assessment of exposure to electromagnetic fields …€¦ · assessment of exposure to...

Paper

ASSESSMENT OF EXPOSURE TO ELECTROMAGNETIC FIELDSFROM WIRELESS COMPUTER NETWORKS (WI-FI) IN

SCHOOLS; RESULTS OF LABORATORY MEASUREMENTS

A. Peyman, M. Khalid, C. Calderon, D. Addison, T. Mee, M. Maslanyj, and S. Mann*

Abstract—Laboratory measurements have been carried out withexamples of Wi-Fi devices used in UK schools to evaluate theradiofrequency power densities around them and the total emit-ted powers. Unlike previous studies, a 20 MHz bandwidth signalanalyzer was used, enabling the whole Wi-Fi signal to be capturedand monitored. The radiation patterns of the laptops had certainsimilarities, including a minimum toward the torso of the userand two maxima symmetrically opposed across a vertical planebisecting the screen and keyboard. The maxima would haveresulted from separate antennas mounted behind the top left andright corners of the laptop screens. The patterns for access pointswere more symmetrical with generally higher power densities ata given distance. The spherically-integrated radiated power (IRP)ranged from 5 to 17 mW for 15 laptops in the 2.45 GHz band andfrom 1 to 16 mW for eight laptops in the 5 GHz band. Forpractical reasons and because access points are generally wall-mounted with beams directed into the rooms, their powers wereintegrated over a hemisphere. These ranged from 3 to 28 mW for12 access points at 2.4 GHz and from 3 to 29 mW for six accesspoints at 5 GHz. In addition to the spherical measurements ofIRP, power densities were measured at distances of 0.5 m andgreater from the devices, and consistent with the low radiatedpowers, these were all much lower than the ICNIRP referencelevel.Health Phys. 100(6):594–612; 2011

Key words: electromagnetic fields; environmental assessment;exposure, population; exposure, radiofrequency; Wi-Fi; WLAN.

INTRODUCTION

DURING THE last few years, the use of wireless local areanetworks (WLAN) has increased rapidly, offering flexi-bility and mobility to users. This has made the technol-ogy popular among a wide range of users, including theeducation sector. In WLAN, the most popular technologyused for the wireless portion of the network known as

Wi-Fi, devices and computers are connected to the LocalArea Network (LAN) wirelessly through antennas trans-mitting and receiving radio waves to establish the wire-less connection. The majority of Wi-Fi devices operate at2.4–2.4835 GHz, while some devices operate in 5.15–5.825 GHz (with three sub bands of A, B, and C).Terminal devices such as laptop computers are known as“clients,” and the point of entry to the wired network isreferred to as an “access point,” usually located within afew tens of meters and in the same building. People usingor in the proximity of Wi-Fi equipment are exposed tothe radio signals emitted from it and will absorb some ofthe transmitted energy in their bodies. Therefore, therecent increase in the popularity of Wi-Fi devices hasproduced concerns about the use of wireless computersystems in schools and the extent to which the pupils areexposed to radio waves emitted from such sources.

There are few recent studies concerning the expo-sure of people to signals from WLAN systems. Thestudies can be categorized into two types: those thatassessed typical public exposure to WLAN in the envi-ronment (Foster 2007; Schmid et al. 2007a) and thosethat examined compliance with exposure guidelines un-der scenarios that produce higher than typical exposures;e.g., with the antennas transmitting continuously andwithin a few cm of the body (Schmid et al. 2007b; Kuhnet al. 2007). The environmental studies found exposureswell within the guidelines set by the International Com-mission on Non-Ionising Radiation Protection (ICNIRP),whereas the pessimistic scenarios found exposures com-parable to those from mobile phones, although still withinexposure guidelines. These guidelines are set to protectagainst known adverse health effects that result from exces-sive absorption of energy from radio signals in body tissues.Nevertheless, Wi-Fi is a rapidly developing technology, andquantitative information on exposure is sparse, particularlyfor installed systems. Moreover, given the precautionaryadvice in some countries regarding the use of mobilephones by children, it is important to quantify the exposurefrom Wi-Fi equipment as used by children in schools and

* Physical Dosimetry Department, Health Protection Agency,Chilton, Didcot Oxon, OX11 0RQ, UK.

For correspondence contact: Azadeh Peyman, Physical DosimetryDepartment, Health Protection Agency, Chilton, Didcot Oxon, OX110RQ, UK, or email at [email protected].

(Manuscript accepted 8 October 2010)0017-9078/11/0Copyright © 2011 Health Physics Society

DOI: 10.1097/HP.0b013e318200e203

594 www.health-physics.com

compare it not only with exposure guidelines, but also withexposures from mobile phone handsets.

In an attempt to fill these gaps, a systematic inves-tigation was carried out on the types of Wi-Fi equipmentin use in UK schools. Extensive laboratory measure-ments were carried out to assess the exposure of pupils tosuch devices. This paper reports the distribution ofelectric field strengths and the calculated radiated powersaround a sample of popular Wi-Fi devices in schoolsduring transmission. Other parts of the project involvecomputer modeling of the energy absorption in chil-dren’s bodies and determining the proportion of the timethat devices transmit during typical school lessons.

MATERIALS AND METHODS

Technical and regulatory standards forWLAN technology

The IEEE802.11x (1999) family of standards spec-ifies the technical configuration of Wi-Fi technology. Theoriginal 802.11 standard specified 1 Mbps and 2 Mbps datarates in the 2.4 GHz band. Standards 802.11a and 802.11gwere later developed to attain data rates up to 54 Mbps (theformer in the 5 GHz frequency band) and avoid overcrowd-ing in the 2.4 GHz band. The latest 802.11n standardprovides for devices that can use up to four channelssimultaneously and deliver rates up to 600 Mbps.

Radio devices are subject to the requirements of theRadio and Telecommunications Terminal Equipment(RTTE) Directive in Europe (Directive 1999/5/EC).Among other requirements, this Directive requires thatequipment not cause harmful interference. Harmonizedstandards allowing a presumption of conformity with thisessential requirement have been written by the EuropeanTelecommunications Standards Institute (ETSI); how-ever, non-compliance with these standards does notallow a presumption that harmful interference will occur.Standard EN 300328 (2006) applies to equipment oper-ating in the 2.4 GHz band and includes a requirement thatthe Equivalent Isotropically Radiated Power (EIRP) doesnot exceed 100 mW. Standard EN 301893 (2007) appliesto devices operating in the 5 GHz bands. The EIRP islimited to 200 mW in Band A (5.15–5.35 GHz) and 1000mW in Band B (5.470–5.725 GHz) [Band C (5.725–5.825 GHz) is not available in Europe]. These emissionstandards do not relate to health effects, although limitingthe emitted power from devices will indirectly have theeffect of limiting exposures.

The RTTE Directive also requires that hazardousradiation not be produced by devices, and “radiation”is taken to include radiofrequency electromagneticfields. Technical standards have been written by the

European Committee for Electrotechnical Standardi-sation (CENELEC) and the International Electrotechni-cal Commission (IEC), which provide a link to exposureguidelines; e.g., from ICNIRP. As above, compliancewith these standards enables a presumption that hazard-ous radiation is not produced, but non-compliance doesnot in itself allow the reverse to be concluded. EN 62311is a generic IEC standard, harmonized under the RTTEDirective, which may permit the assessment of Wi-Fidevices. EN 62209-2 (2010) is a more specific standardfor assessing exposure from devices used in close prox-imity to the body, which may be linked into the harmo-nized RTTE framework in due course.

ICNIRP guidelines contain reference levels ex-pressed as values of electric field strengths and powerdensity that can be compared with measured or calcu-lated values. The ICNIRP reference level values for thegeneral public limit the electric field strength and powerdensity at frequencies between 2 and 10 GHz to 61 Vm�1 and 10 W m�2, respectively. The reference levels arenot intended as limits, but are designed such that com-pliance with them should ensure compliance with morefundamental basic restrictions. These are expressed interms of specific energy absorption rate (SAR), measuredin W Kg�1, in the body tissues. Exposures in relation tothe basic restrictions are being assessed through com-puter modeling in another part of this project (Findlayand Dimbylow 2010).

Parameters that affect people’s exposureThe main parameters that should be considered in

assessing the exposure of people to WLAN sources aresummarized below.

Output power. Power absorption in body tissues isproportional to the output power of a device for a givenexposure scenario. Radio wave power density (W m�2) isalso proportional to output power, and electric field strengthis proportional to the square root of output power.

Frequency. Radio waves penetrate less into bodytissues as frequency increases. The electric field compo-nent of a wave penetrating into the body reduces to 36%of its initial value after a distance known as the skindepth. The skin depths of tissues with low water content,such as fat (31.5 mm) and bone (16.2 mm), are greaterthan those with higher water content, such as muscle (7.7mm) and skin (8.4 mm).†

Distance. The antennas inside laptops and otherWi-Fi devices are usually small (a few cm in diameter)

† The skin depth data are calculated for a frequency of 2.45 GHzand based on conductivity of tissues reported in Gabriel (1996).

595Exposure to electromagnetic fields from Wi-Fi in schools ● A. PEYMAN ET AL.

www.health-physics.com

compared to the distance from the device to the user.This usually means that the exposure reduces veryrapidly with increasing distance, broadly according to theinverse square law (far field region). Therefore it isimportant to note that users of laptops tend to be tens ofcm away from them (except for the hands), which isfurther away than mobile phones are normally used fromthe head.

Antenna configurations. Ideally WLAN antennaswould radiate their power equally in all directions, but realantennas tend to radiate preferentially in certain directionsand have nulls in others. The extent to which the antennasdirect the radiated power toward the user is a key factorwhen assessing exposure from Wi-Fi devices.

Most modern Wi-Fi devices are equipped with adiversity function involving the provision of severalantennas within or protruding from their body shell. Thisallows switching of individual bursts to the most appro-priate antenna for optimal performance; e.g., if theantenna in use begins to experience a fading signal.Because the antennas are located at different positionswithin the device, the radiation pattern of the devicechanges according to which antenna is in use.

Duty factor considerations. The international ex-posure guidelines from ICNIRP are based on powerabsorption in the body tissues averaged over 6 min.According to Foster (2007), since WLAN devices transmitshort pulses (bursts) and the duty factor (the ratio of thepulse duration to the pulse repetition period) of transmissionof access points is usually low, the time-averaged exposureshould be far below that produced by a continuous sourceoperating at the same peak power level.

The duty factor increases when data are transmittedto the WLAN device, and the increase in duty factordepends in turn on the rate of data transmission. For agiven data rate, a higher order modulation scheme (onewith more bits encoded per symbol) reduces the dutyfactor, which consequently can lead to lower exposure tothe user.

Data rates and quality of the signal. The maxi-mum data rate for a given variant of the 802.11 standardcan ideally be achieved if the Wi-Fi devices are closeenough to the access point. However, the throughput fallsto a lower rate as the distance increases due to signalattenuation. The data throughput is also affected by otherfactors such as reflections from surrounding objects andnetwork congestion, making it very difficult to predictthe radio conditions when the orientation of devices ischanged.

WLAN in UK schoolsMany devices with Wi-Fi connectivity options are

available in schools, including laptops, interactive whiteboards, thin clients, wireless slates, tablet PCs, andwireless microphones. Some schools have obtained wire-less voting systems to assist in the evaluation of lessons.The widespread use of wireless location and asset track-ing systems and wireless closed circuit television(CCTV) cameras in schools is also predicted. It is alsonotable that smart phones are becoming increasinglypopular among pupils.

Discussions with the British Education Communi-cations and Technology Agency (BECTA) and some ofthe major suppliers of educational resources to the UKschools revealed that laptops are the most popular wire-less devices used in schools, where up to 32 can be usedin a single classroom simultaneously. The number ofaccess points used in each school depends on the size andstructure of the building. A typical small primary schoolcan have four or five access points, while in largerschools the network can consist of up to 100 accesspoints.

According to BECTA (2007), the most widely usedWi-Fi standard in UK schools was 802.11g, wheretwo-thirds of primary and 80% of secondary schools hadequipment operating at this level. Either in combinationor exclusively, a third of primary and 46% of secondaryschools utilized 802.11b. The 802.11a standard is alsoused in some schools, usually for administration andteacher training. Overall, there seems to be no particulartrend in selecting a wireless standard in schools. It isexpected that Wi-Fi use in schools has increased stillfurther since these data were collected.

Devices under testDue to the popularity of laptops in classrooms and

the fact that the majority of Wi-Fi exposure would comefrom these client devices, it was decided that the exper-imental measurements would focus on the exposure ofchildren to laptops and access points.

Several laptop manufacturers and models were iden-tified from BECTA’s list of recommended suppliersthrough the so called “Infrastructure Services Frame-work.” Although specific makes and models of laptopsand access points were tested, the interest of the projectis in devices in general, and the models tested areregarded as examples of equipment used in schools;hence, the particular makes and models are not listed.

Tables 1 and 2 present lists of laptops and accesspoints selected for measurements in this study. All thelaptops listed in Table 1 had a conventional design withthe screen hinged from the back of the keyboard exceptfor one (LT13), which was in a rectangular Tablet PC

596 Health Physics June 2011, Volume 100, Number 6


format. The hinged laptops were all opened to an angle of115 degrees for the measurements, which seems to be thetypical screen orientation among most of the users. Theselected access points used in this study were eithercircular with internal antennas or rectangular with eitherinternal or multiple external antennas.

Configuration of WLAN devicesThe laptops and access points selected for assess-

ment in this study had various configuration options,network types and modes of operation based on the IEEEstandards of 802.11a, b, g and n.

The 802.11g standard (which is compatible with theearlier 802.11b standard) was tested in all the laptopswhere available, representing the most widely usedWi-Fi standard in UK schools. To have quantitativeinformation for devices working in the 5 GHz region, the802.11a standard was also considered in this study. Someof the devices also offered the 802.11n standard, but as itremained to be ratified when the testing was performed,no measurements were performed with this standard, andthe option was disabled throughout the study. Bluetoothtransmitters were also disabled in each laptop before

Table 1. List of laptops selected for testing in this study.a

DeviceID

Wi-Fimodes

Dimensionsb

(cm)

Available frequency channels

Power levels2.4 GHz 5−6 GHz

LT01 b/g 34 � 24 � 4 Not selectable Not available Not selectableLT02 b/g 36 � 27 � 4 Not selectable Not available Not selectableLT03 b/g 36 � 27 � 4 1−13 Not available 25, 50, 75, 100%LT04 a/b/g/n 37 � 27 � 4 1−11 36−48 1−5LT05 a/b/g/n 37 � 28 � 4 1−11 36−48 1−5LT06 b/g 34 � 25 � 4 1−14 Not available Not selectableLT07 a/b/g 32 � 24 � 4 1−11 36−48 1−6LT08 a/b/g 30 � 25 � 4 1−11 36−48 1−5LT09 a/g/n 33 � 23 � 3 1−14 36−64, 100−140, 149−165 27, 50, 75, 100%LT10 a/b/g/n 33 � 25 � 4 1−11 36−48 1−5LT11 a/b/g 26 � 17 � 5 1−14 36−64, 100−140, 149−165 25, 50, 75, 100%LT12 b/g 23 � 17 � 4 Not selectable Not available Not selectableLT13 b/g 23 � 12 � 3 Not selectable Not available Not selectableLT14 a/b/g 40 � 29 � 5 1−11 36−48 1−5LT15 b/g 28 � 24 � 4 1−13 Not available 1−6

a Note 1: Channel 1 in the 2.4 GHz band is 2.412 GHz, and successive channels are in 5 MHz steps up to Channel 14 at 2.484 GHz.Note 2: Only every fourth channel is available from 36 to 64 in the 5 GHz band. Channel 36 is 5.180 GHz, and the increments are20 MHz upto channel 64, which is 5.320 GHz.Note 3: Only every fourth channel is available from 100 to 140 in the 5 GHz band. Channel 100 is 5.500 GHz, and the incrementsare 20 MHz up to channel 140, which is 5.700 GHz.Note 4: Only every fourth channel is available from 149 to 165 in the 5 GHz band. Channel 149 is 5.745 GHz, and the incrementsare 20 MHz up to channel 165, which is 5.825 GHz.b Dimensions are presented as Length � Width � Depth.

Table 2. List of access points selected for testing in this study.a

DeviceID

Wi-Fimodes

Dimensionsb

(cm)

Available frequency channels

Power levels2.4 GHz 5−6 GHz

AP01 b/g/n 21 � 16 � 4 1−13 Not available Not selectableAP02 a/b/g 23 � 16 � 4 1−11 36−48 1−5 (each step doubles power)AP03 a/b/g/n 17 � 6 1−13 36−48 Any level to maximumAP04 b/g/n 22 � 15 � 4 1−13 Not available Not selectableAP05 b/g/n 17 � 17 � 4 1−13 Not available Not selectableAP06 b/g 14 � 14 � 3 1−13 Not available Not selectableAP07 a/b/g 10 � 10 � 4 1−13 36−48, 36−40, 100−140 Any level to maximumAP08 b/g 22 � 16 � 4 1−13 Not available 1−5 (each step doubles power)AP09 a/b/g/n 21 � 18 � 5 1−11 36−48 Not selectableAP10 a/b/g 19 � 19 � 3 1−13 Not selectable 7 steps to maximumAP12 a/b/g 17 � 5b 1−13 Not available Any level to maximumAP13 b/g 23 � 18 � 4 1−13 36−48 Not selectable

a Note 1: The shape of AP3 and AP12 were similar to a typical smoke detector.Note 1: AP1, AP4 and AP9 had 3 external antennas.Note 2: AP2, AP5 and AP7 had 2 external antennas.Note 3: AP6 had one external antenna.Note 4: AP3, AP8, AP10, AP12 and AP13 had internal patch antennas.b Dimensions are presented as Length � Width � Depth.



performing the measurements in order to minimize thechannel interference and to eliminate the unwantedsignals.

Laboratory measurement setupThe main objectives of the laboratory measurements

were to establish the angular distribution of electric fieldstrength around each laptop and access point, integratethe power flowing through a surface enclosing eachdevice to determine the radiated power, identify theangular directions in which the electric field strength wasat a maximum, and finally to measure the electric field

strength at these angles as a function of distance. For this,a new experimental system was constructed at HPA’sChilton site, as described in the following sections. Thearrangement of equipment used is also shown in Fig. 1.

Anechoic chamber and manual positioningdevice (goniometer). An anechoic chamber was specif-ically made for this study, within which the electric fieldstrength around each device under test was measured.The room has dimensions of 3.6 m � 2.4 m � 2.4 m withwooden walls that are lined internally with a foam-basedradiofrequency absorber material (Emerson & Cuming,

Fig. 1. The arrangement of different instruments required for the E-field measurement from a laptop: (a) inside theanechoic chamber and (b) the schematic arrangement.



ECCOSORB AN79, N.V. Nijverheidsstraat 7A, B-2260Westerlo, Belgium) specified to have a reflection coef-ficient of �20 dB at Wi-Fi frequencies.

The device under test was mounted on a manualpositioning device, known as a goniometer, allowing thedevice to be rotated in two orthogonal planes andpermitting the measuring antenna to sample the radiationpattern at any angle. The distance of the measuringantenna could also be varied to profile electric fieldstrength as a function of distance.

All of the measurements during this work weremade in the far field region with respect to the sources.At 2.4 GHz, the wavelength is about 12.5 cm, whichmeans the reactive near field extends to around 2 cmfrom the source (based on the usual �/2� criterion, where� is the wavelength). X-raying and dismantling some ofthe laptops showed that the internal Wi-Fi antennas areno more than around 5–10 cm in size, while the antennasof access points were up to around 10 cm in size. Hence,the radiating near field extends no further than around 16cm at 2.4 GHz (based on the usual 2D2/� criterion, whereD is the maximum source dimension). Repeating thesecalculations for 5 GHz gives distances of around 1 cmand 33 cm for the extent of the reactive and radiating nearfield regions, respectively.

As apparent from Fig. 1, the manual positioningsystem consists of a pillar supporting a turntable onwhich a horizontal cradle is mounted. The turntableallows for azimuth rotation, and the cradle allows forelevation. The goniometer is made of non-conductingmaterial with low dielectric constant, while still provid-ing the necessary rigidity. The pillar and the turntablewere made of Perspex (�r � 3.4�3.8), the cradle wasmade of polycarbonate (�r � 2.7�3.1) and the bearingsand screws were made of nylon (�r � 3�4). Perspexpillars were used to raise the laptop to the appropriateheight above the cradle so its hinge was aligned with theaxis of rotation. A Perspex clamp across the keyboardsecured the laptop in place.

Electric field antennas. An Austrian Research Cen-tres (ARC) Seibersdorf Precision Conical Dipole (PCD)8250 broadband antenna (Seibersdorf Labor GmbH,Forschungszentrum, 2444 Seibersdorf, Austria) was cho-sen to make measurements of electric field strength in the2.4 GHz band. The Siebersdorf PCD is a broadbandantenna covering the frequency range of 80 MHz to 3GHz. The length of the PCD is 13 cm, which makes itpossible to carry out field measurements at a distance of39 cm and more from the source as recommended byIEEE C95.3 (1991). When connected to a signal ana-lyzer, the PCD antenna is capable of measuring electricfield strengths down to 10 mV m�1.

Antenna pattern measurements, balance tests andnoise floor tests were performed at 2.4 GHz–2.5 GHz toensure that the characteristics of the antenna were ade-quate for field measurements at these frequencies. Forantenna pattern measurements, the variation of the an-tenna’s response as a function of angle was examined.All the measurements were performed in the purposelybuilt anechoic chamber, and the results showed noasymmetry at either 2.4 or 2.5 GHz.

The balance test was also carried out by measuringthe difference in output value using a vector signalanalyzer when the antenna was rotated by 180 degrees.The results revealed that the large majority of thedifferences were within �0.5 dB, the acceptance marginset by the UK National Physical Laboratory (NPL).

For Wi-Fi devices operating in the 5 GHz band, aQ-par Angus QSH12N10S horn antenna (Q-par AngusLtd., Barons Cross Laboratories, Leominster, Hereford-shire HR6 8RS, UK) with a frequency range of 3.9 to 5.9GHz was used. The aperture of the horn antenna is 11.8cm diameter, which makes it possible to carry out fieldmeasurements at a distance of 35 cm or more from thesource. The horn antenna was calibrated to nationalstandards at the required frequencies by the NationalPhysical Laboratory, and it was connected to the signalanalyzer via a 5 m co-axial cable. The loss of the cablewas measured using traceable equipment and was 5.8 dBat 5.2 GHz.

Antenna pattern measurements were also performedat 5.2 GHz to estimate the beam width of the hornantenna. These showed that it had a narrower beam widththan that of the PCD antenna at 2.5 GHz. Therefore, toensure that the antennas of even the largest laptop werewithin the main beam of the horn antenna, the distancebetween the device under test and the antenna wasincreased from 1 m to 1.5 m.

Vector signal analyzer (VSA). Almost all of theprevious studies on the exposure assessment from Wi-Fiequipment have used spectrum analyzers as their maintool to analyze the signals (Foster 2007; Schmid et al.2007a and b). However, in the authors’ opinion, spec-trum analyzers were not ideally suited to the purpose ofthis project. This is because their maximum bandwidth istypically only 5 MHz and therefore significantly lessthan that of Wi-Fi signals, meaning that post hoc inte-grations of the power spectrum have to be performedusing special functions such as band power measure-ment. Long sweep times have to be used in order toaverage out the stochastic nature of the Wi-Fi signal andgain a repeatable measurement, even when transferringdata at a notionally uniform rate.



Instead, for this study, an Agilent N9020A MXAsignal analyzer (Agilent Technologies UK Ltd., 610Wharfedale Rd., IQ Winnersh, Wokingham, Berkshire,RG41 5TP, UK) with a frequency range from 20 Hz to8.4 GHz was used. It has a bandwidth of 25 MHz,allowing the detection of the whole WLAN signal. It wasalso possible to capture individual Wi-Fi bursts in thetime domain and then analyze them to identify the burstpower, modulation scheme, and many other parameters.

LanTraffic software. In order to generate andcontrol the Wi-Fi signal from the laptops, a wirelessnetwork was required to establish either on ad hoc orinfrastructure mode. In this experiment, the infrastructurenetwork consisted of a laptop and an access pointhardwired to a computer. The advantage of this setup isthat it provides much more control over the structure ofthe radio signal than is possible simply by transferringfiles or video using everyday software. LanTraffic soft-ware (zti telecom, ZTI, 1 Boulevard d’Armor, BP 20254,22302 Lannion, France) was used to control and monitorthe traffic between the device under test (the WLANtransmitter, or sender) and the WLAN receiving device(also called receiver; e.g., an access point in the case oflaptop measurements).

Wireless transmission measurement software. Avisual basic program was developed to allow communi-cation with the signal analyzer control software. Theprogram uploads the specified settings and extracts thefollowing parameters from each burst:

● Position of laptop: either an angular position (�90° �� 90°, �180° � � � 180°) or a distance (0.5 m to1.9 m);

● Time: allows synchronization of the measured datawith the transmission data logged by LanTraffic;

● Burst power: the main parameter for determining theRMS (Root Mean Squared) electric field strengthduring the burst;

● Band power: the spectral power of the burst, which intheory should be roughly equal to the burst power.This is measured to allow filtering out of acknowl-edgement signals emitted by the WLAN receivingdevice;

● Modulation scheme: allows monitoring of the qualityof the transmitted WLAN signal; and

● Error Vector Magnitude (EVM): gives a quantitativemeasure of the quality of the transmitted WLAN data.

Calibration check of the measurement setupAlthough the MXA signal analyzer has a traceable

manufacturer’s calibration, a procedure was followed tomonitor the overall uncertainty associated with the field

strength measurements. An Agilent E4438C vector sig-nal generator (Agilent Technologies UK Ltd., 610Wharfedale Rd., IQ Winnersh, Wokingham, Berkshire,RG41 5TP, UK) was used together with an AgilentE4418B single channel power meter (Agilent Technolo-gies UK Ltd., 610 Wharfedale Rd., IQ Winnersh, Wok-ingham, Berkshire, RG41 5TP, UK) to provide anadditional route to traceability and confidence checksthroughout the measurement session.

The Agilent E4438C signal generator has a fre-quency range from 250 kHz to 4 GHz and is capable ofgenerating arbitrary waveforms, which emulate a varietyof wireless communication signals that use frequencyand time multiplexing, including those of IEEE 802wireless LANs. Agilent Signal Studio WLAN N7617Bsoftware was used to generate a waveform that emulatesa Wi-Fi signal of a given modulation scheme, dutyfactor, amplitude and burst duration, which can then beuploaded onto the signal generator. The amplitude of thegenerated signal was �25 dBm, which is higher than thatmeasured from the laptops. The duty factor was maxi-mized so that the burst power was equal to the time-averaged power when measuring with the power meter.

The Agilent E4418B single channel power meteris compatible with all 8480 series sensors and providesaverage power measurements from �30 to �20 dBm.A highly accurate sensor head, the 8482A (100 kHz–4.2 GHz), was used to cover the 2.4 GHz frequencyregion. The sensor head was calibrated to nationalstandards to check the accuracy of the MXA withtraceable equipment.

The integrity of the measured data was assessed bycomparing the readings from MXA to those of the powermeter at the beginning and the end of each measurementsession. The differences were generally small (�0.5 dB,and the maximum recorded was 0.8 dB).

Since the E4438C signal generator had an upperfrequency range of 4 GHz, this procedure could not beperformed at 5 GHz. However, the low offset valuesfound at 2.4 GHz provided overall confidence in thecalibration of the MXA signal analyzer.

Evaluation of electric field around devicesunder test

The main objective of this study was to establish theangular distribution of the electric field around eachdevice under test and quantify the electric field strengthas a function of distance in the direction of maximumemission.

Angular distribution of electric fields aroundlaptops. The electric field strength was measured in two



orthogonal directions: x and y. The electric field projec-tion along the line connecting the center of the goniom-eter and the measuring antenna (defined as the z-axis)was considered negligible given the distance between thesource and the antenna (1 or 1.5 m).

The electric field strength around each of the laptopswas measured at a matrix of points located on the surface ofa sphere enclosing the laptop. Several combinations ofazimuth and elevation rotation on the goniometer wereconsidered, covering positions of �90° � � � 90° and�180° � � � 180° in 30° steps. The azimuth (�) andelevation (�) angles refer to the spherical coordinate system,with the z-axis coinciding with the hinge of the laptop. Ashift in angles was made to ensure that � � 0° and � � 0°correspond to the front of the laptop in the direction of thetorso of the user (Fig. 2). For access points with externalantennas, the feed point of the central external antenna wascentered in the goniometer with the antennas aligned withthe main body of the access point.

The distance between the measurement antenna andthe device under test was set as a compromise to achieveenough sensitivity to detect the incident signal whilediminishing the amplitude of the element of the signalreflected by the walls. For the PCD antennas, themeasuring distance was set at 1 m. However, due to thedirectional characteristics of the horn antenna, the mea-surement distance was increased to 1.5 m.

Careful considerations were made to determine thebest position for the access point to ensure stable andreliable communication with the laptop without causinginterference with the measurements. Placing the accesspoint anywhere in the anechoic chamber generated asignal at the measuring antenna comparable to that fromthe laptop, thereby triggering the detection of a burst. Itwas also noticed that transmission was optimized whenthe access point was not too close to the laptop. Althoughtests showed it was possible to place the access point

outside the anechoic chamber, which would drop its signalat the measuring antenna to negligible levels, the quality ofthe transmission could not always be guaranteed. Thereforethe solution adopted was to place the access point in thecorner of the anechoic chamber furthest from the laptop,enclosed with absorbing foam blocks. The absorbers en-closing the access point were placed 10–20 cm away fromthe device to avoid overheating. This setup allowed a good,repeatable and stable transmission and minimized thechance of the signal analyzer detecting access point signals.Despite the measures taken to avoid interference, occasion-ally readings were recorded by the signal analyzer fromaccess points, and these were removed during the post-measurement data analysis.

Angular distribution of electric fields aroundaccess points. The access points assessed in this studycould be broadly categorized into two types: Those of thefirst type were typically circular-shaped with internalpatch antennas and a ground plane behind the antennas.This type of access point was designed to be mounted onwalls, with the field distribution directed away from thewall. The second type of access point had multiple (typi-cally 1 to 3) protruding monopole antennas generating anomni-directional radiation pattern about their axes, withnulls in the direction to which each antenna pointed and alsoin the broadly opposite direction. Since most of the accesspoints tested in this study had multiple monopole or patchantennas, it was expected that the radiation pattern of thesedevices had directional characteristics depending on thegeometry of the ground plane inside their body-shell inrelation to the antenna structure.

Initial tests on the first type of access point revealedoccasional blocking of the signal by the ground plane,resulting in the failure of the communication link. Thismade it practically impossible to perform a complete setof spherical measurements, and therefore hemispherical

Fig. 2. Radiation pattern coordinate system for (a) laptops and (b) access points.



measurements were made assuming that the field distri-bution would be similar or lower behind the main body(or ground plane) of the devices under test, according tothe symmetry of the devices. The measurements weremade in 30° steps at each angular position for �90° � �� 90° and �90° � � � 90°. The mounting plate on thegoniometer was also modified so that the access pointscould be positioned centrally during the measurements.

Electric field strength as a function of distanceAt this stage, the devices under test were positioned

at the angles that produced the maximum field strength,and measurements were carried out as a function ofdistance, ranging from 0.5–1.9 m in 10 cm steps.

Practical considerations

Signal stability tests and warming up issues.Initial measurements revealed that the burst power valuesrecorded by the signal analyzer drifted by 15% in the first30 min after the transmission was started. This reduced toless than 3% after 2 h of transmitting, emphasizing theneed for adequate equipment warm-up times. As a result,a 30 min warm-up time was allowed for all equipment(including the laptops) prior to measurements.

Number of data points collected. For each deviceunder test and at each position, 50 samples of electricfield strength were extracted. This number of sampleswas determined by the maximum practical duration of ameasurement session. A larger number of samples wasnot needed since burst power measurements were quitestable for most of the devices. Most burst power mea-surements associated with a given power level had a

standard deviation of less than 10%, with some laptopshaving most of their burst power measurements withstandard deviations below 1%.

Duty factor considerations. The proportion of thetime during which devices transmit in normal use is animportant factor to consider when assessing the expo-sure. The amount of data transferred, the modulationscheme chosen by the Wi-Fi device, and retransmissionof the signal due to weak or lost packets are amongparameters that affect the duty factor. To investigate thedependence of the duty factor on these aspects of the Wi-Fisignal, a series of measurements was carried out in aGigahertz Transverse Electro-Magnetic (GTEM) cell (ETS-Lindgren, Unit 4, Eastman Way, Pin Green Industrial Area,Stevenage, Hertfordshire, SG1 4UH, UK) where laptopscould be isolated from the influence of external radiosignals. The output power was monitored with a spectrumanalyzer connected to the GTEM cell. A network connec-tion was also established and maintained between a trans-mitting laptop and a receiving access point. The set-up forduty factor measurements is shown in Fig. 3.

The laptops were configured to transmit at 22 Mbpsusing two User Datagram Protocol (UDP) ports tosustain a high duty factor transmission. A dual direc-tional coupler was connected to the input port of theGTEM cell in order to separate the transmission of thelaptop from the beacon pulses of the access point. Amanual stepped attenuation function (0–23 dB in 1 dBsteps) was also used to simulate progressively increasingdistances between the access point and the laptop andreplicate the attenuation produced by the radio environ-ment. The results showed that quality and strength of the

Fig. 3. Block diagram of the GTEM measurement system used for duty factor measurements.



radio signals had a direct effect on the data transmissionrate and the duty factor, as apparent from Fig. 4. For agiven radio environment (i.e., for a fixed attenuation),a reduction in data rate by the user would result in acorresponding reduction in the duty factor. However,when the data rate was reduced by the Wi-Fi devices due

to the poor radio conditions and weaker signal strength,the effect was opposite and the duty factor was increased.This was because it takes longer to send a given amountof data using a lower order modulation scheme. Thelowest possible data rate was the 1 Mbps offered in the802.11b standard, and the highest duty factor recordedwas 96%.

Measurement uncertainty. All the measurementdevices used in this study either had a traceable manu-facturer calibration or were calibrated to national stan-dards by the UK National Physical Laboratory. However,as with any other physical quantity, the measurement ofelectric field strength is associated with a number ofuncertainties. Therefore, to have a realistic view of themeasured quantities, a simple uncertainty budget wasdrawn up, taking into consideration both random andsystematic sources of errors. Different elements of un-certainty that would affect the measurement setup wereidentified and given an appropriate statistical distribu-tion, and the total combined uncertainty values were thencalculated. The uncertainty budget presented in Table 3for the electric field strength measurement was drawn in

Fig. 4. Effect of attenuation on duty factor and data throughput(error bars represent the range of values gathered in the experi-ment). The dotted and dashed lines represent the data, while thesolid lines are six-order polynomial best-fit curves.

Table 3. Elements of uncertainty and the total combined uncertainty associated with the measurements of electric fieldstrength a: at 2.4 GHz b: at 5 GHz. The nomenclature and methodology were taken from NIST (1994) and ISO (1995).To obtain uncertainty values for power density the values in this table should be doubled.a

Source Parameter

Specificuncertainty

(dB) DistributionDivision

factorStandard

uncertainty

VSA Frequency response 0.67 Normal 2 0.34VSA Display reading uncertainty 0.10 Rectangular 1.73 0.06VSA Drift 0.26a Rectangular 1.73 0.15Antenna Calibration uncertaintyb 1.20 Normal 2 0.60Antenna Uncertainty caused by imbalance 0.50 Rectangular 1.73 0.29Mismatch Uncertainty caused by reflections 2.00 U shaped 1.41 1.42Repeatability Limited repeatability 1.80 Normal 1 1.80

Combined standard uncertainty [dB] 2.42Expanded uncertainty (K � 1.96) [dB] 4.73

b

Source Parameter

Specificuncertainty

(dB) DistributionDivision

factorStandard

uncertaintyVSA Frequency response 0.92 Normal 2 0.46

VSA Display reading uncertainty 0.10 Rectangular 1.73 0.06VSA Drift 0.26c Rectangular 1.73 0.15Antenna cable Calibration uncertainty 0.10 Normal 2 0.05Antenna cable Cable attenuation frequency interpolation 0.03 Rectangular 1.73 0.02Antenna Calibration uncertainty 0.80 Normal 2 0.40Mismatch Uncertainty caused by reflections 2.90 U shaped 1.41 2.06Repeatability Limited repeatability 0.50 Normal 1 0.50

Combined standard uncertainty [dB] 2.21Expanded uncertainty (K � 1.96) [dB] 4.33

a The value quoted is from preliminary measurements that showed a drift of 3% within a timeframe of 2 h after the system was warmedup for 30 min.b Uncertainties from the cable connecting the antenna to the VSA have been included in the antenna calibration.c The value quoted is from preliminary measurements that showed a drift of 3% within a timeframe of 2 h after the system was warmedup for 30 min.



accordance with established guidelines (NIST 1994; ISO1995; CENELEC 2008).

Processing of the experimental dataThe experimental data obtained during the electric

field strength measurements were subjected to somepost-processing as mentioned below.

Filtering of the data. Some of the recorded datawere not considered reliable and were excluded forspecific reasons. In some instances, the signal analyzercould not lock onto a burst, resulting in a large EVMvalue (�25%). Occasional detection of the access pointsignal, overloading of the signal analyzer and freezing ofthe data acquisition program also occasionally resulted inthe recording of spurious data. As a rule of thumb, anyrecorded burst power value which was larger than �5dBm (or 0 dBm for measurements as a function ofdistance) has been removed from the main data set.

Splitting of the data into several groups. Prelim-inary measurements of electric field strength showed thatin a given set of samples, there could be two or threedistinct levels of burst power values (Fig. 5). This wasdue to switching between two or more antennas in someof the devices under test. Therefore, during the dataanalysis, the samples were grouped and tested for inde-pendence using the Excel inbuilt t test algorithm.

Synchronization of the LanTraffic data with thesignal analyzer measurements. In order to check thequality of the transmission during a typical measurementsession, the data rate recorded by the LanTraffic as a

function of time and the time at which the samples wererecorded by the signal analyzer were plotted against eachother.

Conversion of the raw data into electric fieldstrength. The electric field strength was calculated usingthe following equation (NPL 2004):

E � V � AF � ATT, (1)

where E is the electric field strength (dB�Vm�1), V is themeasured voltage (dB�V), AF is the antenna factor(dBm�1), and ATT is the cable attenuation (dB). Substi-tuting for V and including the attenuation of cable for themeasurement antenna in the antenna factor, eqn (1) canbe rewritten as:

E � 106.9897 � P � AF, (2)

where P is the measured power (dBm).

Calculation of power density and radiated power.The power density was derived using eqn (3):

S �E2

377(3)

where the unit of S is W m�2, and E has now beenconverted to linear units.

It is also useful to calculate the equivalentisotropically-radiated power (EIRP) of an antenna forcomparison with the emission limits set by standardsbodies. EIRP is the power that would have to be emittedif the antenna were isotropic in order to produce a powerdensity equal to that observed in the direction of maxi-mum gain of the actual antenna. In this study, the EIRP

Fig. 5. Burst power samples fluctuating among three levels due to antenna diversity; data at two different angulardirections are presented.



was calculated using the maximum measurement ofpower density:

EIRP � 4�r2Smax(r), (4)

where EIRP is in units of W, r is the distance to theantenna in meters, and Smax(r) is the maximum powerdensity measured at that distance in W m�2.

For real antennas, which are not isotropic, the EIRPis greater than the true radiated power by an amountknown as the antenna gain. Therefore it is more realisticto estimate the total radiated power by integrating thepower flowing through a spherical surface enclosing thedevice. To calculate the integrated radiated power (IRP),the power density at each measurement point is multi-plied by the area of the spherical surface that itrepresented:

Stot � �i

�j

A��i, �jS��i, �j, (5)

where S(�i, �j) is the measured power density at eachposition, and A(�i, �j) is the area represented by thatsample.

Eqn (5) takes into account emissions from all theexisting antennas in the device simultaneously, while inreality, at a given moment in time only one antenna istransmitting. Therefore, at certain angles the IRP calcu-lated from eqn (5) consists of emissions from oneantenna, while at other angles it takes into accountemissions from a different antenna. In general, thecalculated radiated power will be from the antenna thatproduces the greatest field strength at the recording site(the position of the measuring antenna), which meansthat the integrated radiated power should be an over-estimate of the true radiated power.

MEASUREMENT RESULTS

Figs. 6a and b show the radiation pattern for selectedlaptops operating at 2.4 and 5 GHz bands, respectively.The laptops shown are those with the highest recordedintegrated radiated power. Similar plots are presented foraccess points in Figs. 7a and b. The left and right edgesof the radiation patterns, shown in these figures, corre-spond to the directions 90 degrees to the sides of thedevice under test.

Table 4 contains the angular direction of themaximum electric field strength measured for eachdevice under test at 1 m (for 2.4 GHz) and 1.5 m (for5 GHz). The corresponding field strength values ateach angle, the equivalent power densities and theEIRP values are also presented. The spherically-integrated radiated power (IRP) for all devices undertest is presented in Table 5.

Figs. 8 and 9 show the power density valuescalculated from maximum electric field strength as afunction of distance for all the laptops and access points,respectively. The solid curves represent the power den-sity calculated from a theoretical antenna having an EIRPequal to 100 mW at 2.4 GHz or an EIRP of 200 mW at5 GHz (the limits in the harmonized standards).

Although in most cases, two to three power densitylevels (corresponding to more than one antenna) wereobserved, for the purpose of clarity only the highestpower density levels are displayed in Figs. 8 and 9.

DISCUSSION

Radiation patternThe laboratory measurements showed that the elec-

tric field strength distributions were not particularlyuniform around the laptops. However, as apparent fromFigs. 6a and b, the radiation patterns for different laptopsgenerally display some common characteristics: a mini-mum in the direction away from the front of the laptop(towards the torso of the user), and two maxima sym-metrically opposed across a vertical plane bisecting thescreen and keyboard. From the outset of the project, twoto three distinct levels of electric field strength wereobserved for most of the laptops. This is due to the factthat laptops typically had more than one antenna: usuallyone mounted at the top left and one at top right corner,behind the monitor screen. This was confirmed by x-rayimages taken from individual laptops. Each of the anten-nas present in a laptop would produce one of the maximamentioned above.

In the case of access points, the power densities alsovaried around the spherical surface with maxima andminima around the contour plot. In particular, AP01 andAP02 showed maxima in directions to the sides of themain body of the access points. It should be pointed out,however, that at these positions the transmitting antennaswere closest to the measuring antenna and were alignedwith the x-y plane of the measuring antenna. AlthoughFig. 7 seems to suggest that power density distributionsof access points were more symmetric than those oflaptops, it is difficult to interpret these patterns as themaxima are smaller than the angular resolution of themeasurements (30°). It should be noted, however, thatthe resolution of the measurements is limited by the sizeof the measurement antenna, the distance between themeasuring antenna and the source, and the duration of themeasurements.

Maximum radiation anglesThe results showed that the directions of the maxima

were variable for most of the laptops, while some



Fig. 6. Radiation pattern for selected laptops operating at (a) 2.4 GHz (b) 5 GHz.



Fig. 7. Radiation pattern for selected access points operating at (a) 2.4 GHz (b) 5 GHz.



(namely LT02, LT03, and LT04) have similar directionsof maximum radiation. The majority of the maximaoccurred at � � �30° and � � �60° (Table 4).

The angular field strength maxima were related totwo main factors: the intrinsic directional characteristicsof the antennas and the position of the transmittingantenna in the device. In the case of the latter, it isimportant to note that the distance to the measurementpoint is taken from the center of the Wi-Fi device (e.g.,laptop) and not to the location of the transmitting antennain the device. Considering that laptop antennas aregenerally on the top edge of the screen at each corner,this implies that the actual distance to the transmittingantenna may differ from the measurement distance (of 1or 1.5 m) by up to around 15 cm.

Variation of electric field strength and powerdensity as a function of distance

Overall, the ensemble of power density plots calcu-lated from maximum electric field strength as a functionof distance (Figs. 8 and 9) broadly follows the expectedinverse-square dependence on the distance.

For the laptops operating at 2.4 GHz, the maximumelectric field strength decreased from 2,893 to 226mV m�1 as the distance increased from 0.5 to 1.9 m.

Table 4. Angular direction, maximum electric field strength and equivalent power density measured at 1 m for devicesusing IEEE 802.11g/b standard (2.4 GHz) and at 1.5 m for devices using IEEE 802.11a standard (5 GHz).

Device ID

IEEE 802.11 g/b standard(2.4 GHz) measurement

distance � 1 m

IEEE 802.11a standard(5 GHz) measurement

distance � 1.5 m

Orientation(�, �)

Electric field(mVm�1)

Power density(mWm�2)

EIRP(mW)

Orientation(�, �)

Electric field(mVm�1)

Power density(mWm �2)

EIRP(mW)

LT01 60, �30 1,045 3 36 — — — —LT02 �60, �150 1,216 4 49 — — — —LT03 �60, �150 1,306 5 57 — — — —LT04 �60, �150 1,048 3 37 0, 180 549 1 23LT05 �30, 90 719 1 17 �30, �150 467 1 16LT06 �30, �90 1,154 4 44 — — — —LT07 �60, 180 1,055 3 37 �90, 150 740 1 41LT08 30, 120 766 2 20 �30, 150 440 1 15LT09 �30, 120 1,009 3 34 �90, 120 254 0.2 5LT10 30, 150 1,054 3 37 30, 180 605 1 28LT11 �90, 150 1,144 3 44 0, �150 770 2 45LT12 30, �120 1,142 3 44 — — — —LT13 90, 60 837 2 23 — — — —LT14 60, 180 970 3 31 0, �30 368 0.4 10LT15 �60, 180 909 2 28 — — — —AP01 90, �60 2,257 14 170 — — — —AP02 �90, 90 2,619 18 229 �90, 90 860 2 56AP03 30, �60 1,987 10 132 �90, 30 1,233 4 114AP04 60, 90 1,160 4 45 — — — —AP05 �90, �90 1,136 3 43 — — — —AP06 90, �30 1,350 5 61 — — — —AP07 90, �90 2,091 12 146 60, 0 1,483 6 165AP08 60, �90 1,402 5 66 — — — —AP09 �90, 60 683 1 16 �90, �90 472 1 17AP10 �30, 0 1,064 3 38 �30, �30 1,185 4 105AP12 �30, �60 1,592 7 85 �90, 30 796 2 48AP13 �30, 0 1,136 3 43 — — — —

Table 5. The spherically integrated radiated power (IRP) for alldevices under test.

Device ID

IEEE 802.11g/bstandard (2.4 GHz)

IRP (mW)

IEEE 802.1astandard (5 GHz)

IRP (mW)

LT01 9 —LT02 17 —LT03 15 —LT04 12 9LT05 5 4LT06 11 —LT07 11 16LT08 9 5LT09 16 1LT10 10 6LT11 9 13LT12 15 —LT13 8 —LT14 11 4LT15 8 —AP01 20 —AP02 24 9AP03 28 29AP04 8 —AP05 6 —AP06 5 —AP07 12 29AP08 9 —AP09 3 3AP10 10 25AP12 14 7AP13 8 —



These correspond to power density values of 22 to 0.13mW m�2, respectively. Similar trends are observed forlaptops operating at 5 GHz.

For access points operating at 2.4 GHz, the maxi-mum electric field strength decreased from 5,716 to 286mV m�1 as the distance increased from 0.5 to 1.9 m. Thecorresponding maximum power density value at 0.5 mwas 87 mW m�2, decreasing to 0.22 mW m�2 at 1.9 m.Similar trends are observed for access points operating inthe 5 GHz band.

Considering the power density measurements atshort distances (less than 1 m at 2.4 GHz and 1.5 m at 5GHz), it is important to recognize that the field valuesmeasured at lesser distances could be lower than the truevalues because the measurement antennas are not isotro-pic and do not fully measure power incident on them atangles away from boresight. The position of the antennain the laptop with respect to the measurement antennas isnot known exactly during measurements and will deviate

from the center of the main beam of the measurementantenna (although within the main beam). Also, themeasurement distance may differ from 0.5 m by up to 15cm with larger laptops. This can result in the measuredfield strength being attenuated by an amount that de-pends on the position of the transmitting antenna withrespect to the measurement antenna. The maximum errorresulting from this is estimated to be less than 0.5 dB forthe 2 GHz measurements and less than 1.2 dB for the 5GHz measurements.

Maximum radiated powerThe spherically-integrated radiated power for lap-

tops ranged from 5 to 17 mW in the 2.4 GHz band andfrom 1 to 16 mW in the 5 GHz band (Table 5). For accesspoints, the hemispherically-integrated powers rangedfrom 3 to 28 mW at 2.4 GHz and from 3 to 29 mW at 5GHz. However, given that some of the access points hadsymmetrical structures with regard to the unmeasured

Fig. 8. Variation of power density as a function of distance for laptops operating at (a) 2.4 GHz and (b) 5 GHz band.The power density calculated from an EIRP equal to 100 mW and 200 mW limits are shown for comparison. The errorbars represent the standard deviation of a given set of samples recorded by the signal analyzer at that position. Forclarity, only the largest set of error bars is shown.

Fig. 9. Variation of power density as a function of distance for access points operating at (a) 2.4 GHz and (b) 5 GHzband. The power density calculated from an EIRP equal to 100 mW and 200 mW limits are shown for comparison.



hemisphere, it might be reasonable to assume that thesedevices had total powers that were double thehemispherically-measured powers. Thus, while the max-imum radiated powers from access points are greater thanfrom laptops, the variation in the values is large and it isdifficult to generalize.

The equivalent isotropic radiated power (EIRP)values recorded at 1 m for laptops operating at 2.4 GHzranged from 17 to 57 mW (Table 4). Laptops operating at5 GHz have EIRP values recorded at 1.5 m in the rangeof 5 to 45 mW. Thus, the EIRPs of laptops were arounda factor of three higher than their IRPs, indicatingantenna gains of around 5 dB.

The equivalent isotropic-radiated power (EIRP) val-ues recorded at 1 m for access points operating at 2.4GHz ranged from 16 to 229 mW (Table 4). Access pointsoperating at 5 GHz have EIRPs recorded at 1.5 m in therange of 17 to 165 mW. Thus the EIRPs of access pointswere around a factor of five higher than their IRPs.Assuming no power was radiated into the unmeasuredhemisphere, this would imply antenna gains of around 7dB, although a perfectly symmetrical pattern with equalpower directed into the unmeasured hemisphere wouldimply antenna gains of around 4 dB.

In considering the EIRP values above 100 mW fordevices operating at 2.4 GHz, it should be noted that themeasurement protocol used in this work did not followthe precise testing schedule required under the emissionstandards mentioned earlier (EN300328/301893). Also,measurement uncertainty (Table 3) could account forsome of the difference. However, the measured EIRPsare consistent with radiated powers of several tens ofmW and antenna gains that are higher than thoseof simple monopoles or patches. While these types ofantennas are mounted on the bodies of the access points,the large size of the complete access point structures inrelation to the wavelength would tend to increase gain.

Comparison with exposure guidelines and mobilephone exposure

All the field strengths recorded in this study are wellbelow the corresponding ICNIRP reference level valuesdefined for the general public (ICNIRP 1998 and 2009).

For all the devices tested, the maximum electricfield strength recorded at 0.5 m was 5,716 mV m�1. Thiscorresponds to a maximum power density value of 87mW m�2. These values are well below the 61 V m�1 and10 W m�2 reference levels set by ICNIRP for protectingthe public against possible health effects from non-ionizing radiation.

It is also useful to compare the exposure levelsarising from Wi-Fi devices to those of mobile phones asanother commonly used wireless communication device.

When comparing exposure from different sources, it isimportant to take into consideration the output power ofdifferent devices, the frequency of operation and also thedistance the device is used from the human body.

On the basis of the measured IRP values, mobilephones have maximum output powers that are larger thanWi-Fi devices. According to standards, a GSM phone(operating at 900 MHz or 1,800 MHz) can producemaximum transmission powers of 1 or 2 W depending onthe frequency of operation. In practice, due to the use ofTime Division Multiple Access (TDMA) technology,GSM mobile phones have a maximum time-averagedoutput power of either 125 or 250 mW. Three-G (3G)phones also transmit continuously with maximum 125mW power.

Adaptive power control reduces the output power ofmobile phones to the minimum necessary to maintaincommunication during calls. For GSM mobile phones,the output power is reduced by 50% on average (Vrijheidet al. 2009), while the reduction is much greater and toaround 1% for 3G phones (Gati et al. 2009). Duringnormal use, the time averaged output power of Wi-Fidevices is also less than the maximum possible becausethe traffic carried is usually much less than the maximumpossible. As demonstrated in this study, although theWi-Fi devices can be configured in a controlled environ-ment to produce a duty factor of 95%, in practice the dutyfactor does not usually exceed 65%. This means that thetime-averaged output power of Wi-Fi devices will gen-erally be lower than that of mobile phones.

The exposure from Wi-Fi devices is also likely to belower than from mobile phones because their antennastend to be further away from the body during normal use.The highest exposures from mobile phones are incurredwhen the phone is used next to the head, the normalposition of use.

CONCLUSION

The electric field strengths from 15 laptops and 12access points, typical of those used in UK schools, wereassessed during Wi-Fi transmission. A signal analyzerwas used instead of a spectrum analyzer to allow thedetection of the WLAN signal in its entirety, hencemaking it possible to capture individual Wi-Fi bursts.

Overall, the radiation pattern around the laptops hadsimilar characteristics; i.e., with a minimum in thedirection away from the front of the laptop (toward thetorso of the user) and two maxima symmetrically op-posed across a vertical plane bisecting the screen and thekeyboard. Also, the electric field distribution aroundaccess points was more symmetric than those around thelaptops.



The maximum electric field strength values re-corded at 0.5 m around laptops and access points were2,893 mV m�1 and 5,716 mV m�1, respectively, indicat-ing that the field strengths from access points weregenerally higher (almost double) compared to those fromlaptops. The results also demonstrated that the electricfield strength reduced rapidly with distance for all Wi-Fidevices. The corresponding maximum power densityvalues for the laptops and access points at 0.5 m were 22mW m�2 and 87 mW m�2, respectively, decreasing to 4mW m�2 and 18 mW m�2 at 1 m distance.

While the position of the user was not explicitlyconsidered in this study and SAR values form the subjectof another part of the project, considering all thepractically-possible angular distributions of electric fieldstrength, this study concludes that all the devices undertest had maximum electric field strength and correspond-ing power density values well below ICNIRP limits of 61V m�1 and 10 W m�2, respectively.

The maximum spherically-integrated radiated powercalculated for the laptops was 17 mW. For practicalreasons and because access points are generally wall-mounted with beams directed into the rooms, theirpowers were integrated over a hemisphere with themaximum recorded value of 29 mW.

This study also found that it may take longer totransmit a given amount of data when the Wi-Fi signal isweak (i.e., higher duty factor), which in turn could resultin higher values of exposure, as has recently beenreported by Verloock et al. (2010).

Although the main finding of this study is that thepower densities around Wi-Fi devices are well within theICNIRP reference level at distances of 0.5 m and more,it is also important to consider the absorption of radio-frequency energy in the body of a person near thedevices. To address this, a parallel study has been carriedout which uses dosimetric modeling techniques to assessthe localized specific energy absorption rates (SARs)arising from Wi-Fi equipment in models of adults andchildren. The results of this further study are published ina separate article (Findlay and Dimbylow 2010).

Acknowledgments—The authors wish to acknowledge the work of StuartBurnett from Technical Services and Compliance group and Chris Gregoryfrom Facilities Services at Health Protection Agency (HPA) for thedevelopment and installation of the manual positioning system. Thanks arealso due to Liz Rance from the Physical Dosimetry Department for hersupport in managing the project. This study was financially supported byHPA’s research and development fund.

REFERENCES

BECTA. Wireless technology in schools. C3 Education for TheNational Education Research Panel (NERP); Bishop’s

Stortford, Hertfordshire, UK: National Education ResearchPanel (NERP); 2007.

CENELEC. Basic standard on measurement and calculationprocedures for human exposure to electric, magneticand electromagnetic fields (0 Hz–300 GHz). Brussels:CENELEC; EN50413; 2008.

Directive 1999/5/EC of the European Parliament and of theCouncil of 9 March 1999 on radio equipment and telecom-munications terminal equipment and the mutual recognitionof their conformity (R&TTE Directive). Brussels: DGEnterprise & Industry; 1999.

EN 300 328 V1.7.1. Electromagnetic compatibility and radiospectrum matters (ERM) wideband transmission systems,data transmission equipment operating in the 2.4 GHz ISMband and using wide band modulation techniques, Harmo-nized EN covering essential requirements under article 3.2of the R&TTE Directive. Sophia Antipolis Cedex, France:European Telecommunications Standards Institute; 2006.

EN 301 893 V1.4.1. Broadband Radio Access Networks(BRAN) 5 GHz high performance RLAN harmonized ENcovering essential requirements of article 3.2 of the R&TTEDirective. 2007.

EN 623 11. Assessment of electronic and electrical equipmentrelated to human exposure restrictions for electromagneticfields (0 Hz—300 GHz). IEC62311:2007, modified; 2007.

EN 62209-2. Human exposure to radio frequency fields fromhandheld and body-mounted wireless communication de-vices—human models, instrumentation, and procedures.Part 2: procedure to determine the specific absorption rate(SAR) for wireless communication devices used in closeproximity to the human body (frequency range of 30 MHzto 6 GHz); 2010.

Findlay R, Dimbylow PJ. SAR in a child voxel phantom fromexposure to wireless computer networks (Wi-Fi). Phys MedBiol 55:N405–N411; 2010.

Foster KR. Radiofrequency exposure from wireless LANSutilizing WI-FI technology. Health Phys 92:280–289; 2007.

Gabriel C. Compilation of the dielectric properties of body tissuesat RF and microwave frequencies. Report N.AL/OE-TR-1996-0037, Occupational and environmental health directorate.Brooks Air Force Base, TX: Radiofrequency Radiation Divi-sion; 1996.

Gati A, Hadjem A, Wiart J. Exposure induced by WCDMAmobile phones in operating networks. IEEE Trans WirelessComm 8:5723–5723; 2009.

ICNIRP. Guidelines for limiting exposure to time-varyingelectric, magnetic, and electromagnetic fields (up to 300GHz). Health Phys 74:494–522; 1998.

ICNIRP. Statement on the “guidelines for limiting exposure totime-varying electric, magnetic and electromagnetic fields(up to 300 GHz).” Health Phys 97:257–259; 2009.

IEEE. IEEE recommended practice for the measurement ofpotentially hazardous electromagnetic fields—RF and mi-crowave. Piscataway, NJ: IEEE Standard Association; IEEEC95.3; 1991.

IEEE. IEEE Standard for Information technology. Telecommu-nications and information exchange between systems. Localand metropolitan area networks. Specific requirements Part11: wireless LAN medium access control (MAC) andphysical layer (PHY) specifications. IEEE Std 802.11;1999.

ISO. Guide to the expression of uncertainty in measurement.Geneva: International Organization for Standardisation;1995.

Kuhn S, Lott U, Kramer A, Kuster N. Assessment methods fordemonstrating compliance with safety limits of wireless



devices used in home and office environments. IEEE TransElectromagnetic Compatibility 49:519–525; 2007.

National Institute of Standards and Technology. Guidelines forevaluating and expressing the uncertainty of NIST measure-ment results. Taylor BN, Kuyatt CE, eds. Washington, DC:United States Department of Commerce; Technical Note1297; 1994.

NPL. A guide to power flux density and field strengthmeasurement. Middlesex London: National Physical Labo-ratory, UK; The Institute of Measurement and Control;2004.

Schmid G, Preiner P, Lager D, Uberbacher R, George R.Exposure of general public due to wireless LAN applica-tions in public places. Radiat Protect Dosim 124:48–52;2007a.

Schmid G, Lager D, Preiner P, Uberbacher R, Cecil S.

Exposure caused by wireless technologies used for short-range indoor communication in homes and offices. RadiatProtect Dosim 124:58–62; 2007b.

Verlook L, Joseph W, Vermeeren G, Martens L. Procedure forassessment of general public exposure from WLAN inoffices and in wireless sensor network testbed. Health Phys98:628–638; 2010.

Vrijheid M, Mann S, Vecchia P, Wiart J, Taki M, Ardoino L,Armstrong BK, Auvinen A, Bedard D, Berg-Beckhoff G,Brown J, Chetrit A, Collatz-Christensen H, Combalot E, CookA, Deltour I, Feychting M, Giles GG, Hepworth SJ, Hours M,Iavarone I, Johansen C, Krewski D, Kurttio P, Lagorio S,Lonn S, McBride M, Montestrucq L, Parslow RC, SadetzkiS, Schuz J, Tynes T, Woodward A, Cardis E. Determinantsof mobile phone output power in a multinational study:implications for exposure assessment. Occup Environ Med66:664–671; 2009.

f f



assessment of exposure to electromagnetic fields …€¦ · assessment of exposure to...

Documents