IEEE Communications Magazine • August 2011 630163-6804/11/$25.00 © 2011 IEEE

INTRODUCTION

Wireless communications are expected to be themajor worldwide cause of energy consumptionwithin a few years, with a devastating impact interms of pollution and energy waste. The highpower needed for operating wireless devicescauses huge greenhouse gas (GHG) emissions inthe atmosphere, while the portion of energyactually traveling on the communication media isone or two orders of magnitude smaller than theenergy consumed by the overall system. Jointlywith the environmental impact of new wirelessdevices (e.g., new towers and base station sites),the highly inefficient use of power might becomea serious menace for the environment. In partic-ular, base stations cause more than 80 percent ofthe operator’s power consumption, which makesthe design of base stations a key element fordetermining both the environmental impact ofwireless networking and the operational expen-

diture. However, the availability of next genera-tion wideband and reliable cellular networks(e.g., LTE and HSPA [1]) can help reduce theGHG footprint by reducing the need for travel-ing and printing documents, as emerged in arecent study by the Australian operator Telstra.

To give an idea of the actual energy and pol-lution cost induced by a base station, considerthat most of the operating base stations continu-ously consume at least 2 kW, mainly generatedby means of fossil fuels, and hence produce asmuch CO2 as a few automobiles in their averageutilization cycle. New generation base stationsreduce their consumption to typically 800 W forsecond-generation (2G) systems (GSM), and aslow as 430 W for 3G systems (UMTS). Manufac-turers are further targeting to halve the powerconsumption of their base stations in two years,and to lower the environmental impact by allow-ing network operators to deploy cheaper multi-technology sites. Manufacturers particularlyfocus on new base station design, which wouldreduce capital expenditures (CAPEX) and oper-ational expenditures (OPEX). To achieve thisgoal, manufacturers and operators point at opti-mizing the architecture of the site and usingrenewable energies, improving the power effi-ciency of the hardware, and deploying smartresource management tools that activate net-work resources only when needed.

Site design is fundamental for efficient usageof energy. For instance, flexible and lightweightsites could be better located to provide uniformcoverage with no need for high transmissionpower. Most important, site design is responsiblefor minimizing ownership and managementcosts, and reducing the need for energy-hungryair conditioning. Clearly, site design depends onthe availability of more efficient, compact, andlow-power-consuming electronic devices. A sub-stantial reduction of the power needed to oper-ate a base station also significantly helps thedeployment of sites totally or partially operatedby means of renewable energies (e.g., solar pan-els, bio-fuels, and eolic turbines), as reported,for example, by Huawei and Ericsson.

In terms of hardware improvements, manu-facturers are replacing existing power amplifierswith new efficient devices using digital predistor-tion (DPD) or envelope tracking for widebandsignals. Noticeably, using these efficient poweramplifiers allow the deployment of new compact

ABSTRACT

Cellular wireless networks are expected toprovide high-quality audio and video serviceswhile enabling fast and low-cost Internet accessto mobile users. The need for green cost-efficientnetworks is twofold: reduce the service price andpreserve the environment. In this work, we dis-cuss the various strategies that help reduce infra-structure costs, power costs, and greenhouse gas(GHG) emissions with no impairments on thequality of network services. These strategiesrange over a wide area from enhancing the elec-tronics, to developing new energy-aware radioaccess protocols, to deploying enhanced basestations with tunable capacity. To reduce bothcapital and operational expenditures, and theGHG footprint, manufacturers propose newcompact installation with lightweight antennasystems, very efficient power amplifiers, and effi-cient hardware and software. The resulting econ-omy can be up to 50 percent or more by reducingthe electricity bill, sparing the use of air condi-tioning, and deploying compact sites that wouldseldom require maintenance. Recent scientificpublications confirm that a very high gain couldbe achieved by optimizing the use of base sta-tions proactively, and huge additional improve-ments could be obtained by optimizing powersaving mechanisms by leveraging traffic statistics.

ENERGY EFFICIENCY IN COMMUNICATIONS

Vincenzo Mancuso and Sara Alouf, INRIA Sophia Antipolis Méditerranée

Reducing Costs and Pollution inCellular Networks

ALOUF LAYOUT 7/21/11 12:29 PM Page 63

IEEE Communications Magazine • August 201164

sites, which can be operated with half the poweror less.

Finally, operators are developing new man-agement tools to reduce the amount of devicesoperating with low or zero load, which has beenshown to be almost as expensive as running atfull capacity [2], say, by switching off unloadedbase stations overnight, while avoiding coverageand capacity degradation. Operators like Huawaiclaim that using radio and computationalresources efficiently might easily turn into a 40percent drop in operational costs.

Recent scientific publications confirm thatpower save and quality of service enhancementsare not conflicting objectives for wireless networkdesign. Power save modes and operational costsoptimization have been studied by means of ana-lytical models and simulations. Sleep mode hasbeen proposed as an interesting solution for lowloaded transmitters, and it has been analyticallyshown that optimal sleep periods can be selectedas a function of the statistical distribution of thepacket arrival over the wireless interface [3, 4].Optimization is carried out in terms of costreduction subject to quality of service constraintssuch as the average or maximum tolerable packetdelivery time. The results of [3] can be used todesign the behavior of an on/off base station dur-ing time intervals in which the expected traffic isstationary and data bursts can be modeled asmemoryless arrivals. In case of more general traf-fic patterns, dynamic programming has been usedto characterize the optimal sleeping interval ateach sleep mode activation epoch [4]. Joint radioresource management strategies have been pro-posed for the cooperation between GSM andUMTS systems on the same base station. Theanalytical study in [5] illustrates benefits offeredby the cooperation of 2G and 3G systems. In par-ticular, intertechnology handover (i.e., from 2Gto 3G and vice versa) enables a substantial reduc-tion of the system blocking probability at noadditional cost, with no need for increasing thecapacity of the network. Furthermore, the deci-sion for a user to select one of the available radiosystems can be performed subject to congestioncontrol and energy aware mechanisms, with theinteresting conclusion that, under low load condi-tions, the higher gain is obtained by letting allusers use one technology only and putting insleep mode the base station components operat-ing on other technology. Also, standardizationbodies like the Third Generation PartnershipProject (3GPP) have started to provide operatorswith suitable tools to implement power savingmechanisms, such as continuous packet connec-tivity (CPC) for always on customers using high-speed data connections [6].

The survey is organized as follows. We illus-trate the green strategies adopted by vendorsand operators. We introduce the architecture ofthe newest base stations. We give insights on theoperational costs and energy consumption of abase station. We conclude the survey.

GREEN DEPLOYMENT STRATEGIESIn this section, we focus on the strategies thataim at reducing operators’ costs and environ-mental impact due to base station deployment

and operation. We illustrate how green compo-nents have been introduced in the design of thecore network and base station sites (possibly tar-geting renewable power sources), in the adop-tion of efficient hardware, in the responsible useof packaging and energy, and in the introductionof power saving features in resource manage-ment.

CORE NETWORK RATIONALIZATIONA contribution to green networking comes fromthe simplification of the core network, and fromthe optimization of the efficiency of soft switchesthrough which data streams are routed and con-trolled. For instance, since 2007, the Australianoperator Telstra and Ericsson have been devel-oping an efficient, compact, high-capacity mobilesoft switch server to simplify the mobile network.The new soft switch comes with blade technology(i.e., using electronic boards or blades), which ishighly modular and reduces the energy use ofthe soft switch by up to 60 percent per sub-scriber. This innovative mobile soft switch is suit-able for both GSM and UMTS networks, andsupports up to eight million subscribers withonly two equipment shelves. Interestingly, notonly does the new technology allow the deploy-ment of fewer soft switches (75 percent less) tosatisfy the traffic demand, but also each newmobile soft switch is easy to upgrade and exhibitsa GHG footprint as little as 10 percent that ofexisting servers. Telstra is planning to replace by2010 the 18 current regional mobile soft switchservers spread across Australia by one mobilesoft switch server cluster using blade technology(deployed at two sites for redundancy anddependability issues). This rationalization shouldreduce equipment floor space by 85 percent, cutenergy use by 75 percent, and proportionatelyreduce the GHG footprint. Similarly, Huaweireports that its large-capacity mobile soft switchserver uses a platform, named ATAE, thatenables a pool technique with results similar tothe Telstra case, and has brought an overall costreduction of 80 percent for the Sichuan mobilenetwork in China.

GREEN BASE STATION SITESHere we show three different examples of sitedesign, each targeting — for different environ-ments — low GHG footprint and low installa-tion, operational, and maintenance costs. Thenwe discuss the utilization of renewable energy.

OPTIMIZED DESIGNThe Tower Tube — The target of Ericsson’sintroduction of a new radio base station site con-cept, Tower Tube, is threefold: reducing pollu-tion causes, making the wireless network morecost efficient, and allowing the integration of thesite with the surrounding environment. TowerTube has a modern design for a 5-m-diameter,40-m-high flexible concrete tower encapsulatingall radio base station equipment and antennas.The concrete itself has a lower environmentalimpact than traditional steel, producing 30 per-cent less CO2 emissions during production andtransportation. Unlike commonly adopted sites,the network equipment is installed at the bottomof the site, and then rised by means of an elava-

A contribution to

green networking

comes from the

simplification of the

core network,

and from the

optimization of the

efficiency of

soft switches

through which data

streams are routed

and controlled.

ALOUF LAYOUT 7/21/11 12:29 PM Page 64

IEEE Communications Magazine • August 2011 65

tor to the top of the tower, next to the antennas.The advantages of this choice are particularlyevident in the reduced loss in the feeder betweenthe radio frequency (RF) unit and the antenna,and in the possibility to exploit the wind as apassive cooling system. Energy saving can bequantified in a 40 percent reduction of opera-tional costs. Additionally, since Tower Tubeoccupies 60–75 percent less space than conven-tional sites, site acquisition is cheaper, whichmeans, lower capital expenditures are required.Even lower operational costs are possible withthe wind-powered version of the Tower Tube,endowed with five-meter blades and a verticalrotor.

The Capsule Site — Finding new locations forradio base stations is challenging, especially inurban and suburban areas, where obtainingbuilding permit for a base station might requirethe design of an unobtrusive site, or also theintegration of the site as a new landmark in thesurroundings. The Ericsson Capsule Site is anexample of base station site design for easydeployment in urban areas. The Capsule Site isan all-in-one solution encapsulating all necessaryequipments. The site cover is made of compos-ite, weighs 800 kg, and can be delivered andinstalled in less than a day.

The Flexi Base Station — An example of effi-cient site design and management is the Flexibase station by Nokia Siemens Networks. Thebenefits are as follows:• Reduce site energy costs up to 70 percent• Reduce size and weight of network equip-

ment by 80 percent• Allow flexible location (i.e., indoor and out-

door) with no need for air conditioning• Shorten antenna feeders• Make new sites ready for future radio tech-

nologiesThe Flexi base station GHG footprint is furtherdiminished by its software-based capacity andcapability upgrades. In fact, Flexi base stationscan be remotely managed, in terms of tuning therunning capacity (i.e., the number of GSM fre-quencies in use), and also remotely upgraded,canceling the need for site visits. The Flexi basestation is also ready for using renewable energysuch as solar or wind power.

Renewable Power Sources — Consideringthe demand for new base station installations inmany regions where an efficient power grid isnot yet available (or where a power grid doesnot exist at all), the possibility to operate wire-less base stations relies on the use of local powergenerators. In the past, power generators havebeen using diesel engines, while the currenttrend is towards the replacement of diesel withbio-fuels, and traditional generators with solar-or wind-powered generators. Unfortunately, acomplete replacement is not possible in shortterms, since many old installations require waytoo much energy to be reliably supplied bymeans of renewable sources. Hence, a consump-tion reduction strategy is strictly necessary toenable the deployment of renewable energy-powered base stations.

In 2008, the GSM Association (GSMA), theassociation gathering nearly 800 worldwidemobile operators, launched a plan for deployingrenewable energy sources for 118,000 new andexisting base stations in developing countries by2012. The expected green gain is to save 2.5 bil-lion liters of diesel and cut CO2 emission up to6.3 million tons per year. Solar energy solutionshave been proposed since 2000, at least to oper-ate in conjunction with diesel generators. Erics-son and Alcatel-Lucent, among other companies,have proposed solar-powered base stations formacro-coverage and low capacity, useful forrural or low-density areas.

Although the 100 percent-solar-powered solu-tion is in the target, and operators like OrangeGuinea targeted at having more than one thou-sand 100 percent-solar-powered base stations bythe end of 2009, many installations are stillhybrid, using solar panels and batteries toreplace only one out of the two diesel generatorsusually deployed. Nevertheless, the gain is high,since a typical site consumes approximately20,000 liters of diesel per year, which can beeffectively halved by means of solar panels andbatteries. Further gain can be obtained by com-bining multiple renewable power sources in thesame site. By using solar and wind power, plusan auxiliary diesel generator, African operatorZain has deployed a site in Dertu, a remote vil-lage in Kenya. The operational cost due to elec-tricity has been reduced by 80 percent for thatsite, compared with using diesel generators only.

ENERGY-EFFICIENT POWER AMPLIFIERSNext generation cellular networks will make alarge use of non-constant envelope signals. Inparticular, the widely accepted adoption oforthogonal frequency-division multiple access(OFDMA) systems will challenge the manufac-turers to design more efficient and accuratewideband orthogonal frequency-division multi-plexing (OFDM) modulators. In an N-subcarrierOFDM, the transmit power is evenly spread overthe subcarriers, and the peak-to-average powerratio (i.e., the ratio between the maximum

Figure 1. 1024-subcarrier OFDM signal amplified with constant input voltagesupply. The power dissipated as heat is huge due to the difference between thesignal envelope and the voltage supply level.

OFDM signalFixed input voltage

Power dissipated as heat

ALOUF LAYOUT 7/21/11 12:29 PM Page 65

IEEE Communications Magazine • August 201166

instantaneous power and the average power) isequal to N [7]. Thereby, the maximum voltageamplitude at the power amplifier is √

—N times the

average voltage amplitude, and it is not conve-nient to amplify the OFDM signal using a con-stant input voltage, which would turn intoconstant power consumption even when the sig-nal is very low (and the unused power is dissipat-ed as heat, Fig. 1).

In order to be power-efficient, the averageinput power at the power amplifier has to bekept as close as possible to the power needed fortransmitting the signal. Possible solutions consistof either distorting the signal or dynamicallyadapting the input voltage of the amplifiers.

DPD and Doherty Power Amplifiers — Digi-tal predistortion (DPD) and Doherty techniquesare used to adapt the signal to the amplifier’scharacteristics and to boost the power emissionwhen the signal level is above a fixed threshold,thus avoiding excessive signal clipping.

The DPD and linearization technique usespre-distortion of the signal to be transmitted andoperates a compensation for nonlinearities inthe final RF output stage of the amplifier. In theprocess, it also improves adjacent channels inter-ference and reduces the error vector magnitude(i.e., it improves the modulation accuracy) andallows some compensation for the distortioncaused by nonlinearities near compression. Thistechnology obtains the best result when DPDand linearization are used as part of a feedbacksystem architecture incorporating active sam-pling of the output signal. This way, in fact, thesystem fully compensates for changes in amplifi-er characteristics with time, temperature, andsignal characteristics [8].

The Doherty power amplifier configurationhas been originally proposed in 1936 [9]; it usestwo amplifying devices driven in parallel, withtheir outputs combined. One carrier amplifier,operating in Class A/B, provides all the outputpower until the power required causes it to enterits nonlinear region. A peaking amplifier, operat-

ing in Class C, provides additional power whenthe carrier enters its nonlinear region. Whileseveral academic papers have quoted impressive-ly high efficiency capabilities for Doherty ampli-fiers [8], in practice, at the frequencies and withthe high output power, which are used in cellularnetworks, the typical efficiency achieved isaround 25 percent to 30 percent [10]. Dohertyamplifiers have a limited bandwidth due to thecomplicated and essentially narrow-band match-ing required between the two internal amplifiers.However, the bandwidth available (few MHz) isadequate enough for cellular systems.

Huawei, the Chinese telecom equipmentprovider, has launched a low-power base stationadopting DPD and Doherty technologies. Interms of energy, this base station can be run withas low as 500 W, thereby helping operators tosave up to 5700 kWh/year, or, equivalently,reduce the CO2 emissions due to energy genera-tion by 1.7 tons of coal per year. Such low-power-demanding hardware can easily rely on renewablepower supplies. Huawei’s solution has been uti-lized over 100,000 GSM base stations since 2007,so helping to save about 570 GWh/year, repre-senting a reduction of CO2 emissions equivalentto burning 170,000 tons of coal.

Envelope Tracking Technology — Envelopetracking was proposed by Bell Labs researchersin 1937, but it has been efficiently implementedonly recently after very fast, low-noise powertransistors have became available (usingLDMOS, GaN, GaAs, etc.). The basic idea ofenvelope tracking is as follows: instead of chang-ing the signal to match the power amplifier char-acteristics, dynamically try to adjust the supplyvoltage of the power amplifier to match theenvelope of the signal to be amplified. Thus,envelope tracking ensures that the output deviceremains in its most efficient operating region(i.e., in saturation). Figure 1 shows the ineffi-ciency of using power amplifiers with constantinput voltage supply. Without envelope tracking,the difference between the constant power fedinto the RF amplifier and the radio-frequencyoutput waveform is dissipated by the power tran-sistor as heat. With envelope tracking, as depict-ed in Fig. 2, the supply voltage tracks the signalenvelope, so the input power closely matches theRF output power. This matching turns into adramatic reduction of the energy dissipated asheat. Performing envelope tracking requires aCPU cost that can consume as much as a fewwatts, which are negligible in comparison to theenergy saved (tens or hundreds of watts) in thecase of high-power amplifiers.

The first practical commercial implementa-tion of envelope tracking is very recent (Nujira,2008), and can make a significant contribution tothe power efficiency of the power amplifiers,improving this from the 15 percent of Class A/Bpower amplifiers to 45 percent nowadays, and,according to vendors, the power efficiency willsoon reach 60 percent using the latest galliumnitride power transistors for RF. Nujira intro-duced and patented High Accuracy Tracking(HAT), an ultra-high-efficiency wideband modu-lator for radio-frequency with power amplifiersthat make use of envelope tracking. It is, there-

Figure 2. Power amplifiers’ input voltage can be made variable when envelopetracking is used. The figure shows the case of OFDM signals with 1024 sub-carriers: very few power is wasted as heat as the difference between the signalenvelope and the voltage supply level is small.

OFDM signal Voltage

Variable input

Power dissipated as heat

ALOUF LAYOUT 7/21/11 12:29 PM Page 66

IEEE Communications Magazine • August 2011 67

by, suitable for a broad class of non-constantenvelope modulators such as those required forOFDM transmissions. HAT modulators for 3G(code-division multiple access, CDMA), high-speed packet access (HSPA), digital video broad-cast (DVB), WiMAX, and Long Term Evoultion(LTE) provide the wireless manufacturer withthe possibility to design smaller cost-effectivebase stations which also have a reduced environ-mental footprint: the first commercial HAT-based RF unit for WCDMA/HSPA/LTE(February 2009) uses 50 percent less energy thantraditional devices. By deploying base stationsusing the HAT modulator, a cellular networkwith 20,000 base stations could save 35 MW,thus saving the operator each year $37 million inenergy costs, which potentially reduces CO2emissions by 130,000 tons. Table 1 comparesClass A/B, Doherty, and HAT power amplifiersin terms of per-amplifier efficiency and per-net-work costs. Doherty amplifiers’ performance isbetween that of traditional Class A/B and HAT.

ENERGY MANAGEMENT TOOLSEnergy-efficient solutions should be able to cor-relate energy consumption and traffic intensity.As a first step in this direction, wireless opera-tors have proposed the adoption of smart soft-ware features for enabling power saving at thebase station. Here, we introduce some examplesof such software.

NectAct SQM — Energy conservation can beobtained through traffic-load-driven capacitymanagement at the base stations. Nokia SiemensNetworks’ NetAct Service Quality Manager(SQM) is a software solution for managing avail-able capacity when traffic load is low. NetActcontrols base station power consumption auto-matically based on preconfigured settings; forexample, NetAct’s power save function allowsservice providers to manage power consumptionduring times of high and low traffic. Traffic pat-terns can be managed from the network down tothe individual base station level. Software likeNetAct SQM works for 2G and 3G systems, andallows power to be saved by tuning the setting ofthe base station automatically. For instance, sys-tem capacity tuning is scheduled every few tensof minutes, and the actual capacity is chosen as afunction of the current traffic load, the history ofthe traffic load, and the estimate of the trafficload during the next hour. Capacity adjustmentsconsist of changing, say, the number of activefrequencies for GSM access networks, which canalso result in turning off all radios if the trafficduring the next hour is expected to fall below aminimum threshold. In particular, the commer-cial base station software NetAct SQM comeswith a strategy named base station power savingat night, which allows the operator to set a timewindow during which the base station capacitycan be reduced. The actual capacity decision ismade in accordance with a traffic profile that isestimated over the hours of the week, and onlyafter a series of preconditions is met:• Current traffic is below a tunable threshold.• Current traffic is within its expected profile

(so traffic estimates can be assumed to notexceed significantly the expected value).

• The traffic expectation for the next hour isbelow the threshold.

• No relevant alarms and events are sched-uled during the next hour.

Smart Power Management — Another soft-ware for the dynamic activation of networkequipment is Smart Power Management (SPM)developed by Nortel. SPM software enables 2Gnetwork operators to switch off the radio net-work equipment when there is no traffic beingprocessed by the system. Nortel estimates poten-tial power savings of up to 33 percent reductionin base station power consumption.

Dynamic Power Save — In February 2009,Alcatel-Lucent launched a new feature calledDynamic Power Save, which promises powerconsumption reduction up to 30 percent throughthe possibility to turn off the power amplifiers inGSM transceivers on the basis of the trafficactivity monitored by the base station. Interest-ingly, the company says the software upgradecan be installed on the (roughly) 500,000 Alca-tel- Lucent base stations deployed over the past10 years.

Standby Mode — Standby operational modeshave been launched since 2007 by Ericsson,whose power saving estimate is about 10 to 20percent, depending on the traffic pattern.

BEST PRACTICES FOR ALOW-CARBON ECONOMY

Remarkable energy economy and GHG reduc-tion can be obtained by adopting responsiblepractices for the setup and management of thenetwork, beginning with the packaging. A clearexample of such green practices can be observedin the green action plan defined by China Mobile —the Chinese cellular operator — for the manage-ment of packaging. This strategy is defined bythe following six R’s: Recovery & Recycle refer tousing recyclable packaging materials that requirelow energy cost during their life-cycle. Right &Reduce refer to using small and lightweight car-ton design, thereby reducing packaging andtransportation costs. Returnable & Reuse refer tothe use of efficient recycling systems in order toextend the life cycle of packaging materials. Fol-lowing these directives, the Chinese manufactur-er Huawei reported an annual cost reduction of12,000 m3 of lumber, 2,700,000 liters of oil, 0.75

Table 1. Efficiency, costs and environmental impact of a 20,000-base-stationnetwork with different power amplifier technologies.

Traditionaltechnology

Dohertytechnology

Envelope tracking(HAT)

Power ampl. effic. 15% 25% 45%

Power consumption 51.7 MW 27.2 MW 16.1 MW

Power cost $54.3 M $28.6 M $17.0 M

CO2 emission 194,600 tons 102,400 tons 60,800 tons

ALOUF LAYOUT 7/21/11 12:29 PM Page 67

IEEE Communications Magazine • August 201168

GWh of energy, and a cut of 6,172,000 tons inCO2 emissions.

The green strategies introduced so far repre-sent the tools network operators can leverage toenable low-cost virtual infrastructure and ser-vices, contributing to reducing the need for trav-eling and so to the rising of a low-carboneconomy. Since 2007, green technologies byEricsson, Nokia Siemens Networks, Alcatel-Lucent, Huawei, Nujira, and other leading com-panies have been used worldwide to reduceenergy costs and correspondingly reduce GHGemissions due to power generation. As an exam-ple, in 2008, Ericsson set its first carbon foot-print target, aiming for a 40 percent reductionover five years, starting with a 10 percent reduc-tion in 2009.

Table 2 summarizes the impact of the differ-ent strategies presented so far. Next, we focus onthe site cost, illustrate the various cost compo-nents incurred when operating a base station,and the potential for reduction of each of thesecosts.

ARCHITECTURE OF AGREEN BASE STATION

The analysis of expenditures on 3G networksite construction shows that base station infra-structure costs make up 30 percent of the totalcost. Other expenses such as the site acquisi-tion depend on the base station architecture.Smaller size and more power-efficient base sta-tions imply reduced investments and opera-tional costs. In order to move further with costreduction, operators are asking for flexiblebase stations, where multiple technologiescould be run simultaneously, multiple carrierscould be operated on demand, and easyupgrades could be performed by simply updat-ing software components.

From the operator’s perspective, runningmultiple radio technologies in a single base sta-tion (multimode base station) means sharedCAPEX for the site and reduced managementexpenses. As detailed later, a base station main-ly consists of the baseband unit, the RF unit,the transport unit, and a control unit, the otherunits being auxiliary to these networking units.To make the installation simpler, more flexible,

and multimode, Huawei, Alcatel-Lucent, andother manufacturers have proposed to locateonly the RF units on site, and to use the opencommon public radio interfaces (CPRIs) tocommunicate with the baseband unit and otherunits like the remote control and transportunits. The base-band unit can be made as smallas a 2U-unit (i.e., with a 19 ft × 3.5 ft frontpanel) and can be conveniently installed inexisting cabinets such as auxiliary power supplycabinet, transmission equipment cabinet, orother remote equipment cabinets. The radio-frequency unit can be made light enough to beinstalled on a tower, a pole, or simply against awall. Thereby, since the radio-frequency unitinstallation is made easy, the conventional feed-er is no longer used, which enables a closerantenna interface and avoids a 3 dB power loss.This base station can be used for both 2G and3G radio access networks, as in the case of theFlexi base station. In particular, the control unitcan be shared between 2G and 3G radio-fre-quency units by simply loading the appropriatesoftware. For example, the previously men-tioned Flexi base station was originally meantfor GSM/EDGE networks, but it has beenupgraded to support all 3GPP technologiesfrom wideband CDMA (WCDMA)/HSPA toLTE, having them all running concurrently in asingle unit (Flexi Multiradio base station, on themarket since early 2009). This upgrade has beenpossible just by updating the base station soft-ware. Another multimode solution was recentlyproposed by the Chinese operator ZTE, whoseUnified Hardware Platform supports GSM,CDMA, UMTS, time-division synchronousCDMA (TD-SCDMA), and LTE. Other multi-carrier RF units are available for CDMA, inwhich one unit can support several carriers(four to eight carriers in current implementa-tions), and multiple carriers can share the sameantenna system without the need of using a sig-nal combiner.

Finally, it is worth mentioning that manufac-turers are developing new software defined radioaimed at easily configuring the radio interfaceand upgrading the transmission technology atzero cost. For example, in 2009 ZTE launchedits software defined radio solution to be used forGSM/CDMA/UMTS base stations, and alsoready to evolve LTE.

Table 2. Impact of different strategies with respect to old-design base stations.

Element Solution CAPEX saving OPEX saving Less CO2

Core network Compact soft switches 75% 90% 80–90%

Base station site Radio equipment next to theantenna > 50% 40–70% 30%

Power amplifiers Envelope tracking — > 50% > 50%

Software Site power save — Up to 40% Up to 40%

Low-carbon practices Packaging, recycling, bio-fuels — > 40% > 50%

Overall impact ~ 70% > 50% > 50%

Energy-efficient

solutions should be

able to correlate

energy consumption

and traffic

intensity. As a first

step in this direction,

wireless operators

have proposed the

adoption of smart

software features for

enabling power

saving at the base

station.

ALOUF LAYOUT 7/21/11 12:29 PM Page 68

IEEE Communications Magazine • August 2011 69

BASE STATIONOPERATIONAL COST ANALYSIS

In this section, we provide the reader with adetailed breakdown of the costs related to theoperation of a base station; in particular, weanalyze the operational costs due to power con-sumption.

OPERATIONAL COSTSA base station is a modular composition of sev-eral units. The resulting base station buildingblocks are depicted in Fig. 3 and introduced inwhat follows.

The antenna system includes one or moreantennas for one or more RF bands in use. Thebaseband process unit is in charge of handlingthe data and voice streams by operating digitalsignal processing, and generating and parsingcontrol signals. The RF unit is the RF modemthat connects the baseband unit to the antennasystem, and includes a digital processing boardand the power amplifiers necessary to drive theantenna radiation. The control unit providesconfiguration tools and management interfacesto the base station equipment, and handlesalarms and operational logs. The transport unitprovides transport services for the data to beexchanged between the base station and the corenetwork, including user generated data, voiceand video streams, and control messages (e.g.,for handling handover procedures within theradio access network). The cooling system isresponsible for control of the ambient tempera-ture so that electronic devices operate depend-ably. The power supply unit has to provide thebase station equipment with continuous and sta-ble electrical power, and can include power gen-erators (e.g., solar panels and/or dieselgenerators) and a voltage rectifier for AC/DCconversion.

Note that a base station might include one ormore RF units for different bands, for different2G/3G technologies, and for multiple coverage

sectors. Also, baseband units could be sharedbetween multiple base stations, thus making itpossible to deploy simple base stations, whichconsist of RF unit, antenna system, and powersupply only.

Base station installation and upgrade expens-es include the acquisition of equipment, itstransport to the site, plus the costs of the siteproperty (or rental) and building the site itself,including towers for the antenna system andwaterproof cases for the equipment. These capi-tal expenditures are strongly dependent on thesize and weight of the base station, and canrange from a few thousand dollars to a few hun-dred thousand dollars. However, base stationcosts are dominated by the operational cost ofthe energy needed to operate the site. For both2G and 3G base stations, vendors report typicalconsumption values on the order of several hun-dreds of watts (0.5 to more than 2 kW). In par-ticular, baseband, control, and transport unitsoperate with very low power, as they requireonly a few tens of watts, resulting in as few as 5percent of the total power consumption. The RFunit consumption is mainly due to the poweramplifiers that have to feed the antenna system.This consumption is of the order of several hun-dreds of watts, with less than 100 W being radi-ated by the antenna. Noticeably, about half ofthe power provided by the amplifiers is wastedby the antenna system on the feeder that con-nects the antenna to the RF unit (typical cableattenuation at 2 GHz is 0.1 to 0.5 dB/m, whichyields a typical 2 to 3 dB loss in traditional basestation sites).

Note that the power needed for digital pro-cessing, operated at both the baseband and RFunits, accumulates as little as 10 percent of theoverall base station consumption. Furthermore,according to Huawei’s data (which do notaccount for voltage rectifiers), 93 percent of thebase station energy is spent on digital processing,transmission, and air conditioning, the remaining7 percent being spent on network devices andcontrol units collocated at the base station site.

Figure 3. Modular composition of a high-power base station. A dual power supply, AC and DC, is usuallyneeded to feed the cooling system and electronic devices, respectively. Only a minor fraction (e.g., 60 W) ofthe total power consumption (e.g., ~2 kW) is radiated on the air.

PowerData/signalingControl

Airinterface

50 WControl unit

50 W

DC150 W

TransportNetworkinterface

500 WCooling systemPower supply

50 WBase-band Radio-frequency

Antenna system

AC

60 W

60 W

1000 W

Base station installa-

tion and upgrade

expenses include the

acquisition of equip-

ment, the transport

to the site, plus the

costs for the site

property (or rental)

and for building the

site itself, including

towers for the

antenna system and

waterproof cases for

the equipment.

ALOUF LAYOUT 7/21/11 12:29 PM Page 69

IEEE Communications Magazine • August 201170

These data also confirm that digital processing,transport, and control operations consume anegligible amount of power in comparison withthe power amplifiers stage within the RF unit.Overall, amplifiers’ consumption can represent60 percent of the total power, and, consideringelectronic and transmission devices only (i.e., notconsidering air conditioning), power amplifiersare responsible for as much as 80 percent ofbase station consumption.

Interestingly, a relevant fraction of the powerconsumption in a base station is due to the airconditioning system [2], and a non-negligiblequote of energy (~10 percent) is wasted due toAC/DC conversion needed to feed the electron-ics present in the digital devices (i.e., the control,transport, baseband, and RF units) [2]. Thepower wasted by auxiliary equipment, such asthe air conditioning and voltage rectifier, can bequantified in at least one forth of the overallbase station power demand, but can grow up tomore than one third of the total power con-sumption.

In summary, the main causes of power con-sumption, and hence pollution, are the poweramplifiers in the RF unit and the air condition-ing system. The former strongly affects the lat-ter, since about half of the power used forcontrol, transport, processing, and transmissionis dissipated as heat by the power amplifiers. Airconditioning consumption also depends on theenvironmental conditions and the design of thebase station spaces, and it can be as power-hun-gry as the power amplifiers. Remarkably, newlyproposed efficient power amplifiers might enable20 percent economy over DPD power amplifiersand 40 percent over traditional amplifiers. Fur-thermore, these new power amplifiers eliminatethe need for AC and allow the use of renewablepower sources, which do not require the pres-ence of power rectifiers, for a total potentialenergy economy up to 50 percent and GHGemissions down to almost zero. Another 20 per-cent of economy can be obtained by optimizingthe control and networking units of the base sta-tion.

Table 3 summarizes the data on operationalcosts available on the manufacturers’ websites(e.g., published by Alcatel-Lucent, Ericsson,Huawei, Motorola, Nortel, Nokia Siemens Net-works, SEI, Nujira, ZTE). The table shows thateach base station component has a differentimpact on the overall operational costs. Controland transport units are considered together,since they are commonly deployed on a singlephysical device. Similarly, specific data about thestandalone baseband unit are not available,while it is possible to obtain the overall con-sumption of the transmission units (baseband,RF, and antennas), the cost due to the poweramplifiers, and data on the antenna system.Hence, we split these cost in three parts:• The cumulative consumption due to digital

processing, which includes all basebandcosts plus the digital processing cost at theRF unit

• The remaining portion of RF consumption(i.e., the power amplifiers consumption),which is dissipated as heat

• The consumption due to antennasThe table also includes costs due to air condi-tioning and voltage rectifiers (AC/DC).

Table 3 also enlightens how technologicalimprovements might dramatically reduce opera-tional costs up to nearly 70 percent. Power sav-ings might be attained by making the electronicand network devices more efficient (more than 20percent of gain can be obtained by consideringthe first three rows in the table), shortening theantenna feeder (another 10 percent reduction ofoverall consumption), removing the need for theair conditioning system by reducing the heat pro-duced by the power amplifiers (up to 25 percentless power), and eliminating the need for voltagerectifiers by using DC generators (e.g., renewablepower supplies based on solar panels, whichmight save an extra 10 percent). Note that effi-cient electronics alone enable as much saving as50 percent or more of the overall consumption. Infact, efficient electronics are necessary for design-ing small boards that can be located close to theantenna and do not require air conditioning.

Table 3. Operational costs repartition for a traditional base station provided with a single radio board, a cooling system, and an AC/DCrectifier, plus control and transport devices. Original costs are compared with possible cost savings due to technology enhancementsand usage of renewable energy.

Cost impact with traditionalbase stations

Relative saving estimate withtechnological enhancements

Absolute saving estimate withtechnological enhancements

Control and transport ≤5% ≤60% ≤3%

Digital processing (baseband+ RF) ≤10% ≤50% ≤5%

Power amplifiers (RF) 30–60% ~40% ~15%

Antenna system 15–20% ~50% ~10%

Air conditioning 10–25% 100% 10–25%

AC/DC ~10% 100% ~10%

Total 100% N/A 50–68%

ALOUF LAYOUT 7/21/11 12:29 PM Page 70

IEEE Communications Magazine • August 2011 71

MEASUREMENTS OFBASE STATION CONSUMPTION

There is quite a lot of data available in the sci-entific literature about the power consumptionof each component of a base station, either for2G base transceiver stations or 3G Node Bs,which is what 3G base stations are called. Incontrast, many base station vendors advertise theaverage power consumption of their equipmentand usually provide aggregate data (i.e., per basestation or even per network consumption data).

In [2], the authors report on a measurementstudy of a typical 2G/3G base station in theVodafone Portuguese mobile network. Theexperimental study shows that for a commonoutdoor base station, operating three RF units(one for GSM in the 900 MHz band, one forGSM in the 1800 MHz band, and one forUMTS), about 76 percent of the power con-sumption is due to the electronics, while about24 percent of the total power is consumed by thecooling system. The same study shows that thecost of the GSM RF unit is roughly 30 percenthigher than the cost of the UMTS RF unit.

Noticeably, for both technologies, the powerconsumption only slightly fluctuates over time [2,Fig. 22]. GSM consumption, in a time window of15 min, is reported to be 0.27 to 0.30 kWh; thatis, it exhibits an average power consumptionranging from 1.08 kW in off-peak hours to 1.20kW during peak hours. On the other hand, theUMTS consumption, in a time window of 15min, ranges from 0.19 to 0.22 kWh, and corre-spondingly the average drained power approxi-mately ranges from 0.76 kW to 0.88 kW during aday. Note also that RF units are DCoperated.With the efficiency of the AC/DC rectifier usedat the tested base station being about 92 per-cent, the actual power consumption due to eachRF unit is actually 8.7 percent higher than thatdiscussed before (Section IV in [2]).

The measurements reported in [2] suggestthat the presence of traffic only slightly affectsthe power consumption at the base station. Amajor reduction in the power consumption isthen possible only by turning off the base stationradio(s).

CONCLUSIONSThroughout the article, we have surveyed thestrategies adopted by base station manufacturersand operators on the road towards a low-costand environment-friendly wireless networking.Most of the current green best practices concernthe rationalization of:• Capital expenditures, by optimizing the base

station site architecture and the distributionof the sites over the targeted coverage area

• Operational expenditures, by minimizingthe energy consumption of electronicdevices and reducing the need for coolingsystems

In addition, since the majority of the operationalexpenditures are due to the electricity consump-tion of RF transceivers, new software-based

management tools have been coming into play,which try to enforce a sleep mode on the equip-ment expected to handle low or no traffic duringoff-peak hours.

Based on the data available on per deviceenergy consumption, the survey strongly suggeststhat high-efficiency electronics might dramatical-ly reduce the consumption of next-generationOFDMA-based systems. However, a predomi-nant portion of the base station cost is incurredas soon as the radio devices are turned on, andthis does not depend on the traffic flowingthrough the devices. Therefore, smart and flexi-ble sleep mechanisms should be provided inorder to make the energy consumption dependon the time-varying traffic load.

REFERENCES[1] E. Dahlman et al., 3G Evolution: HSPA and LTE for

Mobile Broadband, Second Edition, Academic Press,Oxford, UK, 2008.

[2] F. Corriea Alegria and F. A. Martins Travassos, “Imple-mentation Details of An Automatic Monitoring SystemUsed on A Vodafone Radiocommunication Base Sta-tion,” IAENG Engineering Letters 16, vol. 4.

[3] S. Alouf, E. Altman, and A. P. Azad, “Analysis of anM/G/1 Queue with Repeated Inhomogeneous Vacationswith Application to IEEE 802.16e Power Saving Mecha-nism,” Proc. QEST, Saint-Malo, France, 2008, pp.27–36.

[4] A. P. Azad et al., “Optimal Sampling for State ChangeDetection With Application to the Control of SleepMode,” Proc. 48th IEEE Conf. Decision and Control(CDC), Shanghai, China, 2009.

[5] L. Saker, S.-E. Elayoubi, and H. O. Scheck, “Energy-Aware System Selection in Cooperative Networks,”Proc. VTC Fall, 2009.

[6] 3GPP TS 25.214, Physical Layer Procedures (FDD),release 8.

[7] S. Shepherd, J. Orriss, and S. Barton, “Asymptotic Limitsin Peak Envelope Power Reduction By Redundant Cod-ing in Orthogonal Frequency-Division Multiplex Modu-lation,” IEEE Trans. Commun., vol. 46, no. 1, 1998, pp.5–10.

[8] S.-C. Jung, O. Hammi, and F. M. G. Ghannouchi,“Design Optimization and DPD Linearization of GaN-based Unsymmetrical Doherty Power Amplifiers for 3GMulticarrier Applications,” IEEE Trans. Microwave Theo-ry and Techniques, vol. 57, no. 9, 2009, pp. 2105–13.

[9] W. H. Doherty, “A New High Efficiency Power Amplifierfor Modulated Wave,” Proc. IRE, vol. 24, 1936, pp.1163–82.

[10] Y.-S. Lee et al., “Highly Linear and Efficient Asymmetri-cal Doherty Power Amplifiers with Adaptively Biascon-trolled Predistortion Drivers,” Proc. IEEE MTT-S Int’l.Microwave Symp. Digest, 2009, pp. 1393–96.

BIOGRAPHIESVINCENZO MANCUSO received his M.Sc. in electronics in 2001and a Ph.D. in telecommunications in 2005 from the Uni-versity of Palermo, Italy. He was with the University ofRoma “Tor Vergata” in Italy, then back to the University ofPalermo as a postdoc, and a visiting scholar at the ECEDepartment of Rice University, Houston, Texas. In June2009 he joined the MAESTRO team at INRIA Sophia Antipo-lis, Frence. Since September 2010 he has been with Insti-tute IMDEA Networks, Spain.

SARA ALOUF received an M.Sc. in computer networking anddistributed systems in 1999 and a Ph.D. in computer sci-ence in 2002, both from the University of Nice SophiaAntipolis, France. During 2003–2004 she was a postdoctor-al fellow at the Free University at Amsterdam, and sinceMarch 2004 she has been with INRIA working as a full-time researcher in the MAESTRO project team. Her researchinterests include modeling and performance evaluation ofcommunication networks.

A predominant por-

tion of the base sta-

tion cost is incurred

as soon as the radio

devices are turned

on, and this does

not depend on the

traffic flowing

through the devices.

Therefore, smart and

flexible sleep

mechanisms should

be provided in order

to make the energy

consumption depend

on the time-varying

traffic load.

ALOUF LAYOUT 7/21/11 12:29 PM Page 71