Int.J.Electron.Commun.(AEÜ)66 (2012) 76–79
ContentslistsavailableatScienceDirect
International
Journal
of
Electronics
and
Communications
(AEÜ)
jo u rn al h om e p a g e :w w w . e l s e v i e r . d e / ae u e
Broadband
impedance
matching
via
lossless
unsymmetrical
lattice
networks
Metin
S¸
engül
∗KadirHasUniversityEngineeringFaculty34083Cibali,Fatih-Istanbul,Turkey
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received16September2010 Accepted11May2011 Keywords: Broadbandnetworks LosslessnetworksLatticenetworks,Impedancematching Synthesis
a
b
s
t
r
a
c
t
Inthispaper,abroadbandimpedancematchingnetwork(equalizer)designalgorithmhasbeenproposed.
Intheequalizer,alosslessunsymmetricallatticenetworkhasbeenutilized.Thebranchimpedancesof
thelatticenetworkareconsideredassinglyterminatedlosslessLCnetworks,sinceitisnotdesiredto
dissipatepowerintheequalizer.Aftergivingthealgorithm,itsusagehasbeenillustratedviaanexample.
© 2011 Elsevier GmbH. All rights reserved.
1. Introduction
Designofbroadbandmatchingnetworksisanessentialproblem formicrowavecommunicationsystemengineers[1].Soanalytic theoryofbroadbandmatching[2],[3]andcomputer-aided-design (CAD)methodsareessentialtoolsforthedesigners[4–6].Butit is wellknownthat analytictheory isdifficult toutilizeeven if thesourceandloadimpedancesaresimple.Therefore,itisalways attractivetouseCADtechniques.AlltheCADtechniquesoptimize thematchedsystemperformance.Butperformanceoptimization ishighlynonlinearwithrespecttoelementvaluesandneedsvery goodinitials [7].Asa result,initialelementvalues arevital for successfuloptimization.
In matching network design problems, ladder networks are used,sincethesestructureshaveverylowsensitivity[8].In addi-tiontheyarepreferablesince theyare unbalanced,thusallthe shuntbranchescanbegrounded.Ifnon-minimum-phasetransfer characteristicsaredesired,thenetworkcomplexityincreases.This addedcomplexityistheresultofright-halfplanezerosofatransfer function,whichcanberealizedonlybyasignal-cancellation pro-cess.Thisrequiresmorethanonepathoftransmissionbetween theinputandoutputports.Inladdernetworks,thiscanbe real-izedbyparallelorbridgedstructures.Withoutthecommonground betweentheinputandoutputports,righthalf-planezeroscanbe realizedbyabridgestructureoftheformshowninFig.1.
Therefore, in this paper, two-port bridge structures are employedinmatchingnetworks.Inthefollowingsections, two-portbridgenetworksandrealfrequencybroadbandmatchingwill
∗ Tel.:+902125336532;fax:+902125335753.
E-mailaddresses:msengul@khas.edu.tr,mtnsngl@gmail.com
besummarized.Then,theproposeddesignalgorithmandan exam-plewillbepresented.
2. Bridgenetworks
When a voltage source of frequency ω0 with the
resis-tance RS is applied to the input port of a bridge network
(Fig. 1) and the branch impedances are related such that [Z1(jω0)Z4(jω0)=Z2(jω0)Z3(jω0)],thevoltagedropacrosstheload
impedanceZLiszero,resultinginazerooftransmissionorazero
ofthevoltagetransferratioat(p=jω0).Abridgenetwork
satis-fyingtheseconditionsissaidtobebalanced.Ifoppositearmsof thebridgehavethesameimpedances [ZA(p)=Z1(p)=Z4(p)] and
[ZB(p)=Z2(p)=Z3(p)],thenetworkissaidtobesymmetrical.Ifthe
bridgeleadsaretwisted,theconfigurationseeninFig.2isobtained, whichisknownasaunsymmetricallatticenetwork.Althoughany physicallyrealizabletransferfunctioncanbesynthesizedbyusing unsymmetricalRLClatticenetwork,asymmetricallatticerestricts therangeofrealizablefunctions.Especially,certainphysically real-izabletransfer functionshavingpoles onthejω-axiscannot be synthesizedbyusingsymmetricallatticestructures[9].
3. Realfrequencybroadbandmatching
Letusconsidertheclassicalsinglematchingproblem(resistive sourceimpedanceandcomplexloadimpedance)showninFig.2.
ThematchingconditionsofthecomplexloadZLtotheresistive
generatorRScanbeformulatedintermsofthenormalized
reflec-tioncoefficientsatports1and2.Theinputreflectioncoefficients fNormcanbedefinedby
1=Zin−RS
Zin+RS
(1) 1434-8411/$–seefrontmatter © 2011 Elsevier GmbH. All rights reserved.
M.S¸engül/Int.J.Electron.Commun.(AEÜ)66 (2012) 76–79 77 ) ( 1 p Z Z 3 ( p ) ) ( 2 p Z Z 4 ( p ) ) ( p Z L E + S R
Fig.1.Generaltwo-portbridgenetwork.
whereZin istheinputimpedanceseenatport1whenport2is
terminatedbytheloadZL.Similarlythereflectioncoefficientatport
2canbedefinedby 2= Zout−ZL∗
Zout+ZL
(2) whereZoutistheimpedanceseenatport2whenport1is
termi-natedbythesourceresistanceRS,andtheupperasteriskdenotes
complexconjugation.Here,2isthenormalizedreflection
coeffi-cientatport2.Sincethetwo-portisconsideredaslossless,wehave ontheimaginaryaxisofthecomplexfrequencyplane
1 2=
2 2. (3)
Then,thetransducerpowergain(TPG)atrealfrequenciescan beexpressedas TPG(ω)=1−
1 2 =1−2 2 . (4)Thegoalinbroadbandmatchingistodesignthelossless net-workN,whichconsistsofthearmimpedancesZ1(p),Z2(p),Z3(p)
andZ4(p),suchthatTPGgivenbyEq.(4)ismaximizedinside a
desiredfrequencyband.Obviously,maximizingTPG(ω)meansto minimizingthemodulusofthereflectioncoefficients
1or2.Inthiscontext,thematchingproblemisreducedtothedetermination ofarealizableimpedancefunctionZinorZout.
LettheequalizerinputimpedanceZinbeexpressedintermsof
itsrealandimaginarypartsontherealfrequencyaxisas
Zin(jω)=Rin(ω)+jXin(ω). (5)
ByusingEq.(5),Eq.(4)andEq.(1)weobtaintransducerpower gain(TPG)intermsoftherealandimaginarypartsoftheinput impedanceZinoftheequalizerNandthesourceresistanceRSas
follows:
TPG(ω)= 4RSRin(ω)
(RS+Rin(ω))2+(Xin(ω))2
. (6)
NamelythematchingproblemconsistsoffindingZin(jω)such
thatTPG(ω)ismaximizedinsideadesiredfrequencyband.OnceZin
isdeterminedproperly,theequalizernetworkNcanbesynthesized directlybyusingthe obtainedimpedance orthecorresponding reflectioncoefficient. ) ( 1 p Z ) ( 4 p Z ) ( p Z L S R E + Z 3 ( p ) Z 2 ( p ) in Z , 1 ρ ρ 2 , Z out N
Fig.2.Unsymmetricallatticenetwork.
4. Rationaleofthematchingprocedure
Foralosslesstwo-portliketheonedepictedinFig.2,thecanonic formofthescatteringmatrixisgivenby[10],[11]
S(p)=
S11(p) S12(p) S21(p) S22(p) = g(p)1 h(p) f(−p) f(p) −h(−p) (7) wherep=+jωisthecomplexfrequencyvariable,and=±1 isaunimodular constant.Ifthetwo-portisreciprocal,thenthe polynomialf(p)iseithereven orodd.In thiscase, =+1iff(p) iseven,and =−1iff(p)isodd.Thus, foralossless,reciprocal two-port=f(−p)
f(p) =±1. (8)
Foralosslesstwo-portwithresistivetermination,energy con-versationrequiresthat
S(p)ST(−p)=I, (9)
whereIistheidentitymatrix.TheexplicitformofEq. (9)is knownastheFeldtkellerequationandgivenas
g(p)g(−p)=h(p)h(−p)+f(p)f(−p). (10) In Eqs.(7) and(10),g(p) is a strictlyHurwitz polynomialof nthdegreewithrealcoefficients,andh(p)isapolynomialofnth degreewithrealcoefficients.Thepolynomialfunctionf(p)includes alltransmissionzerosofthetwo-port.
ConsiderthebridgenetworkseeninFig.1orFig.2.Sinceitis notdesiredtodissipateanypowerintheimpedancesZ1(p),Z2(p),
Z3(p)andZ4(p),theymustcontainonlyinductorsandcapacitors.
Alsotheirterminationsmustbeeithershortoropen,notresistive terminations.Sotheseimpedancesaresinglyterminatedlossless LCnetworks[12].
In[12],ithasbeenshownthatforalosslesssinglyterminated network,theinputreflectioncoefficientcanbeexpressedas S11(p)=±
g(−p) g(p) =˛
g(−p)
g(p) , (11)
where ˛=+1 and ˛=− 1 corresponds to an openand short termination,respectively.Sotheprocedureproposedin[12]can beused todesign theimpedancesZ1(p),Z2(p),Z3(p) andZ4(p).
Thenthefollowingalgorithmisproposedtodesignthebroadband impedancematchingnetworkbyusingunsymmetricallattice net-works.
4.1. Algorithm 4.1.1. Inputs
• ωi(actual)=2fi(actual);i=1,2,...,Nω:measurementorcalculation
frequenciesselectedarbitrarily.
• Nω:Totalnumberofmeasurementorcalculationfrequencies.
• ZL(actual)(jωi)=RL(actual)(ωi)+jXL(actual)(ωi);i=1, 2,...,Nω:
mea-suredorcalculatedloadimpedancedataatNωfrequencypoints.
• RS:Givensourceresistance.
• fNorm:Normalizationfrequency.
• R0:Normalizationresistance,usually50.
• nk;k=1,2,3,4:Desirednumberofelementsinthearmsofthe
bridgenetwork.
• ˛k=±1;k=1,2,3,4:Desiredterminationtypeofthearmsofthe
bridgenetwork.
• gk(p);k=1,2,3,4:Initializedpolynomialg(p)describingthearm
impedancesofthebridgenetwork.
• T0:Desiredflattransducerpowergainlevel.
• ı:Thestoppingcriteriaforthesumofthesquareerrors.Formany practicalproblems,itissufficienttochooseı=10−3.
78 M.S¸engül/Int.J.Electron.Commun.(AEÜ)66 (2012) 76–79
4.2. Computationalsteps
Step1:Ifthegivenloadimpedanceandfrequenciesareactual values, not normalized, then normalize the frequencies with respecttofNormandsetallthenormalizedangularfrequencies
ωi=
fi(actual)
fNorm
Normalizetheloadimpedancewithrespecttonormalization resistanceR0overtheentirefrequencybandas
RL= RL(actual) R0 ,XL= XL(actual) R0 .
Itshouldbenotedthat iftheloadis specifiedasadmittance data, then the normalization resistance R0 multiplies the real
andimaginarypartsoftheadmittancedata(i.e.,GL=GL(actual)R0,
BL=BL(actual)R0).
Butifthegiven loadimpedance andfrequencies arealready normalized,thengotothenextstepdirectlywithoutany normal-izationprocess.
Step2:Calculatetheinputimpedancevaluesofthearmsofthe bridgenetworkas Z(k)(jωi)= 1+S(k),11(jωi) 1−S(k),11(jωi) , (12) k=1,2,3,4 and i=1,2,...,Nω
whereS(k),11(jωi)=˛kggkk(−jω(jωi)i) istheinputreflectioncoefficient
ofthearmsofthebridgenetwork.
Step3:Calculatetheinputimpedanceofthebridgenetwork whenport2isterminatedbytheloadimpedanceZLviathe
follow-ingequation(seeFig.2) Zin= N D where N=Z1(Z4ZL+Z3ZL+Z2Z3+Z3Z4+Z2Z4)+Z2Z4ZL+Z2Z3ZL+Z2Z3Z4 and D=Z1(ZL+Z2+Z4)+Z2Z3+Z2ZL+Z4ZL+Z3Z4+Z3ZL.
Theinputimpedanceexpressionhasbeenobtainedbyusing −to−Ytransformationequations[13].
Step4:CalculatetransducerpowergainviaEq.(6)asfollows
TPG(ωi)=
4RSRin(ωi)
(RS+Rin(ωi))2+(Xin(ωi))2
whereRin(ωi)=Real (Zin(jωi)) andXin(ωi)=Imag (Zin(jωi)).
Step5:Calculatethesumofthesquarederrorvia
ıc= Nω
i=i ε(jωi) 2 whereε(jωi)=T0−TPG(ωi). Step6:Ifıc≤ı,synthesizeS(k),11(p)=˛kggk(−p)k(p) andobtainthe
arm networks of the bridge, then stop. Otherwise, change ˛k
(terminationtypes)andgk(p)(initializedpolynomials)viaany
opti-mizationroutineandgotoStep2.
5. Example
Inthissection,anexamplewillbegiventoillustratethe pro-posedalgorithm.Hereallthecalculationswillbemadebyusing
) ( 1 p Z ) ( 4 p Z L R S R E + ( ) 2 p Z ) ( 3 p Z L C L L
Fig.3. Thesourceandloadterminations,RS=1,LL=1,CL=3,RL=1(normalized).
Table1
Calculatedloadimpedancedata(normalized).
ω RL XL 0.1 0.9174 −0.1752 0.2 0.7353 −0.2412 0.3 0.5525 −0.1972 0.4 0.4098 −0.0918 0.5 0.3077 0.0385 0.6 0.2358 0.1755 0.7 0.1848 0.3118 0.8 0.1474 0.4450 0.9 0.1206 0.5743 1.0 0.1000 0.7000
normalizedvalues.Afterdesigningthematchingnetwork,all
com-ponentscanbede-normalizedbyusingthegivennormalization
frequency(fNorm)andresistance(R0).
Thesourceresistanceandtheloadimpedancewhichisselected
asaseriesinductorandaparallelconnectionofacapacitoranda
resistorinnormalizedvaluescanbeseeninFig.3.
Sincethegivensourceandloadterminationshavenormalized elementvalues,thereisnoneedanormalizationprocess.InTable1, thecalculatedloadimpedancevaluesaregiven.
Theselectedinitialcoefficientsofthepolynomials(gk(p)),the
alphaconstants(˛k)andthedesiredflattransducerpowergain
level (T0) are as follows, g1=[ 4 2 3 ], g2=[ 2 4 3 ], g3=
[ 3 5 2 ],g4=[ 1 2 4 ],˛1=+1,˛2=−1,˛3=−1,˛4=−1,and
T0=0.7,respectively.
After running the proposed algorithm, the fol-lowing polynomial coefficients and alpha constants are obtained, g1=[ 6.0437 23.1923 3.1920 ], g2=
[ 6.3061 7.7312 0.2542 ], g3=[ 13.1356 6.4255 0.0907 ],
g4=[ 1.3511 13.3529 12.2343 ], ˛1=+1, ˛2=−1, ˛3=−1,
˛4=−1,respectively.
After synthesizing the corresponding reflection coefficients
S(k),11(p)=˛kggkk(−p)(p)
thebridgenetworkseeninFig.4isreached. TheobtainedtransducerpowergaincurveisgiveninFig.5.
Actualelementvaluescanbeobtainedbyde-normalization.In thiscase,actualelementvaluesaregivenby
ActualCapacitor= (NormalizedCapacitor/2fNorm)
R0 , 1 L L R S R E + L C L 1 C 2 L 3 L 4 L 2 C 3 C 4 C
Fig. 4.Designed matching network, L1=0.2606, C1=7.2658, L2=30.4138,
M.S¸engül/Int.J.Electron.Commun.(AEÜ)66 (2012) 76–79 79
Fig.5.Transducerpowergain.
ActualInductor=
NormalizedInductor 2fNorm R0,ActualResistor=(NormalizedResistor)R0.
Sincethematchingnetwork isdesignedbyusingnormalized values,thecutofffrequency ofthenetworkis ω=1(seeFig.5). Afterde-normalizationprocess,itshiftstothegivennormalization frequency,sincefi(actual)=ωifNorm.
AscanbeseeninFig.5,anearlyflattransducerpowergaincurve isobtainedwithintherequiredfrequencybandatthedesiredflat gainlevel(T0=0.7).
6. Resultsandconclusion
An algorithm has been proposed to design broadband impedancematchingnetworksvialosslessunsymmetricallattice networks.Sinceitisnotdesiredtodissipatepowerintheequalizer, thearmimpedancesofthelatticenetworkareselectedassingly terminatedlosslessLCsections.Inthepaper,singlematching prob-lem(resistivesourceimpedanceandcomplexloadimpedance)has beenconsidered.Butthesameprocedurecanbeusedeasilyfor dou-blematchingproblems(complexsourceimpedanceandcomplex loadimpedance).
Intheexample,thedesiredflattransducerpowergainlevelis selectedas0.7.Ascanbeseenfromthetransducerpowergain graph,anearlyflatgaincurvearoundthislevelhasbeenobtained.
Itisshownthattheproposedmethodgeneratesverygood ini-tialstoimprovethematchedsystemperformancebyoptimizing theelementvalues.Therefore,itisexpectedthattheproposed algo-rithmcanbeusedasafront-endforthecommerciallyavailableCAD toolstodesignbroadbandmatchingnetworksforcommunication systems.
AppendixA. Supplementarydata
Supplementarydataassociatedwiththisarticlecanbefound,in theonlineversion,atdoi:10.1016/j.aeue.2011.05.005.
References
[1] YarmanBS.Broadbandnetworks.WileyEncyclopediaofElectricaland Elec-tronicsEngineering;1999.
[2]YoulaDC.Anewtheoryofbroadbandmatching.IEEETransactionsonCircuit Theory1964;11:30–50.
[3]Fano RM. Theoretical limitations on the broadband matching of arbitrary impedances. Journal of Franklin Institute 1950;249: 57–83.
[4] Awr:Microwaveofficeofappliedwaveresearchinc.www.appwave.com. [5]Edl/ansoftdesignerofansoftcorp.www.ansoft.com/products.cfm. [6]Adsofagilenttechologies.www.home.agilent.com.
[7]YarmanBS,S¸engülM,Kılınc¸A.Designofpracticalmatchingnetworkswith lumped-elementsviamodeling.IEEETransactionsonCircuitsandSystemsI: RegularPapers2007;54(8):1829–37.
[8]ChenWK.PassiveandActiveFilters.NewYork:Wiley;1986.
[9]YengstWC.ProceduresofModernNetworkSynthesis.NewYork:The Macmil-lanCompany;1964.
[10]BelevitchV.ClassicalNetworkTheory.SanFrancisco:HoldenDay;1968. [11]Aksen,A.Designoflosslesstwo-portwithmixed,lumpedanddistributed
elementsforbroadbandmatching.Dissertation.Bochum,Germany:Ruhr Uni-versity,1994.
[12] S¸engül M. Reflectance-based foster impedance data modeling. Frequenz Journal of RF Engineering and Telecommunications 2007;61(7–8): 194–6.
[13] NilssonJW.Electriccircuits.NewYork:Addison-WesleyPublishingCompany; 1993.
MetinS¸engülreceivedB.Sc.andM.Sc.degreesinElectronicsEngineeringfrom
˙IstanbulUniversity,Turkeyin1996and1999,respectively.HecompletedhisPh.D. in2006atIs¸ıkUniversity, ˙Istanbul,Turkey.Heworkedasatechnicianat ˙Istanbul Universityfrom1990to1997.HewasacircuitdesignengineeratR&DLabsof thePrimeMinistryOfficeofTurkeybetween1997and2000.Between2000and 2008,hewasalectureratKadirHasUniversity, ˙Istanbul,Turkey.Dr.S¸engülwas avisitingresearcheratInstituteforInformationTechnology,Technische Univer-sitätIlmenau,Ilmenau,Germanyin2006forsixmonths.Heworkedasanassistant professoratKadirHasUniversitybetween2008and2010.Currentlyheis serv-ingasanassociateprofessoratKadirHasUniveristy.Dr.S¸engülisworkingon microwavematchingnetworks/amplifiers,datamodelingandcircuitdesignvia modeling.