Bioelectromagnetics 26:125^137 (2005)

A Numerical and Experimental Comparisonof Human Head Phantoms for ComplianceTesting of MobileTelephone Equipment

Andreas Christ,1* Nicolas Chavannes,2 Neviana Nikoloski,1 Hans-Ulrich Gerber,2

Katja Pokovic¤ ,2 and Niels Kuster1

1Foundation for Research on InformationTechnologies in Society (IT’IS),Zu« rich, Switzerland

2Schmid & Partner EngineeringAG, Zu« rich, Switzerland

A new human head phantom has been proposed by CENELEC/IEEE, based on a large scaleanthropometric survey. This phantom is compared to a homogeneousGenericHead Phantomand threehigh resolution anatomical head models with respect to specific absorption rate (SAR) assessment.The head phantoms are exposed to the radiation of a generic mobile phone (GMP) with differentantenna types and a commercial mobile phone. The phones are placed in the standardized testingpositions and operate at 900 and 1800 MHz. The average peak SAR is evaluated using both experi-mental (DASY3near field scanner) and numerical (FDTD simulations) techniques. The numerical andexperimental results compare well and confirm that the applied SAR assessment methods constitute aconservative approach. Bioelectromagnetics 26:125–137, 2005. � 2005 Wiley-Liss, Inc.

Key words: dosimetry; FDTD methods; measurement techniques

INTRODUCTION

In the past, numerous human head models havebeen proposed for use in the testing of mobile telecom-munications equipment (MTE) for compliance withsafety standards. Head phantoms both for the numericaland for the experimental assessment of the specificabsorption rate (SAR) have been suggested by differentresearch groups. Since the possible health risks ofmobile phones have become an issue of public concern,the mobile phone industry as well as several govern-mental regulatory bodies have been emphasizing theneed for a standardized procedure for the compliancetesting of MTE which meets the highest requirementswith respect to accuracy and repeatability. It is ob-vious that a human head model which is well definedin terms of shape and dielectric parameters representsthe cornerstone of any testing procedure, for example[IEEE, 2003; IEC, 2004]. Furthermore, the assessedexposure should be higher than the maximum humanexposure occurring under normal operational condi-tions. Since the maximum exposure is not known, basicrequirements for phantoms used in the compliancetesting procedures of handheld devices have been out-lined in Kuster et al. [1997] and were incorporated byIEEE [2003].

. The peak spatial average SAR shall be a conserva-tive estimate of the actual value expected to occur in

the heads of a significant majority of persons duringnormal usage of wireless handsets.

. The test results shall not unnecessarily overestimatethe peak SAR expected in actual users in order toprevent unnecessary inhibition of the advancementof new mobile communications technologies.

. The phantom shall enable high repeatability, allowstable and repeatable device positioning for peakSAR measurements and be effective for verify-ing repeatability and reproducibility among inter-laboratory comparisons.

. The phantom shall be practical for routine com-pliance testing.

. The phantom shall satisfy these criteria for con-temporary and future handset designs and be

�2005Wiley-Liss, Inc.

——————Grant sponsor: Schmid and Partner Engineering AG (SPEAG),Switzerland; Grant sponsor: Swiss Commission for Technologyand Innovation (KTI); Grant sponsor: Mobile ManufacturersForum (MMF), Belgium.

*Correspondence to: Andreas Christ, Foundation for Research onInformation Technologies in Society (IT’IS), Zeughausstr. 43,CH-8004 Zurich. E-mail: christ@itis.ethz.ch

Received for review 2 September 2003; Final revision received 14October 2004

DOI 10.1002/bem.20088Published online in Wiley InterScience (www.interscience.wiley.com).

unbiased with respect to any particular handset de-sign or shape, i.e., handset designs which producelower assessed SAR values should correspondinglyresult in reduced exposure in real-world situations,and vice-versa.

The objective of this study was to provide sup-porting scientific evidence that SAM meets the abovementioned requirements when compared to the expo-sure of actual human anatomies, including children.In order to substantiate the numerical results, the SARin the homogeneous liquid filled phantoms was alsodetermined experimentally.

PROPOSED STANDARD PHANTOMS

The Generic Twin Phantom [Kuster et al., 1997]and the Specific Anthropomorphic Mannequin (SAM)[IEEE, 2002] have been proposed and widely used forcompliance testing. Both phantoms are based on aseries of studies about the dependence of the absorptionupon internal anatomy, head size and shape (adultvs. child), tissue parameters, phone, accessories, etc.[Kuster and Balzano, 1992; Hombach et al., 1996;Meier, 1996;Meier et al., 1997;Kuster, 1998; Schonbornet al., 1998; Drossos et al., 2000; Kuster, 2001]. Thefindings of these studies led to the conclusion that aconservative approach is achieved if the phantomsatisfies the following requirements.

. The H-field generated by the handset at the interfaceof the shell and the head tissue-simulating liquid(HSL) is higher or equal to the H-field generated atthe skin of any user during intended operation of thephone. In other words, the separation between phoneand tissue-simulating liquid should be equal orsmaller than in the real world.

. HSL leads to higher absorption than any combina-tion of head tissues [Drossos et al., 2000].

Specific Anthropomorphic Mannequin

In order to maximize consumer trust in com-pliance testing procedures, the Standardization Co-ordinating Committee 34 of IEEE (SCC34-SC2)recognized the necessity to base the phantom on alarger survey of human heads [Gordon et al., 1989].Consequently, it developed and proposed the SAM.Theshape of SAM (rightmost in Fig. 1) corresponds to the90th percentile of data collected regarding the adultmale head. The material of the shell was selected aslossless and as thin as technically feasible, i.e., 2 mm.At the ear reference point, the thickness is increasedto 6 mm, the average thickness of the compressed ear.The ear itself is shaped such that it permits the accurateand repeatable positioning of the phone under test. Thedielectric parameters of the tissue-simulating liquidfor SAM were derived from worst case considera-tions [Drossos et al., 2000]. They are er¼ 41.5 and s¼0.97 S/m at 900MHz and er¼ 40.0 and s¼ 1.40 S/m at1800 MHz.

The SAM phantom has also been adopted byCENELEC [CENELEC, 2001a] as well as ARIB[ARIB, 2002] and IEC [IEC, 2004] for the compliancetesting of handheld phones. Lacking better criteria,CTIA has further proposed the use of SAM for radiatedpower andRF receiver performance tests [CTIA, 2001].

Generic Twin Phantom

Prior to the introduction of the SAM phantom bythe standardization bodies, the Generic Twin Phantom(second from the right in Fig. 1) was formerly used forMTE compliance testing by different research groups,manufacturers, and test houses. Although the model isobsolete, it has been included since a large number ofphones was authorized based on measurements withthis phantom [Federal Communications Commission,2003].

For the development of theGenericTwinPhantom,the head dimensions of 52 adult volunteers (male and

Fig. 1. Anatomicalandhomogeneousphantomsat same scale. (From left to right: 3-year-old child(3YC),HR-EF-1, adultmale, genericphantom,SAM).

126 Christ et al.

female) were measured in an area of 160� 150 mm2

covering the region around the cheek, the ear, and thetemporal bone. For the measurements, a planar devicewas used which slightly compressed the ear similar toduring usage of a mobile phone. The shape of thephantom was designed such that the distance betweenthe handset surface and the inner surface of the shellwould always be smaller than for 90% of the peopleinvestigated. The physical phantom shell is construct-ed of Ureol (er¼ 3.7) with a thickness of 2.7 mm. It isfilled with brain tissue-simulating liquid (er¼ 41.0,s¼ 0.85 S/m at 900MHz and er¼ 40.1, s¼ 1.71 S/m at1800 MHz). The ear of the Generic Twin Phantom isrepresented by a lossless spacer with a radius of 15 mmand a thickness of 2mmpositioned at the location of theopening of the auditory canal. This leads to an overallshell thickness of 4.7 mm in the ear region. A detaileddescription of the Generic Twin Phantom can be foundin Kuster et al. [1997].

HIGH RESOLUTION ANATOMICALHEAD MODELS

Numerical Representation

The spatial peak SAR obtained with the twohomogeneous phantoms was compared with the spatial

peak SAR obtained with three high-resolution anato-mical human models. For all these models, at least skintissue, muscle, fat, bone, CSF, gray and white brainmatter, blood, cartilage, vitreous humor, lens and eyesclera were distinguished. The tissue parameters asprovided by Gabriel et al. [1996] were assigned to allanatomical head models independent from their age(Table 1). For the comparison of the different head sizesbased on gender and age, Figure 1 shows the modelstogether with the two standard phantoms at the samescale.

All anatomical models have been derived fromimages (MRI or photographs) of slices through thebody. Since the models need to be rotated into thetesting positions and high grid resolution is required inthe ear region, a pre-discretized representation usingcubical cells (voxels) would lead to decreased accuracydue to multiple discretization. Therefore, a particulardata format has been developed which preserves the de-tails of the source images and keeps the computationalexpenses for storage and manipulation (e.g., rotation orscaling) low. First, all boundaries of the tissue regionsof all images are identified by a biologist and markedwith the help of an image processing software (Fig. 2).The coordinates of these boundaries are stored in thenumerical model. A rotation/translation matrix canbe applied on the boundary coordinates such that the

TABLE 1. Parameters of the Tissue Types of the Anatomical Head Models

Tissue type

900 MHz 1800 MHz

r (kg/m3)er s (S/m) er s (S/m)

Blood 61.36 1.54 59.37 2.04 1060Bone (cancellous) 20.79 0.34 19.34 0.59 1810Brain (gray matter) 52.73 0.94 50.08 1.39 1040Brain (white matter) 38.89 0.59 37.01 0.91 1040Cerebellum 49.44 1.26 46.11 1.71 1040Cerebro-spinal fluid 68.64 2.41 67.20 2.92 1010Cornea 55.24 1.39 52.77 1.86 1170Ear (average skin and cartilage) 42.03 0.82 39.54 1.24 1055Fat 5.46 0.05 5.35 0.08 920Gland 59.68 1.04 58.14 1.50 1040Lens 46.57 0.793 45.35 1.15 1100Lower jaw 20.79 0.34 19.34 0.59 1810Middle brain (gray matter) 52.73 0.94 50.08 1.39 1040Muscle 55.03 0.94 53.55 1.34 1040Skin 41.41 0.87 38.87 1.18 1010Skull 20.79 0.34 19.34 0.59 1810Spinal cord (gray matter) 52.73 0.94 50.08 1.39 1040Spine 20.79 0.34 19.34 0.59 1810Thalamus (gray matter) 52.73 0.94 50.08 1.39 1040Tongue 55.27 0.94 53.57 1.37 1040Upper jaw 20.79 0.34 19.34 0.59 1810Ventriculus lateralis 68.64 2.41 67.2 2.92 1010Vitreous humor 68.90 1.64 68.57 2.03 1010

Comparison of Human Head Phantoms 127

model can be freely positioned in the simulation grid.Thus, the tissue parameters can be assigned accordingto the actual grid coordinates and resolution.

European Female

The High Resolution European Female headphantom (HR-EF-1, second from the left in Fig. 1)was generated from magnetic resonance images (MRI)taken of a 40-year-old female volunteer. During theMRI scan, the ear was slightly compressed againstthe surface of the head in order to simulate the correctshape of the pinna while using a mobile phone. Thecompressed pinna has an approximate thickness of4 mm. The MRI data consist of 121 different slices. Inthe ear region, the slices have a thickness of 1 mm; allother slices have a thickness of 3 mm. The high resolu-tion of the ear region allows the accurate representationof the pinna as well as the anatomical details of theinner ear. Fifteen different tissues are distinguished.The parameters are derived according to Gabriel et al.[1996].

The reference point for the positioning of thespeaker of the phonewas chosen to be 15mmabove andbehind the auditory canal opening (see ExposurePositions).

Adult Male

The adult male head phantom (center of Fig. 1) isbased on the dataset provided by the Visible HumanProject [Visible Human Project, 1996]. The slices weretaken normal to the body axis at distances of 2 mm(�0.2 mm) considering more than 100 different tissueor organ types. The ear reference point is set 15 mmabove and behind the auditory canal opening (see Ex-

posure Positions). The outer range of the pinna is cut offsuch that the overall thickness of the ear does not exceed6 mm.

The body of the phantom is rather full figured.Comparatively thick layers of fat and connective tissuealso occur in the region of the neck and the nape behindthe ear (approximately 3.5 mm). For all these materials,fat tissue parameters were assigned.

Three-Year-Old Child

The model of the 3-year-old child (3YC) is alsobased on MRI data (leftmost in Fig. 1). The slices weretaken parallel to the sagittal plane at distances of 1.1mm(�0.2 mm). The distance of the ear reference pointto the speaker position is reduced to 11 mm in orderto take into account the smaller size of the child’s ear(see Exposure Positions). The thickness of the pinnais approximately 4 mm. Although 32 tissue types areconsidered in the MRI scans, the same 15 tissue para-meters as for the HR-EF-1 model are applied.

EXPOSURE CONDITIONS

Mobile Phones

For this study, the exposure of the head modelsdescribed above to the radiation from four mobilephones has been compared (Fig. 3). Three of these aregeneric mobile phone (GMP) models, which consistof an electrically and geometrically well defined metal-lic box, which can be equipped with three differentantennas.

Themetallic boxof thegeneric phonemodelshas asize of 140� 40� 16 mm with a Plexiglas coat-ing (er¼ 3.7) on the side which faces the cheek of

Fig. 2. Generation of an anatomical phantom: original image (left), image with marked tissueboundaries (center), numericalmodel (right).

128 Christ et al.

the phantom head (Fig. 3). The ear reference point(loudspeaker position) is assumed to be centered 10mmbelow the top of the phone. Unwanted surface currentson the feeding cablewere reduced by integrating a tunedstub into the case of the phone at the input connector andby placing ferrites around the feeding cable. The fourthphone is a commercially available tri-band model. SeeChavannes et al. [2003] for details of the numericalmodeling and simulation. The details of the phones arelisted below:

. GMP-M900: 900 MHz l/4 monopole antenna,80 mm length, 1.2 mm radius.

. GMP-M1800: 1800 MHz l/4 monopole antenna,36.85 mm length, 1.2 mm radius.

. GMP-H900: 900 MHz helical antenna, 6.5 mmdiameter, seven turns with 78 pitch angle, 0.6 mmwire radius.

. T-250: Commercially available tri-band phone(GSM, DCS, PCS) by Motorola, Inc. The CADdata set was provided by the manufacturer, and asample was purchased for measurements.

Exposure Positions

In this study, only the two standard positionswere evaluated for left hand usage. Due to the antennamounting positions, it can be expected that the maxi-mum exposures occur on the left side of the phantom.

The spatial peak SAR values averaged over 10 g1 wereassessed and compared for the following test positions:TheGeneric TwinPhantomwas exposed in the ‘‘touch’’and ‘‘1008’’ positions as defined in [CENELEC, 1998].SAM was evaluated in the ‘‘touch’’ and ‘‘158 tilted’’positions as described in [IEEE, 2003].2

The speaker of the phone is placed at the earreference point (ERP). The location of the ERP is de-fined to be 15 mm behind the entrance to the ear canal(EEC) as shown in Figure 5.3 of IEEE [2003]. Thispoint was chosen by IEEE [2003] because it representsthe location of optimum acoustical coupling and there-fore themost likely position for the speaker of the phoneduring normal usage. For the anatomical phantoms, thephonewas placed in positions, which correspond to thisdefinition but consider the smaller head and ear sizes ofthe 3YC.

In all cases, the phone was rotated such that theline connecting the speaker and microphone coincidedwith the plane through the ear reference points and the

Fig. 3. Mobilephones (from left to right):GMP-M900,GMP-M1800,GMP-H900,T250.

________1Most international standards [CENELEC, 2001b; ARIB, 2002;IEEE, 2002; IEC, 2004] define an averaging mass of 10 g. Thisvalue is used throughout the paper in lieu of the averaging mass of1 g as defined by [IEEE, 1999, 2003]. The averaging mass of 10 gis considered to be representative because only qualitative results(with respect to the SAR in SAM) are discussed.2In the remainder of this paper, both the ‘‘1008’’ and the ‘‘158tilted’’ positions will be referred to as the ‘‘tilted’’ position.

Comparison of Human Head Phantoms 129

center of the closed mouth (reference plane). Figure 4and Table 2 define the tilting angles of themobile phone,as well as the distances between antenna feedpoint andtissue/liquid.

METHODS

FDTD Simulations

All simulations were performed with an in-housefinite difference time domain (FDTD) code [Tafloveand Hagness, 2000] that was jointly developed byacademic and industrial groups. The software inte-grates a CAD environment, which enables the freepositioning, moving, and tilting of the head phantommodels. Nonuniform meshes can be generated auto-matically, according to the requirements of thegeometry.

In order to minimize uncertainties due to stair-casing effects with the conducting structures, thephonemodel is always oriented along the grid axes andthe head phantoms are tilted instead. The phantomswere discretized in different nonuniform mesheswith grid steps between 0.3 and 10 mm in freespace. In a series of simulations, the required gridresolution was determined such that discretizationinfluences on the numerical results could be exclud-ed. Inside the phantoms, the grid step was reduced tobe less than l/10. In the ear regions of the phantomthe maximum grid step was additionally limited to0.3–0.5 mm (see also Validation below). This enablesthe correct representation of the phantom shell andthe anatomical details of the pinna and minimizes thenumerical errors at the location where the SARmaximum is expected. The overall grid dimensionswere 390� 330� 550 mm3. Depending on the phoneposition and head model, between 5 and 12 milliongrid cells were required for the simulations. The cal-culation domain was truncated with PML absorbingboundary conditions.

The helical antenna is simulated with a novelsubgrid algorithm based on cubical spline interpolation[Chavannes, 2002]. The algorithm allows a local meshstep refinement of 1:2. Thus, the helix is discretizedwith amesh step size of 0.3mm,which is a quarter of itswire diameter. The application of the subgrid algorithmtherefore allows very accurate prediction of the fieldsaround the antenna.

SAR Averaging

For the assessment of the averaged SAR, analgorithm compliant with the procedure defined inIEEE [2001] has been implemented [Frohlich et al.,2002]. According to IEEE [2001], each voxel is con-sidered to be the center or sampling point of a cubicalvolume which is expanded until the required averagingmass is reached within 10%. This sampling point isassigned the SARvalue averaged over the cube volume,and its status is marked. In detail, the averaging algo-rithm works as follows:

Fig. 4. Definition of the mobile phone tilting angles at the SAMphantom: (a) side, (b) top, (c) frontalview.

130 Christ et al.

. Voxels for which cubes enclosing the requiredmass exist are assigned the cube average, and theirstatuses are set to ‘‘valid.’’ These cubes are allowedto contain a maximum limit of 10% of air.

. Each voxel, which contributes to at least one aver-aging volume is assigned the status ‘‘used.’’ It isset to the highest averaged value to which itcontributed.

. Voxels which are neither ‘‘valid’’ nor ‘‘used’’ areassigned the status ‘‘unused’’ and are treatedseparately.

The remaining unused voxels are expanded with-out the restriction of the 10% limit of included air.Further, the averaging cube is constructed with thevoxel on the cube’s surface. A more complete des-cription of the algorithm can be found in Frohlich et al.[2002].

RESULTS

The results for themaximum10 g averaged spatialpeak SAR values obtained from the tested positionsare summarized in Figure 5 for the 900 and 1800 MHz

bands. All values were normalized to those of SAM(100%). In this first comparison, the absorption in thepinna of all anatomical human phantoms was not con-sidered in the averaging process.

For the different phones and frequencies, differentlocations of the spatial peak SAR could be observed(Figs. 6 and 7). The location of the cube containing thespatial peak SAR depends on the phone design but canalso be different between phantoms for the same phone(Figs. 6 and7). The locations andmagnitudes dependonthe local anatomy. Since the safety guidelines recom-mend upper limits for the spatial peak SAR independentof the location and tissue, it is not relevant whetherthe phantom used for compliance testing accuratelyrepresents the location but that the magnitude is alwaysconservative.

For the GMP at 900MHz in the ‘‘touch’’ position,the averaged SAR occurs in the cheek for both themonopole and the helical antennas because of the highsurface currents on the phone case, which are veryclose to the lossy tissue. In the ‘‘tilted’’ position andat 1800 MHz (both positions), the maximum SAR isalways at the antenna feedpoint. The only exceptionsfor this are the adult male at 900 MHz and the HR-EF1

TABLE 2. Tilting Angles of the Generic Mobile Phone (GMP) for ‘‘Touch’’ and ‘‘Tilted’’Exposure Positions (see Fig. 4) and Distances Between Liquid and Feedpoint

HR-EF-1 3YC Adult male Generic SAM

f 678 738 64.58 658 618y 108 7.78 �2.88 08 3.98C (touch) 738 78.58 87.48 738 868dliq.-feed (touch) 20.5 mm 17.8 mm 16.0 mm 19.5 mm 15.5 mmC (tilted) 988 93.58 102.48 1008 1018dliq.-feed (tilted) 13.0 mm 15.0 mm 16.2 mm 14.0 mm 14.5 mm

Fig. 5. Comparisonofthe10gmaximumspatialpeakspecificabsorptionrate (SAR)obtainedby thedifferent phantomsinthe two testedpositionsexcludingthepinna.Allresultswerenormalized to thespatialpeakSARassessed for SAM.

Comparison of Human Head Phantoms 131

at 1800MHz in the ‘‘touch’’ position. These differencesare due to the individual head anatomies as shown inFigures 6 and 7.

Figure 8 shows the 10 g averaged SAR distribu-tion for the T-250 phone in the ‘‘touch’’ position for900 MHz and 1800 MHz. The maxima are alwayslocated in the phantom cheek, with the exception of the3YC phantom at 1800 MHz. In the ‘‘tilted’’ position,

the maxima are in the cheek for 900 MHz and in the‘‘touch’’ position for 1800 MHz.

The following conclusions can be drawn:

. For the standard positions evaluated, SAM alwaysprovided the highest spatial peak SAR. It must benoted that the ear of the adult female was stronglycompressed, i.e., to only 4 mm thickness; the pinna

Fig. 6. Ten-gram average SAR distributions for SAM and the three anatomical phantoms in the‘‘touch’’position for the genericmobile phone (GMP) withmonopole antenna.0 dB corresponds to15 W/kgnormalized to1Wantennaoutput power.The square indicates the location of the averagedSARmaximum.

Fig. 7. Ten-gram average SAR distributions for SAM and the three anatomical phantoms in the‘‘tilted’’position for theGMPwithmonopoleantenna.0 dB correspondsto15W/kgnormalized to1Wantennaoutput power.Thesquare indicatesthelocationof theaveragedSARmaximum.

132 Christ et al.

of the ear of the adult male was cut to a thicknessof 6 mm, and the child’s face proportions and earthickness corresponded to an age of 3 years.

. At 900 MHz, the T-250 mobile phone shows theH-field maximum in the region of the display/keyboard (Fig. 8 and Chavannes et al. [2003]). Forthese cases, the overestimation in the ‘‘touch’’ posi-tion by the SAM phantom is relatively small. Figure 5shows the comparison of the worst case overestima-tion for the different phones; the highest SAR valuesgenerally occur in the ‘‘tilted’’ position. If the peakSAR is induced at the location of the display orkeyboard, the relevance of the distance between theantenna feedpoint and the tissue is secondary. Thesecases require more detailed study and discussion.

. In addition, the results confirm that the SAR valuesare not directly linked to the head size. The largesthead (adult male, Fig. 1) shows values which do notnecessarily fall below those of the HR-EF1. TheSAR in the 3YC head is generally in the same orderof magnitude as the SAR in the HR-EF1. Majordifferences only occur (e.g., for GMP-M 1800 MHzin Fig. 5) if the location of the maximum averageis further away from the head surface and thereforefrom the antenna feedpoint (HR-EF1 at 1800 MHz

in Fig. 7). These changes are mainly due to the dif-ferent tissue distributions and anatomies of the threehead phantoms (e.g., higher content of connectivetissue with lower conductivity in the adult male).

Figure 9 shows the spatial peak SAR averagedover 10 g evaluated for the pinna only.3 These dataindicate that if the values for the head tissue are met, thevalues for the pinna treated as an extremity are intri-nsically met as well, because the values do not exceedthose assessed by SAM for the head tissue bymore thana factor of two. If the pinna is not treated as an extremity,SAM provides reliable data with underestimation ofless than 1 dB for the investigated cases (Fig. 10).

VALIDATION

FDTD is a well establishedmethod, and the appli-ed code has been extensively validated. Uncertaintiesdue to staircasing [Burkhardt, 1999] and dispersion

Fig. 8. Ten-gram average SAR distributions for SAM and the three anatomical phantoms in the‘‘touch’’position for theT-250mobilephone.0 dB correspondsto15W/kgnormalized to1Wantennaoutput power.Thesquare indicatesthelocationof theaveragedSARmaximum.

________3The masses of the pinnae of the anatomical phantoms are: 10.7 g(HR-EF1), 3.7 g (3YC), 12.2 g (adult male). For the pinna of the3YC phantom the average SAR over the whole pinna is used,because its mass is less than the required 10 g.

Comparison of Human Head Phantoms 133

[Christ, 2003] have been studied in detail. The first hasno relevance for near field assessments, whereas thelatter can be well controlled by reducing the meshstep. In our study, the minimum mesh step size inthe region of the ear and cheek was refined until theSAR deviations from a high resolution grid (mesh stepsize of 0.15 mm) were less than 1%. The results showthat a step size of 0.5 mm and below is sufficient(approximately 0.7% deviations); this step size wastherefore used in all simulations. For grid steps of2 mm, the SAR deviation is about 5% compared tothe finest grid resolution applied (0.15 mm). For de-creasing mesh step size, the deviations tend towardzero. It must be noted that these uncertainties are notgenerally valid and may be several times larger de-pending on the source.

The uncertainty, which cannot be intrinsicallyassessed is that of the modeling of the transmitter. This

is especially crucial, since the exposure depends on thefeedpoint impedance and the current distribution onand in the device [Kuster andBalzano, 1992]. The lowerthe physical definition of the phone, the larger the un-certainty becomes. However, there is no procedure forassessing this uncertainty numerically, so that the totaluncertainty of simulations cannot be determined.More-over, experimental evaluations have the advantage thatthe actual transmitter is tested (i.e., uncertainty is zero)and that the other uncertainties can be assessed. There-fore, validation of the numerical results through mea-surements is always recommended.

In this study, the simulations were validated bymeasurements using SAM (Fig. 11) as well as theGeneric Twin Phantom (Fig. 12). The measurementdata were obtained with the DASY3 near field scanner,which is the successor of the dosimetric asses-sment system described in Schmid et al. [1996]. The

Fig. 9. Comparisonofthe10gmaximumspatialpeakSARobtainedby thedifferent phantomsinthetwo tested positions in the pinna.All resultswere normalized to the spatial peak SARassessed forSAM.

Fig. 10. Comparisonofthe10gmaximumspatialpeakSARobtainedby thedifferentphantomsinthetwo tested positions including the pinna. All results were normalized to the spatial peak SARassessed for SAM.

134 Christ et al.

measurements were conducted according to the latestguidelines for compliance testing [IEEE, 2003]. Themeasurement uncertainty for these evaluations (95%confidence interval) was determined to be 26.8% forDASY3, according to the procedure described inMethods of the same standard.

The deviation between simulations and measure-ments was generally less than 12%. In cases involvinghigh deviations, large differences in the feedpointimpedance were observed, which accounts for someof the variation [Kuster and Balzano, 1992]. Table 3provides the simulated and measured impedances forSAM and the Generic Twin Phantom. In summary, the

agreement between measured and simulated results iswithin the uncertainties of the applied methods forspatial peak SAR assessments, thereby validating thenumerical simulations.

CONCLUSIONS

Within the bounds of this study, the exposurevalues assessed with SAM following the protocol asdefined in IEEE [2003] and CENELEC [2001b] werecompared to those of several anatomical headmodels aswell as of the Generic Twin Phantom. Four differentphones were evaluated in the ‘‘touch’’ and ‘‘tilted’’

Fig. 11. Measuredandsimulated spatialpeak SARin the SAMphantom.Thevaluesarenormalizedto an antenna output powerof1W.The error bars indicate the 95% confidence interval of themea-surements.Theuncertaintyofthecomparisonisdifficult toassessandisestimatedtobebetter than20%.

Fig. 12. MeasuredandsimulatedspatialpeakSARintheGenericTwinPhantom.Thevaluesarenor-malized toanantennaoutputpowerof1W.Theerrorbarsindicatethe95%confidenceintervalofthemeasurements.Theuncertaintyof the comparisonisdifficult toassessandisestimated tobebetterthan 20%.

Comparison of Human Head Phantoms 135

positions on the left side of the phantom. The resultssupport the following conclusions:

. Compared to the original method for MTE com-pliance testing based on the Generic Twin Phantom,the SAM phantom and the new liquid parametersbased on worst case considerations represent a moreconservative approach.

. The study confirms the findings of Schonborn et al.[1998] that smaller heads do not result in increasedSAR values and that all differences can be explainedby differences in distance and anatomical variationsin tissue structures. These findings contradictGandhi et al. [1996].

. SAM provides a conservative estimate for the head-tissue exposure caused by the investigated phones inthe two standard positions.

. The study supports the assumption that the spatialpeak SAR is intrinsically met for the pinna whentreated as extremity.

. SAM also provides reasonable exposure data if thepinna is included in the averaging process.

Additional studies will become necessary fordevices operating at higher frequencies, should newtissue parameters become established or in considera-tion of different phone designs.

REFERENCES

ARIB. 2002. Specific absorption rate (SAR) estimation for cellularphone (standard version 2). Association of Radio Industriesand Businesses (ARIB), Tokyo, Japan.

Burkhardt M. 1999. Contributions toward uncertainty assessmentsand error minimization of FDTD simulations involvingcomplex dielectric bodies. Ph.D. thesis, Diss. ETHNr.13176,Zurich.

CENELEC. 1998. Considerations for evaluation of human exposureto electromagnetic fields (EMFs) from mobile telecommu-nication equipment (MTE) in the frequency range 30MHz–6GHz, European Committee for Electrotechnical Standardi-zation (CENELEC), Central Secretariat, rue de Stassart 35,1050 Brussels, Belgium.

CENELEC. 2001a. Product standard to demonstrate the complianceof mobile telephones with the basic restrictions related to

human exposure to electromagnetic fields (300 MHz–3 GHz), European Committee for Electrotechnical Standar-dization (CENELEC), Central Secretariat, rue de Stassart35, 1050 Brussels, Belgium.

CENELEC. 2001b. Basic standard for the measurement of specificabsorption rate related to human exposure to electromagneticfields frommobile phones, European Committee for Electro-technical Standardization (CENELEC), Central Secretariat,rue de Stassart 35, 1050 Brussels, Belgium.

Chavannes N. 2002. Local mesh refinement algorithms for enhanc-ed modeling capabilities in the FDTD method. Ph.D. thesis,Diss. ETH Nr. 14577, Zurich.

Chavannes N, Tay R, Nikoloski N, Kuster N. 2003. RF designof mobile phones by TCAD: Suitability and limitations ofFDTD. IEEE Antennas Propagation Mag 45:52–66.

Christ A. 2003. Analysis and improvement of the numerical pro-perties of the FDTD algorithm, Ph.D. thesis, Diss. ETHNr.15057, Zurich.

CTIA. 2001. Method of measurement for radiation RF powerand receiver performance: Draft Revision 1.0, CellularTelecommunication and Internet Association CertificationPrograms, 1250 Connecticut Avenue NW, Suite 800,Washington, DC 20036.

Drossos A, Santomaa V, Kuster N. 2000. The dependence ofelectromagnetic energy absorption upon human head tissuecomposition in the frequency range of 300–3000MHz. IEEETrans Microw Theory Tech 48:1988–1995.

Federal CommunicationsCommission. 2003. Equipment authoriza-tion system generic search, 445 12th Street. SWWashington,DC, 20554. Available from: http://gullfoss2.fcc.gov/prod/oet/cf/eas/reports/GenericSearch.cfm

Frohlich J, Cherubini E, Kuster N. 2002. Implementation of the newIEEE standard for computational dosimetry. In: 24th AnnualMeeting of the Bioelectromagnetics Society, June, 2002,Quebec, Canada, pp 167–168.

Gabriel C, Gabriel S, Corthout E. 1996. The dielectric properties ofbiological tissues: I. literature survey. Phys Med Biol 41:2231–2249.

Gandhi OP, Lazzi G, Furse C. 1996. Electromagnetic absorption inthe human head and neck for mobile telephones at 835and 1900 MHz. IEEE Trans Microw Theory Tech 44:1884–1897.

GordonCC,Churchill T, ClauserCE,Bradtmiller B,McConville JT.1989. Anthropometric survey of U.S. army personnel:Methods and summary statistics 1988. U.S. Natick, MA:Army Natick Research Center, Technical Report NATICK/TR-89/044.

Hombach V, Meier K, Burkhardt M, Kuhn E, Kuster N. 1996. Thedependence of EM energy absorption upon human headmodeling at 900 MHz. IEEE Trans Microw Theory Tech 44:1865–1873.

TABLE 3. Measured and Simulated Feedpoint Impedance [X] of the Experimental Phantoms

SAM Generic twin

Measured Simulated Measured Simulated

900 MHz monopole touch 68.5� j13.4 60.0� j15.3 71.8� j9.6 72.2� j26.8900 MHz monopole tilted 65.3� j18.1 46.9� j20.0 60.2� j18.8 54.8� j31.01800 MHz monopole touch 42.5� j16.6 40.8� j11.3 49.5� j21.5 46.3� j25.71800 MHz monopole tilted 46.1� j13.0 42.9� j9.0 44.2� j16.3 42.8� j18.1900 MHz helix touch 37.6þ j22.2 37.1þ j39.1 34.6þ j12.5 38.2þ j36.6900 MHz helix tilted 39.9þ j38.6 32.5þ j58.7 36.8þ j43.9 38.7þ j77.4

136 Christ et al.

IEC. 2004. Procedures to measure the specific absorption rate(SAR) for hand-held mobile wireless devices in thefrequency range of 300 MHz–3 GHz). IEC Tech Committee106:IEC62209 Pt.1.

IEEE. 1999. IEEE Std. C95.1-1999, IEEE standard for safetylevels with respect to human exposure to radio frequencyelectromagnetic fields, 3 kHz to 300 GHz. New York,NY: Institute of Electrical and Electronics Engineers (IEEE),Inc.

IEEE. 2001. IEEE P1529/D0.0, recommended practice for de-termining the spatial-peak specific absorption rate (SAR)associated with the use of wireless handsets—computationaltechniques. New York, NY: Institute of Electrical andElectronics Engineers (IEEE), Inc.

IEEE. 2002. IEEE Std. C95.1-2002, IEEE standard for safety levelswith respect to human exposure to radio frequency electro-magnetic fields, 3 kHz to 300 GHz. New York, NY: Instituteof Electrical and Electronics Engineers (IEEE), Inc.

IEEE. 2003. IEEE Std. 1528-2003, recommended practice fordetermining the peak spatial-average specific absorptionrate (SAR) in the human head from wireless communica-tions devices—measurement techniques, 19December 2003.New York: Institute of Electrical and Electronics Engineers.

Kuster N. 1998. In: Carlo GL, editor. Compliance testing of hand-heldmobile communications equipment. In:Wireless phonesand health. Boston/Dordrecht/London: Kluwer AcademicPublishers.

Kuster N. 2001. In: Claude Legris, editor. Latest progress in ex-perimental dosimetry for human exposure evaluations and

for characterization and optimization of exposure setupsused in biological experiments. In: Communication mobile—effects biologiques. Paris, France: CADAS, Academie desSciences. pp 43–69.

Kuster N, Balzano Q. 1992. Energy absorption mechanism bybiological bodies in the near field of dipole antennas above300 MHz. IEEE Trans Vehicular Technol 41:17–23.

Kuster N, Kastle R, Schmid T. 1997. Dosimetric evaluation ofmobile communications equipment with known precision.IEICE Trans Commun E80-B:645–652.

Meier K. 1996. Scientific bases for dosimetric assessments incompliance tests. Ph.D. thesis, Diss. ETH Nr.11722, Zurich.

Meier K, Kastle R, Hombach V, Tay R, Kuster N. 1997. The de-pendence of EM energy absorption upon human headmodeling at 1800 MHz. IEEE Trans Microw Theory Techn45:2058–2062.

Schmid T, Egger O, Kuster N. 1996. Automated E-field scanningsystem for dosimetric assessments. IEEE Trans MicrowTheory Techn 44:105–113.

Schonborn F, Burkhardt M, Kuster N. 1998. The difference of EMenergy absorption between adults and children. Health Phys74:160–168.

Taflove A, Hagness SC. 2000. Computational electromagnetics:The finite-difference time-domain method. 2nd edition.Boston and London: Artech House, Inc.

Visible Human Project. 1996. The Visible Human Project,U.S. National Library of Medicine, 8600 Rockville Pike,Bethesda, MD 20894. Available from: http://www.nlm.nih.gov/research/visible/visible_human.html

Comparison of Human Head Phantoms 137