MWJournal Jan 2008

8/3/2019 MWJournal Jan 2008

1/6

T

odays handsets have to meet toughtechnical requirements. Mobile phoneshave to deal with an ever-increasing

number of services, while at the same time thecost of the systems is being reduced. R&D inthe mobile phone industry copes with this sit-uation by continuously improving the mobilephone efficiency in order to accommodateservice enhancements in the mobile network.Thus, we are moving towards mobile designsthat are not only becoming thinner, smallerand more complex with every generation, butalso have to perform with the same or evenbetter performance, and in more frequencybands.

In addition to maximizing the antenna-ac-cepted power of the handsets, the effects onthe antenna performance from surroundingobjects such as the human body have to be

studied and considered. Homogeneous mod-els are used when measuring the effects onantenna performance and represent a conservative estimation of antenna losses and dissi-pated power. The performance of the antennaand the entire system may be quantified usingsets of technical requirements for both passiveand active modes.

In passive mode, the antenna performanceis often measured by the antenna efficiency,which is subdivided into the radiation efficien-cy and the return-loss efficiency. In active

ELECTROMAGNETIC

SIMULATION OF

MOBILE PHONE

ANTENNA PERFORMANCEThe telecommunications sector is making great advances aimed at delivering an

even stream of high-tech devices, covering the significant consumer demands in

this sector. Electromagnetic (EM) simulation is becoming an increasinglyimportant tool in the design flow, not only at the antenna level but also at the

phone and environmental levels. This article compares simulated results with

measurements for several steps in the phone design chain.

OMID SOTOUDEHSony Ericsson Mobile CommunicationsKista, Sweden

TILMANNWITTIGCST GmbHDarmstadt, Germany

Reprinted with permission ofMICROWAVE JOURNAL from the January 2008 issue.2007 Horizon House Publications, Inc.


2/6

mode, the entire system efficiency isdefined by the total radiated power(TRP) on transmitting (Tx), and the

total isotropic sensitivity (TIS) on re-ceiving (Rx). The active performanceof the handset is often measured us-ing exact and time-consuming proce-dures. These have to be conductedseveral times during the developmentphase of the device. Furthermore,the product has to be developed to acertain stage before any measure-ment or reliable prediction of anten-na performance is possible.

Numerical simulation of terminalantennas has been actively discussed

in literature. There has been consider-able development in both hardware

and software recent-ly and it is importantto continuously up-date the field withthe latest develop-ments. This articleconsiders the accu-racy of calculation ofmobile phone mod-els by comparingthese calculationswith measurements.The models used aretaken from the re-

cently released Sony Ericsson M600mobile phone1 and measurementswere conducted at Sony Ericssons test

facilities. The emphasis is on the accu-racy of calculations using the presentstate-of-the-art.

SIMULATION TECHNOLOGY

For all simulations, the 3D EMtool CST MICROWAVE STUDIO

(CST MWS),2 based on the finite in-tegration technique (FIT),3was used.This method represents a consistenttransformation of the analyticalMaxwell equations into a set of matrixequations, the Maxwell Grid Equa-

tions (MGE). Thereby, essentialphysical properties of the analyticalequations, such as energy conserva-tion and passivity, are maintained indiscrete space.4 The MGEs can besolved by various techniques, fromstatics to the optical regime, in thetime and frequency domains.

In the microwave range, the fre-quency domain approach is very flexi-ble in the choice of discretization(tetrahedral or Cartesian meshes maybe used), but it requires the solution

of a large linear system of equationsfor each frequency step. The time do-main solver, in contrast, if used on aCartesian mesh, only requires matrix

vector multiplications and thereforescales linearly with the number ofmesh cells, both in terms of memoryrequirements and in simulation time.This offers an advantage for complexsimulation models such as mobilephones; however, traditional time do-

main methods like FDTD can exhibitpoor convergence properties.

Within the frame of FIT, advancedmeshing techniques such as the Per-fect Boundary Approximation (PBA)

and Thin Sheet Techniques (TST)can be implemented. Still based on aCartesian mesh, the geometry is de-scribed conformally, but all the ad-vantages of a time-domain approach,like memory efficiency and broad-band results, are maintained. To im-prove the efficiency of certain simula-tions even further, a stable multi-levelsub-gridding scheme can be imple-mented.

In order to calculate radiation effi-ciency correctly, it is important tomodel the skin depth of lossy metalaccurately. FIT can be extended toconsider the frequency dependencyfor permeability, permittivity andconductivity in one broadband simu-lation. Taking dispersion into accountis especially important when model-

ing tissue-simulating materials.

APPLICATION NOTE

TABLE I

CONVERGENCE STUDYFOR THE ANTENNA LEVEL CASE IN FIGURE 1

Run 1 Run 2 Run 3

No. of mesh-cells 221,610 383,160 660,539

Simulation time (s) 450 690 1130

Max. difference 0.067 0.014

Radiation efficiency 1 GHz 0.894 0.901 0.906

Radiation efficiency 2 GHz 0.908 0.924 0.923

v Fig. 1 Analysis phases of the mobile phone study.

v Fig. 2 The converged mesh for theantenna simulation.

MEASUREMENTSIMULATION

SIMULATIONMEASUREMENT

0

5

10

15

20

253210

3.02.52.01.5FREQUENCY (GHz)

FREQUENCY (GHz)(a)

(b)

1.00.5

0

2

4

6

8

10

RETURN

LOSS(dB)

RADIATION

EFFICIENCY(dB)

v Fig. 3 Simulation vs. measurement at theantenna level; (a) return loss and (b) antennaefficiency.


3/6

then 0.02 over the complete band (0to 3 GHz). Additionally, the conver-gence of the radiation efficiency atthe two bands was evaluated.

The total convergence took 38minutes. For further simulations andstructure optimizations, the third run

can be skipped, as the mesh setup inthe second run with around 383,000mesh cells delivers well convergedresults. This means that all furthersimulations in the optimizationprocess will have a run time of only12 minutes. Details including individ-ual simulation times can be takenfrom Table 1.

The converged mesh is shown inFigure 2. The bent planar parts ofthe PIFA antenna, the antenna carri-er and the PCB can be seen clearly.As the bendings are not aligned withthe Cartesian mesh, this would cause

APPLICATION NOTE

ANALYZED STRUCTURES

The EM simulations conducted onthe M600 mobile phone was done onseveral system complexity levels.These are shown in Figure 1, fromleft to right with the least complexstructures, such as the simple anten-na design and PCB, to the completephone structure containing severalhundred components and finally theentire system using the head of theStandard Anthropomorphic Model(SAM),5,6 a homogeneous hand mod-

el and the full phone.In real mobile phones, many ob-jects in the vicinity of the antenna el-ement have to be considered becauseof their effect on the performance ofthe system. Therefore, great caremust be taken when selecting andsetting the correct simulation para-meters for the relevant objects.

The analysis of antenna perfor-mance in the vicinity of bodies oftencomprises assumptions and simplifi-cations. In general, homogeneous

bodies are utilized in order to mea-sure the most conservative behaviorof the system. Because of this, homo-geneous head and hand models areused for simulations as well.

Throughout this work, the antennaeffects were quantified by calculatinginput impedance and antenna effi-ciency.

SIMULATION AT ANTENNA LEVEL

At the antenna level, the designand optimization of the antenna itselfis the prime goal. Results of interestare typically the return loss, radiationefficiency and radiation pattern of theantenna at various frequencies. Eventhough the transient solver is used for

the simulations, frequency quantitieslike near and far fields can be evalu-ated easily at numerous frequenciesduring one transient run due to fieldmonitors based on discrete Fouriertransforms (DFT). Realistic loss val-ues were chosen for both the metallicand dielectric objects.

A convergence study indicates thatthe model converges to an accuratesolution. Starting with a relativelycoarse mesh of 221,000 mesh cells,and refining it using an energy-based

criterion, the final solution isachieved in only three steps. The cri-terion for stopping the study is thatthe maximum difference of an S-parameter between two runs is less

v Fig. 4 The radiation pattern of an antenna for two GSM bands.

v Fig. 5 Full phone model containing phone plastics (red and blue)and metallic parts (copper and gold).

3210

0

5

10

15

20

0

2

4

6

8

10

FREQUENCY (GHz)(a)

3.02.52.01.51.00.5

FREQUENCY (GHz)(b)

MEASUREMENTSIMULATION

SIMULATION MEASUREMENT

RETURN

LOSS(dB)

RADIATION

EFFICIENCY(dB)

v Fig. 6 Comparison of simulation andmeasurement for the full phone simulation;(a) return loss and (b) antenna efficiency.

11 2 1 2 1

PCBTOUCHSTONE 6.104 nH

1.941 pF3.055 nH6.371 nH

2.743 nH

8.8 pF

6.86 pF

v Fig. 7 Network simulation including matching network and the3D results of the antenna.


4/6

tion of the phone geometry for theinvestigated frequencies.

The structure was imported intoCST MWS using the STEP interface.No additional healing was necessarybefore conducting the simulations, which is an important pre-requisite

for efficient industrial design andworkflow. The simulation of the fullmobile phone consisted of 594,000mesh cells (again after a convergencestudy as described in the previoussection). The total simulation for theconverged model took 19 minutes.

The simulation of the return lossand radiation efficiency of the fullphone is compared with measure-ments in Figure 6. As the completephone is now considered, the fre-quencies are shifted down to thewell-known mobile phone bands. Theresults again agree well with respectto resonance frequencies, bandwidthsand radiation efficiency, but some dif-ferences occur in the return loss forthe upper frequencies.

These can be explained by the un-certainty regarding the antennas ex-act feeding point during measure-ment and measurement de-embed-ding. Additionally, the properties ofthe various materials used in the

phone model may be inexact.Alongside radiation efficiency andreturn loss, global quantities such asthe TRP and the TIS are relevant forthe simulation of the completephone. These values require the con-sideration not only of the antennacharacteristics but also the amplifier,the signal transmission inside theprinted circuit board and the match-ing network.Figure 7shows the sim-plified setup of such a circuit in CSTDESIGN STUDIO (CST DS) in-

cluding an idealized source, touch-stone file describing the PCB trans-mission, matching network and a mi-crostripline to feed the antenna. Thesimulation delivers system S-parame-ters, system near and far fields, andfrom these, the TRP value. The TRP value of this idealized setup gives23.27 dBm at 1.8 GHz and amplifierpower of 0.25 W.

The near field is also of signifi-cance for the complete phone, as in-teraction with other electromagnetic

devices such as hearing aids (hearingaid compatibility, HAC) might occur.Near-field information can be pre-dicted accurately by means of simula-

APPLICATION NOTE

significant problems in simple stair-case methods; however, due to thethin sheet technique, mesh cells can

be intersected by metallic sheets. To-gether with the PBA technique, theshown grid, although it might lookrelatively coarse, delivers fully con-verged results.

The converged simulation resultsof the antenna compared with mea-surements are shown in Figure 3.

The results are in agreement forboth the return loss and the radia-tion efficiency. Finally, the radiationpattern is shown in Figure 4 fortwo GSM frequencies. Since theplastic housing is not considered forthis study, the resonances are slight-ly shifted to ~1 and ~2 GHz.

SIMULATION AT PHONE LEVEL

After the antenna design is com-plete, the next step is to include it inthe complete phone. This enables the

evaluation of the coupling effects ofneighboring objects such as the bat-tery, camera, flash capacitors, etc., aswell as the influence of dielectric ma-terials such as the housing and dis-play screen.

The phone is subdivided intoroughly 60 components, each consist-ing of hundreds or even thousands ofindividual facets(see Figure 5theback cover and battery lid are hiddenfor the picture). The componentsused for the simulation are chosen

based on their influence on electro-magnetic fields (which is controlledby both dimensions and location), inorder to give an accurate simplifica-

E

(a) (b)

H E H

1747MHz

894MHz

v Fig. 8 Electric and magnetic near fields on the back of the phone; (a) simulation and (b)measurement.

FREQUENCY (MHz) r (S/m)835 41.5 0.90

900 41.5 0.97

1800-2000 40.0 1.40

FREQUENCY (GHz)

(a)

(b)

50

40

30

20

10

03210

EPS EPS EPS RE (LIST) EPS IM (LIST)

ELECTRICDISPERSION:

2NDORDERMODEL(FIT)

v Fig. 9 Tabulated values for the tissueproperties by IEEE 1528 (a) and the broad-band material fit used by CST MWS (b).


5/6

to measurable data, numerical simu-lation grants insight into previouslyunseen electromagnetic detail.

Within one simulation run, all-im-portant quantities such as return loss,radiation efficiency, near and far

fields, and loss monitors (all at variousfrequencies) can be obtained. Ad-vanced mesh technology significantlyreduces the simulation time, bringingit down to a few minutes for an anten-na simulation. Even complex automat-ic optimization runs become feasible.The increasing efficiency and reliabili-ty of simulation, which reduces designcost risk, is recognized as indispens-able in the industry. s

References

1. www.sonyericsson.com.2. www.cst.com.3. T. Weiland, A Discretization Method for

the Solution of Maxwells Equations for Six-Component Fields, Electronics and Com-munication, (AE), Vol. 31, 1977, p. 116.

4. T. Weiland, Time Domain Electromag-netic Field Computation with Finite Dif-ference Methods, Int. J. Num. Modeling,Vol. 9, 1996, pp. 295319.

5. IEEE Standard 1528-2003, www.ieee.org.6. IEC Standard 62209-1, www.iec.ch.

Omid Sotoudehwas awarded his MSc degreein satellite engineering from the SwedishInstitute of Space Physics in 2001 and his PhD

degree in electrical engineering from ChalmersUniversity of Technology, Gothenburg, in2005. He then joined Sony Ericsson MobileCommunications AB in Stockholm, where hehas worked on research into antennas andpropagation, especially in the field ofnumerical modeling.

Tilmann Wittig received his Dipl-Ing degreein electrical engineering in 1998 and his PhDdegree in the field of numerical simulation ofelectromagnetic fields in 2003, both from theUniversity of Technology, Darmstadt,Germany. He joined CSTs technical team inGermany in 2004, where he currently works asa senior application engineer for EM fieldsimulations.

APPLICATION NOTE

tion.Figure 8 compares the normal-ized E and H near fields at a distanceof 10 mm from the back of the phonein a free space configuration. The

blue outline represents the positionof the phone. (Measurements wereperformed by Ericsson Research,Stockholm, Sweden.)

SIMULATION AT BODY LEVEL

A final test for the mobile phone isto evaluate it in the presence of a hu-man body, with particular emphasison the head and hand. In accordancewith IEC and IEEE standards,5,6 theStandard Anthropomorphic Model isused as the head model. The fre-

quency dependent dielectric proper-ties of the tissue simulating liquid arealso defined by this standard and canbe modeled as a dispersive materialin the simulation tool. Figure 9shows the required values tabulatedand realized in the simulation tool.

The simulation of this phone, withhead and hand models, required 4.24million mesh cells and had a simula-tion run time of 1.58 hours on a PC(dual core dual CPU, 2 GHz, 8 GBRAM). The number of mesh cells canbe reduced using the sub-griddingscheme. When applied, a very finemesh is created inside the phone, acoarser one in the head and a very ba-sic one in the vacuum (see Figure10). Using sub-gridding reduces thenumber of mesh cells to only 922,000and the simulation time to 44 minutes.

If the simulation is started withoutsub-gridding on a hardware accelera-tion card, the simulation time can bereduced even further to only 27 min-

utes. Excluding the meshing time,

the solver speed-up using sub-grid-ding is 3.6. With the accelerator card,a speed-up factor of 5.8 can beachieved compared to a non-acceler-ated workstation.

Such a simulation can give impor-tant insight into how the SAM phan-

tom or the homogeneous body mod-els affect the performance of the mo-bile phone. The radiation pattern isobviously affected, but the radiationefficiency is also influenced by thehead and hand.Figure 11 shows theradiation pattern of the phone; a sig-nificant difference is visible in com-parison to the plain antenna far fieldsfrom Figure 4.

As previously mentioned, the radi-ation efficiency is influenced by thebody models. Table 2 shows the an-

tenna efficiency calculations for onlythe phone, the phone placed at theright cheek of SAM and the phoneheld at SAMs right cheek with ahand model present.

Finally, full-SAM simulations arevery useful to predict dissipated pow-er. This quantityas a measured val-ueis another important design is-sue and a requirement for certifica-tion. However, a simulation allowsthe designer to control this power at amuch earlier design stage.

CONCLUSION

This article has shown what is cur-rently possible in the world of ad-vanced 3D EM simulation. Through-out all steps of a mobile phones ter-minal antenna developmentfromantenna design, through full phoneoptimization up to investigating theinfluence of, and impact on a bodymodelsimulation and measurementhave been compared and shown to be

in very good agreement. In addition

v Fig. 10 Sub-grid setup for a mobilephone simulation in the presence of headand hand.

v Fig. 11 The modified radiation patternfor the two GSM bands in the presence ofhead and hand.

TABLE II

ANTENNA EFFICIENCY CALCULATIONS

Frequency 897.4 MHz 1747.6 MHz

Phone 100% 100%

Phone head 23% 48%(6.3 dB) (3.2 dB)

Phone head 6% 7%hand (12.2 dB) (11.5 dB)


6/6

SonyEricssonM600s

imulatedusingCSTMICROWAVE

STUDIO2008

Documents

MWJournal Jan 2008