UNIVERSITY
OF TRENTO
DEPARTMENT OF INFORMATION AND COMMUNICATION TECHNOLOGY
38050 Povo – Trento (Italy), Via Sommarive 14
http://www.dit.unitn.it
S
YNTHESIS OF A
T
HREE
-D
IMENSIONAL
T
RIBAND
(L1-L2
GPS
AND
W
I
-F
I
)
P
REFRACTAL
T
REE
A
NTENNA
R. Azaro, E. Zeni, T. Gazzini, R. Dallapiccola, A. Massa
November 2006
and Wi-Fi) Prefra tal Tree Antenna
Renzo Azaro, Member, IEEE, Edoardo Zeni, Tommaso Gazzini, Ramona Dallapi ola,
Andrea Massa, Member, IEEE
Department of Information and Communi ation Te hnologies,
University of Trento,Via Sommarive 14,38050 Trento - Italy
Tel. +39 0461 882057,Fax +39 0461 882093
E-mail: andrea.massaing.unitn.it, {renzo.azaro,edoardo.zeni}dit.unitn.it
and Wi-Fi) Prefra tal Tree Antenna
Renzo Azaro, Member, IEEE, Edoardo Zeni, Tommaso Gazzini, Ramona Dallapi ola,
Andrea Massa, Member, IEEE
Abstra t
In this letter, the synthesis of a 3D wire antenna with a tree-like shape and
op-erating at three dierent frequen y bands (
L1 − L2 GP S
bands andW i
− F i
) ispresented. The geometri al parameters of theantennaaredetermined bymeans of
a parti leswarm-based optimization pro edureexploitingthedegrees of freedom of
thereferen etreegeometry for ttingtheele tri al proje t onstraints. Theresults
of thenumeri al simulation areshownand ompared withthose fromexperimental
measurements for assessingthe ee tiveness ofthesynthesispro edure.
Key Words - Antennas Synthesis, Fra tal Antennas, Tree-like Antennas, Multiband
Be ause of the growing demand of miniaturized and multiband antennas for ele troni
devi es employing multiple wireless standards, several resear hes have been devoted at
analyzingthe ele trodynami behaviorof fra taland pre-fra talgeometries[1℄[2℄[3℄. As
a matter of fa t, fra tal and pre-fra tal geometries present very interesting features in
terms of multiband behaviorand overall size.
Besides planargeometries [4℄[5℄[6℄, easilybuilt on diele tri substrates by meansof
stan-dard and low- ostphotolitographi te hniques, three-dimensional fra taltree-likeshapes
have been investigated [7℄. As a matter of fa t, despite their omplex realization, these
stru tures present veryinteresting ele trodynami properties thanksto the large number
ofdegrees offreedominthe hoi eofthevaluesoftheirdes riptivegeometri parameters.
In [7℄, Petko and Werner showed that fra tal tree geometries an be fruitfully employed
as end-loads for the miniaturizationof onventional dipoles or monopoleantennas.
However, although a multi-band behavior an be realized, the spe tral distribution of
multiple resonant frequen ies is stri tly related to the geometri al parameters of the
self-similar stru ture with xed relationships among the frequen ies whatever the
fra -tal shape. Therefore, the allo ationof several working frequen y bands inthe frequen y
spe trum turns out to be a omplex task with onstrained hoi es. In order to
over- ome su h a limitation, the insertion of rea tive loads in the fra tal tree stru ture has
beenproposed in[7℄atthe ost of anin reased omplexityin the onstru tion pro ess of
the devi e.
Anotheree tivemethodologyforobtainingamultibandbehavior,without onstraintson
resonan efrequen ies,isbasedontheintrodu tionofperturbationsbothinsizeandlength
of the fra tal geometry[8℄[9℄[10℄[11℄. A ording tothis guideline, this paperpresents the
design of a three-dimensional pre-fra tal tree antenna operating at three dierent
fre-quen y bands(
GP S L
1
-L
2
bandsandW i
− F i
). The synthesis of the pre-fra talgeome-try is arried out through a numeri aloptimizationpro edurebased on aparti leswarm
optimizer(
P SO
)[12℄[13℄[14℄, whi hoptimizesthe geometri alparameters des ribing thetree geometryavoiding the insertionof any lumped load toobtaina multibandbehavior.
three-and measurementson anexperimental prototype.
2 Triband Pre-Fra tal Tree Antenna Design and
Opti-mization
The synthesis of the three-dimensional geometry of the multiband prefra tal tree has
been formulated as an optimization problem by xing suitable onstraints in terms of
gain values and impedan e mat hing at the input port (Return Loss values) in both
L
1
and
L
2
frequen y bands of theGP S
system ( entered respe tively atf
1
= 1575.42
MHzand at
f
2
= 1227.60
MHz) as well as in theW i
− F i
band ( entered atf
3
= 2440.00
MHz).As far as the impedan e mat hing is on erned, a Return Loss lower than
−10 dB
isrequired. Moreover, the radiation hara teristi s of the three-dimensional antenna are
required to satisfy the following onstraints. Con erning
L
1
andL
2
GP S
bands, anhemispheri al overage isneeded withgainvalues suitablefor ommer ial
GP S
re eivers.More in detail, gain values greater than
G
min
= 3.0 dBi
atθ
= 0
o
and greater than
G
min
= −4.0 dBi
atθ
= 70
o
by assuming the presen e of a lownoise
30 dB
preamplierstage at the input port of the
GP S
re eiver be ause of the weakness of theGP S
signalsat the earth surfa e. Furthermore, taking into a ount the performan es required by a
ommer ial mobile transmitter/re eiver in the
W i
− F i
band, the following onstraintshave been dened: gain values greater than
G
min
= −6.0 dBi
atθ
= 70
o
and an average
gain greater than
G
min av
= −10.0 dBi
in the angular range70
o
< θ <
90
o
. Finally, in
order to minimizethe overall size of the antenna, the
3D
stru ture is required tobelongto a volume having dimensions
λ
0
4
×
λ
0
4
×
λ
0
4
, beingλ
0
4
∼
= 6 cm
at the lowest resonan efrequen y (
f
2
= 1227.60 MHz
).A ording to [7℄[15℄, the self similar stru ture of the prefra tal tree has been derived by
repeatedlyapplyingthe so- alledHut hinsonoperatorup tothestage
i
= 3
. Asshown inFig. 1, startingfrom atrunk of length
l
0
lyingalong thez
-axis ina artesian oordinateEx ept the trunk and the nal bran hes, every bran h is onne ted to a jun tion at its
ownterminations. Thus, itisatthe sametime hildandparent ofotherbran hes. These
stru tures an be grouped in dierent lasses depending on: (a) the numberof bran hes
R
that lead oa jun tion,(b) thebentangleξ
b
with respe t tothe parent dire tion,and( )the angle
ξ
s
that separates the onse utive bran hes leadingo the samejun tion. Inthis paper, in order to simplify the building pro edure, a four bran hes (
R
= 4
) fra taltree hara terized by
ξ
b
= 90
o
and
ξ
s
= 90
o
has been onsidered. Therefore, the resulting
fra tal shape is omposed by bran hes lying alongthe axesof the oordinatesystem and
the antenna stru ture is uniquely determined by: (i) the trunk length
l
0
and (ii) thelengths
l
i,j
of the bran hes,i
being the iteration index andj
(j
= 1, ... , R
i
) denotes the
j
-th bran h generated at thei
-th stage (i
= 1, ..., I
) as pi torially shown in Fig. 1(b)(
i
= 1
), Fig. 1( ) (i
= 2
), and Fig. 1(d) (i
= 3
). Summarizing, theI
-th order prefra taltree is onstituted by
L
=
P
I
i=1
R
(I−i+1)
+ 1
segments (the bran hes and the trunk) ofir ular ross-se tion
r
= 0.5 mm
in radius.Under these assumptions, the des riptive parameters of the pre-fra tal tree (trunked at
I
= 3
fortuningthreedierentresonan efrequen ies)havebeenoptimizedbyminimizingthe following ost fun tion
ℑ γ =
M −1
X
m=0
N −1
X
n=0
T −1
X
t=0
max
0,
G
min
{t∆θ, n∆φ, m∆f } − Φ {t∆θ, n∆φ, m∆f }
G
min
(1)+
V −1
X
v=0
max
0,
Ψ {v∆f } − V SW R
max
V SW R
max
whereγ
= {l
0
, l
i,j
, i
= 1, ... I = 3, j = 1, ... , R
i
}
,
∆f
is the sampling frequen y step inthe
L1
,L2
andW i
− F i
bands,∆θ
and∆φ
are the sampling angular steps of the gainfun tion,
Φ
γ = Φ {t∆θ, n∆φ, m∆f}
is the gain fun tion of the radiating stru tureγ
omputed at(
θ
= t∆θ
,φ
= n∆φ
,f
= m∆f )
,andΨ
γ = Ψ {v∆f}
istheV SW R
valueat
f
= m∆f
. Moreover, sin e besides ele tri al onstraints some other onditions havebeen stated for limitingthe physi alsize of the antenna, a set of geometri al onstraints
have been dened toavoidthe synthesis of unfeasible stru tures.
(
M oM
) ele tromagneti simulator[17℄ forminimizing(1), thus determiningthe antennastru ture. More in detail,starting fromthe trialsolutions
γ
(k)
p
(p
∈ [1, P ]
andk
∈ [1, K]
beingtheindexoftheparti leinthesolutionsswarmandtheiterationindex,respe tively)
iteratively generatedby the
P SO
, the prefra talgenerator determines the orrespondingantenna stru ture. Then, the
V SW R
and gain values are omputed by means of theM oM
simulator, whi h takes into a ount the presen e of a referen e ground plane ofinnite extent. The iterative pro ess ontinues until
k
= K
orℑ
opt
≤ η
,
η
being theonvergen e threshold and
ℑ
opt
= min
k
n
min
p
h
ℑ
γ
(k)
p
io
.3 Numeri al Analysis and Experimental Validation
Con erning the
P SO
implementation, the following setup has been used: a swarm ofP
= 8
trial solutions, a thresholdη
= 10
−3
, and a maximum number of iterationsequal
to
K
= 2000
. The remainingP SO
parameters have been set taking into a ount theguidelines proposed in [13℄ and a ording tothe rules dened in [16℄.
As an illustrative example of the optimization pro ess, Fig. 2 shows the evolution of
the behaviors of the
V SW R
in orresponden e with the optimal solution determined atsu essive iterations of the optimization pro ess. As it an be observed, starting from a
ompletely mismat hed onguration (
k
= 0
), the solution improves (k
= 10, k = 50
)until an antenna design is found (
k
= k
conv
) that ts the ele tri al spe i ations. Asfar as the geometri onstraints are on erned, the synthesized stru ture turns out to be
adequatesin eitsmaximumdimensionsareequalto
3.28 [mm]
alongthex
-axis,5.17 [mm]
along the
y
-axis,and3.82 [mm]
alongthez
-axis.From the omputationalpointofview, Figure3showsthe plotoftheoptimalvalueofthe
ost fun tion
ℑ
opt
versus the iterationnumber
k
. As it an be noti ed, the optimizationrequired
k
conv
= 75
with aCP U
-time at ea h iteration of about5 sec
(P entium IV
,1800 MHz
,512 MB
RAM).Thanks to the solution estimated at the end of the
P SO
-based optimization, aproto-type of the three-dimensional antenna has been built (Fig. 4). In order to arry out
on-90 × 140 [cm
2
]
and the Return Loss values have been measured with a s alar network
analyzerby pla ingthe measurementsetupinsideanane hoi hamber. Figure5shows a
omparison between omputed and measured values that points out asatisfa tory
agree-ment amongsimulations and experiments.
A similar omparative evaluation has also been performed for the radiation properties
of the antenna. More in detail, Figure 6 shows measured and simulated gain fun tions
in the horizontal plane [
θ
= 0
o
- Fig. 6(a)℄ and in a verti al plane [
φ
= 0
o
- Fig. 6(b)℄.
As required, the antenna prototypeallows anhemispheri al overage inthe
L
1
-L
2
GP S
bands and, onsidering the
30 dB
preampli ation stage at the input port, the gainval-ues turn out to be ompliant with the requirements (
Φ
GP S
{θ = 0
o
} > G
min
{θ = 0
o
} =
3 dBi
andΦ
GP S
{θ = 70
o
} > G
min
{θ = 70
o
} = −4 dBi
). Similar on lusions hold truefor the
W i
− F i
frequen y band (Φ
W i−F i
{θ = 70
o
} > G
min
{θ = 70
o
} = −6 dBi
andΦ
Av GP S
{70
o
< θ <
90
o
} > G
min Av
{70
o
< θ <
90
o
} = −3 dBi
).4 Con lusions
Thedesignandtheoptimizationofathree-dimensionaltribandprefra taltreeantennahas
beendes ribed. Theantenna stru turehas been synthesized byoptimizingthelengths of
thebran hesofaprefra taltreewithasuitableparti leswarm-basedalgorithminorderto
omply withgeometri alandele tri alrequirements. Aprototypeofthe tribandantenna
has been built and an exhaustive set of omparisons between measured and simulated
antenna parameters has been arriedout.
A knowledgments
This work has been supported in Italy by the Progettazione di un Livello Fisi o
'In-telligente' per Reti Mobili ad Elevata Ri ongurabilità, Progetto di Ri er a di Interesse
[1℄ D. H.Wernerand R.Mittra,Frontiersin ele tromagneti s.Pis ataway: IEEEPress,
2000.
[2℄ J. Gianvittorio and Y. Rahmat-Samii, "Fra tals antennas: a novel antenna
minia-turization te hnique, andappli ations,"IEEE Antennas Propagat. Mag.,vol.44,pp.
20-36, Feb. 2002.
[3℄ D. H. Werner, R. L. Haupt, and P. L. Werner, Fra tal antenna engineering: the
theory and design of fra tal antenna arrays, IEEE Antennas Propagat. Mag., vol.
41, pp. 37-59, O t. 1999.
[4℄ M. Sindou, G. Ablart, and C. Sourdois, "Multiband and wideband properties of
printed fra tal bran hed antennas," Ele tron. Lett., vol.35, no. 3, pp. 181-182,Feb.
1999.
[5℄ C. Puente, J. Claret, F. Sangues, J. Romeu; M.Q. Lopez-Salvans, and R. Pous,
Multibandpropertiesof afra taltree antennagenerated by ele tro hemi al
deposi-tion, Ele tron. Lett., vol.32,no. 25,pp. 2298-2299, De . 1996.
[6℄ D.H.Werner,A.R.Bretones,andB.R.Long,Radiation hara teristi softhin-wire
ternary fra tal trees, Ele tron. Lett., vol.35,no. 8,pp. 609-610, De .1999.
[7℄ J. S. Petko and D. H. Werner, Miniature re ongurable three-dimensional fra tal
tree antennas, IEEE Trans.AntennasPropagat.,vol.52,no. 8,pp. 1945-1956,Aug.
2004.
[8℄ C.Puente,J.Romeu,R.Pous,X.Gar ia,andF.Benitez,Fra talmultibandantenna
based on the Sierpinskigasket, Ele tron. Lett., vol. 32,pp. 1-2, Jan. 1996.
[9℄ M. Fernandez Pantoja,F. Gar ia Ruiz, A.Rubio Bretones, R.Gomez Martin,J. M.
Gonzales-Arbesù, J.Romeu,and J.M.Rius,GAdesignofwirepre-fra talantennas
and omparisonwithothereu lideangeometries, IEEEAntennasWirelessPropagat.
Gomez Martin, J.M. Gonzales-Arbesù, J. Romeu,J. M. Rius, P. L.Werner,and D.
H. Werner,GAdesignof smallthin-wire antennas: omparisonwithSierpinskitype
prefra tal antennas, IEEE Trans.Antennas Propagat., vol.54, pp. 1879-1882, Jun.
2006.
[11℄ R. Azaro, F. De Natale, M. Donelli, E.Zeni, and A. Massa, Synthesis of a
prefra -tal dual-band monopolar antenna for GPS appli ations, IEEE Antennas Wireless
Propagat. Lett., vol.5, pp. 361-364,2006.
[12℄ J. Kennedy,R.C. Eberhart,and Y.Shi, Swarm Intelligen e.San Fran is o: Morgan
Kaufmann Publishers, 2001.
[13℄ J. Robinson and Y. Rahmat-Samii,"Parti le swarm optimization in
ele tromagnet-i s," IEEE Trans. Antennas Propagat., vol. 52,pp. 397-407,Feb. 2004.
[14℄ M.DonelliandA. Massa,"A omputationalapproa hbasedonaparti leswarm
op-timizerformi rowaveimagingoftwo-dimensionaldiele tri s atterers,"IEEETrans.
Mi rowave Theory Te hn., vol.53, pp. 1761-1776, May 2005.
[15℄ D. H. Werner, P. L. Werner, and D. H. Chur h, Geneti ally engineered multiband
fra tal antennas, Ele tron. Lett., vol. 37,pp. 1150-1151, Sep. 2001.
[16℄ R. Azaro, F. De Natale, M. Donelli, A. Massa, and E. Zeni, "Optimized design of
a multi-fun tion/multi-bandantenna for automotive res ue systems," IEEE Trans.
Antennas Propagat., vol.54, no. 2,pp. 392-400,Feb. 2006.
[17℄ R.F.Harrington, FieldComputationbyMomentMethods.Malabar,Florida: Robert
•
Figure 1. Geometryofthe Tri-BandPrefra talTreeAntenna: (a)i
= 0
,(b)i
= 1
, ( )i
= 2
, and (d)i
= I = 3
.•
Figure 2. SimulatedReturnLossattheinputportofthe Tri-BandPrefra talTreeAntenna at dierent iterations of the optimization pro ess: (a)
k
= 0
, (b)k
= 10
,( )
k
= 50
,and (d)k
= k
conv
.•
Figure 3. Behaviorof the ost fun tionversus the iterationnumber.•
Figure 4. Photograph of the prototype of the Tri-BandPrefra tal Tree Antenna.•
Figure 5. Comparison between measured and simulated ReturnLoss values.•
Figure 6. Computedandmeasuredvaluesofthegainfun tioninthe(a)horizontalplane (
θ
= 0
o
)and inthe (b) verti alplane (