power consumption of 217 W,which agrees again excellent (a standard deviation of 1.4%)with the range of power consumptions in Table II. Thepower consumption obtained with our model correspondswith the average power consumption for weekdays duringthe following hours: between 5 AM and 6 AM, 9 AM and10 AM, 12 PM and 1 PM, 2 PM and 3 PM, 8 PM and 9 AM,and 10 PM and 11 PM. This also validates our model. Theload factor F8 for the time between 8 AM and 9 AM is setto be 1 as this is the time interval with the highest powerconsumption (average of 0.44 Erlang). Table III shows theminimum, maximum and average Erlang for different peri-ods. The same tendency as for the macrocell base station,that is, during the week and during daytime more users areactive, can be noticed. The other load factors can be foundin Table II.

For the weekend, the fluctuations in power consump-tion are small as shown in Table II. Therefore, again,only one average load factor F is determined for thesehours. The average power consumption for the weekend is

225 W, which corresponds with F equals to 0.98 (averageof 0.33 Erlang).

2.4. Range of a base station

2.4.1. Link budget.

To determine the coverage of a base station, a link bud-get has to be calculated. A link budget takes into account allof the gains and the losses from the transmitter through themedium to the receiver. For this determination, the max-imum allowable path loss PLmax to which a transmittedsignal can be subjected while still being high enough atthe receiver has to be calculated. It includes all the possi-ble elements of loss associated with interactions betweenthe transmit and the receive antennas and is defined as theratio of the transmitted power to the received power of thesignal [18]. Table IV summarizes the link budget parame-ters for the coverage calculations of both the macrocell andmicrocell base stations for all technologies considered.

Some parameters of Table IV have the same value forboth the macrocell and the microcell base stations becausethese parameters are either technology dependent (such asthe frequency and bandwidth), or mobile station dependent(such as the antenna gain and the feeder loss of the mobilestation) or based on environmental assumptions (such asthe fade margin). The values for the input power PTx ofthe antenna and the antenna gain of the base station dif-fer between a macrocell and a microcell base stations. Amicrocell base station has a typical PTx of 2 W; also, 6 Wis possible. In this study, a PTx of 2 W is used, whichcorresponds with 33 dBm [19]. As mentioned earlier, amicrocell base station supports only one sector; therefore,an omnidirectional antenna is used. The antenna gain for



Table IV. Link budget table for a macrocell and microcell base stations for the technologies considered.


Mobile HSPA LTE Mobile HSPA LTEParameter WiMAX WiMAX

Frequency (MHz) 2500 2100 2600 2500 2100 2600Maximum input power 35 43 43 33 33 33of base station PTx (dBm)Effective input power 35 24.7 43 33 13.8 33of base station PTCH

Tx (dBm)Antenna gain of base station (dBi) 16 17.4 18 6 5 4Antenna gain of mobile station (dBi) 2 0 0 2 0 0Soft handover gain (dB) 0 1.5 0 0 1.5 0Feeder loss of base station (dB) 0.5 0 2 0.5 0 2Feeder loss of mobile station (dB) 0 0 0 0 0 0Fade margin (dB) 10 10 10 10 10 10Interference margin (dB) 2 2 2 2 2 2Bandwidth (MHz) 5 5 5 5 5 5Receiver SNR (dB) [6, 8.5, 11.5 [�3.1, 0.1, 3.4 [�1.5, 3, 10.5 [6, 8.5, 11.5 [�3.1, 0.1, 3.4 [�1.5, 3, 10.5

15, 19, 21]a 6, 7.1, 9.6 14, 19, 23 15, 19, 21]a 6, 7.1, 9.6 14, 19, 2315.6]b 23, 29.4]c 15.6]b 23, 29.4]c

Number of used subcarriers 360 — 301 360 — 301Number of total subcarriers 512 — 512 512 — 512Noise figure of mobile station (dB) 7 9 8 7 9 8Implementation loss 2 0 0 2 0 0of mobile station (dB)Duplexing TDD

(a)[1/2 QPSK, 3/4 QPSK, 1/2 16-QAM, 3/4 16-QAM, 2/3 64-QAM, 3/4 64-QM].(b)[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].(c)[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].

this type of antennas varies from 4 to 6 dB depending onthe technology.

2.4.2. Transmission modes.

In mobile telecommunication systems, the mobile sta-tions typically support different transmission modes. Thesetransmission modes are determined by a coding rate anda modulation scheme. For wireless communication, thebinary bit stream has first to be translated into an analoguesignal, which can be performed by using a modulationscheme such as Quadrature Phase Shifting Keying (QPSK)and 16-Quadrature Amplitude Modulation (QAM) or64-QAM. Furtermore, a coding rate is used for ForwardError Correction, which is responsible for the correction oferrors occurred. The coding rate indicates how many re-dundant bits will be added per number of information bits.The modulation scheme and the coding rate have an impor-tant role in the determination of the physical bit rate. Eachtransmission mode is characterized by a receiver Signal-to-Noise Ratio (SNR), which represents the required SNRat the receiver for a certain Bit Error Rate (here, a BitError Rate of 10�6 is assumed). It is thus a determiningfactor for the required signal power that indicates whatthe signal power level is for just acceptable communica-tion quality. Table IV gives an overview of the differenttransmission modes supported by each technology and the

corresponding receiver SNR. The higher the modulationscheme and coding rate, the higher the required receiverSNR, the higher the bit rate and the lower the range, whichresults in a lower energy efficiency as illustrated in [5, 20].

2.4.3. Interference.

To take the interference into account, a special margin,the interference margin, is introduced in the link budget(Table IV). This margin represents the interference a usermight experience because of other users in the same cell,as well as the interference caused by neighbouring cells.Increasing coverage is the main objective in the compar-ison, and thus, the same interference margin (i.e. 2 dB)is assumed for both the macrocell and the microcell basestations as carried out in [19]. When the purpose is toincrease throughput, the microcell base station will bemainly located within the cell covered by the macrocellbase station (the so-called umbrella structure), and a moreexhaustive study of the inference will be needed [21].

2.4.4. Scenario.

Once PLmax is known, the range can be determined byusing a propagation model. For the macrocell base station,the Erceg C model is used [22], whereas for the micro-cell base station, the Walfisch–Ikegami (WI) model is more



Table V. Suburban scenario under consideration.

Parameter Macrocell base station Microcell base station

Area type Suburban SuburbanHeight of the base station hBS (m) 30 6Height of the mobile station hMS (m) 1.5 1.5Coverage requirement 90% 90%Propagation model Erceg C W-IShadowing margin (dB) 13.2 12.8Building penetration loss (dB) 8.1 8.1

suitable [23]. The characteristics of the scenario consideredare listed in Table V. A suburban area is assumed with aheight of 1.5 m for the mobile stations. For the macro-cell base stations, a height of 30 m is assumed, whereasfor the microcell base station, a height of 6 m is chosen,which corresponds with the height of the roof-gutter of atwo-storied house (i.e. 3 m per floor). Both base stationsare placed outdoor. Furthermore, for the mobile stations, anindoor residential scenario is assumed. The mobile stationsconsidered are laptops equipped with a Wireless NetworkInterface Card.

3. OPTIMIZATION TOOL FORGREEN WIRELESS ACCESSNETWORKS

The model described in Section 2 is used in the deploymenttool GRAND for design and optimization of green wire-less access networks. The purpose of the tool is to covera specific area, the target area, as energy-efficiently (i.e.with minimal power consumption) as possible. This toolis based on the work presented in [5]. The tool can opti-mize existing networks of operators, or it can support thedeployment of a new network. The investigation of the lat-ter is the focus of this study. In [5], optimization of existingnetworks is discussed.

The algorithm for determination of the energy-efficientnetwork is based on a genetic search algorithm. A flowchart of our algorithm can be found in Figure 4.

Firstly, the target area is loaded in shapefile format(Block 1 in Figure 4). A shapefile stores non-topologicalgeometry and attribute information for the spatial features

in the data set [24]. The geometry of a feature is storedas a shape comprising a set of vector coordinates. Theshapefile contains, amongst others, information about theposition, the form and the height of the buildings in the tar-get area. An example can be found in Figure 5 showing athree-dimensional shapefile of Ghent, Belgium. With thisinformation, a set of possible locations for the base stationpositions is determined (Block 2 in Figure 4). A base sta-tion can be placed on each building with a height equal toor higher than the height chosen by the user. This thresh-old can be chosen freely; however, the lower this threshold,the larger the set of possible base station locations and thusthe longer the calculation time. The antenna of both themacrocell and the microcell base stations will be placed atthe height of the building.

Secondly, a first set or population of solutions is gener-ated by using mutations (Block 3 in Figure 4). Each solu-tion of this population is a possible network to cover thetarget area. The size of the population is defined by the userself. Initially, all base stations are inactive. Through vari-ous mutations different solutions are created. Six differentmutations are defined in the GRAND tool:

� Make an inactive base station active or an active basestation inactive.

� Add 1 dBm to or remove 1 dBm from the input powerof the antenna of an active base station (maximumvalue equals one of those in Table IV; the minimumvalue is 0 dBm).

� Change the cell type of the base station (a macro-cell base station becomes a microcell base station andvice versa).

Figure 4. Flow chart of the GRAND algorithm.



(a) (b)

Figure 5. Shapefile of the target area for the GRAND tool (a) and a subset of possible locations for the base stations (b).

Which base station and how it will be adapted are deter-mined at random. The mutations are also used in step 4(Figure 4) of the algorithm.

Next, this initial population will be improved throughgenerations (Block 4 in Figure 4). A new generation pop-ulation is created by mutating solutions of the previousgeneration, by selecting the best solutions in the previousgeneration (on the basis of a fitness function (Block 5 inFigure 4)) and by the crossover operator until the chosenpopulation size is reached again. The new population isthen used in the next iteration of the algorithm until thestop condition is met (Block 7 in Figure 4). More infor-mation about the selection procedure, the fitness function,the crossover operator and the stop condition, can be foundin [5]. Note that the model described in Section 2 is usedfor determining the solution fitness, which measures the

quality of the solution (Block 6 in Figure 4) in terms ofcoverage and power consumption.

Once the stop condition is met, the algorithm enters itslast stage (Block 8 in Figure 4). It was chosen to use aheuristic, that is, a genetic search algorithm, as it is veryhard and time consuming to solve the problem in a deter-ministic way. Because of the heuristic nature of the algo-rithm, it is possible that some base stations cover a pieceof the target area that is already covered by other base sta-tions. The purpose of the last step of the algorithm is toremove this kind of overlap.

The final result of the tool is a shapefile where the activebase stations are indicated as well as the coverage of eachbase station. As it is assumed that the antenna transmits thesame amount of power in each direction, the coverage areaof each base station is visualized by a circle. Furthermore,

1590

1600

1610

1620

1630

1640

1650

1660

1670

1680

0:00

1:00

2:00

3:00

4:00

5:00

6:00

7:00

8:00

9:00

10:0

011

:0012

:0013

:0014

:0015

:0016

:0017

:0018

:0019

:0020

:0021

:0022

:0023

:00

Time [hours]

Pow

er c

onsu

mpt

ion

Pel

[W]

Pow

er c

onsu

mpt

ion

Pel

[W]

Macrocell weekdayMacrocell weekend day

(a)

360

362

364

366

368

370

372

374

376

378

380

0:00

1:00

2:00

3:00

4:00

5:00

6:00

7:00

8:00

9:00

10:0

011

:0012

:0013

:0014

:0015

:0016

:0017

:0018

:0019

:0020

:0021

:0022

:0023

:00

Time [hours]

Microcell weekdayMicrocell weekend day

(b)

Figure 6. Evolution of the power consumption during a weekday and the weekend for a HSPA macrocell (a) and microcell(b) base stations.



the tool also returns a file with technical information abouteach base station (e.g. input power of the antenna, heightof the antenna and power consumption of the base station)

4. RESULTS

4.1. Evolution of the power consumptionduring the day

In this section, it is investigated how the power consump-tion of a macrocell and a microcell base stations fluctu-ates during a weekday and a weekend day. Figure 6(a)presents the power consumption of a HSPA macrocell andFigure 6(b) the power consumption of a HSPA microcellbase station as a function of the time of the day for both aweekday and a weekend.

Figure 6 shows that the power consumption Pel fora microcell base station (maximum 377 W) is signifi-cantly lower than for a macrocell base station (maximum1673 W). The power consumption of the macrocell basestation is thus 4.4 times higher than for the microcellbase station because of the higher power consumption ofits air conditioning, its higher input power PTx of theantenna and the presence of a microwave link and multiplesectors (Section 2).

For a weekday, the lowest power consumption (i.e.1609 W) for a macrocell base station is obtained between3 AM and 5 AM. The highest power consumption (i.e.1673 W) is found between 10 AM and 11 AM. For themicrocell base station, the highest power consumption (i.e.377 W) is reached between 8 AM and 9 AM, and the low-est power consumption (i.e. 362 W) between 1 AM and2 AM. However, the course of the power consumption fora microcell base station varies more than for the macrocellbase station. This is due to the fact that when there is too

much traffic on one of the neighbouring macrocell base sta-tions, some of the traffic is diverted to the microcell basestation. In general, the fluctuations in power consumptionare limited for the microcell base station compared withthe macrocell base station (Section 2.3.2).

4.2. Comparison of the energy efficiencybetween macrocell and microcell basestations

The energy efficiency of a macrocell and a microcell basestations is here investigated for the technologies consid-ered and for different bit rates. The suburban scenario ofTable V is considered. To make a fair comparison betweenthe macrocell and the microcell base stations, the powerconsumption accumulated over 1 week for both base sta-tions is considered. Pel in Equation (1) will thus corre-spond with the total power consumption for 1 week ofthe macrocell and the microcell base stations, respectively.Figure 7 presents the power consumption PCarea per cov-ered area as a function of the bit rate for both the macrocelland the microcell base stations in a 5 MHz channel.

With our assumptions, Figure 7 shows that the macro-cell base station is more energy efficient (i.e. a lowerPCarea) than the microcell base station despite its higherpower consumption (Section 4.1). PCarea is about fourto 18 times lower for a macrocell base station than fora microcell base station (for e.g. 0.27 mWh/m2 versus1.52 mWh/m2 for mobile WiMAX and 3.8 Mbps) becauseof the higher range obtained by the macrocell base station.The range for the macrocell base station is four to eighttimes higher than for a microcell base station dependingon the technology considered because of its higher inputpower PTx, its higher antenna gain and its higher antennaheight (Table IV).

0 2 4 6 8 10 12 14 16 180

0.5

1

1.5

2

2.5

Bit rate [Mbps] Bit rate [Mbps]Pow

er c

onsu

mpt

ion

PC

area

per

cov

ered

are

a [m

Wh/

km2 ]

Pow

er c

onsu

mpt

ion

PC

area

per

cov

ered

are

a [m

Wh/

km2 ]

LTE macroHSPA macroMobile WiMAX macro

(a) (b)

0 2 4 6 8 10 12 14 16 180

5

10

15

20

25LTE microHSPA microMobile WiMAX micro

Figure 7. Comparison of the power consumption PCarea per covered area of a macrocell (a) and microcell (b) base stations in a 5-MHzchannel for different technologies.



For the microcell base station, mobile WiMAX is themost energy efficient technology for bit rates higher than3.8 Mbps (1.91 mWh/m2 for about 5.5 Mbps versus3.96 mWh/m2 for LTE and 6.39 mWh/m2 for HSPA).Mobile WiMAX performs better because of its higherantenna gain for both the base and the mobile stations(Table IV). In general and for the case considered here,HSPA is less energy efficient (higher value for PCarea)than mobile WiMAX and LTE because of its lower effec-tive input power of the antenna P TCH

Tx . P TCHTx is the power

reserved by the base station for the traffic channels andis lower for HSPA because it uses Wideband Code Divi-sion Multiple Access, whereas mobile WiMAX and LTEboth use Scalable Orthogonal Frequency Division MultipleAccess [5].

LTE is the most energy efficient for bit rates between 2.8and 3.8 Mbps (PCarea is 1.30 mWh/m2 for about 2.5 Mbpsversus 3.28 mWh/m2 for HSPA). Bit rates below 2.8 Mbpsare only supported by HSPA (PCarea is 2.43 mWh/m2 for1.3 Mbps) and not by mobile WiMAX or LTE.

For the macrocell base station, mobile WiMAX is themost energy efficient for bit rates higher than 11.5 Mbps(versus 3.8 Mbps for a microcell base station) because ofits lower receiver SNR resulting in a higher range. For bitrates between 2.8 and 11.5 Mbps, LTE is the most energyefficient, which is also due to its lower receiver SNR forthis range of bit rates.

4.3. Application: prediction of the powerconsumption of a wireless access networkin Ghent

In this section, it is investigated how much electrical poweris needed to cover a part of the city of Ghent, Belgium.

The target area is shown in Figure 5(a) and has a surface of13.3 km2. A coverage requirement of 90% is assumed. Foreach technology considered, a new network is deployed,once with only macrocell base stations and once with bothmacrocell and microcell base stations. The set of possi-ble locations for the base stations is all the buildings inthe target area with a height equal to or higher than 10 m,which results in 13 437 possible locations. Some of theselocations are shown in Figure 5(b). To make a fair com-parison between the different technologies, a bit rate of(approximately) 5 Mbps is assumed in a 5 MHz chan-nel. This corresponds with the 2/3 QPSK, the 3/4 QPSKand the 1/2 16-QAM modulation for mobile WiMAX, LTEand HSPA, respectively. Furthermore, the algorithm willstop when 1000 generations are generated or when thesimulation lasts longer than 1 hour.

Figure 8 gives a visual overview of the energy-efficientnetwork with only macrocell (a) and both macrocell andmicrocell base stations (b) for mobile WiMAX resultingfrom the GRAND tool. Numerical results for all consideredtechnologies are summarized in Table VI.

Table VI shows that a higher energy efficiency (lowerPCarea) is obtained when microcell base stations are intro-duced because of the lower total power consumption forthe same coverage. The lowest total power consumptionfor combined macrocell and microcell base stations isfound for LTE (81.9 kW), followed by mobile WiMAX(83.8 kW) and HSPA (178.0 kW). For HSPA, the energyefficiency gain is very limited as the coverage of the net-work with both macrocell and microcell base stations ismuch lower than when only macrocell base stations areused (94.9 % versus 97.7 %).

Figure 9 shows the time behaviour of the power con-sumption by illustrating the evolution of PCarea duringa week day for the network with only macrocell base sta-tions and for the network with both the macrocell and the

(a) (b)

Figure 8. Energy-efficient network with only macrocell base stations (a) and with both macrocell and microcell base stations (b) formobile WiMAX resulting from the GRAND tool.



Table VI. Comparison between the networks with only macrocell base stations and with both macrocell and microcell base stationsfor the technologies considered (5 Mbps in a 5-MHz channel).

Mobile WiMAX Mobile WiMAX LTE LTE HSPA HSPAParameter macro macro + micro macro macro + micro macro macro + micro

Number of macrocells (�) 78 66 52 50 114 109Number of microcells (�) — 4 — 2 — 3Average total 97.2 83.8 83.2 81.9 184.8 178.0power consumption (kW)Coverage (%) 97.6 97.4 99.3 99.1 97.7 94.9Average PCarea (kW/km2) 7.5 6.5 6.3 6.2 14.2 14.1

0 5 10 15 20 253.5

4

4.5

5

5.5

6

6.5

7

7.5

8

Hour of the day [h] Hour of the day [h]

Hour of the day [h]

PC

area

[kW

/km

2 ]

PC

area

[kW

/km

2 ]

PC

area

[kW

/km

2 ]

Mobile WiMAX macroMobile WiMAX macro + micro

(a)

0 5 10 15 20 256.1

6.15

6.2

6.25

6.3

6.35

6.4

6.45

6.5

6.55LTE macroLTE macro + micro

(b)

0 5 10 15 20 2513.8

13.9

14

14.1

14.2

14.3

14.4

14.5

14.6

14.7

14.8HSPA macroHSPA macro + micro

(c)

Figure 9. Evolution of PCarea of the networks as a function of the hour of the day with only macrocell base stations and with bothmacrocell and microcell base stations for mobile WiMAX (a), LTE (b) and HSPA (c).

microcell base stations. PCarea of each network followsthe curve of the macrocell base station’s power consump-tion (as given in Figure 6(a)). For the networks with bothmacrocell and microcell base stations, this is due to the factthat there is a much larger amount of macrocell base sta-tions present in the network compared with the microcellbase stations. Furthermore, the power consumption of the

microcell base stations varies less compared with that of amacrocell base station as mentioned before.

5. RELATED WORK

As increasingly more attention is drawn to the power con-sumption in wireless access networks, the topic starts to



appear in literature [16, 19, 25, 26]. In [25], a model is pro-posed where the power consumption is characterized bythree parameters: the transmission power, the efficiency ofthe transmission chain and a parameter that contains all theother power needed. [16, 26] both present a model wherethe power consumption is a function of the load. In [19],the authors proposed a power consumption model for bothmacro and micro sites. Our study differs from these studiesin several aspects. First, each component of the base stationis modelled individual, whereas the other studies combinedifferent components in one parameter, which makes itdifficult to investigate the influence of the different compo-nents on the total power consumption, as well as possibledependencies between the components. Second, our studycompares the energy efficiencies of different technologies,whereas the other studies focus only on one wireless tech-nology. Finally, our work is based on real measurementstaken from an operation network, which also gives a goodindication of the accuracy of the predicted values.

[17] also proposes a power consumption model as afunction of the traffic load in an operational 3G Univer-sal Mobile Telecommunications System (UMTS) network,whereas our study combines measurements on a UMTSmicrocell base and HSPA macrocell base stations. How-ever, this study agrees with our conclusion that traffic(or load) variations are not followed by similar variationsin power consumption. This conclusion is also confirmedby [16, 26].

Finally, in [27], a similar deployment tool is dis-cussed, which designs single frequency terrestrial DigitalVideo Broadcasting networks while minimizing the car-bon emissions of the network. However, GRAND developsnetworks for a broader range of wireless access technolo-gies. Furthermore, GRAND allows also to develop hybridmacrocell and microcell base stations.

6. CONCLUSION

In this paper, a power consumption model is proposed forboth macrocell and microcell base stations and validated bytemporal measurements. The evolution of the power con-sumption during the day is investigated by means of thesemeasurements. With this model, the energy efficiency ofthree different wireless technologies is considered, namelymobile WiMAX, HSPA and LTE. This model is then usedin the deployment tool GRAND, which allows to designan energy-efficient network for a pre-defined target. Thistool allows us to investigate the influence of the introduc-tion of microcell base stations in the network on the energyefficiency and the power consumption.

With the assumptions made, it was found that a micro-cell base station consumes, in general, about 4.4 times lessthan a macrocell base station. However, a macrocell basestation is more energy efficient than a microcell base sta-tion as a higher coverage is obtained with a macrocell basestation. Furthermore, the measurements agree excellentlywith the proposed models.

A higher energy efficiency is obtained when microcellbase stations are introduced in the network because a lowertotal power consumption is obtained (for all technologies)for the same coverage. It is thus concluded that although asingle microcell base station has a lower energy efficiencythan a macrocell base station, it is useful to introducemicrocell base stations in the network.

In the future, both cell zooming and sleep modes shouldbe added to both the model and the GRAND tool. Cellzooming is a technique whereby the cell size is adjustedaccording to traffic load, user requirements and channelconditions. When sleep modes are introduced in the net-work, base stations can be switched off during periods oflow or no traffic. For both cell zooming and sleep modes,a traffic model needs to be composed. Furthermore, wewill also extend the GRAND tool to make optimizationstowards capacity possible.

ACKNOWLEDGMENTS

The work described in this paper was carried out withthe support of the IBBT-project GreenICT. W. Joseph is apost-doctoral fellow of the FWO-V (Research FoundationFlanders).

REFERENCES

1. Koutitas G, Demestichas P. A review of energy effi-ciency in telecommunication networks, In 17th Telecom-munications Forum TELFOR 2009, Serbia, Belgrade,November 24; 2–7.

2. IEEE Computer Society and the IEEE Microwave The-ory and Techniques Society. Part 16: Air Interface forFixed and Mobile Broadband Wireless Access Systems:Amendment 2: Physical and Medium Access ControlLayers for Combined Fixed and Mobile Operation inLicensed Bands and Corrigendum 1. IEEE: New York,USA, February 2006.

3. LTE: 3rd generation partnership project: technical spec-ification group radio access network: evolved univer-sal terrestrial radio access (E-UTRA): user equipment(UE) radio transmission and reception (TS 36.101 v9.1.0Release 9) 3GPP, September 2009.

4. 3rd generation partnership project: technical speci-fication group radio access network: physical layeraspects of UTRA high speed downlink packet access(Release 4), TR 25.848 v4.0.0, 3GPP, March 2001.

5. Deruyck M, Tanghe E, Joseph W, Martens L. Modellingand Optimization of Power Consumption in WirelessAccess Networks, Vol. 34. Elsevier Computer Commu-nications: Oxford, United Kingdom, November 2011.2036–2046.

6. Correia LM, Zeller D, Blume O, Ferling D, Jading Y,Gódor I, Auer G, Van der Perre L. Challengesand enabling technologies for energy aware mobile



radio networks. IEEE Communications Magazine 2010;48(11): 66–72.

7. Pfannenberg, 2010. http://www.pfannenbergusa.com/catalog/catalog-downloads/2010_Catalog.pdfLast accessed: December 2010.

8. Raab FH, Asbeck P, Cripps S, Kenington PB,Popovic ZB, Pothecary N, Sevic JF, Sokal NO. In RFand microwave power amplifier and transmitter tech-nologies - part 1. High Frequency Electronics: Bedford,United Kingdom, 2003; 22–36.

9. Ophir RF, 2010. model 5303009, 5303025, 5303075,5303018A, 5303023.

10. Power-One S series, 2010.11. Daikin FUQ125BW13/R2Q125D7413, FHQ35BW1B,

2010.12. Allgon Microwave - AMR Transcend PLUS, 2010.13. Ceragon Networks - FibeAIR 1500P Family, 2010.14. Trangobroadbandnetworks - TrangoLINK - Giga, 2010.15. Joseph W, Verloock L, Tanghe E, Martens L. In-situ

measurement procedures for temporal RF electromag-netic field exposure of the general public. Health Physics2009; 96(5): 529–542.

16. Micallef G, Mogensen P, Scheck H.-O. Cell size breath-ing and possibilities to introduce cell sleep mode,In European Wireless Conference 2010, April 2010;111–115.

17. Peng C, Lee S-B, Lu S, Luo H, Li H. Traffic-drivenpower saving in operational 3G cellular networks,In 17th Annual International Conference in MobileComputing and Networking (MobiCom), Las Vegas,Nevada, USA, 2011, DOI: 10.1145/2030613.2030628.

18. Saunders S. Antennas and Propagation of WirelessCommunication Systems. Wiley: Hoboken, UnitedStates, 1999.

19. Richter F, Fehske AJ, Fettweis GP. Energy effi-ciency aspects of base station deployment strategies for

cellular networks, In IEEE 70th Vehicular TechnologyConference Fall (VTC 2009-Fall), Anchorage, Alaska,September 2009; 1–5.

20. Deruyck M, Vereecken W, Tanghe E, Joseph W,Pickavet M, Martens L, Demeester P. Power consump-tion in wireless access networks, In European WirelessConferece, Lucca, Italy, April 2010; 924–931.

21. Bhaumik S, Narlikar G, Chattopadhyay S, Kanugovi, S.Breathe to stay cool: adjusting cell sizes to reduceenergy consumption, In First ACM SIGCOMM Work-shop on Green Networking, New Delhi, India, 2010;41–46.

22. Erceg V, Greenstein L, Tjandra S, Parkoff S, Gupta A,Kulic B, Julius A, Bianchi R. An empirically based pathloss model for wireless channels in suburban environ-ments. IEEE Journal on Selected Areas in Communica-tions July 1999; 7(7): 1205–1211.

23. ESRI. Commission of the European communicationsand COST telecommunications, In Cost 231 FinalReport, Digital Mobile Radio: Cost 231 View on theEvolution Towards 3Rd Generation Systems, 1999; 153.

24. ESRI. ESRI Shapefile Technical Description - WhitePaper, 1998.

25. Jada M, Hämäläien J, Jäntti R, Hossain MMA. Powerefficiency model for mobile access network, In The 21stAnnual IEEE International Symposium on Personal,Indoor and Mobile Radio Communications: Workshopon Wireless Green (PIMRC 2010, W-GREEN), Istanbul,Turkey, September 2010; 316–321.

26. Auer G, Giannini V, Desset C, Gódor I, Skillermark, P,Olsson M, Imran MA, Sabella D, Gonzalez MJ,Blume, O, Fehske A. How much energy is needed torun a wireless network. IEEE Wireless Communications2011; 18(5): 40–49.

27. Koutitas G. Low carbon network planning, In EuropeanWireless Conference, 2010; 411–417.


Documents

Power consumption model for macrocell and microcell base stations