UNIVERSITY
OF TRENTO
DIPARTIMENTO DI INGEGNERIA E SCIENZA DELL’INFORMAZIONE
38123 Povo – Trento (Italy), Via Sommarive 14
http://www.disi.unitn.it
A PSO-DRIVEN SPLINE-BASED SHAPING APPROACH FOR
ULTRA-WIDEBAND (UWB) ANTENNA SYNTHESIS
L. Lizzi, F. Viani, R. Azaro, and A. Massa
January 2011
Wideband (UWB) Antenna Synthesis
Leonardo Lizzi, Federi oViani,Renzo Azaro, and Andrea Massa
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,
{leonardo.lizzi,federi o.viani, renzo.azaro}dit.unitn.it
Wideband (UWB) Antenna Synthesis
Leonardo Lizzi, Federi oViani,Renzo Azaro, and Andrea Massa
Abstra t
This paper deals with the synthesis of UWB antennas by means of a PSO-driven
spline-based shaping approa h. In order to devise a reliable and ee tive solution,
su h a topi is analyzed a ording to dierent perspe tives: 1) representation of
the antennashapewithasimple ande ient des ription,2) denitionofasuitable
des ription of the UWB Tx/Rx system, 3) formulation of thesynthesisproblem in
terms of an optimization one, 4) integration of the modelling of the UWB system
into a omputationally e ient minimization pro edure. As a result, aninnovative
synthesis te hnique based on the PSO-based iterative evolution of suitable shape
des riptorsis arefully detailed. Toassesstheee tiveness oftheproposedmethod,
asetofrepresentative numeri al simulationsareperformedandtheresultsare
om-pared with themeasurements from experimental prototypesbuilt a ording to the
designspe i ations omingfromtheoptimizationpro edure. Tofo usonthemain
advantages and features of the proposed approa h, omparisons with the results
obtained witha standard parametri approa h arereported, aswell.
Keywords: AntennaSynthesis,Ultra-Wide-Band,SplineRepresentation,Parti leSwarm
Ultra-wideband(UWB)isaterm ommonlyusedtodes ribeawirelesste hnology where
very short low-power time pulses, whi h o upy a very large frequen y bandwidth, are
transmitted/re eived. Su h a ommuni ation te hnique allows high transmission
data-rates, multipath immunity, low probability of inter ept, and low power onsumption.
Moreover, it enables the possibility for a single system to simultaneously operate in
dif-ferent ways (e.g., as a ommuni ation devi e, a lo ator or a radar) [1℄. Sin e the US
Federal Communi ation Commission (FCC)revision on UWB transmissions in 2002 [2℄
and the Ele troni Communi ationsCommittee (ECC)de ision inEuropeabout the use
of UWB te hnologies[3℄, thedesign ofUWB systems has drawn a onsiderableattention
be omingakey topi in the framework of wireless ommuni ations.
Asfarastheradiatingsub-systemis on erned,unlike onventionalnarrow-bandsystems
where modulatedsinusoidalwaveforms o upy smallportionsof the frequen y spe trum,
theassumptionofauniformbehavioroftheantennaisnolongerreliablewhendealingwith
UWB frequen y bandwidths and the distortion of the transmitted time-domain pulses
should be taken into a ount and arefully prevented. Consequently, UWB antennas
turn out to be riti al omponents of the whole system and their ele tri al parameters
need an a urate optimization in a wide range of frequen ies to minimize/redu e the
signal distortions. In order to properly address su h an issue, suitable te hniques are
ne essary sin e a standard synthesis pro ess aimed at determining lassi al
frequen y-domain parameters of both the transmitting antenna and the re eiving one (e.g., gain,
radiation patterns, ree tion oe ients, and polarization) is not enough to ensure the
distortionlessof the UWB systemand the orre ttransmission/re eption oftime-domain
pulses. More spe i ally, the design of an UWB antenna requires ustomized synthesis
te hniques to satisfy the UWB requirements [4℄ as well as proper analysis tools for an
a urate des riptionof the antennabehaviors inthe time domain[5℄.
Generallyspeaking, twomain methodologiesfortheantennadesign anbeusually
re og-nized. The former, indi atedas parametri approa h, onsiders areferen e shapedened
by a xed numberof geometri des riptors to be optimized to satisfy the proje t
gular stru ture [9℄, diamond [10℄[11℄, and bow-tie [12℄. On the other hand, the so- alled
building-blo kapproa h synthesizesthe antennageometrythroughasuitable ombination
of elementarybuilding-blo ks asshown in [13℄ and [14℄.
In this paper, by exploiting the best features of both approa hes, an innovative UWB
antenna synthesis method,whosepreliminaryresultshavebeen presented in[15℄,is
are-fullydes ribed andanalyzed. InordertodesignUWB antennasthatmeetthe userneeds,
both the simpli ity in des ribing some geometri al hara teristi s (e.g., the feedline
ex-tension, the groundplane,and substrate dimensions) of the parametri approa hand the
exibility of a motherstru ture-based method are ne essary toallowa fastand ee tive
synthesispro edure. Moreover, thereisthe needtointegratethesynthesiste hniquewith
an analysis method aimed at evaluating not only the impedan e bandwidth and
radia-tion propertiesbut alsotaking intoa ount the ee ts ofthe propagation hannelonthe
UWB system. Towards this end, both the transmitting antenna and the re eiving one
are onsideredto simulate the whole system.
The outline of the paper is as follows. Se tion 2 gives a des ription of the proposed
approa h fo using on the representation of the antenna shape as well as on the analysis
method. As for the assessment, a set of numeri aland experimentalresults are reported
and dis ussed (Se t. 3). Final omments and on lusions follow inSe t. 4.
2 Mathemati al Formulation
TheUWBantennasynthesisproblemisformulatedintermsofanoptimizationonewhere
a set ofunknown representativedes riptors are tunedthrough aniterativepro ess aimed
at ttingsuitablerequirements/ onstraints onthe ele tri albehavior inthe UWB
band-width [2℄[3℄. Towards this end, a spline-based[16℄ shapegenerator and a PSO pro edure
[17℄are usedtodene the onvergent su essionof trialsolutions. By usingaset of
on-trolpointsandgeometri aldes riptorsto ode theantennashaperepresentation, aset of
trialgeometriesisevaluatedatea hiterationandupdatedthroughthePSO strategyuntil
a suitablemat hing between estimated and required spe i ations is obtained. As far as
University of Trento and implemented following the guidelines in [19℄ and [20℄, is used
to simulate both the transmitting antenna and the re eiving one (assumed of identi al
shapes) as well asthe propagation hannel.
Forthesakeof larity,thekey-pointsoftheproposedapproa haredetailedinthefollowing
sub-se tions.
2.1 UWB Antenna Shape Representation
The rst issue, addressed in dealing with UWB antenna synthesis, is on erned with a
suitable and exible representation of the antenna shape. To this end, there is the need
to provide an antenna geometry generator (
AGG
) able to model a wide set of planar pat h stru tures as a fun tion of a smallnumber of optimization variables. A ordingly,a ombination of parametri des riptors and spline urves [16℄ (assumed as building
blo ks)seems to be a good solution. In parti ular, some antenna features are expressed
intermsofthevaluesofgeometri alparametersinxedranges. Inparti ular,thefeedline
isassumedofre tangularshapeaswellasboththegroundplaneandthesubstrate(Fig. 1).
Thosepartsarefullydeterminedbyxingtheirextensions(i.e.,theirlengthsandwidths).
The des ription of the antenna geometry is ompleted by a spline-based representation
aimed at oding omplex ontours by means of a limitedset of ontrol points.
By assuming a symmetry with respe t to the
y
− z
plane, only one half of the physi al stru ture of the antennahas tobemodeled. Let{ϕ
l
; l = 1, ..., L; L = 4}
(1)bethesetofparametri des riptors. Morespe i ally,
ϕ
1
isthesubstratelength,ϕ
2
isone half of the substrate width,ϕ
3
is one half of the feed-line width, andϕ
4
is the length of the groundplane. Moreover,ϕ
(min)
l
≤ ϕ
l
≤ ϕ
(max)
l
,ϕ
(min)
l
andϕ
(max)
l
beingxed onstantsthat denethe range of admissiblevariation of the
l
-th des riptor. As for the remaining part of the antenna geometry, let∆ (t) = {x (t) , y (t)}
be the mathemati al des ription ofthe B-splinepat hperimeterasafun tionofthe urvilinear oordinatet
. Su ha urve turns out to be the linear ombinationofM
-degree polynomialsδ
n,M
asfollows∆ (t) =
N
X
n=1
M
X
m=0
δ
n,M
(P
n−M+m,t
) ,
M
= 3
(2)where
P
n
= (x
n
, y
n
)
,n
= 1, ..., N
,denotes then
-th ontrolpointwhose Cartesian oordi-nates are(x
n
, y
n
)
. Inorder toavoidthe generationof unrealisti stru tures, thefollowing onstraints are imposed, aswell:x
1
= ϕ
3
andx
N
= 0
.A ordingtosu harepresentation,atrialantennashape
A
istheresultoftheappli ation of the operatorAGG
A
= AGG ξ
where
ξ
= {(x
n
, y
n
) , n = 1, ..., N; ϕ
l
; l = 1, ..., L} = {ξ
j
, j
= 1, ..., J; J = 2 × N + L}
(3)odes the setofvariables,tobeoptimizedtofullltheUWB proje trequirements,whi h
unequivo ally identies the orresponding geometri almodelof
A
.2.2 UWB Antenna System Chara terization
These ondsteptobefa ed onsistsinidentifyingafastandreliablenumeri alpro edure
to evaluate the ele tri behaviorof ea h trialshape. Be ause of the very large frequen y
band of operationof UWB systems and the transmission/re eption of short time pulses,
a natural approa h would onsider a time-domain des ription. However, from an
exper-imental point-of-view, a frequen y domain analysis ould be preferable be ause of the
higher a hievable measurement a ura y[21℄.
Takingintoa ountthese onsiderations,atimedomainrepresentationisinitiallyadopted
to dene the requirementsof the transmitting-re eiving antennasystem. Then,these
re-quirementsaretranslated inthe frequen y domainusing the s atteringparameters
rep-resentationtodealwithphysi alquantitiesexperimentally-dete tableinastraightforward
and a urate way.
In order to have a distortionless behavior, the frequen y domain transfer fun tion (i.e.
responsewithalineardependen eonthe frequen yor,analogously, onstantgroupdelay.
Undergoodimpedan emat hing onditions[22℄[23℄,thetransferfun tion anbe
approx-imated with the s attering oe ient
s
21
. Notwithstanding su h a omputation usually needsthe numeri almodellingof thewholetransmitting/re eivingantennasystem withavery expensive and time- onsuming omputationalpro edure. Tosimplify the numeri al
model and to redu e the omputationalburden, let us onsider a simple method
analo-gous to the so- alled Pur ell antenna gain measurement method [24℄. More spe i ally,
the two-antenna system [Fig. 2(a)℄ is substituted by an antenna and a ree ting surfa e
lo atedjustinfrontofthetransmittingdevi eatadistan eequaltoonehalfoftheTx/Rx
distan e [Fig. 2(b)℄. Although su h an operation annot always be performed (sin e an
image antenna is not generally identi al to the original one even though a perfe tly
re-e tingsurfa e ofinnitearea is onsidered[25℄),howeverthe antennasystem undertest
fullysatisesthemethodhypothesesbe auseofthe axialsymmetryofthereferen eshape
(Fig. 1). Consequently, after simple manipulations,the mutual s attering oe ient
s
21
an beexpressed as followss
21
= s
12
= s
11
− s
M P
11
(4) wheres
11
=
b
1
a
1
a
2
=0
is the inputree tion oe ient of the originalantenna system [Fig.
2(a)℄, while
s
M P
11
=
b
′
1
a
′
1
isthe same quantity, but in the image onguration[Fig. 2( )℄.Moreover, underthe assumptionsofantennaswith negligiblestru tural s attering
ontri-butions[21℄, the evaluationof
s
11
an be arriedoutby onsideringa singleantennathat radiates infree spa e [Fig. 2(d)℄s
11
⋍ s
0
11
=
b
0
a
0
.
(5)A ordingly, itappears that
s
21
= s
0
11
− s
M P
11
.
(6)In su h away, the whole system an be modeledbymeans of a singleantenna instead of
from the value of
s
11
with and withoutthe presen e of aninnite metalli plate.2.3 PSO-Based Optimization of the UWB Antenna Shape
On e an ee tivemethodto hara terize atrialshape of the
U W B
antenna is available, there is the need to dene an evolution strategy able toguide the synthesis pro ess to anal design that fully satisesthe proje tguidelines and onstraints dened by the user.
To this end, let us reformulate the synthesis problem as an optimization one where a
suitable ost fun tion
Ω (A)
is minimizeda ording toaP SO
-based strategyb
A
= arg min
A
{Ω (A)}
(7)
to determine the nal shape
A
b
= AGG
bξ
of theU W B
antenna.In general, evolutionary algorithms use the on ept of tness to represent how well a
parti ularsolution omplywiththedesignobje tive. Thedegreeoftnesstotheproblem
athandofea htrialsolutionisequaltothe orrespondingvalueofthe ostfun tion. Sin e
the design obje tive is to synthesize a
U W B
system hara terized by good impedan e mat hing onditions and by distortionless properties, the ost fun tionΩ
is a ordingly dened.Let
f
1
andf
2
be the lowest and highest frequen y of the band of interest, respe tively. As far as the impedan e mat hing is on erned, the following onstraint is imposed|s
11
(f )| ≤
s
(d)
11
(f )
f
(min)
≤ f ≤ f
(max)
(8)wherethe supers ript
(d)
denotes the targetvalueof the designspe i ation. Moreover,
by assuming that (8) holds true and that
H
≃ s
21
, the onditions for a distortionless system anbereformulatedintermsof: 1)a onstraintonthemagnitudeofs
21
[Condition (a) - Flat amplitude response℄△ |s
21
| ≤ △
(d)
|s
21
|
(9)where
△ |s
21
| , max
an-△τ ≤ △
(d)
τ
(10)where
△τ = max
f
(min)
≤f ≤f
(max)
{τ (f )}−min
f
(min)
≤f ≤f
(max)
{τ (f )}
,τ
(f ) = −
1
2π
d
df
{∠s
21
(f )}
. Starting from these onditions on the s attering parameters, the ost fun tion, whi hmaps the designspe i ations into atness index, isdened asfollows
Ω (A) =
Z
f
(max)
f
(min)
Ω
11
(A) H {Ω
11
(A)} df + Ω
21
(A) H {Ω
21
(A)} + Ω
GD
(A) H {Ω
GD
(A)}
(11) where
Ω
11
(A) =
|s
11
(f )| −
s
(d)
11
(f )
s
(d)
11
(f )
(12)Ω
21
(A) =
△ |s
21
| − △
(d)
|s
21
|
△
(d)
|s
21
|
(13)Ω
GD
(A) =
△τ − △
(d)
τ
△
(d)
τ
(14)H
being the Heaviside fun tionH {Ω} =
1 Ω ≥ 0
0 Ω < 0
.
(15)As far as the
P SO
-based sampling of the solution spa e is on erned, ea h parti le of the swarm of dimensionR
odes a andidate solutionξ
r
(r
= 1, ..., R
) of the optimiza-tionproblem. Atea hiterationk
(k
= 1, ..., K
)ofthe minimizationpro ess,ea hparti le moves losertoitsownbestpositionfoundsofarζ
(k)
r
= arg min
k=1,...,K
n
Ω
h
AGG
ξ
(k)
r
io
(indi ated aspersonalbest position)aswellastowards thepositionofthe best parti leof
the swarm,
ς
(k)
= arg min
r=1,...,R
n
Ω
h
AGG
ζ
(k)
r
io
(referred to as global best position) [17℄ a ording to the rules of evolution of the PSO [26℄. Starting from a swarm ofpar-ti les randomly hosen (i.e., a random hoi e of the ontrol points of the initial shapes
when
Ω
AGG ς
(k)
≤ η
conv
,η
conv
being a user-dened threshold. Then,bξ = ς
(k)
andb
A
= AGG ς
(k)
. 3 Numeri al ResultsIn this se tion,asele ted set ofnumeri alresults fromseveral experimentsis reported to
showthe behaviorof thesynthesis methodaswellastoassess itsreliabilityand e ien y
in ttingthe designrequirements.
Whatever the test ase or experiment, if it is not spe ied, the proje t onstraints have
been xed to
s
(d)
11
(f )
= −10 dB
,△
(d)
|s
21
(f )| = 6 dB
, and△
(d)
τ
= 1 nsec
to guaranteesuitable performan es of the synthesized
U W B
antenna systems. As for the sto hasti optimizer, a population ofR
= 7
parti leshas been used and the onvergen e threshold has been xedtoη
conv
= 10
−3
withthe maximumnumberof iterationsequalto
K
= 400
. Moreover, the values of theP SO
ontrol parameters have been xed toC
1
= C
2
=
2.0
andw
= 0.4
. They have been sele ted a ording to the suggestions given in the referen eliteraturewherethey havebeenfound toprovidegoodperforman e. Foramoredetailed study of onvergen e hara teristi s in orresponden e with dierent values of
these parameters, please referto [27℄.
The rst test ase is about the synthesis of an
U W B
antenna operating in the rangef
(min)
= 6 GHz
-
f
(max)
= 9 GHz
and ompliant with the guidelines of the Ele troni
Communi ations Committee (ECC) [3℄. Su h an experiment is aimed at analyzing the
behavior of the optimization pro ess both in terms of onvergen e and evolution of the
antenna shape. Startingfromarandomlygenerated perimeter[
k
= 0
-Fig. 3(a)℄ hara -terizedbyN
= 4
ontrolpoints,whi hdoesnottatallthedesign obje tives(Fig. 4)as onrmedbythe valueof the ost fun tionΩ
(i.e.,Ω
AGG ς
(0)
∼
= 5
),the trialsolution
improves until the onvergen e onguration [Fig. 3(d)℄ is found. In order to point out
the mainadvantagesofthe proposedspline-basedmethodoveraparametri optimization
approa h, the same problem(in terms of user requirements) has been addressed by
on-sidering the referen e ir ular geometry shown in Fig. 5. The same PSO algorithmhas
been used to modify the unknown parameters
ξ
= {(x
1
, y
1
) ; r; ϕ
l
, l
= 1, ..., 4}
. Figure 6 and Figure 7 show the evolution of the trial geometry and the orresponding s atteringmaximum number of iterations (
k
= K
) has been performed, the onvergen e solution does not t the whole set of proje t requirements (△ |s
21
(f )| > 6 dB
). Con erning the omputationalissues,the plotof the ost fun tionin orresponden e withthe parametriapproa his omparedinFig. 8withthatofthe proposedapproa h. Besides thewiderset
ofpossiblesolutions,thespline-basedte hniqueturnsouttobemoree ientinsampling
the solutionspa e sin e abetter solutionin rea hed in just
k
conv
= 28
iterations.Letusnowanalyzethedependen eoftheiterativepro essonthedimensionofthesolution
spa e. Sin e the generation of the spline geometry stri tly depends on the number of
ontrol points that also denes the dimension of the solution spa e, some simulations
have been performed by varying
N
. As a general rule, a smallnumber of ontrol points wouldde reasethedimensionofthesolutionspa eallowingafastersear h,butatthe ostofaredu ed apa ityofrepresenting omplex ontours. Onthe ontrary,alargernumber
of ontrol points would allow the des ription of a wider set of geometries and of more
omplex antenna shapes, even though with a higher omputationalburden and the need
to enlargethe dimension of the swarm tofully exploit the additionaldegrees of freedom.
In order to better understand the behavior of the optimization in orresponden e with
dierent values of
N
, letus onsider the followingquality indexesF
11
=
1
∆f
Z
f
(max)
f
(min)
|bs
11
(f )| df
(16)F
21
=
1
∆f
Z
f
(max)
f
(min)
(
|bs
21
(f )| − |s
21
(f )|
|bs
21
(f )|
)
df
(17)F
GD
=
1
∆f
Z
f
(max)
f
(min)
(
|b
τ
(f )| − |τ (f )|
|b
τ
(f )|
)
df
(18)wherethesupers ript
¯
indi atesthemeanvalue,overthefrequen yband∆f = f
(max)
−
f
(min)
, of the s attering parameters at the onvergen e [b
s
ij
= s
ij
b
A
℄. Figure 9 shows the values of the quality indexes versusN
as well as the plot ofk
conv
. As expe ted, the number of iterations needed to nd a onvergen e solution in reases withN
sin e the solution spa e be omes larger. However, the quality indexes do not proportionallyturns out to be at
N
= 7
ontrol points. In the following, this value will be assumed as referen e.After the analysis on the optimization pro ess and on the sensitivity of the method to
the antenna shape des riptors, the se ond part of this se tion is devoted to present the
resultsfromthenumeri alandexperimentalassessment. Fortheexperimentalvalidation,
a set of antenna prototypes has been built a ording to the results from the numeri al
simulations by meansof a photo-lithographi printing ir uitte hnology and onsidering
an Arlon substrate (
ε
r
= 3.38
) of thi kness0.78 mm
as diele tri support. Con erning the measurements,ea hprototypehas been fedwitha oaxialline,equipped withaSMAonne tor, onne ted at the input port (Fig. 1). Moreover, an Anritsu Ve tor Network
Analyzer (37397D Lightning) has been used to olle t the data in a non- ontrolled
envi-ronmentand the measurementsof the parameter
s
21
havebeenperformedby onsidering a distan e ofd
= 15 cm
between two identi al (Tx/Rx) antennaprototypes[28℄. Be ause of the operatingfrequen ies of UWB systems and the dimensions of the prototypes, theantennas an be reasonably onsidered inthe far eld region of ea h other.
The se ond test ase deals with the synthesis of a
U W B
antenna that operates in the frequen y range4 GHz 6 f 6 9 GHz
and omplieswith the FCC standard. The values of the antennades riptors obtained by theP SO
-based optimizationpro edureare given in Tab. I and the prototype of the antenna is shown in Fig. 10. As for the ele triperforman es,Figure11showsthesimulatedandmeasuredbehaviorsofthe
s
ij
parameters and ofthegroupdelayτ
. Asit an beobserved, theantennats theproje trequirements fromf
= 4 GHz
up tof
= 9 GHz
(i.e., within the FCC mask [2℄) and the bandwidth turnsouttobeof△f = 5 GHz
(thefra tionalbandwidthbeingequalto76.9
%a ording to [2℄) from simulated as well as measured data. In parti ular,|bs
11
(f )| < −10 dB
[Fig. 11(a)℄, the widest variation of the magnitude ofs
b
21
is less than5 dB
(i.e.,△ |bs
21
(f )| <
△
(d)
|s
21
(f )|
) [Fig. 11(b) and Tab. I℄, and△b
τ
= 0.5 nsec < △
(d)
τ
[Fig. 11(d)℄. Given
the separation
d
= 15 cm
between theantennas,thea hievedvalues fors
b
21
arereasonable and in agreement with the values obtained from the Friis formula. As regards to thedependen e of
|bs
21
(f )|
onthe distan e between the twoantenna prototypes, the in rease of the separation distan e turns out in a shift of the plot towards lower values. Asa onsequen e, the value of
△ |bs
21
(f )|
does not present signi ant variations with the distan e.As far as the group delay is on erned, although in a reasonable a ordan e with
simu-lations, measured data present larger u tuationsprobably ausedby the non- ontrolled
environmentwhere the experimentalmeasurementshave been arried out. Nevertheless,
theatbehaviorsofthemagnitudeof
s
21
andofthegroupdelay,orequivalentlythelinear trend of the phase ofs
21
[Fig. 11( )℄,assure that the synthesized antenna is suitable for UWB ommuni ations where atransmitted signal should not be distorted.The lastexample is on erned with amore hallenging problemwhere awider frequen y
band is required (i.e.,
∆f = 6 GHz
,f
(min)
= 4 GHZ f
(max)
= 10 GHZ
) and a further
onstraint on the dimension of the antenna is imposed for a more easy integration in
ommer ial produ ts (i.e.,
ϕ
(max)
1
= 35 mm
andϕ
(max)
2
= 15 mm
), even though a smaller threshold on the amplitude ofs
21
is xed (△
(d)
|s
21
(f )| = 10 dB
).The des riptive parameters and the main ele tri al indexes of the synthesized antenna
are summarized in Tab. II. Moreover, the plots of the ele tri al parameters over the
whole frequen y range are reported in Fig. 12. As expe ted, the antenna performan es
generally degrade ompared to that of the previous example due to the in reased band.
As amatteroffa t, thedesigned antenna has afra tionalbandwidthof
85.7
%. However, the maximum variationofτ
isof just about0.15 ns
(simulateddata) and the dimensions ofthe antennaare verymoderate(ϕ
b
1
= 33.3 mm
andϕ
b
2
= 12.2 mm
). On theotherhand,△ |bs
21
(f )|
turns out tobe slightly greaterthan the desiredvalue (△ |bs
21
(f )| = 12 dB
vs.△
(d)
|s
21
(f )| = 10 dB
)sin ethemaximumnumberofiterationsK
= 400
hasbeenrea hed andthe onvergen e riterionhas notbeensatised. Probably,agreaterdimensionoftheswarm and alarger number of iterationswould be enoughto arefullymat halsosu h a
onstraint, but su h asetup would require very expensive omputations.
The omparisonbetween thesimulationresultandthedata measuredfromtheprototype
(Fig. 13) further onrm the reliability of the synthesis pro edure. As a matter of fa t,
there is ana eptable agreement between the two plots (Fig. 12) despitesome ree tion
In this paper, an innovative synthesis approa h based on the use of a spline-based
rep-resentation for
U W B
antennas has been presented. The nal shape of the antenna is determined by means of aP SO
-based optimization pro edure that exploits a suitable and omputationally e ient ele tromagneti representation of the whole Tx/Rx UWBsystem.
Theassessmenthasbeen ondu tedondierenttest ases. Firstly,theproposedte hnique
has been testedby onsideringnumeri alexamplestoshowthefeatures andthe behavior
of the iterative pro edure. In order to point out some of the main advantages of the
proposed method, the same test ase has been addressed with a parametri approa h,
as well. Se ondly, we have analyzed the dependen e of the synthesis method on the
des riptors of the antenna shape. Finally, the synthesis results have been veried with
some omparisons with the experimentaldata measured from the antenna prototypes.
The A on lusion, whi h is derived from all the numeri al and experimental results
ob-tained,isthat theproposed approa hrepresents avery promisingmethodology forUWB
antennadesignbe auseofitsexibilityindealingwithdierentuser-denedrequirements
[1℄ L. Yang and G. B. Giannakis, Ultra-wideband ommuni ations, IEEE Signal
Pro- essing Mag.,pp. 26-94,Nov. 2004.
[2℄ US Federal Communi ation Commission, First Report and Order, Revision of
part 15 of the ommission's rules regarding ultra-wideband transmission systems,
FCC02-48, Apr. 2002.
http://hraunfoss.f .gov/edo s_publi /atta hmat h/FCC-02-48A1.pdf
[3℄ Ele troni Communi ations Committee, Draft ECC/DEC/(06)AA, "On the
har-monised onditions for devi es using UWB te hnology in bands below 10.6 GHz,"
ECC/DEC/(06)AA, 2006.
[4℄ Z. N. Chen, X. H. Wu, H. F. Li, N. Yang, and M. Y. W. Chia, Consideration for
sour e pulses and antennas inUWB radio system, IEEE Trans.Antennas Propag.,
vol. 52,no. 7, pp. 1739-1748, Jul.2004.
[5℄ X.Qing,Z.N.ChenandM.Y.W.Chia,Chara terizationofultrawidebandantennas
using transfer fun tions, Radio S i., vol.41,RS1002, pp. 1-10, 2006.
[6℄ C. C. Lin,Y. C. Kan,L. C. Kuo,and H.R. Chuang, "Aplanartriangularmonopole
antenna for UWB ommuni ation," IEEE Mi rowave Wireless Compon. Lett., vol.
15, no. 10,pp. 624-626, O t. 2005.
[7℄ J. Liang, C. C. Chiau, X. Chen and C. G. Parini, "Study of a printed ir ular dis
monopoleantenna for UWB systems," IEEE Trans. Antennas Propag., vol. 53, no.
11, pp. 3500-3504, Nov. 2005.
[8℄ Y. Ren and K.Chang,"An annular ring antenna for UWB ommuni ations,"IEEE
Antennas Wireless Propag. Lett., vol. 5,pp. 275-276,2006.
[9℄ J.Jung,W.Choi,andJ.Choi,"Asmallwidebandmi rostrip-fedmonopoleantenna,"
rounded diamond antennas for ultrawide-band ommuni ations," IEEE Antennas
Wireless Propag. Lett.,vol.3, pp. 249-252,2004.
[11℄ X.H.WuandZ.N.Chen,DesignandoptimizationofUWBantennasbyapowerfull
CADtool: PULSEKIT, Pro . IEEE AntennasPropag.Int.Symp.,vol.2,pp.
1756-1759, Jun. 2004.
[12℄ K.Kiminami,A.Hirata,andT.Shiozawa,"Double-sidedprintedbow-tieantennafor
UWB ommuni ations,"IEEE Antennas Wireless Propag. Lett.,vol.3,pp. 152-153,
2004.
[13℄ J. M. Johnsonand Y. Rahmat-Samii,"Geneti algorithmsand methodof moments
(GA/MOM) forthe design ofintegratedantennas," IEEE Trans.Antennas Propag.,
vol.47, no. 10,pp. 1606-1614,O t. 1999.
[14℄ F. J.Villegas,T. Cwik, Y.Rahmat-samii,andM.Manteghi,"Aparallel
ele tromag-neti geneti -algorithm optimization (EGO) appli ation for pat h antenna design,"
IEEE Trans.Antennas Propag., vol. 52,no. 9, pp. 2424-2435, Sep. 2004.
[15℄ L. Lizzi, F. Viani, R. Azaro, and A. Massa, Optimizationof a spline-shaped UWB
antenna by PSO, IEEE Antennas Wireless Propag. Lett., vol. 6,pp. 182-185, 2007.
[16℄ C. De Boor,A Pra ti al Guide to Spline. Springer, New York, 2001.
[17℄ J.Robinsonand Y.Rahmat-Samii,"Parti leSwarmOptimizationin
Ele tromagnet-i s," IEEE Trans.Antennas Propag., vol. 52,no. 2, pp. 397-407,Feb. 2004.
[18℄ R. F. Harrington, Field Computation by Moment Methods. Malabar, FL: Robert E.
Krieger, 1987.
[19℄ C. A. Balanis, Advan ed Engineering Ele tromagneti s. New York, NY: John Wiley
& Sons, 1989.
[20℄ S. Rao, D. Wilton, and A. Glisson, "Ele tromagneti s attering by surfa es of
Sons, 1982.
[22℄ C. G.Montgomery,R.H.Di ke,andE.M.Pur ell,Prin iplesof Mi rowaveCir uits.
New York: M Graw-Hill,1948.
[23℄ R.E.Collin,Foundationsfor Mi rowaveEngineering.New York: Wiley-IEEEPress,
2000.
[24℄ S. Silver, Mi rowave AntennaTheory and Design. New York: M Graw-Hill,1949.
[25℄ D. M. Kerns, Plane-Wave S attering-Matrix Theory of Antennas and
Antenna-AntennaIntera tions. Nat.Bur. of Stand.Monog.162.Boulder: NationalBureau of
Standards, 1981.
[26℄ J.Kennedy,R.C. Eberhart,and Y.Shi, Swarm Intelligen e.San Fran is o: Morgan
Kaufmann Publishiers, 2001.
[27℄ F. van den Bergh, An analysis of parti le swarm optimizers, Ph. D. dissertation,
Dept. Comput. S i., Univ. Pretoria,Pretoria, SouthAfri a, 2002.
[28℄ S. H. Lee, J. K. Park, and J. N. Lee, A novel CPW-fed ultra-wideband antenna
•
Figure 1. Des riptive parameters of the spline-based antennarepresentation.•
Figure 2. S hemati representation of (a) the system onstituted by a ouple of identi al antennas, (b) the antenna and ree ting surfa e onguration, ( ) theequivalent image onguration, and (d) the single antenna radiatinginfree-spa e.
•
Figure 3. Test Case 1 (f
(min)
= 6 GHz
,
f
(max)
= 9 GHz
, ECC standard,
N
= 4
) - Evolution of the antenna shape: (a)k
= 0
, (b)k
= 10
, ( )k
= 19
, and (d)k
= k
conv
= 28
.•
Figure 4. Test Case1 (f
(min)
= 6 GHz
,f
(max)
= 9 GHz
,ECC standard,N
= 4
) -Plotof (a)s
11
AGG ς
(k)
, (b)
s
21
AGG ς
(k)
, ( )
∠s
21
AGG ς
(k)
,and (d)
τ
AGG ς
(k)
versus the frequen y atdierent iterationsof the optimization
pro ess.
•
Figure 5. Test Case 1 - Parametri approa h - Des riptive parameters for the ir ular monopolereferen e geometry.•
Figure 6. Test Case 1 - Parametri Approa h (f
(min)
= 6 GHz
,
f
(max)
= 9 GHz
,
ECC standard) - Evolution of the antenna shape: (a)
k
= 0
, (b)k
= 100
, ( )k
= 300
, and (d)k
= K = 400
.•
Figure 7. Test Case 1 - Parametri Approa h (f
(min)
= 6 GHz
,
f
(max)
= 9 GHz
,
ECCstandard)-Plotof(a)
s
11
AGG ς
(k)
,(b)
s
21
AGG ς
(k)
,( )
∠s
21
AGG ς
(k)
, and (d)
τ
AGG ς
(k)
versusthe frequen y atdierentiterationsof the optimiza-tion pro ess.
•
Figure 8. Test Case1 (f
(min)
= 6 GHz
,
f
(max)
= 9 GHz
,ECC standard,
N
= 4
) -Comparison between parametri approa h and spline-based approa h - Plot of theoptimal value of the ost fun tion
Ω
AGG ς
(k)
and orresponding terms versus
the iteration number
k
.•
Figure 9. Test Case1 (f
(min)
= 6 GHz
,
f
(max)
= 9 GHz
,ECC standard)-
points
N
.•
Figure 10. Test Case 2 (f
(min)
= 4 GHz
,
f
(max)
= 9 GHz
, FCC standard,
N
= 7
) - Antenna prototype: (a)front view and (b)ba k view.•
Figure 11. Test Case 2 (f
(min)
= 4 GHz
,
f
(max)
= 9 GHz
, FCC standard,
N
= 7
) - Numeri al and measured values of (a)|bs
11
|
, (b)|bs
21
|
, ( )∠bs
21
, and (d)τ
b
versus the frequen yf
.•
Figure 12. Test Case 3 (f
(min)
= 4 GHz
,f
(max)
= 10 GHz
,ϕ
(max)
1
= 35 mm
,ϕ
(max)
2
= 15 mm
, FCC standard,N
= 7
) - Numeri al and measured values of (a)|bs
11
|
, (b)|bs
21
|
, ( )∠bs
21
, and (d)b
τ
versus the frequen yf
.•
Figure 13. Test Case 3 (f
(min)
= 4 GHz
,f
(max)
= 10 GHz
,ϕ
(max)
1
= 35 mm
,ϕ
(max)
2
= 15 mm
, FCC standard,N
= 7
) - Antenna prototype: (a) front view and(b) ba k view.
Table Captions
•
Table I. Test Case 2 (f
(min)
= 4 GHz
,
f
(max)
= 9 GHz
, FCC standard,
N
= 7
) -Geometri des riptors and ele tri parameters of the synthesized antenna.•
Table II. Test Case 3 (f
(min)
= 4 GHz
,f
(max)
= 10 GHz
,ϕ
(max)
1
= 35 mm
,ϕ
(max)
2
= 15 mm
, FCC standard,N
= 7
) - Geometri des riptors and ele triϕ
1
ϕ
2
z
ϕ
3
y
x
ϕ
4
P
3
P
N −1
P
1
P
2
..
..
..
Port
Input
P
N
a
2
a
1
[S]
b
1
b
2
d
Metallic Plate
a
′
1
b
′
1
d/2
(a) (b)a
′
1
b
′
1
a
′
2
b
′
2
J
J
J
J
[S
M P
]
a
0
b
0
( ) (d)ϕ
1
=
7
.9
7
cm
ϕ
2
= 1.59 cm
ϕ
2
= 2.33 cm
ϕ
1
=
3
.6
7
cm
(a) (b)ϕ
2
= 2.46 cm
ϕ
1
=
3
.5
7
cm
ϕ
2
= 2.47 cm
ϕ
1
=
3
.5
7
cm
( ) (d)-25
-20
-15
-10
-5
9
8.5
8
7.5
7
6.5
6
|s
11
| [dB]
Frequency [GHz]
k = 0
k = 10
k = 19
k = k
conv
-42
-40
-38
-36
-34
-32
-30
-28
9
8.5
8
7.5
7
6.5
6
|s
21
| [dB]
Frequency [GHz]
k = 0
k = 10
k = 19
k = k
conv
(a) (b)-180
-90
0
90
180
6
6.5
7
7.5
8
8.5
9
Arg(S
21
) [deg]
Frequency [GHz]
k = 0
k = 10
k = 19
k = k
conv
0.8
0.7
0.6
0.5
0.4
9
8.5
8
7.5
7
6.5
6
τ
[ns]
Frequency [GHz]
k = 0
k = 10
k = 19
k = k
conv
( ) (d) 4 -L. Lizzi et al. , "A Spline-based Shaping Approa h for Ultra-Wideband ..." 23ϕ
1
ϕ
2
z
ϕ
3
y
x
ϕ
4
P
1
Port
Input
r
ϕ
2
= 1.8
cm
ϕ
1
=
7.
5
cm
ϕ
1
=
7.
2
cm
ϕ
2
= 1.9
cm
(a) (b)ϕ
1
=
7.
2
cm
ϕ
2
= 1.9
cm
ϕ
1
=
7.
1
cm
ϕ
2
= 2.0
cm
( ) (d)-20
-15
-10
-5
0
6
6.5
7
7.5
8
8.5
9
|s
11
| [dB]
Frequency [GHz]
k = 0
k = 100
k = 300
k = K
-50
-45
-40
-35
-30
-25
6
6.5
7
7.5
8
8.5
9
|s
21
| [dB]
Frequency [GHz]
k = 0
k = 100
k = 300
k = K
(a) (b)-180
-90
0
90
180
6
6.5
7
7.5
8
8.5
9
Arg(S
21
) [deg]
Frequency [GHz]
k = 0
k = 100
k = 300
k = K
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
6
6.5
7
7.5
8
8.5
τ
[ns]
Frequency [GHz]
k = 0
k = 100
k = 300
k = K
( ) (d) 7 -L. Lizzi et al. , "A Spline-based Shaping Approa h for Ultra-Wideband ..." 2610
-3
10
-2
10
-1
10
0
10
1
10
2
0
50
100
150
200
250
300
350
400
Ω
[arbitrary unit]
Iteration No., k
Spline-based Approach
Parametric Approach
0
50
100
150
200
250
300
350
400
4
5
6
7
8
9
10
0.4
0.8
1.2
1.6
2
2.4
2.8
k
conv
F
11
, F
21
, F
GD
[x 10
-1
]
Number of Control Points, N
k
conv
F
11
F
21
F
GD
(b)
-35
-30
-25
-20
-15
-10
-5
0
9
8.5
8
7.5
7
6.5
6
5.5
5
4.5
4
|s
11
| [dB]
Frequency [GHz]
Simulated Data
Measured Data
-50
-45
-40
-35
-30
-25
-20
-15
-10
9
8.5
8
7.5
7
6.5
6
5.5
5
4.5
4
|s
21
| [dB]
Frequency [GHz]
Simulated Data
Measured Data
(a) (b)-180
-90
0
90
180
4
4.5
5
5.5
6
6.5
7
7.5
8
8.5
9
Arg(s
21
) [deg]
Frequency [GHz]
Simulated Data
Measured Data
2
1.5
1
0.5
0
-0.5
-1
9
8.5
8
7.5
7
6.5
6
5.5
5
4.5
4
τ
[ns]
Frequency [GHz]
Simulated Data
Measured Data
( ) (d) 11 -L. Lizzi et al. , "A Spline-based Shaping Approa h for Ultra-Wideband ..." 30-30
-25
-20
-15
-10
-5
10
9
8
7
6
5
4
|s
11
| [dB]
Frequency [GHz]
Simulated Data
Measured Data
-50
-45
-40
-35
-30
-25
-20
10
9
8
7
6
5
4
|s
21
| [dB]
Frequency [GHz]
Simulated Data
Measured Data
(a) (b)-180
-90
0
90
180
4
5
6
7
8
9
10
Arg(s
21
) [deg]
Frequency [GHz]
Simulated Data
Measured Data
2
1.5
1
0.5
0
-0.5
-1
10
9
8
7
6
5
4
τ
[ns]
Frequency [GHz]
Simulated Data
Measured Data
( ) (d) 12 -L. Lizzi et al. , "A Spline-based Shaping Approa h for Ultra-Wideband ..." 31(b)
Control Points Coordinates [mm℄
x
1
y
1
x
2
y
2
x
3
y
3
x
4
y
4
5.4 51.6 6.9 57.6 9.0 60.6 7.0 65.1x
5
y
5
x
6
y
6
x
7
y
7
2.7 67.3 2.0 64.5 0.0 56.0 Geometri Variables [mm℄ϕ
1
ϕ
2
ϕ
3
ϕ
4
69.2 10.0 5.4 10.9 Performan e△f [GHz]
△ |s
21
| [dB]
△τ [ns]
5 (4-9) 5 0.5Control Points Coordinates [mm℄