SUITABILITYOFTHERSP2PROSOFTWARE-DEFINEDRADIOFORPRECOMPLIANCE
RADIATEDEMISSIONSTESTING
By
WilliamSteversII
ATHESIS
Submittedto
MichiganStateUniversity
inpartialful˝llmentoftherequirements
forthedegreeof
ElectricalEngineeringMasterofScience
2020
ABSTRACT
SUITABILITYOFTHERSP2PROSOFTWARE-DEFINEDRADIOFORPRECOMPLIANCE
RADIATEDEMISSIONSTESTING
By
WilliamSteversII
Thisthesisdemonstratestheabilitytouseanexistingsoftware-de˝nedradioforaccurateRF
measurementsrelatedtotheprecompliancetestingofradiatedemissions.Thesoftware-de˝ned
radiochosenforthisworkistheSDRplayRadioSpectrumProcessor2Pro(RSP2pro).TheRSP2pro
waschosenbecauseofitsabilitytoprovideabsolutereceivedpowermeasurementcapabilitiesata
pricemuchlowerthanatraditionalspectrumanalyzer.
TheabsolutepowerlevelandfrequencyperformanceoftheRSP2prohasbeencharacterized
bydirectcomparisonwithanelectromagneticcompatibilityanalyzerandtheresultsareshown.A
methodforautomatingmeasurementstakenwiththeRSP2procompanionsoftwareisdescribed.
TheRSP2proisalsousedtoperformprecomplianceradiatedemissionstestswithanelectromag-
neticcompatibilityantennaandtheresultsarediscussed.Throughoutthisworkasampleofthree
RSP2prounitsiscomparedandvariationsinmeasurementsacrossthethreeunitsisexamined.
Acommonvery-highfrequency"rabbittelevisionantennaandanultra-highfrequency
"bowtie"televisionantennaareusedwiththeRSP2protocreateacompleteradiatedemissions
precompliancetestsetup.Aprocedureforcharacterizingtheseantennasanddeterminingtheir
suitabilityforradiatedemissionsprecompliancetestingisalsodescribed.TheRSP2protestsetup
formeasuringradiatedemissionsiscomparedtoatraditionalprecompliancetestsetupconsistingof
anelectromagneticcompatibilityspectrumanalyzerandanelectromagneticcompatibilityantenna.
Thiscomparisonincludesadiscussionoftherelativecost,measurementaccuracy,andoverall
convenienceofusingeachtestsetup.
Copyrightby
WILLIAMSTEVERSII
2020
ForUncleMike
iv
ACKNOWLEDGMENTS
Thisthesisistheresultofseveralyearsofresearchandwouldnothavebeenpossiblewithoutthe
supportofsomanyofmycolleaguesandfriends.IamsorrythatIcouldnotmentioneveryone
here,butIhavetokeepthisbrief!
First,IwouldliketothankScottSpryandthewonderfulteachersattheUticaCenterforScience
andIndustry.CSIiswhereIwas˝rstintroducedtoelectricalengineeringandIhavebeenfascinated
witheveryaspectofthis˝eldeversince.TheinstructionandguidanceIreceivedatCSIduring
thattimeinmylifelaidthefoundationfortheengineerIamtoday.
IwouldalsoliketothankBrettLarimoreandtherestoftheC2000teamformakingmy
experiencesatTImemorable.Brettwasagreatmentor,andIwastaskedwithinterestingprojects
thattaughtmehowtobeanelectricalengineer.IshouldalsomentionthatIreallymisstheweather
downintheHoustonarea(eventhoughIonlyeverexperienceditintheSummer).
SpecialthankstoAnton,Michael,andalloftheotherstudentsfromtheElectromagnetics
ResearchGrouphereatMSU.Youallmademefeelwelcomemy˝rstsemesterandsupportedme
throughmylastsemester,andIamsoverythankfulforthat.
Thankyoutomyparentsfortheirsupportandforenablingmetopursuemyinterestsfroma
youngage.Also,aquickshoutouttomysisterOlivia,forkeepingmesaneoverthelastfewmonths,
andtomybestfriendMatt,whoinspiredmetoseethisthingthroughtotheend.Thankyouboth
somuch.
IwouldliketothankmycommitteeandtheotherprofessorsintheMSUECEdepartment
fortheircollectiveinstructionthroughoutmytimeatMSU.Mostofall,IwouldliketothankDr.
Rothwellforhisfriendship,guidance,andpatienceoverthelastseveralyears.Thisresearchstarted
withhisassertionthatanyonecouldwriteathesisiftheywantedto.AlthoughIdoubtedhimatthe
time,thisdocumentisproofthathewasright!
v
TABLEOFCONTENTS
LISTOFTABLES
.......................................
vii
LISTOFFIGURES
.......................................
viii
KEYTOSYMBOLSANDABBREVIATIONS
........................
xvi
CHAPTER1INTRODUCTION
...............................
1
CHAPTER2BACKGROUND
................................
2
2.1ElectromagneticCompatibilityandRadiatedEmissions...............2
2.2EMCRegulationsandPrecomplianceTesting....................11
2.3Software-de˝nedRadio(SDR)............................21
2.4TheRadioSpectrumProcessor2Pro(RSP2pro)...................27
CHAPTER3PERFORMANCEOFTHERSP2PRO
....................
33
3.1EquipmentInformation................................33
3.1.1AgilentE7401AEMCAnalyzer.......................34
3.1.2EMCAntennas................................36
3.1.3HP8657ASignalGenerator.........................42
3.2CharacterizationProcedureandMeasurementAutomation.............50
3.3SignalLevelMeasurementPerformance.......................60
3.4FrequencyMeasurementPerformance........................89
CHAPTER4USINGTHERSP2PROFORRADIATEDEMISSIONSTESTING
.....
98
4.1RadiatedEmissionsMeasurementProcedure.....................98
4.1.1NarrowbandSource:WhipAntenna.....................105
4.1.2BroadbandSource:CustomEquipmentUnderTest(EUT).........109
4.2NarrowbandResults..................................122
4.3BroadbandResults..................................125
CHAPTER5ACOMPLETEPRECOMPLIANCETESTSETUP
.............
131
5.1AntennaSelection,Modi˝cations,andProcedure..................131
5.2VHF"RabbitEar"AntennaPerformance.......................138
5.3UHF"BowTie"AntennaPerformance........................145
5.4ComparisontoaMoreTraditionalTestSetup....................150
CHAPTER6CONCLUSIONS
................................
159
APPENDICES
.........................................
160
APPENDIXAEUTCIRCUITDIAGRAMSANDBILLOFMATERIALS
.....
161
APPENDIXBDIGITALAPPENDIXCONTENTSLIST
..............
163
BIBLIOGRAPHY
........................................
165
vi
LISTOFTABLES
Table2.1:TheadvantagesoffourSDRdigitalsignalprocessingarchitectureswithexamples.23
Table2.2:TheRSP2prohardwarespeci˝cationsthatarerelevanttothisthesis.All
informationtakenfrom[42]and[41]........................30
Table3.1:ParametersoftheEMCanalyzerstate˝lesthatwereusedtocontrolthesettings
oftheEMCanalyzer.................................35
Table3.2:DescriptionoftheninedatasetsthatwereusedtocharacterizetheEMC
analyzerandtheRSP2pro..............................51
Table3.3:SummaryoftheSDRunosettingsusedforalloftheRSP2promeasurements
presentedinthisthesis................................56
Table4.1:InputdatapointsusedbytheCWradiatedemissionssourceshowninFigure4.6.105
Table4.2:DetailsforthefrequenciesthatwerecreatedbythecustomEUTtotestradiated
emissionsacrosstheCISPR-22ClassBfrequencyband..............112
Table4.3:Theconversionofthevoltagesmeasuredbythecurrentprobeintothepredicted
maximumradiatedemissions,
j
ˆ
2
<0G
j
,usingEquations4.1and4.2with
distancesof10mand1m..............................121
Table4.4:MeasuredradiatedemissionsfromthecustomEUTat10mcomparedtothe
predictedradiatedemissions.............................130
Table4.5:MeasuredradiatedemissionsfromthecustomEUTat1mcomparedtothe
predictedradiatedemissions.............................130
Table5.1:ComparisonoftheradiatedemissionsmeasuredfromtheEUTshowninFigure
5.18withseveraldi˙erenttestsetups........................156
Table5.2:ComparisonoftheradiatedemissionsmeasuredfromtheEUTshowninFigure
5.18withseveraldi˙erenttestsetups........................156
vii
LISTOFFIGURES
Figure2.1:ThefourmainEMCissuesthatcane˙ectanelectronicproduct..........3
Figure2.2:AnexampleofacurrentloopcreatedbyasignalonaPCB............4
Figure2.3:Theelectric˝eldscreatedbythedi˙erential-modecurrentalongparallel
conductors......................................6
Figure2.4:Theelectric˝eldscreatedbythecommon-modecurrentalongparallelconductors.7
Figure2.5:AnexampleofaLEDsignthatcreatesEMI....................8
Figure2.6:(a)RadiatedemissionscreatedbytheLEDsignfromFigure2.5and(b)a
controlsamplemeasured1kmawayfromtheLEDsign..............9
Figure2.7:Radiatedemissionlimitsmeasuredatadistanceof10metersfortheEuropean
Union(solid)andtheUnitedStates(dashed)....................12
Figure2.8:Theantennafactorcanbeusedtoeasilyconvertameasuredvoltageback
intotheelectric˝eldstrength,whichcanthenbedirectlycomparedtothe
regulatorylimits..................................13
Figure2.9:Thedi˙erencebetweenthedisplayedoutputtraceofregularpeakdetection
andquasi-peak(QP)detectionwhenmeasuringaspurioussignalwithalow
annoyancefactor[17]................................15
Figure2.10:Atraditionalprecomplianceradiatedemissionsmeasurementtestsetupusing
anEMCanalyzerandashieldedRFchamber...................17
Figure2.11:Alow-costprecompliancetestsetupformeasuringradiatedemissionsinan
uncontrolledRFenvironmentsuchasahallwayoro˚cespace[25].......18
Figure2.12:Blockdiagramofatypicalsuperheterodynespectrumanalyzerasdescribed
in[26]........................................19
Figure2.13:Generalizedblockdiagramforadigitallow-IFSDRreceiverasdescribedin[32].22
Figure2.14:Therelationshipsbetweendevelopmente˙ort,cost,powerconsumption,and
datathroughputforthemostcommonSDRdigitalsignalprocessingarchi-
tectures(adaptedfrom[29])............................23
viii
Figure2.15:TheB200minifromEttusResearch
Š
isacompactFPGA-basedSDRfron-
tendwithtransmitandreceivecapabilitiesoverawidefrequencyrange.....25
Figure2.16:Severalsoftware-centricSDRsthatwereconsideredforthisthesis.The
hardwarecharacteristicsshownwereretrievedfromthemanufacturerwebsites
([35][36][37][38]).................................26
Figure2.17:TheRSP2profromSDRplayisasoftware-centricSDRreceiverwithmany
uniqueRFsignalconditioningcapabilities.....................27
Figure2.18:Asimpli˝edblockdiagramshowingthemaincomponentsoftheRSP2pro
SDR(adaptedfrom[39]and[40]).........................28
Figure2.19:InternalviewoftheRSP2pro;the10bandselectableRF˝lternetworkcan
beseenonthelower-centerofthePCB,whiletheRFtunerICcanbeseenon
thelower-right....................................28
Figure2.20:TheSDRunouserinterface............................31
Figure3.1:TheAgilentE7104AEMCanalyzerthatwasusedforallmeasurementcom-
parisonswiththeRSP2pro.............................34
Figure3.2:(a)TheAgilent11955Aand(b)11956AEMCantennasand(c)theirantenna
factors[44].....................................37
Figure3.3:TheBicoLOG30100XEMCantenna.......................39
Figure3.4:(a)TheBicoLOG30100Xantennafactorand(b)thegainofitspreampli˝er.
Theantennafactorshownin(a)isnegativebecauseofthelargegainofthe
preampli˝ershownin(b)[47]...........................40
Figure3.5:TheHP8657Asignalgenerator...........................42
Figure3.6:FlowchartforthePythonscriptthatwasusedto˝ndthefrequencyo˙setsof
theuncalibratedHP8657Asignalgenerator....................43
Figure3.7:Plotofthe
5
>
associatedwitheachvalueof
5
8=
fromEquation3.1........45
Figure3.8:FlowchartforthePythonscriptthatwasusedtocharacterizethefrequency
driftofthesignalgeneratorovera24-hourperiod.................46
Figure3.9:Measuredsignalgeneratorfrequencydriftfor
C
=
0
to
C
=
12
hours........47
Figure3.10:Measuredsignalgeneratorfrequencydriftfor
C
=
13
to
C
=
24
hours.......48
ix
Figure3.11:Theoutputofthesignalgeneratorwithafrequencysettingof500MHzand
asignallevelof50dB
`
V..............................49
Figure3.12:DetailedviewofthesignalinFigure3.11showinghownarrowbandthesignal
generatoroutputis.................................49
Figure3.13:ThecharacterizationtestsetupconsistingoftheEMCanalyzer,signalgener-
ator,RSP2pro,andlaptop.Herethesignalgeneratorisshownconnectedto
theRSP2pro.....................................52
Figure3.14:DiagramshowingtheRFsignalchainsusedforthecharacterizationtestsetup..52
Figure3.15:Flowchartforthesemi-automatedprocedurethatwasusedtocollecttheEMC
analyzercharacterizationdata...........................54
Figure3.16:Flowchartforthesemi-automatedprocedurethatwasusedtocharacterizethe
RSP2prounits....................................57
Figure3.17:TheRFgainoptimizerthatispartoftheAutoItscriptoutlinedinFigure3.16.
AnoutlineoftheAutoItcodeisshownontheleftandtheSDRunowindow
withtheRFgaincontrolandADCoverloadindicatorisshownontheright...59
Figure3.18:(a)TheVSWRand(b)theinsertionlossoftheSMAcablethatwasusedfor
allofthemeasurementspresentedinthisthesis..................65
Figure3.19:AverageEMCanalyzervoltagemeasurementfor:(a)DataSets1,2,and3,(b)
DataSets4,5,and6,and(c)DataSets7,8and9.Eachdatapointrepresents
theaverageof50measurements..........................66
Figure3.20:RSP2proandEMCanalyzervoltagemeasurementstandarddeviationfor:(a)
DataSet1,(b)DataSet2,and(c)DataSet3.Eachdatapointrepresentsthe
standarddeviationfor50measurements......................68
Figure3.21:RSP2proandEMCanalyzervoltagemeasurementstandarddeviationfor:(a)
DataSet4,(b)DataSet5,and(c)DataSet6.Eachdatapointrepresentsthe
standarddeviationfor50measurements......................69
Figure3.22:RSP2proandEMCanalyzervoltagemeasurementstandarddeviationfor:(a)
DataSet7,(b)DataSet8,and(c)DataSet9.Eachdatapointrepresentsthe
standarddeviationfor50measurements......................70
Figure3.23:AverageRSP2provoltagemeasuremento˙setfor:(a)DataSet1,(b)DataSet
2,and(c)DataSet3.Eachdatapointrepresentstheaverageof50measurements.73
x
Figure3.24:AverageRSP2provoltagemeasuremento˙setfor:(a)DataSet4,(b)DataSet
5,and(c)DataSet6.Eachdatapointrepresentstheaverageof50measurements.74
Figure3.25:AverageRSP2provoltagemeasuremento˙setfor:(a)DataSet7,(b)DataSet
8,and(c)DataSet9.Eachdatapointrepresentstheaverageof50measurements.75
Figure3.26:RSP2proSNRmeasurementstandarddeviationfor:(a)DataSet1,(b)Data
Set2,and(c)DataSet3.Eachdatapointrepresentsthestandarddeviation
for50measurements................................79
Figure3.27:RSP2proSNRmeasurementstandarddeviationfor:(a)DataSet4,(b)Data
Set5,and(c)DataSet6.Eachdatapointrepresentsthestandarddeviation
for50measurements................................80
Figure3.28:RSP2proSNRmeasurementstandarddeviationfor:(a)DataSet7,(b)Data
Set8,and(c)DataSet9.Eachdatapointrepresentsthestandarddeviation
for50measurements................................81
Figure3.29:ThevalueofEquation3.15using(a)DataSet1and2,(b)DataSet4and5,
and(c)DataSet7and8..............................83
Figure3.30:ThevalueofEquation3.15using(a)DataSet2and3,(b)DataSet5and6,
and(c)DataSet8and9..............................84
Figure3.31:RSP2proSNRmeasurementfor:(a)DataSet1,(b)DataSet2,and(c)Data
Set3.Eachdatapointrepresentstheaverageof50measurements........85
Figure3.32:RSP2proSNRmeasurementfor:(a)DataSet4,(b)DataSet5,and(c)Data
Set6.Eachdatapointrepresentstheaverageof50measurements........86
Figure3.33:RSP2proSNRmeasurementfor:(a)DataSet7,(b)DataSet8,and(c)Data
Set9.Eachdatapointrepresentstheaverageof50measurements........87
Figure3.34:RSP2proandEMCanalyzerfrequencymeasurementstandarddeviationfor:
(a)DataSet1,(b)DataSet2,and(c)DataSet3.Eachdatapointrepresents
thestandarddeviationfor50measurements....................90
Figure3.35:RSP2proandEMCanalyzerfrequencymeasurementstandarddeviationfor:
(a)DataSet4,(b)DataSet5,and(c)DataSet6.Eachdatapointrepresents
thestandarddeviationfor50measurements....................91
Figure3.36:RSP2proandEMCanalyzerfrequencymeasurementstandarddeviationfor:
(a)DataSet7,(b)DataSet8,and(c)DataSet9.Eachdatapointrepresents
thestandarddeviationfor50measurements....................92
xi
Figure3.37:AverageRSP2profrequencymeasuremento˙setfor:(a)DataSet1,(b)Data
Set2,and(c)DataSet3.Eachdatapointrepresentstheaverageof50
measurements....................................94
Figure3.38:AverageRSP2profrequencymeasuremento˙setfor:(a)DataSet4,(b)Data
Set5,and(c)DataSet6.Eachdatapointrepresentstheaverageof50
measurements....................................95
Figure3.39:AverageRSP2profrequencymeasuremento˙setfor:(a)DataSet7,(b)Data
Set8,and(c)DataSet9.Eachdatapointrepresentstheaverageof50
measurements....................................96
Figure4.1:(a)Theoutdoorareathatwasconsideredforradiatedemissiontestingand(b)
theambientRFsignalsmeasuredattheoutdoorareafrom30MHzto1000
MHz(averageof3trials)..............................100
Figure4.2:(a)Theundergroundtunnelthatwasconsideredforradiatedemissiontesting
and(b)theambientRFsignalsmeasuredintheundergroundtunnelfrom30
MHzto1000MHz(averageof3trials)......................101
Figure4.3:(a)Theindoorhallwaythatwasultimatelychosenforradiatedemissiontesting
and(b)theambientRFsignalsmeasuredinthehallwayfrom30MHzto1000
MHz(averageof3trials)..............................103
Figure4.4:TheradiatedemissionstestsetupconsistingoftheEMCanalyzer,BicoLOG
30100XEMCantenna,RSP2pro,andradiatedemissionssourceat(a)10m
and(b)1m.....................................104
Figure4.5:DiagramshowingtheRFsignalchainsusedfortheradiatedemissionstestsetup.104
Figure4.6:DetailedviewofthewhipantennaarrangementusedtosimulateCWnarrow-
bandradiatedemissionsfromanEUT.......................106
Figure4.7:(a)ThewhipantennausedtocreateCWnarrowbandradiatedemissions
attachedtoametalsurfaceand(b)itsmeasuredVSWRfrom30MHzto1000
MHz(averageof5measurements).........................107
Figure4.8:(a)ThewhipantennausedtocreateCWnarrowbandradiatedemissionsona
non-metallicsurfaceand(b)itsmeasuredVSWRfrom30MHzto1000MHz
(averageof5measurements)............................108
Figure4.9:Simpli˝edblockdiagramofthecustomEUTconstructedtocreatewideband
radiatedemissionsatmultiplefrequencies.RefertoAppendixAforthe
customEUTcircuitdiagrams............................111
xii
Figure4.10:DetailedviewofthesourcesideshowninFigure4.9...............111
Figure4.11:(a)Thetestsetupusedtomeasurethetime-domaincharacteristicsofthe
customEUTand(b)thetestsetupwiththecurrentprobethatwasusedto
measurethefrequency-domaincharacteristicsofthecustomEUT.........113
Figure4.12:TimeandfrequencycharacteristicsoftheEUTwhensettothebasefrequency
5
=
40
MHz.Thesigni˝cantharmonicsof
5
havebeenmarkedwithan"x"
andtheirvaluesareshown.............................115
Figure4.13:TimeandfrequencycharacteristicsoftheEUTwhensettothebasefrequency
5
=
50
MHz.Thesigni˝cantharmonicsof
5
havebeenmarkedwithan"x"
andtheirvaluesareshown.............................116
Figure4.14:TimeandfrequencycharacteristicsoftheEUTwhensettothebasefrequency
5
=
120
MHz.Thesigni˝cantharmonicsof
5
havebeenmarkedwithan"x"
andtheirvaluesareshown.............................117
Figure4.15:TimeandfrequencycharacteristicsoftheEUTwhensettothebasefrequency
5
=
160
MHz.Thesigni˝cantharmonicsof
5
havebeenmarkedwithan"x"
andtheirvaluesareshown.............................118
Figure4.16:DetailedviewofhowthecustomEUTwasarrangedduringbroadbandradi-
atedemissiontests.................................121
Figure4.17:AverageRSP2provoltagemeasuremento˙setfromEquation3.13forthe
inputdatapointsshowninTable4.1measuredatadistanceof10m.Each
datapointrepresentstheaverageofthreemeasurements..............123
Figure4.18:AverageRSP2provoltagemeasuremento˙setfromEquation3.13forthe
inputdatapointsshowninTable4.1measuredatadistanceof1m.Eachdata
pointrepresentstheaverageofthreemeasurements................124
Figure4.19:AverageRSP2provoltagemeasuremento˙setfromEquation3.13forthe
customEUTradiatedemissionfrequenciesshowninTable4.3measuredata
distanceof10m.Eachdatapointrepresentstheaverageofthreemeasurements.127
Figure4.20:AverageRSP2provoltagemeasuremento˙setfromEquation3.13forthe
customEUTradiatedemissionfrequenciesshowninTable4.3measuredata
distanceof1m.Eachdatapointrepresentstheaverageofthreemeasurements..128
Figure5.1:(a)TheVHF"rabbitear"televisionantennaand(b)theUHF"bowtie"tele-
visionantenna....................................132
Figure5.2:CADmodelsfortheantennaholdersthatareshowninFigure5.1........133
xiii
Figure5.3:(a)Theantennaholder'sdualattachmentpointsand(b)theantennaholder
withbothcameratripodadaptersattached.....................133
Figure5.4:CADmodeloftheantennaclipthatwasusedto˝xtheangularpositionof
rabbitearantennaelementsto180
°
.........................134
Figure5.5:(a)Two4:1televisonbaluns(theoneontheleftwasused),(b)internal
componentsofthebalun,(c)balunwiththescrew-terminalsremoved,(d)the
˝nalmodi˝ed4:1balun,and(e)theconnectionbetweenthemodi˝edbowtie
antennaandthemodi˝edbalun...........................136
Figure5.6:(a)Removalofthemountingpost,(b)theexposedantennaelementconnec-
tors,and(c)the˝nalmodi˝edrabbitearantenna.................137
Figure5.7:DiagramshowingtheRFsignalchainsusedduringthetelevisionantenna
comparisontestsetup................................138
Figure5.8:MeasuredVSWRoftheVHFrabbitearantennawith(a)
ˆ!
=0.375m,(b)
ˆ!
=0.667m,and(c)
ˆ!
=0.991m.Eachtraceistheaverageof˝vetrials...139
Figure5.9:TheamountofpotentialmismatcherrorwiththeEMCanalyzer
"ˆ
ˆ
forall
threeoftherabbirearantennaelementlengths
ˆ!
thatweretested........140
Figure5.10:RadiatedemissionsfromthecustomEUTsetto40MHzmeasuredbyusing
(a)theBicoLOG30100XEMCantennaandbyusingtherabbitearantenna
with(b)
ˆ!
=0.375m,(c)
ˆ!
=0.667m,and(d)
ˆ!
=0.991m.........142
Figure5.11:RadiatedemissionsfromthecustomEUTsetto50MHzmeasuredbyusing
(a)theBicoLOG30100XEMCantennaandbyusingtherabbitearantenna
with(b)
ˆ!
=0.375m,(c)
ˆ!
=0.667m,and(d)
ˆ!
=0.991m.........143
Figure5.12:RadiatedemissionsfromthecustomEUTsetto120MHzmeasuredbyusing
(a)theBicoLOG30100XEMCantennaandbyusingtherabbitearantenna
with(b)
ˆ!
=0.375m,(c)
ˆ!
=0.667m,and(d)
ˆ!
=0.991m.........144
Figure5.13:MeasuredVSWRoftheUHFbowtieantennain(a)thehorizontalorientation
and(b)theverticalorientation.Eachtraceistheaverageof˝vetrials......146
Figure5.14:TheamountofpotentialmismatcherrorwiththeEMCanalyzer
"ˆ
ˆ
forthe
horizontalandverticalorientationsofthebowtieantenna.............147
Figure5.15:RadiatedemissionsfromthecustomEUTsetto120MHzmeasuredbyusing
(a)theBicoLOG30100XEMCantennaand(b)thebowtieantennainthe
horizontalorientation................................148
xiv
Figure5.16:RadiatedemissionsfromthecustomEUTsetto160MHzmeasuredbyusing
(a)theBicoLOG30100XEMCantennaand(b)thebowtieantennainthe
horizontalorientation................................149
Figure5.17:(a)Thetraditionalprecompliancetestsetupand(b)thelow-costtestsetup....150
Figure5.18:DetailedviewoftheEUTshowingthegold5

resistiveload..........151
Figure5.19:DiagramshowingtheRFsignalchainsof(a)thetraditionaltestsetupand(b)
thelow-costtestsetup................................151
Figure5.20:(a)ThebackgroundmeasurementwiththeEUTo˙and(b)theradiated
emissionsfromtheEUTnear41MHz.......................152
Figure5.21:(a)ThebackgroundmeasurementwiththeEUTo˙and(b)theradiated
emissionsfromtheEUTnear41MHz.......................153
Figure5.22:(a)ThebackgroundmeasurementwiththeEUTo˙and(b)theradiated
emissionsfromtheEUTnear50MHz.......................154
Figure5.23:(a)ThebackgroundmeasurementwiththeEUTo˙and(b)theradiated
emissionsfromtheEUTnear50MHz.......................155
FigureA.1:ElectricalcircuitforthesourcesideofthecustomEUTfromSection4.1.2....161
FigureA.2:ElectricalcircuitfortheloadsideofthecustomEUTfromSection4.1.2.....162
xv
KEYTOSYMBOLSANDABBREVIATIONS
ADC
GPIB
CW
SNR
TCXO
ASIC
FPGA
DSP
GPP
IQ
A/D
RBW
IF
OATS
QP
CISPR
EU
FCC
dBFS
LED
EMI
EUT
EMC
SDR
RSP2pro
analog-to-digitalconverter
general-purposeinterfacebus
continuouswave
signal-to-noiseratio
temperature-compensatedoscillator
application-speci˝cintegratedcircuit
˝eld-programmablegatearray
digitalsignalprocessor
generalpurposeprocessor
in-phaseandquadrature
analog-to-digital
resolutionbandwidth
intermediatefrequency
openareatestsite
quasi-peak
ComitéInternationalSpécialdesPerturbationsRadio
EuropeanUnion
FederalCommunicationsCommission
dBfull-scale
light-emittingdiode
electromagneticinterference
equipmentundertest
electromagneticcompatibility
software-de˝nedradio
RadioSpectrumProcessor2Pro
xvi
CHAPTER1
INTRODUCTION
ThisthesisinvestigatesusingtheSDRplayRadioSpectrumProcessor2Pro(RSP2pro)software-
de˝nedradio(SDR)toperformaccurateRFmeasurementsrelatedtotheprecompliancetestingof
radiatedemissions.
Chapter2givesanoverviewofelectromagneticcompatibility(EMC)withafocusonradiated
emissionsmeasurements,regulations,andprecompliancetechniques.ThebasicsofSDRimple-
mentationandtypicalSDRimplementationsarealsodescribed.Themethodologyforselectingthe
RSP2proforthisresearchisexplained.AdescriptionoftheRSP2proanditscompanionsoftware,
SDRuno,isprovided.
Chapter3assessesthesuitabilityoftheRSP2proforprecomplianceradiatedemissionstestingby
directlycomparingitsmeasurementcapabilitiestoanEMCanalyzer.Theprocessforcharacterizing
boththeabsolutepowerandfrequencyperformanceoftheRSP2proisexplainedandtheresults
areexamined.AmethodforautomatingmeasurementswithSDRunoisalsodescribed.
TheRSP2proisusedforprecomplianceradiatedemissionstestsinChapter4.Adetailed
descriptionofthecustomcircuitservingastheequipmentundertest(EUT)isgiven.Radiated
emissionsmeasurementsusingtheRSP2proandanEMCantennaarecomparedwithmeasurements
usinganEMCanalyzerandthesameEMCantenna.
Chapter5describestheuseoftwocommontelevisionantennasforradiatedemissionsmea-
surementsintheRFrangeof30MHzto1000MHz.Adetaileddescriptionofeachantennais
givenandtheresultsobtainedusingtheseantennasarecomparedtoresultsobtainedwithanEMC
antenna.TheutilityoftheRSP2protestsetupformeasuringradiatedemissionsisthencompared
toatraditionalprecompliancetestsetupconsistingofanEMCanalyzerandanEMCantenna.
Chapter6istheconclusionofthisthesisandincludesbothasummaryoftheresultsanda
proposalforfutureworkrelatedtotheuseoftheRSP2proformoreaccurateradiatedemissions
measurements.
1
CHAPTER2
BACKGROUND
Anelectronicproductisanyelectronicsystemthatisintendedforusebyanend-userinadomestic
orindustrialenvironment.Section2.1summarizestheimportanceofEMCandoneofitsmain
areas,electromagneticinterference(EMI)intheformoftheradiatedemissionscreatedbyan
electronicproduct.Governmentalregulationsrelatedtoradiatedemissionsandthedi˙erences
betweeno˚cialEMCtestingproceduresandprecompliancetestingproceduresareexplainedin
Section2.2.Thisincludesadiscussionaboutexisting"low-cost"precompliancetechniques.An
introductiontoSDRtechnologyisgiveninSection2.3.Finally,adescriptionoftheRSP2proand
itscompanionsoftware,SDRuno,isgivenSection2.4.
2.1ElectromagneticCompatibilityandRadiatedEmissions
Anelectronicproductmeetselectromagneticcompatibility(EMC)requirementswhenitdoes
notinterferewithitssurroundingRFenvironmentwhilealsobeingabletoacceptelectronic
interferencewithoutlosingproperfunctionality[1].TheformerdescribesminimizingtheEMI
createdbytheproductitselfwhilethelatterdescribesensuringthattheproductisimmunetoEMI.
AccordingtoPaul[1],alloftheEMCproblemsrelatedtoanelectronicproductcanbesummarized
intothefourgeneralizedcasesshowninFigure2.1.Conductedemissionsandsusceptibility
describeEMIthatisgenerated(emissions)orreceived(susceptibility/immunity)bytheproduct
throughitspowercord,whileradiatedemissionsandsusceptibilitydescribeEMIthatisgenerated
orreceivedbytheproductintheformofRFradiation.Designingforimmunityisgenerallybuilt
intothedevelopmentofanelectronicproductbecauseitisapparenttothedesigners(andthe
end-user)whenaproductisnotoperatingcorrectly[2].Thissame"self-regulation"doesnotapply
totheEMIcreatedbytheproduct,astheproductmaynotinterferewithitselfandwilloperate
normally.EmissionsfromaproductarethereforemorelikelytogounnoticeduntilanEMCissue
isdiscoveredmuchlaterinthedesignprocess.Theinsidiousnatureofradiatedandconducted
2
emissions,combinedwiththeneedforanenvironmentasfreefromEMIaspossible,iswhymany
agenciesaroundtheworldsetcompliancestandardsfortheacceptableamountofEMIcreatedby
anelectronicproduct.Thesecompliancestandardsmustbemetbeforeaproductcanbeplaced
onthemarket.Whileconductedemissionsareanimportantpartofthecomplianceprocess,the
techniquesandproceduresinthisthesiswillbefocusedsolelyonradiatedemissions.Thisis
becauseradiatedemissionsarebothpoorlyunderstoodbymostelectronicproductdesignersand
arethemostdi˚cultpartofpassingano˚cialcompliancetest[3].
Figure2.1:ThefourmainEMCissuesthatcane˙ectanelectronicproduct.
Radiatedemissionshaveaveryspeci˝cde˝nitionwithinEMCandarenottobeconfused
withintentionalRFsignalssuchasthosefromaradiostationorwirelessinternetrouter.While
intentionalradiatorscancauseEMIissues,radiatedemissionsinanEMCcontextmostcommonly
referstotheunintentionalRFsignalsthatoriginatefromthedigitalcircuitswithinanelectronic
product[1].Whenmeasuringradiatedemissionsforcomparisontoregulatorylimits,thestrength
oftheelectric˝eldatagivendistancefromtheelectronicproductismeasured.Theunitsofthis
measurementareindB
`
V/m(decibelsrelativetoonemicrovolt,permeter).Beforereviewingthe
3
compliancestandardsandmeasurementpracticesrelatedtoradiatedemissions,itisimportantto
˝rstunderstandwherethesetypesofemissionscomefrom.
Anymetalliccomponentthatiscapableofcarryingelectriccurrenthasthepotentialtobecome
aradiatingstructurethatwillcreateEMI[1].CommonexamplesofsuchcomponentsincludePCB
traces,ribboncables,andtheproductenclosure.Whenthesecomponentsbecomelargeenough
relativetothewavelengthofthecurrenttheyarecarrying,thecomponentscanproduceradiated
emissions[4].Mostofthecurrentinaconductortravelsinaclosedloopliketheoneshownin
Figure2.2.Herethetotalcurrentthatoriginatesfromtheintegratedcircuitpincircledinmagenta
Figure2.2:AnexampleofacurrentloopcreatedbyasignalonaPCB.
createsaloopconsistingofthepinkpaththatgoestotheyellowLEDandtheorangereturnpath
throughthegroundcopper˝llregionshowninlightorange.Thiscurrentloopexistsbecausea
predominantportionofthecurrentcontinuesalongthepathofleastimpedancewithintheground
regiontothepointfromwhichitoriginated[5].Thistypeofcurrentiscalledadi˙erential-mode
currentbecauseitisreturningtoitssource,creatingaclosedloop.TheloopshowninFigure2.2is
ratherlargebecausetheelectronicproductshownusesatwo-layerPCB.Thisloopcouldbemade
4
considerablysmalleriftherewasaseparategroundlayer,butthisaddsmanufacturingcostand
increasesthecomplexityofthedesign.Notethattheloophereisonlybeingusedasanexampleof
howthecurrenttravelsinthecircuit;theactualswitchingcurrentonthepathshowniswellbelow
50kHzandisnotaconcernforcreatingradiatedemissions[5].Thisiswhytwosignallayersare
su˚cientforthisPCB.
APCBtraceaboveagroundplaneandtheindividualwiresinaribboncablearebothexamplesof
parallelconductorsthatarecommonlyfoundinelectronicproducts.Therearetwotypesofelectric
currentthatexistalongparallelconductorsthatcontributetoradiatedemissions:di˙erential-mode
currentandcommon-modecurrent[6].Di˙erential-modecurrentsareeasilydescribedwithcircuit
theoryandexistinequalbutoppositedirectionsalongparallelconductors(recallthecurrentloop
showninFigure2.2).Thesecurrentsaretheidealsignalsbeingcarriedbytheconductorsina
circuit[7].
Common-modecurrentsaremoredi˚culttopredictbecausetheyexistinthesamedirection
alongparallelconductorsandarecreatedbytheparasiticimpedanceswithintheconductorsin
thecircuit[2].Whenacableisconnectedtoadigitalcircuit,acommon-modecurrentcanexist
alongthecableandwillcauseittobehavelikeanantennathate˙ectivelyradiateselectric˝elds.
Common-modecurrentsdonotonlyexistinthecablesconnectedtoanelectronicproduct.For
example,common-modecurrentsareamajorsourceofnoiseinmicrocomputercircuitsthatdonot
useexternalmemory[5].Specialattentionisgiventothecablingofelectronicproductsbecause
theycaneasilycausethecommon-modecurrentsfromwithintheproducttoescapeintotheRF
environment.Thisisespeciallytrueifthecableconnectorsareeitherpoorlybondedtotheproduct's
metallicenclosureoriftheproductusesanonmetallicenclosure[8].
Thetotalcurrentinaconductorcanberepresentedasthesumofthedi˙erentialandcommon-
modeparts.Therelationshipbetweenthecurrents
˚
1
and
˚
2
carriedbytwoparallelconductorsand
thecommon-modeanddi˙erential-modecurrent,
˚
˘
and
˚
ˇ
,respectively,is
˚
1
=
˚
˘
¸
˚
ˇ
,(2.1)
˚
2
=
˚
˘

˚
ˇ
.(2.2)
5
Bysubtracting(2.2)from(2.1),anequationforthedi˙erential-modecurrent
˚
ˇ
alongtwoparallel
conductorscanbeexpressedas
˚
ˇ
=

˚
1

˚
2


2
.(2.3)
Thetotalcurrents
˚
1
and
˚
2
subtractin(2.3)andthusmostoftheradiatedemissionscreatedby
thedi˙erential-modecurrentcanceloutwhen
˚
1
and
˚
2
aresimilarinbothphaseandmagnitude.
Theradiatedemissionsarenotexactlyzeroevenwhen
˚
1
isidenticalto
˚
2
becausetheparallel
conductorsareseparatedbysomedistance.ThisisillustratedinFigure2.3,wherethedi˙erential
currentisshowninmagentaandtheparallelconductorsareseparatedbyasmalldistance
3_
where
_
isthewavelengthofthesignalcreatingthecurrent.
Figure2.3:Theelectric˝eldscreatedbythedi˙erential-modecurrentalongparallelconductors.
Forsimplicity,onlytheelectric˝eldcreatedparalleltotheribboncablehasbeenconsidered.
Thesecondconductorisclosertotheobservationpoint
A
byadistanceof
3
,sothemagnitude
oftheelectric˝eldcreatedbythesecondconductor,
j

ˆ
2
j
,isslightlylargerthanthemagnitude
oftheelectric˝eldcreatedatthispointbythe˝rstconductor,
j

ˆ
1
j
.Thevalueofthenet˝eld
6

ˆ
C>C0;
=

ˆ
2


ˆ
1
createdbythedi˙erential-currentat
A
issmall,butnonzero.Di˙erential-mode
currentsalsocreateanelectricpotentialwhentheytravelthroughtheparasiticimpedanceswithin
theconductorsofacircuit[2].This"common-modepotential"thencreatesthetroublesome
common-modecurrentsthatradiateelectric˝eldsmuchmoree˙ectively.
Toshowthatcommon-modecurrentscreatemuchlargerradiatedemissions,weaddandsubtract
(2.2)and(2.1)toobtainanequationforthecommon-modecurrent
˚
˘
alongtwoparallelconductors.
Thisequationcanbeexpressedas
˚
˘
=

˚
1
¸
˚
2


2
.(2.4)
Unlikebeforewiththedi˙erential-modecurrent,the˝eldsproducedbythecurrents
˚
1
and
˚
2
now
addtogether.Thee˙ectofthisadditionisshowninFigure2.4.Here,thecommon-modecurrent
Figure2.4:Theelectric˝eldscreatedbythecommon-modecurrentalongparallelconductors.
isshownincyanandtheparallelconductorsareagainseparatedbyasmalldistance
3_
where
_
isthewavelengthofthesignalcreatingthecurrent.Justlikebefore,onlytheelectric˝eldcreated
paralleltotheribboncableisconsidered.Thesecondconductorisclosertotheobservationpoint
7
A
byadistanceof
3
,sothemagnitudeoftheelectric˝eldcreatedbythesecondconductor,
j

ˆ
2
j
,
isslightlylargerthanthemagnitudeoftheelectric˝eldcreatedatthispointbythe˝rstconductor,
j

ˆ
1
j
.However,nowthevalueofthenet˝eldis

ˆ
C>C0;
=

ˆ
2
¸

ˆ
1
,whichislargerthanthe˝elds
createdbytheindividualportionsofthecommon-modecurrent(aslongas
3_
).Thisis
signi˝cantbecauseitshowsthatevensmallcommon-modecurrentscancreateelectric˝elds.The
electric˝eldsshowninFigure2.3andFigure2.4areprovidedonlyforillustrativepurposesand
representaverysimplisticwayofmodelingtheelectric˝eldscreatedbydi˙erential-modeand
common-modecurrents.Moredetailedmethodsforpredictingtheelectric˝eldscreatedbythese
currentscanbefoundin[1],[2],and[7].InSection4.1ofthisthesis,onesuchmethodisusedto
predicttheradiatedemissionscreatedbythecommon-modecurrentsinthecablingofthecustom
equipmentundertest(EUT)thatwasdesignedtoexceedregulatorylimits.
RadiatedemissionscancauseavarietyofissuesinthelocalRFenvironmentwhenleft
unchecked.Light-emittingdiode(LED)deviceslikethesignshowninFigure2.5havecome
underscrutinyinrecentyearsbecausetheycancreatesporadic,broadbandradiatedemissionsthat
interferewithotherelectronicproducts[9],[10],[11],[12].
TheLEDsignshowninFigure2.5generatesconsistentbroadbandemissionscenteredata
frequencyofapproximately107.1MHz[13].Theseemissionsarepowerfulenoughtointerfere
Figure2.5:AnexampleofaLEDsignthatcreatesEMI.
8
withFMradiobroadcastsandshortrangetransmissionsfromvehiclekeyfobs[13].Asampleof
theseradiatedemissionsareshowninFigure2.6.Inthis˝gure,thex-axisisthefrequencyofthe
signalinMHzandthey-axisfortheblackspectrumdisplayistheRFsignalmagnitudeinrelative
units,or"dBfull-scale"(dBFS).They-axisforthebluewaterfallplothasunitsofsecondsand
hasunitsofsecondsandshowshowthemeasuredsignalvarieswithtime,withthemostrecent
measurementbeingshownatthetop.
(a)
(b)
Figure2.6:(a)RadiatedemissionscreatedbytheLEDsignfromFigure2.5and(b)acontrol
samplemeasured1kmawayfromtheLEDsign.
Figure2.6ashowsameasurementoftheRFspectrumneartheLEDsign.Theradiatedemissions
fromthesignarecircledinred.Theothertwopeakstowardstheleftofthespectrumdisplayare
thesignalsfromnearbyFMradiostations.NoticethattheradiatedemissionsfromtheLEDsign
arestrongerthantheFMradiosignals.Thismeasurementwastakenfromamovingvehiclethat
passedbythesignatadistanceofapproximately15meters.Theapproachofthevehicletowards
9
thesignisshownbythebluewaterfallplot.Theintensityoftheradiatedemissionsappearedand
thengreatlyincreasedabouthalfwaythroughthemeasurementwhichindicatesthatthevehiclewas
approachingthesourceoftheradiatedemissions(theLEDsign).
Figure2.6bshowsacontrolmeasurementofthesamepartoftheRFspectrumapproximately
onekilometerawayfromtheLEDsign.Theradiatedemissionsarecompletelyabsentinthecontrol
samplewhiletheFMradiosignalsarestillpresent.Thisisbecausetheradiatedemissionsare
beingcreatedbytheLEDsign.Sincetheradiatedemissionsarestrongenoughtocausenoticeable
interferenceintheareaneartheLEDsign,theregulatorylimitsaremostlikelybeingviolatedand
thesignshouldbereportedtotheappropriateauthorities.
LargeLEDsignsarenottheonlyunusualsourceofthistypeofinterference.Therearecases
wherevehiclekeyfobsandcellphoneshavebeena˙ectedbynon-compliantswitch-modepower
suppliesfoundinothertypesofelectronicsigns[14].TherearealsocaseswhereEMIfromradiated
emissionscanhavedangerousconsequences.InAugust2019,anopen-heartsurgerybloodpump
wasthesubjectofamajorproductrecallafterforty-˝vereportsofinjuriesrelatedtothesystem
malfunctioningwheninthepresenceofEMI[15].Caseslikethisillustratehowimportantitis
forelectronicproductdesignerstofullyunderstandthee˙ectsofradiatedemissionsandwhyitis
equallyimportantforregulatoryagenciestoenforcestrictemissionlimits.
10
2.2EMCRegulationsandPrecomplianceTesting
Di˙erentorganizationsfromindividualcountriesdevelopandenforcestandardsforwhatcon-
stitutesanacceptableamountofradiatedemissionsfromanelectronicproduct[6].IntheUnited
States,theregulatoryorganizationistheFederalCommunicationsCommission(FCC).Asummary
oftheregulatoryorganizationsintheAmericas,Europe,China,Taiwan,andJapancanbefound
in[16].Thisreferencealsoincludesasummaryoftheregulatoryproceduresusedineachofthese
regions.
Whileeachcountrymayhaveitsownregulatoryproceduresrelatedtoradiatedemissions,many
countriesareslowlymovingtowardsthestandardsthatwerecreatedbytheComitéInternational
SpécialdesPerturbationsRadio(CISPR)in1985[2].Thesestandardswereformallyadoptedby
theEuropeanUnion(EU)asEN55022,butaremorecommonlyreferredtoasCISPR-22[17].
Thesetypesofstandardsareoftenupdatedtoimproveharmonization.Forexample,in2017EN
55022wascombinedwithEUEMCdirectiveEN55032[18].Whilethischangedidimprove
standardharmonization,italsoreclassi˝edmanyelectronicproductsthatwereoriginallyexempt
fromtesting.Thisreclassi˝cationrequiredmanyproductdesignerstoretroactively˝xthedesigns
ofexistingproductsbeforetheycouldcontinuetobemarketedintheEU.Situationslikethiswill
continuetohappenasmorecountriesmovetowardsuni˝edEMCstandards,soitisimportantfor
designerstoensurethattheirproductsmeetthemostubiquitousstandardsbeforetheendofthe
designprocess.FocusingonCISPR-22andsimilarEUEMCstandardsisgoodpracticebecause
theyareconsideredthe"mostlikelyvehicleforaccomplishing[international]harmonization"[2].
TheCISPR-22standardappliestoradiatedemissionsinthefrequencyrangefrom30MHz
to1000MHzanddesignatestwobroadcategoriesforelectronicdevices:ClassAandClassB.
TheClassAstandardsspecifymeasurements30mfromtheEUTandareappliedto"industrial"
devices,whiletheClassBstandardsspecifymeasurements10mfromtheEUTandareappliedto
"residential,commercial,andlight-industrial"devices[16].Thevaluesofthelimitsareidenticalfor
bothcategories,butthedi˙erenceinmeasurementdistancemakestheClassBlimitsmorestringent.
TheFCCusesidenticallynamedcategoriesthataremuchbroaderinscope:FCCClassAdevices
11
aredesignatedas"non-residential"andFCCClassBdevicesaredesignatedas"residential"[16].
JustliketheCISPR-22classi˝cations,theFCCClassBlimitsaremorestringent.Figure2.7shows
theradiatedemissionlimitsforClassBdevicesinboththeEUandtheUnitedStates.
Figure2.7:Radiatedemissionlimitsmeasuredatadistanceof10metersfortheEuropeanUnion
(solid)andtheUnitedStates(dashed).
TheFCCspeci˝esthattheradiatedemissionsfromClassBdevicesaretobemeasuredata
distanceof3m,sothevaluesinFigure2.7havebeenadjustedto10mforabettercomparisonto
theCISPR-22ClassBlimits.Fromthiscomparisononecanseethatthelimitsbetweenthetwo
standardsareveryclose.
ThelimitshaveunitsofdB
`
V/mbecausethestandardsapplytothestrengthoftheelectric˝eld,
notthevoltagemeasuredbythereceiver.Toconvertbetweenthemeasuredvoltageandtheelectric
˝eldstrengthofthereceivedsignal,EMCantennasareprovidedwithaspecialparameterknown
astheantenna(ortransducer)factor.Thisparameterisde˝nedovertheentireusablefrequency
12
rangeoftheantenna.Theantennafactor
˙
hasunitsof1/mandisde˝nedastheratiobetween
theelectric˝eldincidentontheantenna
ˆ
andthevoltageatitsterminals
+
[19]:
˙
=
ˆ

+
(2.5)
Adirectresultofthisde˝nitionisthatasmallerantennafactorindicatesthatanantennaismore
sensitive(andantennasensitivityisespeciallyimportantformeasurementscloseto1m)[2].The
manufacturerwilloftenprovidetheantennafactorindecibelunits.Thisisconvenientbecausethe
receivedvoltage
+
canthenbemeasuredinunitsofdB
`
Vandaddedtothedecibelantennafactor
˙
tocalculatethereceivedelectric˝eldstrength
ˆ
inunitsofdB
`
V/m:
ˆ

3`+
š
<

=
+

3`+

¸
˙

3
1
š
<

(2.6)
Therelationshipbetweentheantennafactorandthemeasuredradiatedemissionstrengthissum-
marizedinFigure2.8.
BeforediscussingCISPR-22further,itisimportanttomentionthatreceiversmustmeetstrict
hardwarerequirementsinordertobeusedforfullcompliancemeasurements.Thesehardware
Figure2.8:Theantennafactorcanbeusedtoeasilyconvertameasuredvoltagebackintothe
electric˝eldstrength,whichcanthenbedirectlycomparedtotheregulatorylimits.
13
requirementsarespeci˝edbyCISPR-16-1[17].Forthepurposesofthisthesis,onlytheCISPR-
16-1requirementsonreceiverdetectorswillbediscussed.Inordertousetheradiatedemission
limitsshowninFigure2.7,theemissionshavetobemeasuredwithaquasi-peak(QP)detector.The
QPdetectoraccountsfortherepetitionrate,orannoyancefactor,ofareceivedsignal[20].This
meansthatintermittentsignalsareweightedlessthancontinuoussignalsandthatthedisplayed
voltagelevelofanintermittentsignalcanbesigni˝cantlylowerthanthepeaksignallevel.The
QPdetectorisimplementedwithashortchargingtimeconstantfollowedbyaslowdischarging
timeconstant[20].Thesetimeconstantsarespeci˝edinCISPR-16-1[17].Thechargingand
dischargingbehavioroftheQPdetectorisillustratedbythedottedbluelineinFigure2.9.
Anormalreceiverusespeakdetectionandwilldisplayintermittentsignalsattheirpeakvalues.
Thee˙ectofeachofthesedetectorsonthetracedisplayedbythereceiverisillustratedinFigure
2.9.Arangeisshownforthepeakdetectorbecausepowerfulintermittentsignalscanover-saturate
thereceiverinputandleadtoaninaccuratemeasurement[17].Noticethatbecauseofthelow
repetitionrateofthesignalshowninFigure2.9,theQPtraceissigni˝cantlylowerthanthepeak
valuesofthesignal.AnotherimportantaspectoftheQPdetectoristhatitrespondstosignals
inalinearfashion;largeamplitude,low-repetitionsignalswillproducethesameoutputaslow
amplitude,high-repetitionsignals[21].Forsignalswithaconstantamplitude,theQPdetector
outputisthesameasthepeakdetector'soutput.QPdetectionrequiresverylongsweeptimes,soit
isofteneasierto˝rstidentifyradiatedemissionswithpeakdetectionandtothenverifyindividual
emissionsforcompliancewithQPdetection.
ItisacceptablefordesignersintheUStofocusontheCISPR-22ClassBlimits(despitetheslight
di˙erencebetweentheFCCandCISPR-22limitsshowninFigure2.7)becausetheUnitedStates
andtheEUhaveamultilateralagreementthatcoversEMCandseveralothermarketsectors[16].
ThismeansthatinmostcasestheFCCandCISPR-22standardsareinterchangeableformanyof
theelectronicproductsthataremarketedinbothregions.Wheremajordi˙erencesoccurareinthe
legalenforcementofthestandards.UnliketheUnitedStates,whichexemptscertainmarketsectors
fromcompliancetesting,theEUhasstrictlyenforceditsEMCstandardsfornearlyallelectronic
14
Figure2.9:Thedi˙erencebetweenthedisplayedoutputtraceofregularpeakdetectionandquasi-
peak(QP)detectionwhenmeasuringaspurioussignalwithalowannoyancefactor[17].
devicessincethemid-1990s[22].Thismeansthatthereisusuallyanapplication-speci˝cradiated
emissionslimitforanelectronicproductthatisenforcedbytheEU.Nevertheless,CISPR-22is
broadenoughtocoveralargemajorityofconsumerelectronicproducts.Thisbroadcoverage
(combinedwiththeEU'sstrictenforcementpolicyandthetendencyforglobalharmonization
towardstheEUEMCstandards)iswhytheCISPR-22ClassBlimitsareusedinChapters3,4,and
5ofthisthesis.
CISPR-22includesdetailedspeci˝cationsforradiatedemissionsmeasurementtechniques[17].
Themostimportspeci˝cationitmentionsistheacceptableleveloftestsiteattenuationwherethe
radiatedemissionsmeasurementisperformed.ThepreferredRFtestenvironmentdescribedby
CISPR-22istheopenareatestsite(OATS)thatconsistsofaverticallyscanning,tuneddipole
antenna,anellipticalareasurroundingthetestsitefreeofmetallicobjects,andametallicground
planeonwhichtheantennaandEUTareplaced[23].Asthenameimplies,thetestsiteisalso
requiredtobelocatedoutdoors.Largesemianechoicchamberscanbecalibratedtomeetthe
normalizedsiteattenuationspeci˝edbyCISPR-22andareanacceptablealternativetoanOATS
forfullcompliancetesting[17].
15
Replicatingtheexactradiatedemissionsmeasurementtechniquesspeci˝edinCISPR-22isonly
necessarywhenperformingafullcompliancetest.Thiskindoftestingisnottypicallyperformed
bytheproductdesigners,butbyanindependentthirdpartythatislicensedtodoso.Failuretomeet
EMCstandardsatthispointinthedesignprocessisextremelycostlyforthreereasons:(1)paying
forathirdpartycompanytotestaproductisexpensiveandoftenrequiresschedulingweeksin
advance,(2)ifaproductfailsthetestitmustberedesigned,and(3)ifaproductfailsthetestitmust
beretesteduntilitmeetsthenecessaryEMCstandards.Thisthirdpointisespeciallyimportant
becausesolelyrelyingonexpensivethirdpartytestingcangreatlyincreasethecostittakestobring
aproducttomarket.ItisbesttoperiodicallytestaproductforEMCcompliancethroughoutthe
designprocesssothatasuccessfulfullcompliancetestcanbeguaranteed[3].Thisformofiterative
testingisreferredtoasprecompliancetesting.
Productdesignersatlargecompanieswilloftenuseaprecompliancetestsetupliketheone
depictedinFigure2.10[6].TheEUTisplaceduponanon-metallicsupport10mawayfrom
theEMCantenna.AspecializedspectrumanalyzerwithQPdetection(referredtoasanEMC
analyzer)isthenusedtomeasuretheradiatedemissions.Thesemianechoicchamberislinedwith
RFabsorbersthatcreateacontrolledRFenvironmentfreefromexternalinterference.Thebest
chambersarelargeenoughtoallowforverticallyscanningtheEMCantenna,butsmalleranechoic
chamberswherethisisnotpossiblecanalsobeused.Measurementsmayalsobeperformedwith
aregularspectrumanalyzerthatisonlycapableofpeakdetection.Compromisessuchastheseare
commonwhenperformingprecompliancemeasurements.
Itisimportanttounderstandthatthegoalofprecompliancetestingisnottorecreatefull
compliancetesting.Rather,thegoalofprecompliancetestingistoaideproductdesignersin
identifyingEMCissuesbeforeaproductissentoutforafullcompliancetest[24].Withthisgoal
inmind,theconceptofignoringtheCISPR-22measurementstandardscanbeextendedtocreate
low-costprecompliancetechniquesthatarestilluseful.
Manyproductdesignersdonothaveon-siteaccesstotheequipmentshowninFigure2.10.In
thesesituations,performingradiatedemissionmeasurementsinanuncontrolledRFenvironment
16
Figure2.10:AtraditionalprecomplianceradiatedemissionsmeasurementtestsetupusinganEMC
analyzerandashieldedRFchamber.
canstillhelptoidentifyproblemsignalswhenthemeasurementiscomparedtoabackground
measurementofthesameRFenvironment.OnecanuseaspectrumanalyzerandcommonVHF
andUHFtelevisionantennastomeasuresignalsinthe30MHzto1000MHzrange[25].Figure
2.11showsanexampleofasimpleon-sitetestsetupformeasuringradiatedemissionsfroman
EUT.Thissetupiseasytoconstructandonlyrequiresasingleoperator,whichmeansthatmany
measurementscanbeperformedthroughoutthedesignprocess.
AmajorissuewiththesetupshowninFigure2.11isthatthedistancebetweentheantennaand
theEUTismuchshorterthan10m.Thismeansthatthemeasurementisbeingtakeninthenear
˝eldoftheantennawherecouplinge˙ectswilldrasticallye˙ectthemeasuredemissionstrength
[2].WhilethisproblemcanbesolvedbyincreasingthedistancebetweentheantennaandtheEUT,
theuncontrolledRFenvironmentandthelackofametallicgroundplanecanintroduceothererrors
asthedistancebetweentheantennaandtheEUTisincreased.Forexample,adistanceof10m
allowsforfrequency-dependenterrorssuchassignalre˛ectionswithobjectsintheuncontrolled
17
Figure2.11:Alow-costprecompliancetestsetupformeasuringradiatedemissionsinanuncon-
trolledRFenvironmentsuchasahallwayoro˚cespace[25].
environment.Theseerrorscanbeverylarge,sometimesrangingfrom+6dBto-25dBforeach
re˛ection[23].Theselargeinaccuraciesaredi˚culttopredictbecausetheyvarygreatlydepending
uponthetypeofuncontrolledRFenvironmentthatisusedfortesting.However,performingmany
radiatedemissionmeasurementsinthesameRFenvironmentallowsforarelativecomparison
betweentheresultsofeachtest.Whencombinedwithiterativechangestotheelectronicproduct,
reducingaparticularemissionbyatleast2or3dBcanbeconsideredasigni˝cantimprovement
[25].Thisisanexampleofhowtheprimarygoalforlow-costprecompliancetestingshouldbe
measurementrepetitionwithgradualdesignimprovement.
ManyofthecompromisesmadebythetestsetupshowninFigure2.11signi˝cantlyreducethe
costofprecompliancemeasurements.However,thistestsetupstillrequiresareceiver.Inmost
casesthisisaspectrumanalyzerwithmeasurementcapabilitiesinthefrequencyrangeof30MHz
to1000MHz.AspectrumanalyzerwillcostlessthenanEMIreceiverorEMCanalyzer,butitwill
nothaveaQPdetector.Anotherissueisthatspectrumanalyzersarestillexpensiverelativetothe
18
restofthetestequipmentshowninFigure2.11.Usedspectrumanalyzerscanrangeinpricefrom
$1000.00to$5000.00,whilenewspectrumanalyzerscancostwellover$10,000.00[25].More
a˙ordableUSB-basedspectrumanalyzersexist,buttheystillgenerallycostwellover$300.00
[25].Itisthisauthor'sopinionthatsuchspectrumanalyzerstypicallyhaveminimalfunctionality
andanon-intuitiveuserinterfacethatcangreatlyincreasethetimeittakestoperformevenbasic
measurements.IfonecouldreplacethespectrumanalyzershowninFigure2.11withacomparable
receiverthatcostssigni˝cantlyless,acompletelow-costprecompliancetestsetupcouldbecreated.
AccordingtoKeysightTechnologies[26],aspectrumanalyzerisadevicethatperformsaFourier
transformandthendisplaysthespectralcomponentsofthemeasuredRFsignalasavoltage.Figure
2.12showstheblockdiagramforagenericsuperheterodynespectrumanalyzerasdescribedin
[26].Asweepgeneratorisusedtocreatethehorizontalmovementofthedisplayedvoltage.This
generatorisalsousedtotunethelocaloscillator.Theorangeshadedareashowsallofthesignal
conditioningthatisnecessarybeforethemeasuredRFsignalcanbedisplayed.
Figure2.12:Blockdiagramofatypicalsuperheterodynespectrumanalyzerasdescribedin[26].
19
TheinputRFsignalis˝rstsentthroughanattenuationstagetopreventthemixerinputfrom
beingoverloaded.ThisattenuatedversionoftheRFsignalisthen˝ltereddependingonthe
frequencyrangeofinterestsothatout-of-bandsignalsarenotdisplayedbythespectrumanalyzer.
Themostimportantpartoftheprocessistheintermediatefrequency(IF)conversionstagewhich
beginsatthemixer.ThedownconversiontoanIFallowsfortheprocessingoftheRFinputsignal
atamuchlowerfrequency.TheIFconversionstageisalsowhatdeterminesthetunablefrequency
rangeofthespectrumanalyzer.
TheprocessingperformedontheIFsignalconsistsofanIFgainstage,anoptionalattenuation
stage,anda˝lterstage.TheseareallshownjustafterthemixerinFigure2.12.TheIF˝lter
iswheretheresolutionbandwidth(RBW)ofthespectrumanalyzercanbeadjusted.TheRBW
de˝nesthesmallestdisplayedfrequencyseparationbetweentwoadjacentsignals[27].Anarrow
resolutionbandwidthallowsfortheaccuratemeasurementofRFsignalsthatareveryclosetogether
infrequencyatthecostofgreatlyincreasingthetracesweeptime.
Thelaststageofthespectrumanalyzermodi˝estheoutputoftheIFstagesothattheproper
signallevelisdisplayed.Thisinvolvesanenvelopedetectorthattrackstheenvelopeofthesignal
whilesuppressinganyinstantaneousvariations[26].Theoutputoftheenvelopedetectorisavideo
signalthatispassedthroughavideo˝lterwhichsmoothsoutthe˝naldisplayedsignaltrace.The
outputoftheenvelopedetectorisalsowhereanEMCanalyzerwouldimplementaQPdetector.
Closelyreplicatingthismeasurementprocesswithalow-cost,o˙-the-shelfsoftware-de˝nedradio
(SDR)receivercangreatlyreducethecostofalow-costprecompliancetestsetup.
20
2.3Software-de˝nedRadio(SDR)
Asoftware-de˝nedradio(SDR)isde˝nedbytheWirelessInnovationForum[28]asa"radio
inwhichsomeorallofthephysicallayerfunctionsaresoftware-de˝ned."AccordingtoGrayver
[29],theidealSDRreceiverwouldbeabletoreceive"anywaveformatanycarrierfrequencyand
anybandwidth"whileconsistingofonlyanantenna,ananalog-to-digital(A/D)converter,anda
digitalprocessor.Suchareceiverisphysicallyimpossibletoconstructwithcurrenttechnology,so
apracticalSDRconsistsofananalogsignalprocessingstagecombinedwithadigitalprocessor
asshowninFigure2.13[30].SDRreceivers(andtransmitters)arepredominantlyusedfor
communicationssystems;in2011theywereutilizedby93%ofthemobileinfrastructuremarket
[31].TheoutputoftheSDRreceivershowninFigure2.13hasbeenshownasafrequencyspectrum
displayforcomparisontoatraditionalspectrumanalyzer.
TheanalogstageoftheSDRreceiverissimilartotheRFinputconditioningstageofthespectrum
analyzershowninFigure2.12.Thekeydi˙erenceisthattheSDR'sanalogstagemustdigitize
thesignalwithanA/Dconverterbeforefurthersignalprocessingcanbeperformed.Thisismost
commonlydoneby˝rstseparatingtheRFsignalintoin-phaseandquadrature(IQ)componentswith
aquadraturemixerandthensamplingeachofthesecomponentswithadedicatedA/Dconverter
[32].AftertheA/Dconversion,theIQsamplesarepassedthroughthedigitaldownconversionstage
showninthegreenshadedareaofFigure2.13.Thistypeofdesignisreferredtoasadigitallow-IF
architecture.Implementingthislastconversionstagedigitallyhasseveraladvantages,including
immunityfromtheoperatingconditionsthatcancorruptmoresensitiveanalogcircuitry[32].
Adigitaldownconversionstagealsoreducessystemcostandincreasessystemintegrationwhen
comparedwithatraditionalanalogdownconversionstage.
TheIQsamplesfromtheA/DconversionstageinFigure2.13canalsobedirectlyconvertedto
DCtoincreasesystemsimplicitybyavoidingadditionaldigitalmixingand˝ltercircuitry[32].This
iscalledadirectconversion(orzero-IF)architectureandiscommonlyusedbySDRreceivers(as
discussedattheendofthissection).Thezero-IFarchitecturehasseveralsigni˝cantdisadvantages
whencomparedtothedigitallow-IFarchitectureshowninFigure2.13.Forexample,leakagefrom
21
Figure2.13:Generalizedblockdiagramforadigitallow-IFSDRreceiverasdescribedin[32].
thelocaloscillatorusedintheanalogstagecancreateaDCo˙setandintermodulationproducts
thatwillappearassignalsonthespectrumdisplay[32].Thesefalsesignalscanmakeidentifying
thesignalsofinterestmuchmoredi˚cultfortheoperator.AnSDRreceiverthatuseseithera
traditionalanaloglow-IForthedigitallow-IFarchitectureshowninFigure2.13isabletoavoid
theseissues.
ThedigitalsignalprocessingstageshowninFigure2.13iswheretheSDRreceivertransforms
theIQsamplesintoameaningfulresult.WhenusinganSDRreceivertomeasurethestrengthof
RFsignals,thisstageiscomparabletotheenvelopedetectorandvideo˝lteringusedbyatraditional
spectrumanalyzer.Thereareseveraldi˙erentdevicearchitecturesthatcanbeusedtoperformthe
digitalsignalprocessinginanSDRreceiver.Decidingonanarchitectureislargelydependentupon
theintendedapplication.ThefourmaintypesdescribedbyGrayver[29]arethegeneralpurpose
processor(GPP),thedigitalsignalprocessor(DSP),the˝eld-programmablegatearray(FPGA),
andtheapplication-speci˝cintegratedcircuit(ASIC).Table2.1showstheprimaryadvantagesof
eachofthesearchitectures.
GPPsarethemostcommontypeofdigitalsignalprocessorusedtoday,butforSDRapplications
theyareunabletoprocesshigh-throughputsignalsontheirown[29].ForSDRapplicationsthat
requirehighdatathroughput,theASICandtheFPGAarethebestchoice[29].However,both
requireasigni˝cantamountofexpertisetodevelopwhencomparedtoaGPP.Themajordesign
22
trade-o˙sbetweenthesefourarchitecturesareshowninFigure2.14.Inthis˝gure,thedevelopment
e˙ortshownalongthex-axisdoesnotaccountforthesavingsfrommass-productionbutinsteadis
representativeofthedesignexpertiserequiredtocreateandprototypeanewdesignwithaparticular
architecture.
Table2.1:TheadvantagesoffourSDRdigitalsignalprocessingarchitectureswithexamples.
Figure2.14:Therelationshipsbetweendevelopmente˙ort,cost,powerconsumption,anddata
throughputforthemostcommonSDRdigitalsignalprocessingarchitectures(adaptedfrom[29]).
23
ForthepurposeofcreatinganSDRreceivercapableofmakingradiatedemissionmeasurements
inthefrequencyrangeof30MHzto1000MHz,aGPParchitecturewouldbeideal.Thisisbecause
anSDRreceiverdesignedforthispurposewouldnothaveanystrictpowerrequirementsandwould
notneedtoprocesshigh-throughputsignals.Infact,modernGPPsarefastenoughtoimplement
allofthenecessarydigitalsignalprocessingforpracticalradios[29].IfaGPPisusedforallofthe
digitalsignalprocessing,theSDRreceiverwillonlyrequireadditionalcircuitryfortheorangeand
greenshadedareasshowninFigure2.13.ThiscircuitrywillbecometheSDRfrontend[32]and
thecombinationoftheSDRfrontendwithaGPPcreatesa"software-centric"SDRreceiver[29].
Onesuchsoftware-centricSDRisshowninFigure2.15.ThisSDRutilizesseveralICs:the
AnalogDevices9364ASICperformstheanalogRFsignalhandlingwhileaXilinx
®
Spartan-
6
®
FPGAperformsboththereceiverdigitaldown-conversionandthetransmitterdigitalup-
conversion[33].The˝nalIQsamplesfromtheB200minireceiveraresentoverUSBtoaGPP
(usuallytheCPUofahostPC).ThisuniqueICarrangementallowstheGPPtofocussolely
ondigitalsignalprocessing,whichisagreatadvantagewhenimplementingcomplexwireless
communicationprotocols[29].However,theB200minihasmanyfeaturesthatareunnecessary
forsimplyperformingRFmeasurements.Thesefeaturesmakeittooexpensiveforreplacingthe
spectrumanalyzerinthelow-costtestsetupshowninFigure2.11.Fortunately,manyothertypes
ofsoftware-centricSDRsareavailablethankstotheirpopularityinthehobbyistradiomarket[34].
Aselectionofsoftware-centricSDRfrontendsareshowninFigure2.16.
NotethatwhileitistechnicallythecombinationoftheSDRfrontendandaGPPthatmakesa
completesoftware-centricSDR,thetermSDRiscommonlyusedinreferencetojustthefrontend
byitself.Throughouttherestofthisthesis,thetermSDRwillrefertotheSDRfrontend.
TheSDRsshowninFigure2.16wereallconsideredaspotentialreplacementsforaspectrum
analyzer.TheyeacharecapableofreceivingRFsignalsinthefrequencyrangefrom30MHzto
1000MHzand(atthetimeofwriting)costsigni˝cantlylessthanaspectrumanalyzer.
TheNESDRSMArtSDRwasthemostcompactmodeltested,butitslowsamplingrateand
simplistichardwaredesignmakeittooinaccurateformakingconsistentRFsignalmeasurements.
24
Figure2.15:TheB200minifromEttusResearch
Š
isacompactFPGA-basedSDRfrontendwith
transmitandreceivecapabilitiesoverawidefrequencyrange.
AnotherissuerelatedtothisSDR'slowsamplerateistheuserexperience;searchingforradiated
emissionsoverarangeof970MHzinroughly1.8MHzincrementsisaveryslowandtedious
process.This1.8MHzspanfurthermakesidentifyingrealsignalsdi˚cultbecauseoftheartifacts
thatarecreatedbythezero-IFtunerICusedbytheNESDRSMArtSDR.
TheHackRFOneistheonlySDRshowninFigure2.16withtransmissioncapabilities.Italso
coversamuchlargerfrequencyrange,muchliketheB200mini.However,bothoftheseproperties
maketheHackRFOneanunsuitablespectrumanalyzerreplacementforthesamereasonsthe
B200miniwasnotconsidered.TheHackRFOnealsohasaconsiderablyworsesignal-to-noiseratio
(SNR)thantheotherSDRsshowninFigure2.16[34].Itisthisauthor'sopinionthattheHackRF
Oneshouldonlybeconsideredforlow-costapplicationsfocusingonRFsignaltransmission.
TheAirspyR2isaninterestingSDRbecauseithasthesamezero-IFRFtunerICusedbythe
NESDRSMArtSDRcombinedwithamoreadvanced12-bitA/Dconverter[36].Thisallowsthe
AirspyR2SDRtohaveamuchmoreusablespanofalmost10MHz.Whilethisfrequencyspan
isstillmuchsmallerthanwhatispossiblewiththetrackinggeneratorofaspectrumanalyzer,it
makesseparatingrealRFsignalsfromthezero-IFartifactsveryeasyfortheoperator.TheAirspy
R2isalsoverycompactandissmallerthantheRSP2pro,HackRFOne,andeventheB200mini.
AlloftheSDRsdiscussedthusfarareonlycapableofmeasuringRFsignalstrengthinrelative
25
Figure2.16:Severalsoftware-centricSDRsthatwereconsideredforthisthesis.Thehardware
characteristicsshownwereretrievedfromthemanufacturerwebsites([35][36][37][38]).
unitsofdBFS.Thismeansthatitisimpossibletoknowtheactualpowerofthereceivedsignal,which
isunacceptableforprecomplianceradiatedemissionmeasurements.AnyoftheseSDRscouldbe
usedtomeasureabsolutesignalstrengthbycalibratingthedBFSreadingwithaknowninputpower,
butano˙-the-shelfSDRcapableofsuchmeasurementswouldbemuchmoreconvenient.
26
2.4TheRadioSpectrumProcessor2Pro(RSP2pro)
TheRSP2proSDRshowninFigure2.17isasoftware-centricSDRcapableofperforming
absolutesignalstrengthmeasurementsthankstoitsproprietarycompanionsoftware,SDRuno.
Figure2.18showsasimpli˝edblockdiagramoftheRSP2proandFigure2.19showshowthese
componentsarearranged.
Figure2.17:TheRSP2profromSDRplayisasoftware-centricSDRreceiverwithmanyuniqueRF
signalconditioningcapabilities.
ThePortAandBsignalchaincontainsmanyoftheRFsignalconditioningfeaturesthatmakethe
RSP2prouniquewhencomparedtoothersoftware-centricSDRs.TheselectableRF˝lternetwork
reducesout-of-bandsignals,justlikeaspectrumanalyzer.ThereisalsoanoptionalAM/FM
notch˝lterforsuppressingpowerfulbroadcastsignals.Thisfeatureisidealfortheuncontrolled
RFenvironmentusedduringalow-costprecompliancetest,wheresuchbroadcastsignalsmay
maskemissionsfromtheEUT.TheRSP2prohasadedicatedRFgaincontrolstage(shownafter
theAM/FMnotch˝lterinFigure2.18)andanLNA.Bothcanbecontrolledbytheoperatorvia
software,althoughinpracticeitisbesttouseautomaticgaincontrolfortheLNA[39].
27
Figure2.18:Asimpli˝edblockdiagramshowingthemaincomponentsoftheRSP2proSDR
(adaptedfrom[39]and[40]).
Figure2.19:InternalviewoftheRSP2pro;the10bandselectableRF˝lternetworkcanbeseenon
thelower-centerofthePCB,whiletheRFtunerICcanbeseenonthelower-right.
28
AllofthereceiverfunctionalityoftheRSP2proishandledbytwoASICs:theMiricsSemicon-
ductor,Inc.,MSi001RFtunerandMSi2500USBinterfaceIC.TheMSi001featuresbothzero-IF
andlow-IFoperation,aswellasadedicatedLNAforeachofitsRFinputs.Itisimportanttonote
thatthelow-IFimplementationusedbytheRSP2proisnotthesameasthedigitallow-IFimple-
mentationshownbythegreenshadedareainFigure2.13.Thisisbecausetheselectionbetween
zero-IFandlow-IFmodeontheRSP2protakesplaceintheRFtunerchip(i.e.theorangeshaded
areainFigure2.13).Thisdi˙erencedoesnotsigni˝cantlyimpacttheoperationoftheRSP2pro
becausebotharchitecturesconditionanRFsignalinthesameway.
Inlow-IFmode,theMSi001˝rstampli˝estheconditionedRFsignalandthenusesaquadrature
mixertoseparatethesignalintoIandQcomponentsatanIF.TheIFIQcomponentsarethen
separatelydigitizedbytheMSi2500'sintegratedA/Dconvertersandthe˝nalIQdatastreamissent
overUSBtoaGPPforfurtherdigitalsignalprocessing.
The˛agshipfeatureoftheRSP2proisitsabilitytomeasuretheabsolutepowerofRFsignals.
ModernSDRsmeasuresignallevelattheinputoftheA/Dconverter[39]wheretheincoming
voltageissampledattwicetherateoftheIQbandwidth.ModernA/Dconvertershavevery
accuratereferencevoltages,sothevoltagereadingattheA/Dinputisveryprecise.However,this
voltageisnotthesameasthereceivedsignalstrengthbutisinsteadtheresultingvoltagefromall
oftheprecedingRFstages.ThisiswhymostSDRscannotmeasuretheabsolutepoweroftheRF
input.ThedesignersoftheRSP2prowereabletosolvethisissuebycreatingacomplexdataset
ofthegaine˙ectsforeveryusablecon˝gurationoftheRFstagesshowninFigure2.18[39].The
appropriateadjustmentisautomaticallymadetothespectrumdisplaybasedonhowtheoperator
hascon˝guredtheRSP2prosopowermeasurementscanbemadeseamlessly.Unfortunatelythese
datasetsarenotavailabledirectly,sotheRSP2procanonlymakeabsolutepowermeasurements
whenitisusedwithitscompanionsoftware(seetheendofthissectionformoreinformation).
TheRSP2procanalsomakeveryaccuratefrequencymeasurements.Itdoesthisbyusinga
precision0.5ppmtemperature-compensatedoscillator(TCXO)asareferencefortheIQconversion
andA/Dsampling.Thisfeature,theabsolutepowermeasurementcapability,andtheallmetal
29
shieldedenclosuremaketheRSP2promorecomparabletoapieceoflaboratoryequipmentthana
hobbyistSDRreceiver[41].Together,theymaketheRSP2prothebestcandidateforsuccessfully
replacingaspectrumanalyzerforlow-costprecompliancemeasurements.Table2.2summarizes
allofthekeyhardwarespeci˝cationsoftheRSP2pro.
Table2.2:TheRSP2prohardwarespeci˝cationsthatarerelevanttothisthesis.Allinformation
takenfrom[42]and[41].
Software-centricSDRsrelyheavilyonaGPPforsignalprocessing,andtheRSP2proisno
exception.Asmentionedpreviously,onemustusetheRSP2pro'scompanionsoftwareinorder
tomakeabsolutepowermeasurements.ThissoftwareiscalledSDRunoandisavailableonthe
SDRplaywebsite[37].Atthetimeofwriting,SDRunoonlysupportsWindowsoperatingsystems.
Figure2.20showstheSDRunouserinterface.Itiscomposedofmultipleblocksthatmimic
theappearanceofatraditionalamateurradioreceiver.Thespectrumdisplayhastwocomponents:
atraditionalspectrumdisplayandawaterfallplot.Thex-axisisthefrequencyofthereceived
30
signalinMHzandthey-axisforthespectrumdisplayisthemagnitudeofthesignalinunitsof
dBm(decibelsrelativetoonemilliwatt).They-axisforthewaterfallplothasunitsofsecondsand
hasunitsofsecondsandshowshowthemeasuredsignalvarieswithtime,withthemostrecent
measurementbeingshownatthetop.
Thewaterfallplotisconvenientwhencomparingbackgroundsignalstotheradiatedemissions
fromtheEUT.BypoweringtheEUTonhalfwaythroughacompletewaterfallcapture,theoperator
canquicklyseeifnewpeakshaveappearedorifthenoise˛oorhasbeena˙ectedbytheEUT.
Thisisespeciallytrueforweakorintermittentsignals,whichcanbehardtonoticewithjustthe
spectrumdisplay.
Figure2.20:TheSDRunouserinterface.
ThepowermeterinSDRunoworksbymeasuringthepeaksignalstrengthwithinthebandwidth
ofthesecondaryspectrumdisplayinthetoprightcornerofFigure2.20.Thebandwidthofthis
secondaryspectraldisplayisalsoshownbythegrayregioninthemainspectrumdisplay.Thepower
metermeasuresthepeakvaluewithintheselectedbandwidth.Forradiatedemissionmeasurements,
31
usingthecontinuous-wave(CW)modeworksbestbecauseitallowsforthenarrowestbandwidth.
CWmodealsoincludesabuttonthatwillmovethereceiverfrequencytothemeasuredpeakvalue.
Thepowerlevel(indBm)andthefrequencyofthemeasuredpeakcanbecontinuouslysavedtoa
csv˝lewherethetimeintervalbetweenindividualmeasurementscanbesetbytheoperator.
BeforetheRSP2procanbeusedasaspectrumanalyzer,theaccuracyofitsmeasurementsmust
beveri˝ed.Thetestsperformedin[39]foundthattheabsolutesignalaccuracyoftheRSP2pro
between1to2dB.Thenextchapterwillfocusonperformingalargenumberofmeasurements
underconditionsthatareclosertotheintendedusecase(i.e.measurementsthatwouldbeexpected
duringaradiatedemissionstest).
32
CHAPTER3
PERFORMANCEOFTHERSP2PRO
Thischapterfocusesondescribingthetestequipmentandproceduresthatwereusedtocharacterize
theRSP2pro'smeasurementaccuracy.Adescriptionofthetestequipmentthatwasusedisgiven
inSection3.1.Section3.2thenoutlinesthesemi-automatedcharacterizationproceduresthatwere
usedtoobtainalargenumberofreadingsfromboththeEMCanalyzerandthreeRSP2prounits.
Sections3.3and3.4showthecharacterizationresultsobtainedfromtheseprocedures.These
resultsarethenusedtoshowthemarginoferrorformeasurementstakenwiththeRSP2pro.
3.1EquipmentInformation
ThetestequipmentdescribedinthissectionwasusedtocharacterizetheRSP2pro.The
importantinformationforeachpieceoftestequipmenthasbeenprovidedhereforconvenience.
MeasurementstakenwiththeAgilentE7410AEMCanalyzerwereusedtoassessthemeasure-
mentperformanceoftheRSP2prothroughoutthisthesis.TheEMCanalyzerwasalsousedinthis
sectiontocalibrateandcharacterizethesignalgenerator.
Theantennafactor(AF)valuesofseveralEMCantennaswereusedtodeterminetheoutput
frequencyandamplitudecombinationsusedbythesignalgeneratorinSection3.2.TheBicoLOG
30100XEMCantennawasalsousedforperformingnarrow-bandandbroad-bandradiatedemission
testswithboththeEMCanalyzerandtheRSP2proinChapters4and5.
TheHP8657Asignalgeneratorwasusedtogeneratecontinuouswave(CW)signalswith
speci˝cfrequenciesandamplitudesforthecharacterizationproceduredescribedinSection3.2.
Thesignalgeneratorwasalsocombinedwithastandardwhipantennatocreatenarrow-band
radiatedemissionsthatweremeasuredbyboththeEMCanalyzerandtheRSP2proinSection4.2.
33
3.1.1AgilentE7401AEMCAnalyzer
TheAgilentE7401AEMCanalyzershowninFigure3.1isasuperheterodynespectrumanalyzer
withseveralspecializedEMCfeatures,includingaCISPR-16-1compliantquasi-peak(QP)detector
[43].Ithasa50

RFinputthatiscapableofmeasurementsinthefrequencyrangefrom9.5kHz
to1.5GHz.Accordingtothedatasheet[43],theE7401Ahasanabsoluteamplitudemeasurement
accuracyof

0.30dB.
TheE7401AEMCanalyzerisprogrammableviatheIEEEgeneral-purposeinterfacebus
(GPIB).ThisallowedfortheautomationofnearlyalloftheEMCanalyzermeasurementspresented
inthisthesis.Toautomatemeasurements,anEMCanalyzerstate˝lewascreatedforeach
measurementtype.Eachstate˝lecanbeloadedbytheEMCanalyzertoquicklypresettheanalyzer
settingsbeforeameasurementistaken.Thesestate˝lesaredescribedinTable3.1.Theleftmost
columnshowsthenameofeachspeci˝cstate˝lewhiletheremainingcolumnsshowtheanalyzer
settingsforthatstate˝le.Thistablewillbereferredtothroughouttherestofthisthesisasnecessary.
The"DefaultSweep"state˝lewasusedwheneveraQPmeasurementwasnecessary.The
variable
5
8=
representsthedesiredcenterfrequencyforaparticularmeasurement.Thisstate˝le
Figure3.1:TheAgilentE7104AEMCanalyzerthatwasusedforallmeasurementcomparisons
withtheRSP2pro.
34
hasthesmallestfrequencyspanandthe˝nestresolutionbandwidth(RBW)becausetheQPdetector
modeoftheE7401Ahasanextremelylongsweeptime,makinglargerspanningQPmeasurements
unrealisticgiventheamountofmeasurementsthatwasnecessaryforthisthesis.Thecombination
ofthesmallspanand˝neRBWmakethe"DefaultSweep"stateidealforcapturingthepeakvalueof
thenarrowbandsignalscreatedbythesignalgeneratorsourcethatwasusedforthemeasurements
inthischapter.The"SpectrumSweepPart1"through"SpectrumSweepPart4"state˝leswere
usedtogetherwheneverafullsweepoftheCISPR-22frequencyrangeof30MHzto1000MHzwas
necessary.Separatingthisfrequencyrangeintofourstate˝lesallowsthesignaltraceresolutionof
afullsweeptobe32,768pointsratherthan8192points.
Thelaststate˝le,"RESweep,"wasusedtogetaninitialmeasurementbeforetheQPmeasure-
mentstakenwiththe"DefaultSweep"state˝lesettings.The"RESweep"statehasamuchlarger
spanandRBWsothatameasuredsignal'speakvaluecanbeacquiredquickly.Onceapeakhas
beenacquired,thefrequencyofthepeakisusedasthecenterfrequencyofthe"DefaultSweep."
ThisallowstheQPmeasurementtoalwaysbecenteredattheappropriatefrequencyforagiven
inputsignal.
Table3.1:ParametersoftheEMCanalyzerstate˝lesthatwereusedtocontrolthesettingsofthe
EMCanalyzer.
35
3.1.2EMCAntennas
AsmentionedinChapter2,radiatedemissionmeasurementsareusuallyperformedwithspecialized
antennascalledEMCantennas.EMCantennasareprovidedwithaparametercalledtheantenna
factor,whichisde˝nedbyEquation2.5andcanbeusedtoconvertthemeasuredvoltageto
thestrengthoftheelectric˝eldincidentupontheEMCantenna(thisisshowninFigure2.8).
Manufacturerprovidedantennafactorsareusuallygiveninatableorchartandwillbelistedat
anumberofuniformly-spacedfrequencies;forexample,atevery10MHzforfrequenciesfrom
30MHzto1000MHz.Figures3.2cand3.4ashowtheantennafactorsofseveraldi˙erentEMC
antennasthatwereusedforthisthesis.
TheEMCantennasshowninFigures3.2aand3.2bareabletocovertheentireCISPR-22
frequencyrangewhenusedtogether.TheAgilent11955AEMCantennashowninFigure3.2ais
abiconicalantennadesignedforreceivingsignalsfrom20MHzto200MHz.Theantennafactor
dataprovidedbyAgilent[44]forthisantennaisin10MHzincrementsandthisdatawasusedto
plotthe11955AantennafactorshowninFigure3.2c.ThisantennaismostsensitivetoRFsignals
measuredinthefrequencyrangefrom50MHzto100MHzasindicatedbytheverylowantenna
factorbetweenthesefrequencies[2].At300MHz,thesensitivityofthe11955AEMCantennais
morethan10dBbelowitspeaksensitivity.Thissharpdecreaseinsensitivityhappensbecauseof
thenullthatdevelopsdirectlyinfrontoftheantenna'sboresightasthisfrequencyisapproached
[45].The11955AEMCantennaisbestsuitedformeasurementsbetween80MHzand200MHz
becauseitiswellmatchedinthisrange,asindicatedbythelowvariationoftheantennafactor
betweenthesefrequencies[45].
TheAgilent11956AEMCantennashowninFigure3.2bisalogperiodicantennawithan
antennafactorthatincreasesnearlylinearlyfrom200MHzto900MHz.Thesharpdecline
observedat900MHzinFigure3.2cismostlikelycausedbythelossesintheantenna'sfeeder
coaxialtransmissionline[45].The11956AantennafactorinFigure3.2cwasplottedusingthedata
providedbyAgilent[44]andisin25MHzincrements.Itshouldbenotedthatthis11956Aantenna
isprovidedwithanantennafactorupto2000MHzforusewithothertypesofEMCmeasurements.
36
(a)
(b)
(c)
Figure3.2:(a)TheAgilent11955Aand(b)11956AEMCantennasand(c)theirantennafactors
[44].
37
TheantennafactorshowninFigure3.2chasbeentruncatedat1000MHzbecausethatisthehighest
signalfrequencyrelevanttothisthesis.
Whilebothoftheseantennasareidealforfullcompliancetests,theirlargeelementsizecanlead
toissueswhenusedinalow-costprecompliancesetupliketheoneshowninFigure2.11.Thisis
becausethemeasurementdistanceusedbythissetupiscomparabletotheelementlengthofthese
antennasandthusthebeamwidthoftheradiatedemissionsfromtheequipmentundertest(EUT)
maynotfullyilluminatetheseantennascompletely[2].Formeasurementdistancescloseto1m,
Ott[2]recommendsanantennathatmeetsthefourfollowingcriteria:
‹
Haveawidebandwidth(e.g.the30MHzto1000MHzrequiredforCISPR-22)
‹
Besmallinphysicalsize(ideallylessthan12incheslong)
‹
Beverysensitive(i.e.anantennafactorlessthan6dB)
‹
Havea(reasonably)˛atfrequencyresponse
Ott[2]acknowledgesthatthesecriterioncanbecontradictory(e.g.asensitiveantennaat30MHz
cannotalsobephysicallysmall).However,bycarefullymakingcompromiseswiththesecriteria,
onecanhaveanEMCantennathatisbettersuitedforlow-costprecompliancetestingatshort
distancesthanthetraditionalEMCantennasliketheonesshowninFigures3.2aand3.2b.A
physicallysmallantennahasamuchbetterchanceofavoidingtheaforementionedillumination
issuethatisinherenttothelargerEMCantennasatdistancesnear1m.Soanexampleofagood
compromisewouldbeforgoinga˛atfrequencyresponseorhighsensitivitytoensureasmall
physicalantennasize.
TheBicoLOG30100XEMCantennafromAaroniaAGshowninFigure3.3isaphysically
small,activeantennathatisdesignedtoreceivesignalsfrom30MHzto1000MHz.Thisantenna
isabletomeettwoofOtt's[2]fourcriteriabycompromisingontheoverallsensitivityandresponse
˛atness.TheBicoLOG30100X'santennafactorisshowninFigure3.4a.Thisdatawasprovided
byAaroniaAG[46]andisin10MHzincrements.UnliketheantennafactorsshowninFigure3.2c,
38
theBicoLOG30100Xantennafactorsareallnegative.ThisisbecausetheBicoLOG30100Xisan
activeantennathatusesabuilt-inpreampli˝ertoachieveitswidebandwidth[47].Thepreampli˝er
gainfortheunitthatwasusedinthisthesisisshowninFigure3.4b.Thisdatawastakenfromthe
preampli˝erusingthemethoddescribedin[48]andrepresentsthevaluesmeasuredbyAaronia
AGforthisspeci˝cpreampli˝er.
TheBicoLOG30100XEMCantennaismostsensitivetomeasurementsinthefrequencyrange
from150MHzto500MHz.Thisisindicatedbythelowantennafactorshownbetweenthese
frequencies[2]inFigure3.4a.OnecanalsoseethattheBicoLOG30100XismuchliketheAgilent
11956Ainthefrequencyrangefrom300MHzto1000MHzinthattheantennafactorsofboth
increaselinearlywiththefrequencyinthisrange.
Figure3.3:TheBicoLOG30100XEMCantenna.
Figure3.4bshowsthatthegainofthepreampli˝erisvery˛atacrosstheentirefrequencyrange,
withtheexceptionofanapproximatereductionof0.5dBinthefrequencyrangefrom800MHz
to1000MHz.This˛atresponsemeansthatthelargevariationintheBicoLOG30100Xantenna
factorshowninFigure3.4aisalmostentirelycausedbytheantennaconstructionitself.Theuseof
apreampli˝erdoescomplicateusingthisantennainthatthereisausablebatterylifeofonlythree
39
(a)
(b)
Figure3.4:(a)TheBicoLOG30100Xantennafactorand(b)thegainofitspreampli˝er.The
antennafactorshownin(a)isnegativebecauseofthelargegainofthepreampli˝ershownin(b)
[47].
40
tofourhours[48].Usingtheantennawhileattachedtothechargerisnotrecommendedbythis
author,asthechargerintroducessigni˝cantbackgroundradiatedemissionsthatcaninterferewith
measurements.
ThesmallphysicalsizeoftheBicoLOG30100Xcombinedwithitsabilitytoperformmeasure-
mentsacrosstheentireCISPR-22frequencybandmakeitidealforperforminglow-costprecom-
pliancemeasurementsatdistancesbetween1to2m.Thisantennawasusedforalloftheradiated
measurementsinChapter4ofthisthesis.Thisantennawasalsousedforthelow-costantenna
characterizationspresentedinSections5.2and5.3.Lastly,thisantennawasusedaspartofthe
moretraditionalprecompliancetestsetupinSection5.4.
41
3.1.3HP8657ASignalGenerator
TheHP8657AsignalgeneratorshowninFigure3.5isasynthesizedsignalgeneratorwithacarrier
frequencyrangefrom100kHzto1040MHzandanamplituderangeof-127dBmto+13dBminto
50

[49].TheHP8657Ahasa50

RFoutputandcanbecontrolledfromacomputerviaitsGPIB
interface.Thisfeaturewasusedtoautomatenearlyallofthemeasurementsinthisthesisinvolving
thesignalgenerator.Allmeasurementsinthisthesisinvolvingthesignalgeneratorrequiredsignal
levelslessthan0dBm,sothepossibleamplitudeerrorcausedbythesignalgeneratorisassumed
tobelessthan1.0dB[49].Thiserrorisnotcriticaltotheresultsofthisthesisbecausethisthesis
isonlyinterestedinthemeasurementaccuracyoftheRSP2prorelativetotheEMCanalyzer.
Figure3.5:TheHP8657Asignalgenerator.
AnimportantlimitationontheaccuracyofthefrequencymeasurementsinSections3.4and4.2
arecausedbythespeci˝cHP8657Asignalgeneratorthatwasusedforthisthesis.Accordingto
themanufacturer[49],thesignalgenerator'soutputcenterfrequencyshouldneverstraymorethan
500Hzawayfromthecenterfrequencysetting.However,theunitusedforthisthesiswasfound
tohaveafrequencyerrormuchgreaterthanthis.Sinceanothersignalgeneratorwasnotavailable,
arudimentarycalibrationprocesswasimplementedby˝rstmeasuringtheactualfrequencyfor
eachdesiredfrequencythatwasgoingtobeusedandthencalculatinghowfaro˙themeasured
frequencywasforeachdesiredfrequency.
42
Thecenterfrequencysetting
5
2
ofthesignalgeneratorrequiredforthedesiredcenterfrequency
5
8=
canberepresentedas
5
2
=
5
8=

5
>
(3.1)
where
5
>
istheunknownfrequencyo˙setfrom
5
8=
thatmustbeaccountedfor.Therewere114
separatevaluesfor
5
8=
requiredforthemeasurementsinvolvingthesignalgeneratorpresentedin
thisthesis.Theprocedurefor˝ndingthevalueof
5
2
correspondingtoeachvalueof
5
8=
was
automatedviathePythonscriptoutlinedinFigure3.6.TheEMCanalyzerandthesignalgenerator
wereautomatedviatheirGPIBinterfacesusingcommandsfromthePyVISAlibrary[50].
Beforetheprocedurebegins,theEMCanalyzerandthesignalgeneratorareallowedtowarm
Figure3.6:FlowchartforthePythonscriptthatwasusedto˝ndthefrequencyo˙setsofthe
uncalibratedHP8657Asignalgenerator.
43
upforonehour.Thesignalgeneratoristhensettoanoutputlevelof50dB
`
Vandthe˝rstofthe
114
5
8=
valuesisloadedfromaninputcsv˝le.TheEMCanalyzerloadsthe"RESweep"state˝le
fromTable3.1andusestheloadedvalueof
5
8=
asthecenterfrequencyofthespectrumsweep.
TheEMCanalyzer'sbuilt-inpeaksearchfunctionthenmarksthepeakvaluewithinthe100kHz
measurementspan.Thefrequencyofthispeakisstoredasatemporaryvariable"x"andusedtoset
thecenterfrequencyofthesignalgeneratorforasecondmeasurement.Thissecondmeasurement
isperformedwiththe"DefaultSweep"state˝lesettings.Thismeansthatthefrequencyspanofthe
secondmeasurementismuchnarrower(3kHz),resultingina˝nerpeakfrequencymeasurement.
Thefrequencyofthispeakmeasurementisstoredasthenewvalueof"x."Thesemeasurements
arethenrepeatedusingthenewvalueof"x"astheinitialinputfrequency.
The˝nalvalueof"x"afterthisrepetitionistheresultoffourmeasurements.The˝nalvalueof
"x"isstoredinanewcsv˝leasthe
5
2
correspondingtothedesiredfrequency
5
8=
andtheentire
processisrepeatedforthenextvalueof
5
8=
fromtheinputcsv˝le.Thisprocedureresultedinaset
offrequenciesthatwasusedforallofthemeasurementsinthisthesisinvolvingthesignalgenerator.
Thismeansthattheactualsignalgeneratorcenterfrequencysettingforthesemeasurementsisnot
thedesiredfrequency
5
8=
,butthefrequency
5
2
fromEquation3.1.Figure3.7showshowthevalue
of
5
>
fromEquation3.1increasedlinearlyasthedesiredfrequency
5
8=
wasincreased.
Thiswholeprocedurehadtobeperformedtwiceindi˙erentsessions,asthevaluesof
5
8=
that
wereonlymultiplesof25MHzwerenotincludedinthe˝rstsession.OnecanseefromFigure
3.7thatbothdatasetsincreaselinearlywith
5
8=
,butthattheconditionsparticulartoeachsession
a˙ectedthevalueof
5
>
.Thiskindofinaccuracyisunavoidableandindicatestheimportanceof
measurementconditionswhenusingthesignalgenerator.
Onesuchmeasurementconditionistheoperatingtemperatureofthesignalgenerator,which
increasesoverthetimethatthesignalgeneratorispoweredon.Asthistemperatureincreases,itcan
causethecenterfrequencyofthesignalgeneratoroutputtovaryovertime.Thecharacterization
measurementresultsinSections3.3and3.4allinvolvedthesignalgeneratorandrepeatingthesame
measurementstoacquirealargenumberofsamplesforeachofthe114valuesof
5
8=
.Sometimes
44
Figure3.7:Plotofthe
5
>
associatedwitheachvalueof
5
8=
fromEquation3.1.
thesemeasurementstookaslongas14hourstocomplete,someasurementsatthesamefrequency
wouldoccurhoursapart.Automatingthesemeasurementsmadethiskindofdatacollection
possible,butthelongmeasurementtimesmeantthattheequipmentwassubjecttoperiodsof
continuousoperationandthuspotentialfrequencydriftfromchangesintheoperatingtemperature.
TheEMCanalyzerhasabuilt-incalibrationroutinethatwasusedtopreventpotentialfrequency
drift,butsucharoutinewasnotanavailableforthesignalgenerator.Tobetterunderstandthe
impactofcontinuousoperation,aprocedurewasusedtocharacterizethefrequencydriftoftheHP
8657Asignalgeneratorovera24-hourperiod.ThePythonscriptusedtoautomatethisprocedure
isoutlinedinFigure3.8.TheEMCanalyzerandthesignalgeneratorwereautomatedviatheir
GPIBinterfacesusingcommandsfromthePyVISAlibrary[50].
Beforetheprocedurestarts,theEMCanalyzerisallowedtowarmupforonehour.The˝rst
frequency
5
2
fromthefrequencytablecreatedbythePythonscriptfromFigure3.6isusedasthe
centerfrequencyoftheEMCanalyzer.Thesignalgeneratoristhensettoaconstantoutputlevel
45
Figure3.8:FlowchartforthePythonscriptthatwasusedtocharacterizethefrequencydriftofthe
signalgeneratorovera24-hourperiod.
of50dB
`
Vandanewcsv˝leiscreatedforthecurrenthour.Thesignalgeneratorfrequencyisset
to
5
2
andtheEMCtraceisstoredtothenewcsv˝le.Thisprocessisrepeateduntiltheendofthe
frequencytableisreached,afterwhichthereisadelayofexactlyonehour.Duringthishourthe
EMCanalyzerandthesignalgeneratorarelefton.
Afteranhourhaspassed,theEMCanalyzeriscalibratedandtheentireprocessisrepeateduntil
therearedatasetsforeachofthe114
5
2
valuesoverafull24-hourperiod.Therecordedtraces
foreachhourcanbecomparedtoseehowthesignalgeneratorcenterfrequencyataparticular
frequencysettingdriftsfromhour-to-hour.
Figure3.9showsthefrequencydriftofthesignalgeneratoratallfrequencysettingsoverthe
˝rsttwelvehoursofa24-hourperiod.Onecanseethatallfrequencysettingshavesigni˝cant
46
driftoverthe˝rsthour,butthenthefrequencydriftfromhour-to-hourgreatlydecrease.Figure3.9
alsoshowsthathigherfrequencieshaveamuchlargerfrequencydrift,whereaslowerfrequencies
becomeverystableafteronlytwohours.Notethattheconstantvalueobservedforthe
C
=
0
hourdatafromapproximately620MHzonwardoccursbecausetheyfrequencysettingsabovethis
frequencydriftedbymorethanthemeasurementbandwidthoftheEMCanalyzer.
Figure3.9:Measuredsignalgeneratorfrequencydriftfor
C
=
0
to
C
=
12
hours.
Figure3.10showsthefrequencydriftforallfrequencysettingsoverthesecondtwelvehours
ofa24-hourperiod.Onecanseethatthefrequencydriftatthelowerfrequencysettingsisalmost
non-existentwhilethefrequencydriftatthehigherfrequencysettingsismuchmoresigni˝cant.
However,thefrequencydriftatallfrequencysettingsisstillwellbelow2kHz.
Therearetwokeyconclusionsfromtheresultsofthefrequencydriftcharacterizationprocess.
The˝rstisthatthefrequencydriftoftheHP8657Aiswellbelow1000Hzifthesignalgeneratoris
allowedtowarmupforatleastonehour.Whilethisamountofdriftisnotwithinthespeci˝cation
[49],itiscloseenoughtohavecon˝dencewhencomparingtherelativefrequencymeasurements
betweentheEMCanalyzerandtheRSP2pro.Forthisreason,awarmuptimeofonehourwas
usedforallofthemeasurementsinvolvingthesignalgenerator.Thesecondconclusionisthat
47
Figure3.10:Measuredsignalgeneratorfrequencydriftfor
C
=
13
to
C
=
24
hours.
extendedperiodsofoperationaltimeincreasethestabilityofthesignalgeneratorcenterfrequency.
However,theamountofstabilityvariessigni˝cantlywiththesignalgeneratorfrequencysetting.
Thisvarianceiswellbelow2kHzatallfrequencies,solongmeasurementtimesdonotsigni˝cantly
a˙ecttheresultspresentedinSection3.4.
Figure3.11showsthefullCISPR-22ClassBspectrummeasuredbytheEMCanalyzerdirectly
connectedtothesignalgeneratorwiththesignalgeneratorsettoanoutputlevelof50dB
`
Vat
afrequencyof500MHz.Fromthis˝gureonecanseethatthesignalgeneratoroutputhasno
unexpectedRFsignals.Figure3.12showsadetailedviewofthe500MHzsignalshowninFigure
3.11.This˝gureshowsthatthesignalgeneraotroutputisextremelynarrowband,witha3-dB
bandwidthofonly300Hz.
48
Figure3.11:Theoutputofthesignalgeneratorwithafrequencysettingof500MHzandasignal
levelof50dB
`
V.
Figure3.12:DetailedviewofthesignalinFigure3.11showinghownarrowbandthesignal
generatoroutputis.
49
3.2CharacterizationProcedureandMeasurementAutomation
AllthreeoftheantennafactorsshowninFigures3.2and3.4awereusedtogeneratedatasetsfor
controllingthesignalgeneratoramplitudeintheprocedurediscussedinSection3.2.Thesespeci˝c
antennaswerechosenbecausetheyrepresentthreecommontypesofEMCantennas.Rearranging
Equation2.5andsolvingforthereceivervoltage
+
yields:
+

3`+

=
ˆ

3`+
š
<


˙

3
1
š
<

.(3.2)
RadiatedemissionsneartheCISPR-22ClassBlimitinFigure2.7iswheretheRSP2proneedstobe
themostaccuratewhencomparedtotheEMCanalyzerbecausethisisthelevelexpectedfroman
EUTthatisbreakingtheregulatorylimit.Testingtheaccuracyinthisregionisachievedbyusing
theCISPR-22ClassBradiatedemissionslimittosetthesignalgeneratorvoltage.Byreplacing
+
inEquation3.2withthesignalgeneratorvoltage
+
6
andreplacing
ˆ
withtheCISPR-22ClassB
radiatedemissionlimitataparticularfrequency
ˆ
;8<8C
,theequationforthesignalgeneratorvoltage
ataparticularfrequencycanberepresentedas
+
6

3`+

=
ˆ
;8<8C

3`+
š
<


˙

3
1
š
<

¸
˘

3

(3.3)
wheretheantennafactor
˙
canbetheantennafactorvalueataparticularfrequencyforoneof
thethreeEMCantennasdescribedinSection3.1.2and
˘
isanadjustmentfactorthatisinitially
setto0dB.Tounderstandtheperformanceatotherpotentialradiatedemissionlevels,severalextra
datasetswerecreatedbyadjustingthevalueof
˘
by

10dBtocreateatotalofninedatasets.All
ninedatasetshavebeensummarizedinTable3.2.Inthistable,
+
6
and
˘
arefromEquation3.3
andtheirentriesrepresentthevaluesofthesevariablesforaparticularantennafactor(or
˙
as
showninEquation3.3).Thecsv˝lesforthesedatasetsareavailableintheDigitalAppendixthat
isincludedwiththedigitalversionofthisthesis(refertoAppendixBformoreinformation).
The˝fthcolumnofTable3.2describesthesourceusedforthevalueof
˙
inEquation3.3.
Forexample,DataSet1wascreatedwiththeantennafactorfortheAgilent11955AEMCantenna
(showninFigure3.2)andavalueof10dBfor
˘
.Thefrequencyrangeandstepsizeforeach
50
Table3.2:DescriptionoftheninedatasetsthatwereusedtocharacterizetheEMCanalyzerand
theRSP2pro.
datasetwasdeterminedbythefrequencycomponentoftheantennafactorparticulartothatdata
set.FromthefourthcolumnofTable3.2,onecanseethatthesignalgeneratoramplitudelevel
+
6
variedoverawiderangeacrosstheninedatasets.Thishighlightshowalargerangeofinput
voltagelevelsweretestedbyusingjustnineinputdatasets.EachdatasetinTable3.2hasbeen
assignedanumberthatisusedwhentheyarereferredtoinSections3.3and3.4.
ThesetupusedtocharacterizetheRFsignalmeasurementaccuracyoftheRSP2proisshownin
Figure3.13.Inthis˝gure,thesetuphasbeencon˝guredforRSP2promeasurementsbutthesame
setupwasalsousedfortheEMCanalyzermeasurements.Thelaptoppicturedwasconnectedtothe
EMCanalyzerandthesignalgeneratorviaaGPIB-to-USBadapter.TheRSP2prowasconnectedto
thelaptopviaUSB.Thetestsetup'stwoRFsignalchainsareshowninFigure3.14.ThesameSMA
cablewasusedtoforboththeEMCanalyzerandRSP2promeasurements.Theonlydi˙erence
betweenthemeasurementchainsshowninFigure3.14wastheadditionalRFadapterthatwas
necessaryforconnectingtheEMCanalyzertotheSMAcable.ThreeseparateRSP2prounitswere
characterizedtoinvestigatehowthemeasurementperformancevariedacrossseveralunits.
51
Figure3.13:ThecharacterizationtestsetupconsistingoftheEMCanalyzer,signalgenerator,
RSP2pro,andlaptop.HerethesignalgeneratorisshownconnectedtotheRSP2pro.
Figure3.14:DiagramshowingtheRFsignalchainsusedforthecharacterizationtestsetup.
52
Forthefourdevicesthatwerecharacterized(theEMCanalyzerandthreeRSP2prounits),the
measurementresultsfromthenineinputdatasetsshowninTable3.2werecollected˝veseparate
times.Eachofthesetimesisreferredtoasa"set"(foratotalof˝vemeasurement"sets"per
inputdataset)andeachofthese"sets"werecollectedatadi˙erenttime,usuallydaysapartfrom
oneanother.Foreachofthese˝vesets,theindividualmeasurementswithinasetwererepeated
tentimes.Eachofthesemeasurementsisreferredtoasa"trial."Thismeansthatatotalof50
measurementswereperformedforeachdatapointofthenineinputsets.Thusforeachdevicethat
wascharacterized,23,850individualmeasurementswereperformedforagrandtotalof95,400
individualmeasurements.Tocollectsuchalargeamountofdata,thecharacterizationprocesswas
automatedasmuchaspossible.Di˙erencesintheuserinterfacesoftheEMCanalyzerandthe
RSP2prounitsmadethecharacterizationprocessforeachslightlydi˙erent.Forthisreason,two
separate,semi-automatedprocessesweredevelopedforperformingcharacterizationmeasurements.
Thesemi-automatedprocedureforperformingtheEMCanalyzercharacterizationmeasure-
mentsisoutlinedinFigure3.15.TheEMCanalyzerandthesignalgeneratorwereautomatedvia
theirGPIBinterfacesusingcommandsfromthePyVISAlibrary[50].Theprocessbeginswiththe
usermanuallyselectingthemeasurementsetnumber,whichstartsatone.Theuserthenstartsthe
automatedPythonscript.
ThePythonscript˝rstcommandstheEMCanalyzertoloadthe"DefaultSweep"setup˝le.The
˝rstofthenineinputdatasetsfromTable3.2isthenloadedandthefrequencyandvoltagesettings
fromthisdatasetareusedtosettheoutputofthesignalgenerator.TheEMCanalyzercaptures
thetracedataandrecordsthemeasuredquasi-peak(QP)value.Thisprocessisrepeateduntildata
hasbeenrecordedforallofthedatapointsfromtheinputdataset.Afterallofthedatapointsare
collected,theyaresavedtoanewcsv˝leandthe˝rsttrialiscomplete.After˝vetrials,theEMC
analyzeriscalibratedbeforethenext˝vetrialsarerecorded.Oncetentrialsarecollected,theEMC
analyzeriscalibratedagainandthewholeprocessstartsoverwiththenextinputdataset.After
collectingtentrialsforeachofthenineinputdatasets,theautomatedPythonscriptendsandthe
usermustmanuallyenterthemeasurementsetnumberandstartthePythonscriptagain.After˝ve
53
Figure3.15:Flowchartforthesemi-automatedprocedurethatwasusedtocollecttheEMCanalyzer
characterizationdata.
54
measurementsetshavebeencollected,theEMCanalyzercharacterizationprocedureiscomplete.
Semi-automatingthisprocesswasrelativelystraight-forwardthankstotheEMCanalyzerand
signalgeneratorGPIBinterfaces.ThisisbecausetheGPIBcommandsinthePythonscriptarejust
mimickingtheuserinputsontheEMCanalyzerandthesignalgeneratorfrontpanels.Toslightly
improvethespeedofeachmeasurement,theEMCanalyzerdisplaywasdisabled.Thispreventsthe
EMCanalyzerfromhavingtowastetimecreatingthetraceonthedisplay.Thefewsecondsthat
thissavedforeachindividualmeasurementresultedinamuchfasteroverallcollectiontimegiven
thelargenumberofindividualmeasurementsthatwereperformed.
Thesemi-automatedprocedureforperformingtheRSP2procharacterizationmeasurementsis
outlinedinFigure3.16.ThesignalgeneratorwasautomatedviaitsGPIBinterfaceusingcommands
fromthePyVISAlibrary[50].TheRSP2prounitsrequireduseoftheSDRunocompanionsoftware
describedinSection2.4.Table3.3summarizestheSDRunosettingsthatwereusedforallofthe
RSP2promeasurementspresentedinthisthesis.Thesesettingswerechosentohaveameasurement
spancomparabletothe"DefaultSweep"usedbytheEMCanalyzer.TheRSP2prowasalsousedin
thelow-IFmode,whichhasfeweroptionsfortheRSP2prosamplingratecomparedtotheoptions
availablewithzero-IFmodeoperation.AsdiscussedinSection2.3,thelow-IFoperationreduces
theamountofintermodulationproductsattheoutputoftheSDRmakingitmoresuitablethanthe
zero-IFmodeinspectrumanalysisapplications.WhileSDRunosupportsseveralcon˝gurations
forthelow-IFmode,itwasimportanttousethesamesettingsforallmeasurementssothataccurate
comparisonscouldbemadebetweentheresultspresentedthroughoutthisthesis.
SincetheSDRunouserinterfaceisdesignedformanualuserinputviaamouseandkeyboard,
aspecializedcontrolsoftwarecalledAutoItwasusedtoautomatetheRSP2promeasurements.
AccordingtothetheAutoItwebsite[51],AutoItisa"scriptinglanguagedesignedforautomating
theWindowsGUI"that"usesacombinationofsimulatedkeystrokes,mousemovementandwin-
dow/controlmanipulationinordertoautomatetasksinawaynotpossibleorreliablewithother
languages."BycombiningacustomAutoItscriptwithSDRuno'sbuilt-inabilitytosavemeasure-
mentdatatoacsv˝le,itwaspossibletocompletelyautomateSDRunomeasurements.TheAutoIt
55
Table3.3:SummaryoftheSDRunosettingsusedforalloftheRSP2promeasurementspresented
inthisthesis.
scriptisoutlinedbytheyellowelementsshowninFigure3.16.
TheprocedureinFigure3.16beginswiththeusermanuallyconnectingthe˝rstRSP2prounit's
PortASMAconnectortothesignalgeneratorasshowninFigure3.14.Theuserthenconnectsthe
RSP2protothehostlaptopandmanuallyselectsthemeasurementsetnumber,whichstartsatone.
Finally,theuserstartstheAutoItscript.
TheAutoItscript˝rststartsaPythonscriptthatcontrolsthesignalgenerator.ThisPython
scriptthenloadsthe˝rstinputdatasetfromthenineshowninTable3.2andthefrequencyand
voltagesettingsfromthisdatasetareusedtosettheoutputofthesignalgenerator.ThePython
scriptthenwaitsfortheAutoitscripttomeasureadatapoint.WhilethePythonscriptiswaiting,
theAutoItscriptopensSDRunoandappliesthesettingsshowninTable3.3.Itthencreatesanew
storagecsv˝leforthecurrenttrial.Next,theAutoItscriptsetstheRSP2protunerfrequencyand
optimizestheRFgain(theRFgainoptimizationprocessisshownseparatelyinFigure3.17).
UsingSDRuno'sbuilt-inpeakmeasurementbutton,AutoItrecordsthemeasuredfrequencyand
amplitudeofthestrongestsignalinthe5kHzmeasurementspan.Thispeaksignalistheoutput
fromthesignalgeneratorforthecurrentdatapoint.Whenthedatapointisrecorded,thePython
scriptdetectsthe˝lechangeinthestoragecsv˝leandupdatesthesignalgeneratorsettingswith
56
Figure3.16:Flowchartforthesemi-automatedprocedurethatwasusedtocharacterizetheRSP2pro
units.
57
thenextvaluesfromtheinputdataset.Thisprocessisrepeateduntiltentrialshavebeencollected.
Aftertentrialshavebeencollected,theAutoItscriptrestartsSDRunoandtheentireprocess
isrepeateduntiltentrialshavebeencompletedforeachofthenineinputdatasets.Afterall
ofthetrialshavebeencollected,theAutoItscriptterminatesthePythonscriptandcloses.The
usermanuallyupdatesthemeasurementnumberandthenstartstheAutoItscriptagain.After
˝vemeasurementsetshavebeencollected,theuserrestartstheentireprocedurewiththenext
RSP2prounit.After˝vemeasurementsetshavebeencollectedforallthreeoftheRSP2prounits,
theRSP2procharacterizationprocedureiscomplete.
Figure3.17showshowtheRFgainwasoptimizedforeachRSP2promeasurement.According
to[40],thebestwaytooptimizetheRFgainsettingistoincreasetheRFgainuntilananalog-to-
digitalconverter(ADC)overloadoccurs.Fromhere,thegaincanbereducedbyoneortwostages
untiltheADCoverloadhascleared.TheyellowADCoverloadindicatorcanbeseenintheSDRuno
windowshowninFigure3.17.ThankstoAutoIt'sabilitytousescreenpixelcolorsasfunction
inputs,theAutoItscriptcandetectwhenanADCoverloadoccursbymonitoringthepixelsnear
theADCoverloadindicator.
TheRFgainoptimizerinFigure3.17startswiththeAutoItscriptsettingtheRFgaintozero.
AutoItthenincreasestheslider(showninthecenteroftheSDRunowindowshowninFigure3.17)
byonestage.Next,thescriptdetectsthescreenpixelcolorneartheADCoverloadindicatortosee
ifanoverloadhasoccurred.Ifthecheckedpixelisnotyellow,thegainstageisincreaseduntilan
ADCoverloadoccursandthepixelisyellow.Thegainsettingisthenreducedbyonestageand
theRFgainhasbeenoptimized.WhentheADCoverloadoccursatthelowestgainsetting,theRF
gaincannotbeoptimized.ThishappenswhentheinputRFsignalistoostrong.ThenumberofRF
gainstagesavailablevarydependingonthetunerfrequency[52],sotheamountoftimeittookto
optimizetheRFgainataparticularfrequencyvariedthroughoutthecharacterizationprocess.
58
Figure3.17:TheRFgainoptimizerthatispartoftheAutoItscriptoutlinedinFigure3.16.An
outlineoftheAutoItcodeisshownontheleftandtheSDRunowindowwiththeRFgaincontrol
andADCoverloadindicatorisshownontheright.
59
3.3SignalLevelMeasurementPerformance
InthissectiontheaccuracyoftheRSP2pro'ssignallevelmeasurementcapabilityisexamined
bydirectcomparisontotheEMCanalyzer.Onekeydi˙erencebetweenthetwodevicesisthe
abilitytodisplaymeasuredRFsignalsindi˙erentunits.TheEMCanalyzerisabletodisplaythe
amplitudeofmeasuredsignalsinavarietyofunitssuchasdBm(decibelsrelativetoonemilliwatt)
anddB
`
V.However,SDRunoisonlycapableofdisplayingtheamplitudeofmeasuredsignalsin
dBm.
AccordingtotheRSP2prodatasheet[42],theRSP2pro'sPortARFinputhasanimpedanceof
50

.Withthisknowledge,SDRunomeasurementsrecordedindBmcanbeconvertedtodB
`
Vso
thattheycanbecomparedtothemeasurementsmadewiththeEMCanalyzer.NotethattheEMC
analyzermeasurementsweremadeinunitsofdB
`
Vbecausethegoalofthisthesisistousethe
RSP2protomeasureradiatedemissions,whichhaveunitsofdB
`
V/m.TheconversionfromdBm
todB
`
VcanbeaccomplishedbyderivinganequationthatrelatesdBmanddB
`
Vforasystem
thatisassumedtohaveapurelyresistiveimpedance,
'
.
AccordingtoPaul[1],thepowerofasignalexpressedindBm,
%
3<
,isde˝nedas
%
3<
=
10log
10

%

10

3

where
%
representsthepowerofthesignalinWatts.Rewritingthisequationintermsof
%
3<
yields
%
=
¹
10

3
º
10
%
3<
š
10
.(3.4)
AlsoaccordingtoPaul[1],thevoltageofasignalexpressedindB
`
V,
+
3`+
,isde˝nedas
+
3`+
=
20log
10

+

10

6

(3.5)
where
+
representsthevoltageinVolts.Rewritingthisequationintermsof
+
3`+
yields
+
=
¹
10

6
º
10
+
3`+
š
10
.(3.6)
Forasystemthatisassumedtobepurelyresistive,thetimeaveragepower
%
isrelatedtotheRMS
voltageofthesignal
+
byOhm'slawandcanbeexpressedas
%
=
+
2

'
.(3.7)
60
ThederivationbeginsbysubstitutingEquations3.4and3.6intoEquation3.7andrearranging
termssothatthe
%
3<
and
+
3`+
termsareonthesamesideoftheequation:
%
=
+
2

'
¹
10

3
º
10
%
3<
š
10
=

¹
10

12
º
10
+
3`+
š
10


'

¹
10

3
º
10
%
3<
š
10

10

+
3`+
š
10
=
10

12

'
10
¹
%
3<

+
3`+
ºš
10
=
10

9

'
.
Thenextstepsinthederivationaretotakethelogarithmofeachsideoftheequationandtothen
usethelogarithmproductandpowerrulestosimplifyeachterm:
log
10
h
10
¹
%
3<

+
3`+
ºš
10
i
=
log
10

10

9

'

¹
%
3<

+
3`+
º

10
=
log
10

1
š
'

¸
log
10

10

9

¹
%
3<

+
3`+
º

10
=
log
10

1
š
'


9
.
The˝nalstepistosimplifytheremainingtermsfurtherandtosolvefor
+
3`+
:
%
3<

+
3`+
=
10log
10

1
š
'


90
+
3`+
=
%
3<

10log
10

1
š
'

¸
90
.(3.8)
61
Equation3.8showsthatfora50

system
+
3`+
=
%
3<

10log
10

1
š
'

¸
90





'
=
50

=
%
3<

10log
10

1
š
50

¸
90
=
%
3<
¸
16
Ł
9897004
ŁŁŁ
¸
90
=
%
3<
¸
106
Ł
9897004
ŁŁŁ
ˇ
%
3<
¸
107
.
ThustheconversionbetweendBmanddB
`
Vfora50

systemcanbeperformedwiththeequation
+
3`+
=
%
3<
¸
107
.(3.9)
ThisequationwasusedtoconvertalloftheRSP2promeasurementspresentedinthisthesisinto
dB
`
VsothattheycouldbedirectlycomparedwiththemeasurementstakenwiththeEMCanalyzer.
Equation3.9isverycommonlyusedwhenperformingradiatedemissionmeasurementsbecause
mostspectrumanalyzersandreceivershavea50

RFinput.ThisequationisalsohowtheHP
8657AsignalgeneratorconvertsitsoutputsignalunitsfromdBmtodB
`
V[49].NotethatEquation
3.9onlyholdstrueifthesystemhasa50

impedance.Ifthesystemhassomeotherimpedance
(forexample,75

tomatchwithsometypesofco-axialcable),Equation3.8canbeusedinstead.
Equation3.9alsoassumesthattheimpedanceofthesystemismatchedto50

atallfrequencies.
WhiletheRSP2prodoeshavea50

RFinput,itsinputwillnotbeperfectlymatchedatall
frequencies.Thisistrueforallinstrumentswitha50

inputoroutput[26].AccordingtoAgilent
[26],themeasurementerror
"ˆ
indBduetotheimpedancemismatchbetweenaspectrumanalyzer
andtheRFsourceitismeasuringcanbecalculatedas
"ˆ
=

20log
10

1





d
0

d
B





(3.10)
where
d
0
and
d
B
arethemagnitudeofthere˛ectioncoe˚cientsforthespectrumanalyzerinput
andtheRFsource,respectively.There˛ectioncoe˚cientmagnitude
d
isrelatedtothevoltage
62
standingwaveration(VSWR)withthefollowingequation:
d
=
¹
VSWR

1
º

¹
VSWR
¸
1
º
.(3.11)
TheE7401AEMCanalyzerhasacharacteristicRFinputVSWRof1.35forsignalsmeasured
from1to1500MHz[43]andtheHP8657Asignalgenerator'sRFoutputhasaVSWRof(atmost)
1.5forpoweroutputlevelsettingslessthan+5dBm[49].UsingEquation3.11tocalculatethe
re˛ectioncoe˚cientforboththeEMCanalyzer(
d
0
)andthesignalgenerator(
d
B
)yields
d
0
=
¹
1
Ł
35

1
º

¹
1
Ł
35
¸
1
º
=
0
Ł
1489362
ŁŁŁ
ˇ
0
Ł
15
,
d
B
=
¹
1
Ł
5

1
º

¹
1
Ł
5
¸
1
º
=
0
Ł
20
.
ThesevaluescanthenbeusedwithEquation3.10tocalculateboththenegativeerrordueto
mismatch
"ˆ

=

20log
10

1
¸
0
Ł
030

=

20
¹
0
Ł
128372
ŁŁŁ
º
=

0
Ł
256744
ŁŁŁ
ˇ
0
Ł
26
dB
andthepositiveerrorduetomismatch
"ˆ
¸
=

20log
10

1

0
Ł
030

=

20
¹
0
Ł
132282
ŁŁŁ
º
=
0
Ł
264565
ŁŁŁ
ˇ
0
Ł
26
dB
63
ThusalloftheEMCanalyzermeasurementswherethesignalgeneratorwasusedasasourcehave
anerrorduetomismatch,
"ˆ
ˆ
,ofatleast
"ˆ
ˆ
=

0
Ł
26
dB.(3.12)
ThislossiscomparabletotheoverallamplitudemeasurementaccuracyoftheEMCanalyzer,which
accordingto[43]isapproximately

0.30dB.
ThereisadditionalsignallosscausedbytheSMAcablethatwasusedwithboththeRSP2pro
unitsandtheEMCanalyzer(refertoFigure3.14).Usinganetworkanalyzer,theS-parametersof
theSMAcableweremeasuredandusedtocomputetheVSWRatbothendsoftheSMAcablefrom
30MHzto1000MHz.TheseareshowninFigure3.18.ThemeasuredS-parameterswerealso
usedtoplottheinsertionlossofthecableinFigure3.18b.Notethatthey-axisshowninFigure
3.18astartsat0.90sothatadetailedviewofthemeasuredVSWRcanbeshown.
OnecanseethatatbothendsoftheSMAcablethereisalmostaperfectimpedancematch
forfrequenciesupto200MHz.ThisisindicatedbytheVSWRbeingveryclosetooneoverthis
frequencyrange.Atfrequenciesabove200MHz,theimpedancemismatchbeginstoincreaseinan
oscillatorymannerthatvariesbetweenaVSWRof1.00and1.04.Theconclusionhereisthatany
impedancemismatcherrorsinthemeasurementspresentedinthisthesisarepredominantlycaused
bythedevicesconnectedtotheSMAcable,nottheSMAcableitself.
Figure3.18bshowsthattheinsertionlossoftheSMAcabledoesbecomesigni˝cantrelative
tothe

0.30dBmeasurementaccuracyoftheEMCanalyzerwhenthesignalfrequencyisgreater
thanorequaltoapproximately600MHz.Theinsertionlossofthecableisgreatestaround1000
MHz,whereithasavalueofalmost0.40dB.Whenperformingcomplianceradiatedemission
measurements,theinsertionlossofanycablesusedforsaidmeasurementsmustbeaccountedfor
beforeusingtheantennafactortoconvertthemeasuredvoltageintotheradiatedemission˝eld
strength.However,themeasurementsinthisthesisarefocusedoncomparingtheRSP2protothe
EMCanalyzerandallofthecomparisonmeasurementswereperformedwiththesameSMAcable.
ThismeansthattheinsertionlossshowninFigure3.18bcanbeignoredwhendirectlycomparing
themeasurementaccuracyoftheRSP2protothemeasurementaccuracyoftheEMCanalyzer.
64
Figure3.18:(a)TheVSWRand(b)theinsertionlossoftheSMAcablethatwasusedforallofthe
measurementspresentedinthisthesis.
TocharacterizetheRFsignalmeasurementaccuracyoftheRSP2pro,theprocessdescribedin
Section3.2wasusedtocollectmeasurementswiththeEMCanalyzerandthreeRSP2prounits.
ThemeasurementsusedavarietyofCWsignalsthateachhadaspeci˝cfrequencyandsignallevel
(refertoTable3.2).Figure3.19showstheaveragevoltagefrom50separatemeasurementsmade
usingtheEMCanalyzerforeachofthenineinputdatasets.
ThesignalsshowninFigure3.19aand3.19brepresentthedatasetscorrespondingtothe
Agilent11955Aand11956Aantennafactors(respectively),whileFigure3.19cshowsthesignals
correspondingtotheBicoLOG30100Xdatasets.Oneachoftheseplots,themiddletraceshows
thevoltagelevelforareceivedsignalthatwouldbeexactlyattheCISPR-22ClassBlimitwhen
measuredwiththatparticularEMCantenna.Theothertracesoneachplotcorrespondtothedata
setsthatwerecreatedbyusing
˘
valuesof

10dBwithEquation3.3.Onecanseethatthesignals
65
Figure3.19:AverageEMCanalyzervoltagemeasurementfor:(a)DataSets1,2,and3,(b)Data
Sets4,5,and6,and(c)DataSets7,8and9.Eachdatapointrepresentstheaverageof50
measurements.
66
inDataSets1through6carrysigni˝cantlylesspowerthanthoseinDataSets7,8,and9.The
abruptshiftinsignalpowerobservedat230MHzonalloftheseplotsisfromtheCISPR-22Class
BradiatedemissionlimitthatwasshowninFigure2.7;at230MHz,thelimitsincreasefrom30
dB
`
V/mto37dB
`
V/m.
ThemeasurementsshowninFigure3.19werealsoperformedusingthreeRSP2prounits.Note
thatthefrequencyofeachdatapointpresentedinthissectionhasbeenadjustedtothedesiredsignal
generatorfrequency
5
8=
sothatthedi˙erencesinthemeasuredsignallevelofeachdatapointcan
bemoreeasilycompared.
Figures3.20,3.21,and3.22showthestandarddeviationofeverydatapointfromtheRSP2pro
andEMCanalyzermeasurementsthatwascollectedusingtheinputdatasetsfromTable3.2.Each
ofthese˝gurescorrespondswithoneofthethreeEMCantennasdiscussedinSection3.1.2(i.e.
each˝gureshowstheresultsfromthreeofthedatasetsinTable3.2).
Figure3.20showsthestandarddeviationresultsforDataSets1,2,and3.Thesecorrespond
withtheAgilent11955Aantennafactorand
˘
valuesof+10dB,0,and-10dB(respectively).One
canseefromFigure3.20a,3.20b,and3.20cthatthestandarddeviationoftheRSP2prounitsis
veryconsistentacrosstheentirefrequencyrange,withanaveragevalueofapproximately0.10
dB
`
VatallfrequenciesacrosstheCISPR-22ClassBband.TheEMCanalyzerhasalower
standarddeviationthantheRSP2prounitsinFigure3.20aandasimilarstandarddeviationtothe
RSP2prounitsinFigure3.20b.Interestingly,Figure3.20cshowsthattheEMCanalyzerhasa
muchlargerstandarddeviationthantheRSP2prounits.ThisismostlikelybecauseDataSet3has
thesecondlowestoverallsignalpowerlevelofthenineinputdatasetsandtheRSP2prohasabetter
receiversensitivity(-147dBm)thantheEMCanalyzer,whichhasadisplayedaveragenoiselevel
ofapproximately-119dBmforthefrequencyrangeshowninFigure3.20[43].
ThissametrendcanbeseeninFigure3.21,whichshowsthestandarddeviationresultsforData
Sets4,5,and6.TheseinputdatasetscorrespondwiththeAgilent11956Aantennafactorand
˘
valuesof+10dB,0,and-10dB(respectively).Allthreeplotsinthis˝gureshowthattheRSP2pro
unitseachhaveastandarddeviationofapproximately0.10dB
`
VwhiletheEMCanalyzerhasa
67
Figure3.20:RSP2proandEMCanalyzervoltagemeasurementstandarddeviationfor:(a)Data
Set1,(b)DataSet2,and(c)DataSet3.Eachdatapointrepresentsthestandarddeviationfor50
measurements.
68
Figure3.21:RSP2proandEMCanalyzervoltagemeasurementstandarddeviationfor:(a)Data
Set4,(b)DataSet5,and(c)DataSet6.Eachdatapointrepresentsthestandarddeviationfor50
measurements.
69
Figure3.22:RSP2proandEMCanalyzervoltagemeasurementstandarddeviationfor:(a)Data
Set7,(b)DataSet8,and(c)DataSet9.Eachdatapointrepresentsthestandarddeviationfor50
measurements.
70
signi˝cantlyhigherstandarddeviation,asshowninFigure3.21c.DataSet6hasthelowestoverall
signalpowerlevelofthenineinputdatasets,andFigure3.21ciswherethestandarddeviationof
theEMCmeasurementsvariesthemost.
Figure3.22showsthestandarddeviationresultsforDataSets7,8,and9.Theseinputdata
setscorrespondwiththeBicoLOG30100Xantennafactorand
˘
valuesof+10dB,0,and-10dB
(respectively).Figure3.19showsthatDataSet7istheinputdatasetwiththehighestoverallsignal
level,andFigure3.22ashowsthattheEMCanalyzerhastheleastamountofstandarddeviation
withthisdataset.DataSets8and9arethesecondandthirdmostpowerfulinputdatasets,and
Figure3.22band3.22cshowthatforthesedatasetstheEMCanalyzermeasurementshaveabetter
standarddeviationthantheRSP2prounitsatallfrequencies.Thismeansthattheinputpowerlevel
hasalargeimpactonthemeasurementconsistencyoftheEMCanalyzer.Surprisingly,theRSP2pro
unitseachappeartohavethesamestandarddeviationvalueof0.10dB
`
Vatallfrequenciesdespite
thehigheroverallpoweroftheseinputdatasets.Figure3.22band3.22calsoshowthatthestandard
deviationoftheEMCanalyzerandtheRSP2proslightlyincreasewithfrequency.However,this
increaseisnegligiblecomparedtotheaveragestandarddeviationobservedatallfrequenciesacross
theCISPR-22ClassBband.
ThemainconclusionfromFigures3.20,3.21,and3.22isthatthestandarddeviationfor
individualRFsignalmeasurementsislessthan0.25dB
`
VforthethreeRSP2prounitsthatwere
tested.ThisperformanceisonlyslightlyworsethantheEMCanalyzer,andinthecaseoflow
powerRFsignalmeasurements,thestandarddeviationperformanceoftheRSP2prounitsisactually
betterthanthatoftheEMCanalyzer.ThisismostlikelyduetotheRSP2pro'sexcellentreceiver
sensitivityatthefrequencyrangesthatweretested.
Figures3.23,3.24,and3.25showtheaveragemeasuredamplitudedi˙erencebetweentheEMC
analyzerandeachofthethreeRSP2prounitsforeverydatapointcollected(usingtheinputdatasets
fromTable3.2).Eachofthese˝gurescorrespondwithoneofthethreeEMCantennasdiscussed
inSection3.1.2(i.e.each˝gureshowstheresultsfromthreeofthedatasetsinTable3.2).For
71
these˝gures,theaverageo˙setvalue
+
3
isde˝nedas
+
3
=
+
'

+
ˆ
(3.13)
where
+
'
representstheaveragevoltagemeasuredbytheRSP2proataparticularinputdatapoint
and
+
ˆ
representstheaveragevoltagemeasuredbytheEMCanalyzeratthatsameinputdata
point.Thevalueof
+
3
hasbeenplottedforallnineoftheinputdatasetsfromTable3.2so
thattheamplitudemeasurementaccuracyofallthreeRSP2prounitscanbecomparedwiththe
amplitudemeasurementaccuracyoftheEMCanalyzer.RecallthattheEMCanalyzeramplitude
hasanabsolutemeasurementaccuracyof

0.30dB[43]foralloftheinputdatapointstested.By
usingtheEMCanalyzerforcomparison,anewmarginoferrorcanbedevelopedformeasurements
performedwiththeRSP2pro.
Figure3.23showstheaverageo˙setvalueofEquation3.13forDataSets1,2,and3.The˝rst
observationisthatmeasurementsfromallthreeRSP2prounitsremainveryconsistentacrossall
threeplots.ThisconsistencyisindicatedbythenearidenticaltrendofeachoftheRSP2prodata
setsshowninallthreeplotsofFigure3.23.AllthreeRSP2prounitsalsohavethesamerelative
measurementaccuracy.TheRSP2pro1measurementsareslightlyhigherthantheEMCanalyzer
measurement,whiletheRSP2pro2measurementsareconsistentlysmallerthantheEMCanalyzer
measurements.Forallthreeplots,onecanseethatthemeasurementsforalloftheRSP2prounits
arewithin1.00dBoftheEMCanalyzermeasurementacrosstheentirefrequencyrange.Eachof
theRSP2prounitsalsoperformednearlyidenticallyatfrequenciesbetween260and300MHz.
Figure3.24showstheaverageo˙setvaluefromEquation3.13forDataSets4,5,and6.While
theplotsshowninFigure3.24aand3.24bshowthatthemeasurementsfromallthreeRSP2pro
unitsareconsistent,thereislessconsistencyobservedinFigure3.24c.Thisplotcorrespondsto
theinputdatasetwiththelowestoverallsignallevel.AllthreeplotsshowthatRSP2prounits1
and3haveaboutthesamemeasurementaccuracyrelativetotheEMCanalyzer,whileRSP2pro2
measurementsareconsistentlylowerthantheEMCanalyzermeasurements.UnlikeFigure3.23,
onecanseefromFigure3.24thatthemeasurementsforallthreeunitsarewithin2.00dBoftheEMC
72
Figure3.23:AverageRSP2provoltagemeasuremento˙setfor:(a)DataSet1,(b)DataSet2,and
(c)DataSet3.Eachdatapointrepresentstheaverageof50measurements.
73
Figure3.24:AverageRSP2provoltagemeasuremento˙setfor:(a)DataSet4,(b)DataSet5,and
(c)DataSet6.Eachdatapointrepresentstheaverageof50measurements.
74
Figure3.25:AverageRSP2provoltagemeasuremento˙setfor:(a)DataSet7,(b)DataSet8,and
(c)DataSet9.Eachdatapointrepresentstheaverageof50measurements.
75
analyzermeasurementacrosstheentirefrequencyrange.Thislargeroverallaveragemeasurement
o˙setcanbeattributedtothelowpowerlevelofDataSets4,5,and6(refertoFigure3.19b).
Figure3.25isthe˝nalo˙setvoltagecomparisonandshowstheaverageo˙setvaluefrom
Equation3.13forDataSets7,8,and9.Figure3.25aand3.25bshownearlyidenticaltrendsforthe
RSP2pro1and3measurements.Intheseplots,theRSP2pro1and3measurementsalsohavenearly
identicalo˙setsfromtheEMCanalyzermeasurementfrom400MHzto600MHzand750MHz
to1000MHz.Figure3.25cshowsasimilartrend,buttheaverageo˙setforbothunitsisactually
betterintherangefrom600MHzto800MHz.AllthreeoftheplotsinFigure3.25highlightthat
themeasurementsfromRSP2pro2weresigni˝cantlydi˙erentfromtheothertwounits.Whilethis
trendcanalsobeobservedinFigures3.23and3.24,thelargerfrequencyrangeinthis˝guremore
clearlyillustratestheextentthisdi˙erence.DespiteRSP2pro2'spoorerperformance,theoverall
averageo˙setforallthreeRSP2prounitsisstillwithin2.00dBoftheEMCanalyzermeasurements.
TheresultsshowninFigure3.23,3.24,and3.25indicatethatforallofthesignallevels
tested,thethreeRSP2prounitswereabletohaveameasurementaccuracywithin

2.00dBof
theEMCanalyzer.Measurementsfromtwooftheunitswereabletoachieveanaccuracyofjust
over

1.50dBrelativetotheEMCanalyzer.Thislevelofaccuracyisacceptableforlow-cost
precompliancetesting,wheretheradiatedemissionmeasurementerrorscausedbytheuncontrolled
RFenvironmentcanrangefrom+6dBto-25dB[23].
WhiletheprocedureinSection3.2hasbeenusedheretocompareseveralRSP2prounitsto
anEMCanalyzer,itcanalsobeusedtocalibratetheRSP2pro.WithinSDRuno,theusercan
manuallyaccountforadditionalRFgainexternaltotheRSP2prounitbyadjustingthemeasured
powerlevelbyaconstantvalueindB[53].Thisfeatureisintendedtoaccountforadditionalgains
andlossesintheRFsignalchainconnectedtotheRSP2pro,butcanalsobeusedtomanually
calibratethemeasuredsignallevelifsaidsignalleveliscomparedtothelevelmeasuredbyanother
calibrateddevice(e.g.aspectrumorEMCanalyzer).However,thisadjustmentisconstantatall
frequencies,soitwouldneedtobemanuallyadjustedforeachmeasurementtoaccountforany
frequency-dependentdi˙erences.Forexample,inFigure3.25theaverageo˙setfortheRSP2pro
76
2measurementsisonly-0.50dBinthefrequencyrangefrom30MHzto200MHzbutisnearly
-2.00dBinthefrequencyrangefrom600MHzto800MHz.
Theaccuracyof

2.00dBobservedhereissu˚cientforreplacingaspectrumanalyzerfor
low-costprecomplianceradiatedemissionmeasurements.Thecharacterizationtestsetupallowed
foraveryhighSNRbecausetheRSP2prounitsandtheEMCanalyzerweredirectlyconnectedto
thesignalgeneratorviaanSMAcable.Thisdirectconnectionminimizedthenoise˛oorforallof
themeasurements,providingahighSNRforeverymeasurement.
AccordingtoHorowitz[54],theSNRofameasuredsignalisaratioindBthatisde˝nedas
SNR
=
10log
10

E
2
B

E
2
=

(3.14)
where
E
B
and
E
=
representthermsvoltagesofthedesiredsignalandthesignalnoisethatisalso
present(respectively).Horowitz[54]explainsthat"thesquaredamplitudes[ofEquation3.14
suggest]aratiointermsofpower,whichistheoriginofthedecibelratiode˝nition"usedfor
SNR.TheSNRofameasuredsignalisdirectlyrelatedtothebandwidthofthesignalandthe
measurementbandwidthusedbythereceiver(i.e.theRSP2pro).Forexample,asignalwitha
narrowbandfrequencyresponsewillcontinuetohavethesamepowerlevelasthemeasurement
bandwidthisincreased.ThismeansthattheSNRofthemeasurementcandecreasebecausethe
receiverampli˝erwillcontinuetoaddnoisepowerasthemeasurementbandwidthbecomeslarger
thanthesignalbandwidth[54].Inthiscontext,themeasurementbandwidthisequivalenttothe
spansettingofaspectrumanalyzerorreceiver.RecallthatFigure3.11fromSection3.1.3shows
thatthesignalgeneratorusedforthecharacterizationmeasurementshadanarrowbandoutputsignal
witha3-dBbandwidthofonly300Hz.Thisbandwidthismuchsmallerthanthespanusedbyboth
theEMCanalyzer(3kHz)andtheRSP2pro(128kHz).
SDRUnoalsorecordstheSNRwhensavingthemeasuredsignalpeakandfrequency.The
SNRdatafromallthemeasurementresultsshowninFigures3.23,3.24,and3.25canbeusedto
understandhowtheSNRrelatestotheRSP2prosignalmeasurementaccuracy.
Figures3.26,3.27,and3.28showthestandarddeviationoftheSNRforeachofthedatapoints
thatwasmeasuredfromtheinputdatasetsshowninTable3.2.Eachofthese˝gurescorrespond
77
withoneofthethreeEMCantennasdiscussedinSection3.1.2(i.e.each˝gureshowstheresults
fromthreeofthedatasetsinTable3.2).
Figures3.26and3.27showthestandarddeviationoftheSNRforDataSets1through6.With
theexceptionofonefrequency(825MHz),theseplotsshowthattheSNRforeverydatapoint
testedhadastandardofdeviationlessthan0.5dB.Figure3.19showsthatthesedatasetsallhave
signaloutputlevelsbelow40dB
`
V.Thisimpliesthatfornarrowbandsignallevelslessthan40
dB
`
V,theRSP2proexperiencesaveryconsistentSNR.Thissamelevelofconsistencydidnot
occurwiththemuchhigherinputsignallevelsfromDataSet7,8,and9.Figure3.28showsthe
standarddeviationfortheSNRwhenmeasuringthesedatasets.
AllthreeplotsinFigure3.28showasigni˝cantamountofdeviationintheSNRmeasurements
fromallthreeRSP2prounits.Thelargestdeviationoccursinthefrequencyrangefrom230MHz
to500MHzinFigure3.28a.Thisfrequencyrangecorrespondswiththelargestinputsignallevels
fromDataSet7,whichistheinputdatasetwiththehighestoverallamplitudelevels(refertoFigure
3.19).ThisshowsthattheSNRobservedbytheRSP2provariesbymorethan1dBwhenthe
measuredsignalisanarrowbandsignalwithamagnitudeabove60dB
`
V.Thefrequencyrange
from30MHzto200MHziswhereDataSet9hasasignallevelbelow40dB
`
V,andherethe
standarddeviationofthemeasuredSNRiscomparabletowhatcanbeobservedinFigures3.26
and3.27.Whilethismayseemlikefurtherevidencethatnarrowbandsignallevelsbelow40dB
`
V
provideamuchmoreconsistentSNR,thisdiscrepancyactuallyoccursbecausetheanalog-to-digital
(A/D)convertersoftheRSP2proarebeingoverloadedwhenmeasuringthesignalshigherthan200
MHz.
AnA/Dconverteroverloadcanbeshownbyplottingthedi˙erencebetweenthemeasured
SNRofaninputsignalandtheSNRofthesameinputsignalreducedby10dB.Thedi˙erence
betweentheSNRmeasurementsshouldbeapproximately10dBbutwillbedi˙erentwhentheA/D
convertersareoverloaded.Thismethodworkswellwiththetestsetupusedforthecharacterization
processbecausethenoise˛oorisminimizedbythedirectconnectiontothesignalsource.Using
theinputdatasetsfromTable3.2,thedi˙erence
ˇ
GH
betweendatasetsthatare10dBapartcanbe
78
Figure3.26:RSP2proSNRmeasurementstandarddeviationfor:(a)DataSet1,(b)DataSet2,
and(c)DataSet3.Eachdatapointrepresentsthestandarddeviationfor50measurements.
79
Figure3.27:RSP2proSNRmeasurementstandarddeviationfor:(a)DataSet4,(b)DataSet5,
and(c)DataSet6.Eachdatapointrepresentsthestandarddeviationfor50measurements.
80
Figure3.28:RSP2proSNRmeasurementstandarddeviationfor:(a)DataSet7,(b)DataSet8,
and(c)DataSet9.Eachdatapointrepresentsthestandarddeviationfor50measurements.
81
usedtocreatesuchplots.Thisdi˙erenceisde˝nedas
ˇ
GH
=
SNR
G

SNR
H
(3.15)
wherexandycorrespondtothedatasetnumbersfromTable3.2.Thisequationwasusedwiththe
datasetsfromTable3.2thatusedthesameEMCantennaandwereseparatedby10dBtocreate
Figures3.29and3.30.
Figure3.29showsthatforDataSets7and8,thevalueofEquation3.15ismuchlessthan10
dBatnearlyeveryfrequency.ThisindicatesthattheA/DconvertersoftheRSP2prounitswere
likelyoverloadedwhenperformingmeasurementswiththeseinputdatasets.Figures3.29aand
3.30showthattheotherdatasetsdidnotoverloadtheA/Dconverterssincethedi˙erenceshownin
these˝guresisclosetotheexpectedvalueof10dB.Figures3.30aand3.30beachshowasimilar
trendforthevalueof
ˇ
23
and
ˇ
56
.Figure3.30cshowsthatthevalueofEquation3.15forData
Sets8and9isonlycloseto10forsignalslessthan200MHz.Thisexplainswhythestandard
deviationoftheSNRinFigure3.28cwasmorecomparabletoDataSets1through6;theA/D
convertersarenotbeingoverloadedforinputdatapointsfromDataSet9thatareatafrequency
below200MHz.
Figures3.31,3.32,and3.33showtheaverageSNRforeachofthedatapointsthatwasmeasured
fromtheinputdatasetsshowninTable3.2.Eachofthese˝gurescorrespondwithoneofthethree
EMCantennasdiscussedinSection3.1.2(i.e.each˝gureshowstheresultsfromthreeofthedata
setsinTable3.2).
Figure3.31showstheSNRformeasurementswhereDataSets1,2,and3weremeasured.
Theplotsinthis˝gureshowthatthemeasuredSNRateveryfrequencywasalmostidenticalfor
eachRSP2prounit.AlloftheSNRvaluesinFigure3.31arebetween24and56dB.Figure
3.32showsthattheSNRfortheDataSet4,5,and6measurementsarealsobetween16and56
dB.Forfrequenciesupto600MHz,allthreeRSP2prounitsobservedalmostexactlythesame
SNRforeverydatapoint.Figure3.32alsoshowsthattheSNRobservedbyeachRSP2provaried
signi˝cantlyat700MHz.
82
Figure3.29:ThevalueofEquation3.15using(a)DataSet1and2,(b)DataSet4and5,and(c)
DataSet7and8.
83
Figure3.30:ThevalueofEquation3.15using(a)DataSet2and3,(b)DataSet5and6,and(c)
DataSet8and9.
84
Figure3.31:RSP2proSNRmeasurementfor:(a)DataSet1,(b)DataSet2,and(c)DataSet3.
Eachdatapointrepresentstheaverageof50measurements.
85
Figure3.32:RSP2proSNRmeasurementfor:(a)DataSet4,(b)DataSet5,and(c)DataSet6.
Eachdatapointrepresentstheaverageof50measurements.
86
Figure3.33:RSP2proSNRmeasurementfor:(a)DataSet7,(b)DataSet8,and(c)DataSet9.
Eachdatapointrepresentstheaverageof50measurements.
87
Figure3.33showstheSNRformeasurementswhereDataSets7,8,and9weremeasured.In
this˝gure,theoverallSNRshownonallthreeplotsismuchhigherthantheSNRshowninFigures
3.31and3.32,witharangespanningfrom40to72dB.AllthreeRSP2prounitsappeartohavethe
sameSNRrelativetooneanother,onlydi˙eringbyafewdBatnearlyeverydatapoint.Thetwo
exceptionstothisareinFigure3.33c,wheretheSNRrecordedbyRSP2pro1and3isapproximately
6dBlowerthantheSNRrecordedbyRSP2pro2at700MHz.Asmentionedpreviously,DataSets
7,8,and9wereallabletooverloadtheA/DconvertersoftheRSP2prounitsthatweretested.
ThemainconclusionfromthissectionisthattheRSP2procanmeasureRFsignalswithan
amplitudeaccuracyof

2.00dBsolongastheA/DconvertersoftheRSP2proarenotoverloaded
bythemeasuredsignal.ThisamountofaccuracycanbemaintainedwithanSNRaslowas16dB.
88
3.4FrequencyMeasurementPerformance
ThefrequencymeasurementperformanceoftheRSP2prowascharacterizedbydirectlycom-
paringtothefrequencymeasurementaccuracyoftheEMCanalyzer.AsexplainedinSection3.1.3,
theEMCanalyzerwasalsousedtocalibratetheoutputfrequenciesofthesignalgenerator.This
meansthattheresultspresentedinthissectionarecompletelyrelativetotheEMCanalyzer,as
anindependentcalibratedfrequencysourcewasnotused.Regardless,thestandarddeviationand
di˙erenceinmeasuredfrequencybetweentheRSP2proandtheEMCanalyzerisstillusefulfor
understandinghowsuitabletheRSP2proisfordeterminingthefrequencyofradiatedemissions
duringalow-costprecompliancetest.Thefrequencyofeachdatapointpresentedinthissection
hasbeenadjustedtothedesiredsignalgeneratorfrequency
5
8=
usingEquation3.1sothatthe
di˙erencesintheactualfrequencymeasuredateachdatapointcanbecompared.
Figures3.34,3.35,and3.36showthestandarddeviationofeverydatapointfromtheRSP2pro
andEMCanalyzermeasurementsthatwerecollectedusingtheinputdatasetsfromTable3.2.Each
ofthese˝gurescorrespondwithoneofthethreeEMCantennasdiscussedinSection3.1.2(i.e.
each˝gureshowstheresultsfromthreeofthedatasetsinTable3.2).
Figure3.34showsthestandarddeviationresultscorrespondingtoDataSets1,2,and3.All
threeplotsinthis˝gureshowthatthefrequencymeasuredforeachdatapointwasextremely
consistentforallthreeoftheRSP2prounitsandtheEMCanalyzer.Thestandarddeviationforthe
RSP2prounitsisbelow20Hz,andtheEMCanalyzeronlyhasoneoutlierwithastandarddeviation
of120Hz.TheresultsfromusingDataSets4,5,and6indicateasimilarlevelofconsistencyand
areshowninFigure3.35.Figure3.35crepresentstheresultsfromusingtheinputdatasetwiththe
lowestoverallsignallevel.Here,theEMCanalyzerbecomeslessconsistentthantheRSP2prounits
asthemeasuredfrequencyincreases.However,theEMCanalyzerstillhasastandarddeviation
thatislessthan60Hzwhichismorethansu˚cientwhenidentifyingthefrequencyofradiated
emissions.
Figure3.36showsthestandarddeviationresultscorrespondingtoDataSets7,8,and9.The
plotsinthis˝gureshowthatallthreeoftheRSP2prounitsremainedveryconsistentinmeasuring
89
Figure3.34:RSP2proandEMCanalyzerfrequencymeasurementstandarddeviationfor:(a)Data
Set1,(b)DataSet2,and(c)DataSet3.Eachdatapointrepresentsthestandarddeviationfor50
measurements.
90
Figure3.35:RSP2proandEMCanalyzerfrequencymeasurementstandarddeviationfor:(a)Data
Set4,(b)DataSet5,and(c)DataSet6.Eachdatapointrepresentsthestandarddeviationfor50
measurements.
91
Figure3.36:RSP2proandEMCanalyzerfrequencymeasurementstandarddeviationfor:(a)Data
Set7,(b)DataSet8,and(c)DataSet9.Eachdatapointrepresentsthestandarddeviationfor50
measurements.
92
thefrequencyoftheinputsignalatallfrequencies.Thisisindicatedbythesmallseparation
betweentheRSP2prodatasetsinallthreeplots,asthelargestseparationbetweentheRSP2prodata
setsintheseplotsislessthan40Hz.Figure3.36bhasthelargestincreaseofstandarddeviation
fortheEMCanalyzerfrequencymeasurement,whichreachesamaximumvalueof150Hz.Thisis
amountofinconsistencyisstillinconsequentialforradiatedemissionmeasurements,whichhavea
centerfrequencyintherangefrom30MHzto1000MHz.
Figures3.37,3.38,and3.39showtheaveragedi˙erencebetweentheEMCanalyzermeasured
frequencyandthemeasuredfrequencyofthethreeRSP2prounitsforeverydatapointcollected
fromusingtheinputdatasetsfromTable3.2.Eachofthese˝gurescorrespondwithoneofthe
threeEMCantennasdiscussedinSection3.1.2(i.e.each˝gureshowstheresultsfromthreeofthe
datasetsinTable3.2).Forthese˝gures,theaveragefrequencyo˙set
5
3
isde˝nedas
5
3
=
5
'

5
ˆ
(3.16)
where
5
'
representstheaveragefrequencymeasuredbytheRSP2proforaparticularinputdata
pointand
5
ˆ
representstheaveragefrequencymeasuredbytheEMCanalyzerforthatsameinput
datapoint.Thevalueof
5
3
directlycomparesthefrequencymeasurementaccuracyofallthree
RSP2prounitswiththefrequencymeasurementaccuracyoftheEMCanalyzerforalloftheinput
datapointstested.
TheresultsshowninFigures3.37,3.38,and3.39showthatthefrequencyaccuracyofthe
RSP2pro(relativetotheEMCanalyzer)wasveryconsistentforalloftheinputdatasets.Every
plotinthese˝guresshowthatthefrequencymeasurementofRSP2pro2wassigni˝cantlymore
inaccuratethanthefrequencymeasurementstakenwiththeothertwounits.However,thisinaccu-
racywasstilllessthan1.6kHzfornearlyeverydatapointthatwastested.Thisfrequencyisvery
smallcomparedtothefrequencyofthesignalsencounteredduringCISPR-22ClassBradiated
emissionstesting,whichhavecenterfrequenciesintherangeof30MHzto1000MHz.Figures
3.38and3.39showthatthevalueof
5
3
forallthreeRSP2prounitsincreasedlinearlywiththe
inputsignalfrequencyforsignalsfrom200MHzto1000MHz.Onecanalsoobservethislinear
behavioroccurringforinputfrequenciesof30MHzto200MHzinFigure3.37.Thevalueof
5
3
in
93
Figure3.37:AverageRSP2profrequencymeasuremento˙setfor:(a)DataSet1,(b)DataSet2,
and(c)DataSet3.Eachdatapointrepresentstheaverageof50measurements.
94
Figure3.38:AverageRSP2profrequencymeasuremento˙setfor:(a)DataSet4,(b)DataSet5,
and(c)DataSet6.Eachdatapointrepresentstheaverageof50measurements.
95
Figure3.39:AverageRSP2profrequencymeasuremento˙setfor:(a)DataSet7,(b)DataSet8,
and(c)DataSet9.Eachdatapointrepresentstheaverageof50measurements.
96
this˝gureislessthan400Hzforinputfrequenciesupto200MHz.Thisisidenticaltotheresults
showninFigures3.38and3.39.
Therearetwomainconclusionsfromtheresultspresentedinthissection.First,thatthe
frequencymeasuredbytheRSP2proisveryconsistentforsignalsfrom30MHzto1000MHz,
varyinglessthan60Hzforalloftheinputdatapointsthatweretested.Thisismostlikelyduetothe
accuracyoftheRSP2pro'stemperature-compensatedoscillator(TCXO).Thesecondconclusionis
thatthevalueof
5
3
increasedlinearlywithanincreaseinthemeasuredfrequency.Thisindicates
thattherewaslikelyaninaccuracywiththefrequencymeasuredbytheEMCanalyzer,notthe
RSP2prounits,becausetheEMCanalyzerwastheonlycommonfactorbetweenallofthe
5
3
valuesshown.
97
CHAPTER4
USINGTHERSP2PROFORRADIATEDEMISSIONSTESTING
HavingestablishedamarginoferrorfortheRSP2pro'sabsolutesignallevelmeasurementsin
thepreviouschapter,theRSP2proisnowusedtomeasureradiatedemissionsfromtwodi˙erent
sourcesandtheresultsaredirectlycomparedtotheEMCanalyzer.Section4.1outlinestheradiated
emissionsmeasurementprocedureusedwithboththeEMCanalyzerandtheRSP2pro.Thissection
includesadescriptionoftheRFenvironmentthatwasselectedfortestingandprovidesadetailed
descriptionofthetwosignalsourcesthatwereusedtocreateradiatedemissions.Section4.2
presentstheresultsoftheradiatedemissiontests,andthereceivedRFpowermarginof

2.00dB
developedfortheRSP2prointhepreviouschapteriscomparedtotheseresults.
4.1RadiatedEmissionsMeasurementProcedure
Beforetheprocedureformeasuringradiatedemissionswasdeveloped,severallocationswere
examinedto˝ndanRFenvironmentsuitableforradiatedemissionstests.Sincetheprimarygoal
ofthisthesisistoshowthattheRSP2procanreplacethespectrumanalyzerusedinalow-cost
precompliancesetupliketheoneshowninFigure2.11,onlyuncontrolledRFenvironmentswere
considered.AnuncontrolledRFenvironmentinthiscontextisde˝nedasanareawheretheRF
signalsexternaltotheEUTbeingevaluatedcannotbecontrolled.TheseexternalRFemissionsare
typicallyfromsourceslikenearbyFMradiostationsorothertelecommunicationsequipmentthat
operateinthefrequencyrangefrom30MHzto1000MHz.Figures4.1a,4.2a,and4.3ashowthe
threeuncontrolledRFenvironmentsthatwereconsideredforradiatedemissiontesting.Foreach
environment,theEMCanalyzerandtheBicoLOG30100XEMCantennawereusedtoperform
ascanofthebackgroundRFenvironmentfrom30MHzto1000MHzusingthefour"Spectrum
Sweep"state˝lesfromTable3.1tocon˝guretheEMCanalyzer.
Figure4.1ashowsthe˝rstRFenvironmentthatwasconsidered:anoutdoorareaonMSU's
campusnearPavilionDrive.ThenearbyMSUObservatory(picturedinthebackgroundofFigure
98
4.1a)wasusedtopowertheEMCanalyzer.Figure4.1bshowsthebackgroundRFnoiseatthis
location.Onecanseethatthenoise˛oorinthisenvironmentisat50dB
`
V,whichissigni˝cantly
higherthantheidealnoise˛oorof20dB
`
Vfromthecharacterizationtestsetup(refertoFigure
3.11).AlsonotablearetheFMbroadcastsignals,thestrongestofwhicharevisiblebetween90
and100MHzwithamplitudesofover100dB
`
V.Figure4.1balsoshowsmanyotherbackground
RFsignals.AccordingtotheNationalTelecommunicationsandInformationAdministrationO˚ce
ofSpectrumManagement[55],thestrongsignalsobservedjustbefore200MHzarebroadcast
televisionchannels7through13,whilethestrongsignalsobservedjustbefore300MHzaremost
likelyfromavarietyofnearbymobiletelecommunicationstowers.Figure4.1balsoshowsstrong
signalsjustbefore600MHz,thestrongestofwhichhasanamplitudelevelabove100dB
`
V.These
correspondtotelevisionbroadcastchannels21through36[55].
ThereareseveralreasonswhytheoutdoorenvironmentshowninFigure4.1awasnotchosen
forradiatedemissionstesting.Onereasonwasasimplelackofconvenience,asweatherandtime
constraintsmadecollectingdataatthelocationdi˚cult.However,themainreasonwasthelarge
amountofbackgroundRFnoisethatwouldhavemadeidentifyingradiatedemissionsfromanEUT
verydi˚cult.TheissueofbackgroundRFnoiseisalsoaproblemforcreatingtheopenareatest
sites(OATS)usedforfullcompliancetesting,asthenumberofambientsignalsfromthingslike
digitalbroadcastsarealwaysincreasing[23].
ThenextRFenvironmentthatwasconsideredisshowninFigure4.2a.Thetunnelshownin
this˝gureconnectsthebasementsofAnthonyHallandtheG.M.TroutFoodScienceandHuman
NutritionBuildingonMSU'scampusandisapproximately2.1mwideand2.5mtall.ThisRF
environmentwasconsideredforradiatedemissiontestingbecauseitslocationgreatlyreducesthe
amountofambientRFsignals,asshowninFigure4.2b.OnecanseethattheFMbroadcast
transmissionsarestillpresent,buttheiramplitudeshavebeenreducedbymorethan30dBfrom
whatwasmeasuredattheoutdoorlocationshowninFigure4.1b.Themeasurednoise˛oorin
Figure4.2bisverylow,onlyslightlyhigherthanthe20dB
`
Vnoise˛oorpresentduringthe
characterizationmeasurementsthatwereshowninChapter3.
99
(a)
(b)
Figure4.1:(a)Theoutdoorareathatwasconsideredforradiatedemissiontestingand(b)the
ambientRFsignalsmeasuredattheoutdoorareafrom30MHzto1000MHz(averageof3trials).
100
(a)
(b)
Figure4.2:(a)Theundergroundtunnelthatwasconsideredforradiatedemissiontestingand(b)
theambientRFsignalsmeasuredintheundergroundtunnelfrom30MHzto1000MHz(average
of3trials).
101
ThisRFenvironmentwasultimatelyrejectedforthreemainreasons.First,theareawasnot
veryaccessibleandmadecollectingmeasurementsdi˚cultbecauseofthelargeamountoftimeit
tooktotransportthetestequipmentrequiredfortheradiatedemissiontests.Second,thenarrow
dimensionsofthetunnelmeantthattherewasanincreasedchanceforstrongsignalre˛ectionsfrom
theEUT.Thesekindofre˛ectionsaredi˚culttopredictandwouldhavecontributedtoalarge
amountofinconsistencyinradiatedemissionsmeasurements.Thelastreasonisthatthislocation
istoounique.Mostproductdesignerswillnothaveaccesstoanundergroundtunnelforperforming
radiatedemissionmeasurements.
Figure4.3ashowsthehallwayintheMSUEngineeringBuildingthatwaschosenforperforming
theradiatedemissionmeasurementspresentedinthisthesis.Thishallwayislocatedonthesecond
˛ooroftheMSUEngineeringBuildingandhasawidthofapproximately3.7mandaceilingheight
ofabout2.5m.OnecanseefromFigure4.3bthatalthoughthenoise˛ooriscomparabletothe
undergroundtunnel,therearemanymoreambientRFsignalspresent(especiallyinthefrequency
rangefrom80MHzto500MHz).Theseadditionalsignalsgreatlyobscurethenoise˛ooracross
mostofthe30MHzto1000MHzfrequencyrange,saveforthesmallregionfrom30MHzto40
MHz.However,thelargestambientRFsignalsarestillreducedbymorethan20dBcomparedto
theresultsfromtheoutsideareashowninFigure4.1b.
TheRFenvironmentshowninFigure4.3awaschosenfortworeasons.Themainreasonis
thattheareaislikelycomparabletothekindofuncontrolledRFenvironmentthatwouldbereadily
availabletoanelectronicproductdesigner.Theothermainreasonisthatthislocationallowedfor
testingat10mwhilealsobeingveryclosetowherethetestequipmentwasstored.Thisproximity
tothetestequipmentstoragelocationgreatlyreducedsetuptime,allowingforagreaternumberof
radiatedemissionteststobeperformed.
OncetheRFenvironmentshowninFigure4.3awasselected,aprocedurewasdevelopedfor
performingconsistentradiatedemissionmeasurementsatadistanceof1mand10m.These
distanceswerechosensothatthee˙ectsoftheuncontrolledRFenvironmentthatwerediscussed
inSection2.2couldbeinvestigated.Figure4.4showsthetestsetupusedfortheradiatedemissions
102
(a)
(b)
Figure4.3:(a)Theindoorhallwaythatwasultimatelychosenforradiatedemissiontestingand(b)
theambientRFsignalsmeasuredinthehallwayfrom30MHzto1000MHz(averageof3trials).
103
measurementprocedureforbothdistances.ThetestsetupconsistedoftheEMCanalyzerandthe
RSP2prounits,whichwereplacedonamovabledesk,andtheBicoLOG30110XEMCantenna,
whichwasmountedonacameratripod.Forbothofthedistancestested,theradiatedemissions
sourcewasplacedonafoamteststandthatwasthesameheightastheEMCantenna.Sincethe
goaloftheradiatedemissiontestprocedurewastocomparetheperformanceoftheRSP2proand
theEMCanalyzer,theantennawasonlyusedatasingle˝xedheight.Whenperforminganactual
radiatedemissiontestwithanEUT,theantennashouldbeverticallyscannedwhenpossiblesothat
thestrongestradiatedemissionscanbefound[2].Figure4.5showstheRFsignalchainthatwas
usedbybothtestsetupsshowninFigure4.4.
(a)
(b)
Figure4.4:TheradiatedemissionstestsetupconsistingoftheEMCanalyzer,BicoLOG30100X
EMCantenna,RSP2pro,andradiatedemissionssourceat(a)10mand(b)1m.
Figure4.5:DiagramshowingtheRFsignalchainsusedfortheradiatedemissionstestsetup.
104
4.1.1NarrowbandSource:WhipAntenna
Threetypesofsignalsthatareverycommonwithelectricalsystemsarecontinuouswave(CW)
narrowbandsignals,repetitivebroadbandsignals,andsingleeventbroadbandsignals[56].The
˝rsttwotypesweretestedfortheresultspresentedinthisthesis.Thissectiondescribesthe
radiatedemissionssourcethatwasdesignedtocreateCWnarrowbandradiatedemissions.Table
4.1describesthefrequencyandamplitudeoftheeleveninputdatapointsthatwereusedbytheCW
sourceshowninFigure4.6.Thesespeci˝cfrequencieswereprimarilyselectedduetotheoutput
limitationsofthebroadbandsourcedescribedinSection4.1.2.
Table4.1:InputdatapointsusedbytheCWradiatedemissionssourceshowninFigure4.6.
TheVHF/UHF"whip"antennashowninFigure4.6wastheradiatedemissionssourcethat
wasusedtotransmittheCWsignalsfromtheHP8657Asignalgenerator.Accordingtothe
manufacturer[57],thisomni-directionalwhipantennaisdesignedfortransmittingRFsignalsin
thefrequencyrangesof136to174MHzand400to520MHz.Thisantennaalsohasamagnetic
basesothatitcanbeattachedtometallicsurfacesliketheexteriorofavehicle[57].Inorderto
understandhowthisantennawouldperformwhentransmittingsignalsatotherfrequenciesinthe
30MHzto1000MHzband,anetworkanalyzerwasusedtomeasureitsVSWR.TheVSWRofthe
antennawasmeasuredwhileitwasattachedtoametallicsurface(Figure4.7a)andonceagainwhen
itwasplacedonthefoamteststand(Figure4.8a).Theresultsofeachmeasurementareshownin
105
Figure4.6:DetailedviewofthewhipantennaarrangementusedtosimulateCWnarrowband
radiatedemissionsfromanEUT.
Figures4.7band4.8b.Fromthese˝gures,onecanseethatusingthemagneticbasegreatlye˙ects
theperformanceofthewhipantenna.
Whenattachedtoametallicsurface,themeasuredVSWRofthewhipantennaapproachesa
valueofoneattwoseperatefrequencies:154MHzand450MHz.Thisistobeexpectedbasedupon
themanufacturerspeci˝cations[57].Anantennaisoftenconsideredtohaveagoodimpedance
matchwithitssourcewhentheVSWRislessthantwo[58],andonecanseefromFigure4.7bthat
overmostofthe30MHzto1000MHzrangethewhipantennahasapoorimpedancematch.This
isespeciallytrueforfrequenciesbelow100MHz,andthisisthereasonwhyonlythreefrequencies
below100MHzweretested(refertoTable4.1).
OnecanseefromFigure4.8bthatthemeasuredVSWRwassigni˝cantlydi˙erentwhenthe
antennawasnotmagneticallyattachedtoametallicsurface.Inthiscon˝guration,thereareseveral
106
(a)
(b)
Figure4.7:(a)ThewhipantennausedtocreateCWnarrowbandradiatedemissionsattachedtoa
metalsurfaceand(b)itsmeasuredVSWRfrom30MHzto1000MHz(averageof5measurements).
107
(a)
(b)
Figure4.8:(a)ThewhipantennausedtocreateCWnarrowbandradiatedemissionsonanon-
metallicsurfaceand(b)itsmeasuredVSWRfrom30MHzto1000MHz(averageof5measure-
ments).
108
othernarrowfrequencyregionswithlowVSWRvaluesandthetwolargestnullsshiftslightlyup
infrequency,withnewcenterfrequencyvaluesofapproximately180MHzand480MHz.During
testing,itwasfoundthattheradiatedemissionscreatedwiththeantennaat480MHzwerecapable
ofoverloadingtheEMCanalyzerandtheRSP2pro,evenatadistanceof10m.Thisiswhythe
inputdatapointfor480MHzshowninTable4.1istheonlyinputdatapointwithanamplitude
of50dB
`
V.NotethatthemeasuredVSWRcanvarygreatlydependingontheorientationofthe
antenna(particularlywiththeantennaspositionrelativetootherobjectsinthenear˝eld)andso
thevaluesshowninbothFigures4.7band4.8bshouldonlybeconsideredtypicalvaluesforthe
speci˝ccon˝gurationsshowninFigures4.7aand4.8a.
4.1.2BroadbandSource:CustomEquipmentUnderTest(EUT)
Repetitivebroadbandradiatedemissionscaneasilybecreatedbycommonelectronicsystemcom-
ponentssuchasdigitalclocks[56].Thesetypesofemissionsareusuallyfromtheharmonics
createdbyanelectricsignalatsomefundamentalfrequency.Theseharmonicsoccuratinteger
multiplesofthefundamentalfrequencyandcanhaveradiatedemissionsthatarelargerthanthose
causedbytheelectricsignalatthefundamentalfrequency.Digitalclocksignalscauseparticularly
largeradiatedemissionsbecauseoftheirfastrisetimes[2].Whilefastrisetimesaredesirablefrom
adigitalsignalprocessingperspective,Ott[2]explainsthat"repetitive,high-frequencysignalswith
largecurrentsandfastrise/falltimeswillhavelargespectracontent."Clocksignalstypicallyhavea
50%dutycycle,whichmeansthattheoddharmonicsofthefundamentalfrequencywilldominate
thespectralcontentoftheclocksignal[4]andthustheoddharmonicswillcreatethestrongest
radiatedemissions.
Thelargespectralcontentofclocksignalscanbeexploitedtocreateacircuitthatiscapable
ofcreatingradiatedemissionsatauser-selectablefrequency.AcustomEUTthatdoesthiswas
createdforuseinthisresearch.Thedesignwasbaseduponasimilarcircuit[7]thatusesa10MHz
squarewaveoscillatortodemonstratethatcommon-modecurrentsarethedominantsourceofthe
radiatedemissionsabove30MHzthatarecreatedbysystemcabling.AsexplainedbyPauland
109
Bush[7],thesourceandloadsidesofthecircuitarepoweredviabatterytoshowthatdevicesnot
connectedtothecommercialpowergridcancreatelargeradiatedemissionsfromcommon-mode
currents.Asimpli˝edblockdiagramofthecustomEUTbasedonthedesignin[7]isshownin
Figure4.9.Figure4.10showsadetailedviewofthesourcesideoftheEUT.Theloadsideofthe
customEUTissimplyalogicinverterwithaninputcapacitanceof15pFanditsownisolated5V
powersupply.Thedesign˝lesforthecustomEUTareavailableintheDigitalAppendixthatis
includedwiththedigitalversionofthisthesis(refertoAppendixBformoreinformation).
ThecustomEUTisabletoproducesignalsatmultipledi˙erentfrequenciesusingahigh-
frequencymultiplierICthatusesaphased-lockloop(PLL)implementation.PLLsaremixed
signaldevicesconsistingofaphasedetector,ampli˝er,andvoltage-controlledoscillator[54].
AccordingtoHorowitz[54],PLLsaretypicallyusedintelecommunicationsdevicestosetthe
frequencyofeachoperatingchannel.InthecustomEUTshowninFigure4.9,thePT7C4511
PLLICisusedwiththreedi˙erentcrystaloscillatorstoproducemultipleoutputfrequencies.The
smalltanrotaryswitchesshowninFigure4.10areusedtoselectthecrystaloscillatorusedby
thePLL(10,20,or30MHz).TheredandwhitetoggleswitchesshowninFigure4.10arethen
usedtocontrolthedigitallogicinputsofthePLLICaccordingtothedatasheet[59].ThePLL
ICusestheseinputstodeterminethefrequencymultiplierappliedtotheinputfrequencytocreate
thedesiredoutputfrequency.Theharmonicfrequenciesofseveraloutputfrequenciesareshown
inTable4.2andwereusedtomeasureradiatedemissionsacrosstheentire30MHzto1000MHz
spectrum.
ThethirdcolumninTable4.2showstheharmonicfrequencies(andthustheradiatedemission
frequencies)thatweremeasured.ThistablealsoshowshowtheEUT'sribboncable(withalength
of41cm)comparestothewavelengthofeachofthemeasuredfrequencies.Radiatedemissions
becomestrongerwhentheconductorsintheEUTbecomecomparabletoawavelength[4],sothe
harmonicfrequenciesabove120MHzshouldproducethestrongestradiatedemissions.
Boththetime-domainandfrequency-domaincharacteristicsofthecustomEUTwerecharacter-
izedusingthesetupsshowninFigure4.11aand4.11b.A1GHzoscilloscopewith500MHzprobes
110
Figure4.9:Simpli˝edblockdiagramofthecustomEUTconstructedtocreatewidebandradiated
emissionsatmultiplefrequencies.RefertoAppendixAforthecustomEUTcircuitdiagrams.
Figure4.10:DetailedviewofthesourcesideshowninFigure4.9.
111
Table4.2:DetailsforthefrequenciesthatwerecreatedbythecustomEUTtotestradiatedemissions
acrosstheCISPR-22ClassBfrequencyband.
wasusedtomeasurethetime-domaincharacteristicsofthesignalcreatedbythePLLICateach
endoftheribboncableconnectingthesourceandloadsidesofthecustomEUT.Acurrentprobe
wasthenusedtomeasurethecommon-modecurrentontheribboncablesothatthemagnitude
oftheharmonicsofthePLLICsignalcouldbedetermined.Thecurrentprobewasplacedatthe
centeroftheribboncablefollowingthemethodusedbyPaulandBush[7],whofoundthatthe
common-modecurrentonaribboncableatVHFandUHFfrequenciesremainsmostlyconstant
alongtheentirelengthofthecable.ThisprocedurewasrepeatedforalloftheEUTPLLfrequencies
showninthe˝rstcolumnofTable4.2.
Figures4.12,4.13,4.14,and4.15showtheresultsfromthecustomEUTcharacterization
procedure.Eachofthese˝guresshowsthetime-domaincharacteristicsatthetopandthefrequency-
domaincharacteristicsatthebottom.Also,themostsigni˝cantharmonicshavebeenmarkedin
eachofthefrequency-domainplots.Thecurrentprobethatwasusedhasavery˛atresponse
from10MHzto500MHz:itstransferimpedancedecreasesbylessthan10dBoverthisentire
frequencyrange.Thisallowsforafaircomparisonbetweenthemagnitudesoftheharmonicsthat
weremeasuredbetween10MHzand500MHz.
Figure4.12showsthecharacteristicsofthecustomEUTwhensettoafrequencyof40MHz.
OnecanseethatthesignalreceivedbytheloadsideoftheEUTissigni˝cantlydistortedwhen
112
(a)
(b)
Figure4.11:(a)Thetestsetupusedtomeasurethetime-domaincharacteristicsofthecustom
EUTand(b)thetestsetupwiththecurrentprobethatwasusedtomeasurethefrequency-domain
characteristicsofthecustomEUT.
113
comparedtothesignalthatwasmeasuredonthesourcesideoftheEUT.Alsonotableisthenearly
3.5nsincreaseintherisetimebetweenthesourceandtheload.Thisindicatesthatthesignalis
becomingveryspreadoutasittravelsalongtheribboncable.Thefrequencycharacteristicsshown
inFigure4.12indicatethatthefundamentalfrequencyof40MHzwasnotthestrongestsignal
presentontheribboncable,butthatthesecondharmonicat80MHzwas.Themagnitudesofthe
markedharmonicsaredisplayedtotherightoftheplot,where
+
5
indicatesthemagnitudeofthe
harmonicfrequencyand
+
G5
indicatesthemagnitudeofthe
G
thharmonic.Thesemagnitudesshow
that(withtheexceptionofthesecondharmonic)theoddnumberedharmonicsweresigni˝cantly
strongerthantheevenharmonics.Thisisbecausethedutycycleofthesignalcomingfromthe
sourcesideoftheEUTisnearly50%.The˝rst,second,andthirdharmonicsshowninFigure4.12
weremeasuredduringthebroadbandradiatedemissionstesting.
Figure4.13showsthecharacteristicsofthecustomEUTwhensettoafrequencyof50MHz.
OnecanseethatthesignalreceivedbytheloadsideoftheEUTissigni˝cantlydistortedwhen
comparedtothesignalthatwasmeasuredonthesourcesideoftheEUT.Therisetimeonthe
sourcesideoftheribboncableisalsosigni˝cantlyshorterthantherisetimeobservedattheload
side.ThefrequencycharacteristicsinFigure4.13showthatthefundamentalfrequencyof50MHz
wasthestrongestsignalpresentontheribboncable.Interestingly,themagnitudesofthemarked
harmonicsthataredisplayedtotherightoftheplotshowthattheevenharmonicsarestrongerat
thefrequenciesabove450MHz.The˝rst,second,andthirdharmonicsshowninFigure4.13were
measuredduringthebroadbandradiatedemissionstesting.
Figure4.14showsthecharacteristicsofthecustomEUTwhensettoafrequencyof120MHz.
OnecanseethatthesignalreceivedbytheloadsideoftheEUTissigni˝cantlyreducedwhen
comparedtothesignalthatwasmeasuredonthesourcesideoftheEUT.However,therisetimeof
thesignalhasincreasedbylessthan1ns.Thisindicatesthattheamplitudeisbeingreducedbysome
othermeans.Table4.2showsthatat120MHz,theribboncableisequivalenttoone-sixthofthe
signalwavelength.Thismeansthattheribboncablecanquitee˙ectivelyradiate120MHzsignals,
andsotheloadsideattenuationislikelyduetoradiation.Thefrequencycharacteristicsshownin
114
Figure4.12:TimeandfrequencycharacteristicsoftheEUTwhensettothebasefrequency
5
=
40
MHz.Thesigni˝cantharmonicsof
5
havebeenmarkedwithan"x"andtheirvaluesareshown.
115
Figure4.13:TimeandfrequencycharacteristicsoftheEUTwhensettothebasefrequency
5
=
50
MHz.Thesigni˝cantharmonicsof
5
havebeenmarkedwithan"x"andtheirvaluesareshown.
116
Figure4.14:TimeandfrequencycharacteristicsoftheEUTwhensettothebasefrequency
5
=
120
MHz.Thesigni˝cantharmonicsof
5
havebeenmarkedwithan"x"andtheirvaluesareshown.
117
Figure4.15:TimeandfrequencycharacteristicsoftheEUTwhensettothebasefrequency
5
=
160
MHz.Thesigni˝cantharmonicsof
5
havebeenmarkedwithan"x"andtheirvaluesareshown.
118
Figure4.14supportthis,asthemagnitudeofthe120MHzsignalisthestrongestobservedyetat
over90dB
`
V.Themagnitudesofthemarkedharmonicsdisplayedtotherightoftheplotshow
that(withtheexceptionofthesecondharmonic)theoddnumberedharmonicsaresigni˝cantly
strongerthantheevenharmonics,againduetothe50%dutycycleoftheEUTsignal.Thesecond,
third,andfourthharmonicsshowninFigure4.14weremeasuredduringthebroadbandradiated
emissionstesting.Notethatthe120MHzsignalmeasuredforradiatedemissionstestingwasthe
thirdharmonicofthe40MHzsignalshowninFigure4.12,asthefundamentalfrequencyofthe
120MHzsignalshowninFigure4.14overloadedtheEMCanalyzerandtheRSP2pro.
Figure4.15showsthecharacteristicsofthecustomEUTwhensettoafrequencyof160MHz.
Interestingly,atthisfrequencythesignalcreatedbythePT7C4511PLLICisalmostperfectly
sinusoidal.Aswasobservedat120MHz,thesignalreceivedbytheloadsideoftheEUTis
signi˝cantlyreducedwhencomparedtothesignalthatwasmeasuredonthesourcesideofthe
EUT.Also,therisetimeoftheloadsidesignalhasincreasedbylessthan0.1ns.Table4.2shows
thatat160MHz,theribboncableisequivalenttoone-˝fthofthesignalwavelength.Thismeans
thattheribboncablecane˙ectivelyradiate160MHzsignals,andsotheloadsideattenuationis
likelyduetothesignalbeingradiated.ThefrequencycharacteristicsshowninFigure4.15support
this,asthemagnitudeofthe160MHzsignalisverystrong(nearly90dB
`
V).Themagnitudes
ofthemarkedharmonicsdisplayedtotherightoftheplotshowthat(withtheexceptionofthe
secondharmonic)theoddnumberedharmonicsarestillstrongerthantheevenharmonicsdespite
thesinusoidalshapeofthe160MHzsignal.The˝rstand˝fthharmonicsshowninFigure4.15
weremeasuredduringthebroadbandradiatedemissionstesting.
FollowingthemethoddescribedbyPaulandBush[7],thecommon-modecurrentmeasured
bythecurrentprobe(
˚
?
)canbeusedtopredicttheradiatedemissionsoftheEUTataspeci˝c
frequency
5
withtheequation
j
ˆ
2
<0G
j
=

6
Ł
28

10

7

˚
?
!5

'

(4.1)
where
!
isthelengthoftheconductors,
'
representsthemeasurementdistanceinm,and
j
ˆ
2
<0G
j
representsthemaximumelectric˝eldcreatedbythecommon-modecurrentsalongtheconductors
119
inunitsofV/m.Notethathere,
˚
?
hasunitsofmicroamps.Tocompare
j
ˆ
2
<0G
j
totheCISPR-22
ClassBlimits,itisconvertedtodB
`
V/musingEquation3.5.
Thevoltagesmeasuredwiththecurrentprobe(showninthebottomofFigures4.12,4.13,4.14,
and4.15)canbeconvertedtothemeasuredcurrent
˚
?
ofEquation4.1usingthetransferimpedance
ofthecurrentprobe.ThetransferimpedancehasunitsofdB

.Using
+
?
torepresentthevoltage
measuredbythecurrentprobeataparticularfrequency,themeasuredcurrent
˚
?
canbecalculated
withtheequation
˚
?
=
+
?

/
)
(4.2)
where
/
)
isthecurrentprobe'stransferimpedanceatthatfrequency.Notethathere
˚
?
hasunitsof
dB
`
A.Alsonotethatthemeasuredcurrent
˚
?
isequivalenttotwicethevalueofthecommon-mode
currentpresentontheribboncable[4].
Table4.3showsthepredictedradiatedemissionsforthecustomEUTatthefrequenciesthat
weretested(thefrequenciestestedareshownincolumnthree).ThePLLfrequencyusedtocreate
eachradiatedemissionfrequencyisshownincolumnoneforreference.Onecanseefromtheresults
inTable4.3thattheradiatedemissionscreatedbythecustomEUTaresigni˝cantlyhigherthan
CISPR-22ClassBlimits.Itwasdiscoveredduringinitialtestingthattheradiatedemissionscreated
bythecustomEUTwereactuallycapableofoverloadingtheEMCanalyzerandtheRSP2pro,so
twoferritecoreswereusedontheribboncabletoreducethecommon-modecurrentsintheribbon
cableanddampentheradiatedemissions.Theferritecoreswereplacedaroundtheribboncable
onthesourcesideoftheEUTasshowninFigure4.16.
Figure4.16alsoshowsthecon˝gurationofthecustomEUTthatwasusedforallofthe
broadbandradiatedemissiontests.ThecustomEUTwasplacedonafoamteststandthatwasthe
sameheightastheEMCantenna.Thecenterofthe41cmcablewasplacedinthecenterofthe
foamteststand.Thesourceandloadboardseachhadtheirownvoltageindicatortoensurethatall
testswereperformedwithbothboardsoperatingat5V.
120
Table4.3:Theconversionofthevoltagesmeasuredbythecurrentprobeintothepredictedmaximum
radiatedemissions,
j
ˆ
2
<0G
j
,usingEquations4.1and4.2withdistancesof10mand1m.
Figure4.16:DetailedviewofhowthecustomEUTwasarrangedduringbroadbandradiated
emissiontests.
121
4.2NarrowbandResults
Figures4.17and4.18showtheaveragemeasuredamplitudedi˙erencebetweentheEMC
analyzerandeachofthethreeRSP2prounitsforeverydatapointcollectedfromusingtheinput
datapointsfromTable4.1attestdistancesof10mand1m(respectively).Eachoftheplotsshown
inthese˝gurescorrespondstooneoftheRSP2prounitsthatwastested.Recallthattheaverage
o˙setvalue
+
3
isde˝nedinEquation3.13as
+
3
=
+
'

+
ˆ
where
+
'
representstheaveragevoltagemeasuredbytheRSP2proatparticularinputdatapoint
and
+
ˆ
representstheaveragevoltagemeasuredbytheEMCanalyzeratthatsameinputdatapoint.
Eachinputdatapointwasmeasuredthreetimesfortheresultspresentedinthissection.Recallthat
theEMCanalyzeramplitudehasanabsolutemeasurementaccuracyof

0.30dB[43]forallofthe
inputdatapointstested.ByusingtheEMCanalyzerforcomparison,themarginoferrorof

2.00
dBdevelopedinSection3.3canbeassessed.EachplotinFigures4.17and4.18alsoshowsthe
measuredSNRateachdatapointusingtherightmosty-axis.Theblackdotsintheseplotsrepresent
themeasuredSNR.
Fromtheresultsforthemeasurementsperformedatadistanceof10mshowninFigure4.17,one
canseethatallthreeRSP2prounitshadanaverageo˙setgreaterthan2.00dB.Somefrequencies
performedwithina

2.00dBmarginoftheEMCanalyzer,butwithallthreeRSP2prounitsthe
radiatedemissionsmeasuredat40,160,and360MHzwereoutsidethismargin.Additionally,
RSP2pro1'smeasurementat50MHzwaso˙bymorethan+10.00dBfromtheEMCanalyzer
measurement.TheSNRdataintheseplotsshowthatateveryfrequencytested,theSNRexperienced
byallthreeRSP2prounitswascomparabletotheSNRobservedinSection3.3.TheaverageSNR
valueshowninFigure4.17wasabout10dBhigherforfrequenciesabove200MHzforallthree
RSP2prounits.
Figure4.18showstheresultsforthemeasurementsperformedatadistanceof1mwiththeCW
radiatedemissionssource.Onecanseethatcomparedtothe10mresultsinFigure4.17,allthree
122
Figure4.17:AverageRSP2provoltagemeasuremento˙setfromEquation3.13fortheinputdata
pointsshowninTable4.1measuredatadistanceof10m.Eachdatapointrepresentstheaverage
ofthreemeasurements.
123
Figure4.18:AverageRSP2provoltagemeasuremento˙setfromEquation3.13fortheinputdata
pointsshowninTable4.1measuredatadistanceof1m.Eachdatapointrepresentstheaverageof
threemeasurements.
124
RSP2prounitshadamuchlargeraverageo˙setfromtheEMCanalyzeratmultiplefrequencies,
especiallybelow80MHzandabove200MHz.Interestingly,thereisaregionbetween60and150
MHzwhereallthreeRSP2prounitshadanaverageo˙setofapproximately2dB.Alsonotableis
thatallthreeRSP2prounitsmeasuredalowerlevelofradiatedemissionswhenthefrequencywas
below60MHzwhilenearly80%ofthemeasuredradiatedemissionsabove80MHzwherelarger
thanwhatwasmeasuredbytheEMCanalyzer.Thelargeroverallaverageo˙setatmostfrequencies
andtheshortmeasurementdistanceof1mindicatesthattheRSP2proADCsweremostlylikely
beingoverloaded.TheSNRdatainFigure4.18showsthatateveryfrequencytested,theSNRwas
about10dBhigherthanwhatwasmeasuredat10m.
4.3BroadbandResults
Figures4.19and4.20showtheaveragemeasuredamplitudedi˙erencebetweentheEMC
analyzerandeachofthethreeRSP2prounitsforeachoftheradiatedemissionfrequenciesshown
inTable4.3attestdistancesof10mand1m(respectively).Eachoftheplotsshowninthese
˝gurescorrespondstooneoftheRSP2prounitsthatwastested.Recallthattheaverageo˙setvalue
+
3
isde˝nedinEquation3.13as
+
3
=
+
'

+
ˆ
where
+
'
representstheaveragevoltagemeasuredbytheRSP2proatparticularinputdatapoint
and
+
ˆ
representstheaveragevoltagemeasuredbytheEMCanalyzeratthatsameinputdatapoint.
Eachinputdatapointwasmeasuredthreetimesfortheresultspresentedinthissection.Recallthat
theEMCanalyzeramplitudehasanabsolutemeasurementaccuracyof

0.30dB[43]forallofthe
inputdatapointstested.ByusingtheEMCanalyzerforcomparison,themarginoferrorof

2.00
dBdevelopedinSection3.3canbeassessed.EachplotinFigures4.19and4.20alsoshowsthe
measuredSNRateachdatapointusingtherightmosty-axis.Theblackdotsintheseplotsrepresent
themeasuredSNR.
Figure4.19showstheresultsfrommeasuringradiatedemissionsfromthecustomEUTata
distanceof10m.OnecanseethattheRSP2promeasurementsabove100MHzaresigni˝cantly
125
lowerthantheEMCanalyzermeasurementsforallthreeunits.AllthreeRSP2prounitsalsoexhibit
averylargeo˙setatafrequencyof40MHz,wheretheydi˙erfromtheEMCanalyzermeasurement
by20dB.WhiletheRSP2prounitsmeasureconsistentlywithoneanother,theyareonlywithinthe

2.00dBmarginatthefrequenciesbetween50and100MHz.ThemeasuredSNRiscomparable
towhatwasseenfromthenarrowbandCWradiatedemissionsat10mshowninFigure4.17.
Thebroadbandradiatedemissionsmeasurementsat1mshowninFigure4.20arebyfarthe
resultswiththemostdeviationfromtheEMCanalyzermeasurementsforallthreeRSP2prounits.
Notethattheplotsinthis˝gurehaveamaximumvalueof0dB,not20dB.Muchliketheresults
from10mshowninFigure4.19,Figure4.20showsthatthemeasurementswereconsistentacross
allthreeunits.However,ateveryfrequency(withtheexceptionof40MHz)allthreeRSP2pro
unitsmeasuredmorethan10dBlowerthantheEMCanalyzer.Inthefrequencyrangefrom50
MHzto200MHz,mostofthemeasurementsfromallthreeunitswere20dBlessthanwhatwas
measuredbytheEMCanalyzer.TheSNRdatashowninthis˝gureisverysimilartotheSNR
thatwasobservedwhenmeasuringthesamebroadbandemissionsat10m.However,itisvery
likelythattheRSP2proADCswereoverloadedwhenmeasuringbroadbandradiatedemissionsat
1m.ThisisimpliedbythelargedeviationfromtheEMCanalyzervoltagemeasurementsatthe
frequenciesthatweretested.
TheresultsinthissectioncanbecomparedtothenarrowbandCWresultsfromSection4.2
toconcludethattheRSP2prounitsaremoreaccuratewhenmeasuringCWradiatedemissions.
ThiscouldbebecausetheRSP2proisdesignedtobearadioreceiver,andsotheSDRunosettings
havebeentailoredtomeasurethecorrectpowerlevelwhenmeasuringnarrowbandcommunication
signalsinsteadofbroadbandnoisesignals.Tofurtherinvestigatethevalidityofthisconclusion,
thepredictionsfromTable4.3canbecomparedtotheactualradiatedemissionsmeasuredbythe
threeRSP2prounitsandtheEMCanalyzer.ByaddingatermfortheinsertionlossfromtheSMA
cabletoEquation2.6,themeasuredvoltagescanbeconvertedinto˝eldstrengthsandcomparedto
thepredictions:
ˆ

3`+
š
<

=
+

3`+

¸
˙

3
1
š
<


˚!

3

(4.3)
126
Figure4.19:AverageRSP2provoltagemeasuremento˙setfromEquation3.13forthecustomEUT
radiatedemissionfrequenciesshowninTable4.3measuredatadistanceof10m.Eachdatapoint
representstheaverageofthreemeasurements.
127
Figure4.20:AverageRSP2provoltagemeasuremento˙setfromEquation3.13forthecustomEUT
radiatedemissionfrequenciesshowninTable4.3measuredatadistanceof1m.Eachdatapoint
representstheaverageofthreemeasurements.
128
where
˚!
representsthemeasuredinsertionlossoftheSMAcableataspeci˝cfrequencyindB.
Table4.4and4.5showthemeasuredradiatedemissionsat10mand1m(respectively)thatwere
convertedfromthemeasuredvoltagesusingEquation4.3.Thesecondtolastcolumnofeachtable
showsthepredictedradiatedemissionsfromTable4.3.Notethatforbrevity,onlythemeasured
voltagefortheEMCanalyzer(
+
ˆ
)isshowninTable4.4and4.5.Thevaluesofthevariablesin
Equation4.3usedtoconvertthemeasuredvoltageintoanelectric˝eldareshownineachrow.The
valuesof
˙
correspondtotheBicoLOG30100Xantennafactor.
Table4.4showsthatthemeasuredradiatedemissionsoftheEMCanalyzer(
ˆ
ˆ
)andallthree
RSP2prounits(
ˆ
'
1
,
ˆ
'
2
,
ˆ
'
3
)atadistanceof10m.Allofthemeasuredradiatedemissionsare
signi˝cantlylowerthanthepredictedradiatedemissions,
j
ˆ
2
<0G
j
.Thiscouldbeforanynumber
ofreasons,buttwomaincausesaretheuncontrolledRFenvironmentandtheferritechokesused
todampentheradiatedemissions.Table4.4showsthat,withtheexceptionof800MHz,allthree
RSP2prounitswerewellwithin3.00dBofoneanotheratallfrequencies.However,theradiated
emissionsmeasuredfromallthreeunitsvariedgreatlyfromthevaluesmeasuredbytheEMC
analyzer.
Table4.5showsthatthemeasuredradiatedemissionsoftheEMCanalyzer(
ˆ
ˆ
)andallthree
RSP2prounits(
ˆ
'
1
,
ˆ
'
2
,
ˆ
'
3
)atadistanceof1m.Theseresultsareverysimilartotheones
showninTable4.4.Allofthemeasuredradiatedemissionsaresigni˝cantlylowerthanthepredicted
radiatedemissions,
j
ˆ
2
<0G
j
.OnecanalsoseefromTable4.5that,withtheexceptionof50MHz,
allthreeRSP2prounitswerewellwithin

3.00dBofoneanotheratallfrequencies.However,the
radiatedemissionsmeasuredfromallthreeunitsvariedgreatlyfromthevaluesmeasuredbythe
EMCanalyzer.
129
Table4.4:MeasuredradiatedemissionsfromthecustomEUTat10mcomparedtothepredicted
radiatedemissions.
Table4.5:MeasuredradiatedemissionsfromthecustomEUTat1mcomparedtothepredicted
radiatedemissions.
130
CHAPTER5
ACOMPLETEPRECOMPLIANCETESTSETUP
ThischapterdiscussesthecombinationoftheRSP2proandtwocommontelevisionantennas,which
togethercreateacompleteprecompliancetestsetup.First,thespeci˝csofhowtheantennaswere
selectedandthemodi˝cationsthatweremadetothemaredescribedinSection5.1.Thatsection
alsoincludesadescriptionofthetestprocedureusedforcharacterizingeachantenna.Theresults
ofthischaracterizationprocedurearediscussedinSection5.2andSection5.3.Finally,Section
5.4assessestheutilityoftheRSP2protestsetupcomparedtotheutilityofthetraditionalsetupby
measuringanEUTwithknownradiatedemissions.Thisassessmentincludesadiscussionofthe
relativecost,measurementaccuracy,andoverallconvenienceofusingeachtestsetup.
5.1AntennaSelection,Modi˝cations,andProcedure
AsmentionedinSection2.2,alow-costprecompliancetestsetupliketheoneshowninFigure
2.11typicallyusessometypeofVHForUHFantennainplaceofatraditionalEMCantenna.While
thishasthemajordisadvantageofnotallowingtheuseofanantennafactor,itdoesgreatlyreduce
thecostofthetestsetup.AccordingtoAndréandWyatt[25],commontelevisionantennaswork
wellintheVHFandUHFfrequencybands.Figure5.1showstwocommontelevisionantennas
thatwerepurchasedonlineforlessthan$20.00.The"rabbitear"antennashowninFigure5.1ahas
adjustableelementsandcanbeusedformeasuringsignalsintheVHFfrequencyrangefrom65to
200MHz[25].Figure5.1bshowsthe"bowtie"UHFantennathatisusableforsignalsfrom300
MHzto800MHz[25].InSections5.2and5.3,theusefulnessofeachantennainthefrequency
rangesdescribedbyAndréandWyatt[25]isassessed.
BeforeeitheroftheantennasshowninFigure5.1wasused,supporthardwarewascreatedso
thatconsistentmeasurementscouldbemadewitheachantenna.Thisincludedacustom-designed
antennaholderthatallowseachantennatobeplacedinahorizontalorverticalorientation.Figure
5.2showstheCADmodelsfortheantennaholders.Twodi˙erentversionswererequired,aseach
131
(a)
(b)
Figure5.1:(a)TheVHF"rabbitear"televisionantennaand(b)theUHF"bowtie"television
antenna.
antennahasauniqueshapethatmustbecradledbytheholdersothattheantennacannotmove.
ThewhiteholderinFigure5.2wasdesignedfortherabbitearantennawhilethegreenholderwas
designedforthebowtieantenna.Whiletheseholderswerespeci˝callydesignedfortheantennas
purchasedforthisthesis,itshouldbenotedthattheseholderswillworkwithotherantennasof
thesamestyle.Thisisbecausemosttelevisionantennasareverysimilarinshapeandsize.The
holdermodelsshowninFigure5.2were3D-printedusingblackPLAmaterial(the˝nalholdersare
shownbeingusedinFigure5.1).Thiswasultimatelyapoorchoice,asPLAhasaverylowmelting
point.Duringseveraloutdoormeasurementtests,theholdersactuallydeformedduetoexposure
tobrightsunlight.Usingboilingwater,theholderswereabletobereshaped.Thisproblemcan
beavoidedbyusingamoreresilientmaterialtoconstructtheantennaholders.Thedesign˝lesfor
eachantennaholderareavailableintheDigitalAppendixthatisincludedwiththedigitalversion
ofthisthesis(refertoAppendixBformoreinformation).
Toallowtheantennastobeplacedineithertheverticalorhorizontalorientation,eachofthe
antennaholderswasdesignwithanL-shapedmountingpointthatusesarecessednuttoattach
astandardcameratripodadapter.Figure5.3showshowthecameratripodadaptersattachtothe
132
L-shapedmountingpoint.TherecessednutscanbeseeninFigure5.3aandthecameratripod
adapterscanbeseeninFigure5.3b.
Thelasthardwareaccessorythatwasdesignedfortheantennaswasaclipthatpreventsthe
angularadjustmentoftherabbitearantennaelements.TheCADmodelforthisclipisshownin
Figure5.4.Theclipforcestheantennaelementsoftherabbitearantennatoremainseparated
by180
°
.Thismakestheantennamorelikeacenter-feddipole.Theclipalsoallowsformore
consistentmeasurementssincetheelementsareabletobeplacedinthesamepositionforeachtest.
Figure5.2:CADmodelsfortheantennaholdersthatareshowninFigure5.1.
(a)
(b)
Figure5.3:(a)Theantennaholder'sdualattachmentpointsand(b)theantennaholderwithboth
cameratripodadaptersattached.
133
Figure5.4:CADmodeloftheantennaclipthatwasusedto˝xtheangularpositionofrabbitear
antennaelementsto180
°
.
Theclipwas3D-printedusingastereolithographyprocess.Adetailedviewofthe˝nalclipcan
beseeninFigures5.6band5.6b.Theresinmaterialusedbythestereolithographyprocessworked
muchmorereliablythanthePLAthatwasusedtocreatetheantennaholders.
Thetelevisionantennaseachcomewithalongtwinleadcablethatisdesignedforbeingattached
directlytoabalun.Abalunisatypeofmatchingtransformerthatgetsitsnamefromacombination
ofthewords"BALanced"and"UNbalanced."Balunsaremostoftenusedtointerfacebalanced
center-fedantennastounbalancedcoaxialcables[60].O˙-the-shelftelevisonbalunsarealso
capableofimprovingtheimpedancematchbetweentelevisionantennasandthe50

impedance
usedbymostspectrumanalyzersandreceivers[3].Thebalunsconsideredforthisthesisareshown
inFigure5.5a.Thesearecommontelevisionbalunsthatusea4:1transformertomatchthe300

televisionantennastothe75

impedanceoftelevisioncoaxialcable.
Figure5.5ashowsthateachbalunusesadi˙erentconstructiontoachievethedesiredmatching
behavior.Thismeansthatitisimportanttocharacterizetheantennasthatareusedwiththespeci˝c
balunthatisavailable.ThebalunshownontheleftsideofFigure5.5awasusedforthisthesis.
134
Figure5.5bshowshowthebalunwasoriginallywired.Thetwoinputsofthe4:1transformer
elementwereconnectedtocapacitorsthatwereattachedtothescrew-terminalsonthebalun.This
waswherethebladesonthetelevisionantennas'twin-leadcableswouldnormallybeattached.
Figures5.5cand5.5dshowhowtheoriginaldesignwasmodi˝ed.Thescrew-terminalswere
removedandthecapacitorleadswerefedthroughtwosmallholesthatweredrilledintothebalun
enclosure.Thismodi˝cationwasinspiredbyWyatt[3],whorecommendsshorteningthetelevision
antennatwin-leadsasmuchaspossibletoimprovetheimpedancematchoftheantennatothebalun.
Figure5.5eshowsthebowtieantennawithitsleadscompletelyremovedandhowtheslotinthe
3D-printedantennaholderallowsthebaluntobedirectlysolderedtotheantenna.
Therabbitearantennarequiredfurthermodi˝cationstobeusedwiththebalunand3D-printed
antennaholder.Figure5.6ashowshowthebodyoftherabbitearantennawas˝rstreducedinsize
byremovingitsplasticmountingrod.This˝gurealsoshowsthewhitetwin-leadcablethatwas
originallyattachedtotheantenna.Figure5.6bshowstheconnectorsofeachantennaelementafter
removingthetwin-leadcable.This˝gurealsoshowshowthe3D-printedclipfromFigure5.4was
attachedtotherabbitearantennabodytolocktheantennaelementsinplace.Lastly,Figure5.6c
showsthewiresthatwereattachedtotheconnectorsinFigure5.6b.Thesewiresaremuchshorter
thantheoriginaltwin-leadcableandallowtherabbitearantennatobesoldereddirectlytothe
modi˝edbalun(likethebowtieantennashowninFigure5.5e).
Aftertheantennasweremodi˝edtobettermatchthe50

impedanceoftheEMCanalyzer
andtheRSP2pro,eachwascharacterizedby˝rstmeasuringtheVSWRwithanetworkanalyzer.
TheVSWRwasmeasuredtodeterminetheimpedancemismatchlossesofthetelevisionantennas
andtodetermineiftheusablefrequencyrangesforeachantennaasdescribedbyAndréand
Wyatt[25]arecorrect.Theantennaswerethenusedtomeasuretheradiatedemissionsfromthe
customEUTdescribedinSection4.1.2atadistanceof1musingthesetupshowninFigure4.4b.
ThesemeasurementswereperformedbysweepingacrosstheentireCISPR-22ClassBfrequency
rangewiththeEMCanalyzerandthefour"SpectrumSweep"state˝lesdescribedinTable3.1.
IdenticalmeasurementswerethenperformedusingtheBicoLOG30100XEMCantennasoadirect
135
(a)
(b)
(c)
(d)
(e)
Figure5.5:(a)Two4:1televisonbaluns(theoneontheleftwasused),(b)internalcomponentsof
thebalun,(c)balunwiththescrew-terminalsremoved,(d)the˝nalmodi˝ed4:1balun,and(e)the
connectionbetweenthemodi˝edbowtieantennaandthemodi˝edbalun.
136
(a)
(b)
(c)
Figure5.6:(a)Removalofthemountingpost,(b)theexposedantennaelementconnectors,and(c)
the˝nalmodi˝edrabbitearantenna.
comparisonofitsperformancecouldbemadewitheachtelevisionantenna.Figure5.7showsthe
RFsignalchainsusedforthesemeasurements.
NotethatitisamisconceptionthattheminimumVSWRofatransmissionlineconnectedto
anantennaoccursatthefrequencywheretheantennaisresonant[58].AminimumVSWRat
aspeci˝cfrequencyonlyindicatesthatthereisminimalreturnlossbetweentheantennaandthe
transmissionline.Forthisreason,thedirectcomparisontotheBicoLOG30100Xantennaisa
muchbetterindicatoroftheresonantfrequenciesofeachtelevisionantenna.Accordingto[58],a
VSWRofasmuchas6:1isstillacceptableforsignalreceptionintheVHFandUHFfrequency
bands(althoughaVSWRoflessthan2:1isgenerallypreferred).
137
Figure5.7:DiagramshowingtheRFsignalchainsusedduringthetelevisionantennacomparison
testsetup.
5.2VHF"RabbitEar"AntennaPerformance
Threedi˙erentelementlengths(
ˆ!
)werechosenfortherabbiteartelevisionantenna:0.375
m,0.667m,and0.991m.Theseelementlengthswerecarefullymeasuredeachtimeameasurement
wasperformedtoensureconsistencybetweenmeasurements.Notethatthelongerelementlengths
madeitimpossibletousetherabbitearantennaintheverticalorientation,andsotheantennawas
onlycharacterizedinthehorizontalorientation.Eachoftheselengthsareshownfromshortest
ˆ!
tolongest
ˆ!
inFigures5.8a,5.8b,and5.8c.
Figure5.8showsthemeasuredVSWRforeachofthethreeelementlengthsthatweretested.
Onecanseethatas
ˆ!
increases,theVSWRimprovesatlowerfrequencies.Thisisbestobserved
intheregionfrom60MHzto80MHz,wheretheVSWRimprovesbymorethanfouras
ˆ!
increases.TheelementlengthsshowninFigures5.8band5.8cperformedbestinthisregionand
theregionfrom100MHzto200MHz.Interestingly,theshortestelementlengththatwastestedhas
aVSWRoflessthanthreebetweenapproximately250and350MHz.However,allthreeelement
lengthsshowedaveryoscillatoryVSWRbehaviorabove200MHz,whichindicatesthattheusable
regionof65to200MHzforthisantennadescribedbyAndréandWyatt[25]isaccurate.
UsingEquations3.10and3.11,themeasuredVSWRofeachelementlengthshowninFigure
5.8canbeusedtoplotthemeasurementerrorduetoimpedancemismatchwhentheantennais
138
(a)
(b)
(c)
Figure5.8:MeasuredVSWRoftheVHFrabbitearantennawith(a)
ˆ!
=0.375m,(b)
ˆ!
=0.667
m,and(c)
ˆ!
=0.991m.Eachtraceistheaverageof˝vetrials.
139
usedwiththeEMCanalyzer(
"ˆ
ˆ
).Figure5.9showsthevalueof
"ˆ
ˆ
forallthreerabbitear
elementlengthstested.Onecanseethatthemismatcherrorislessthan2.00dBacrosstheentire
frequencyrange,indicatingthatthemodi˝edbalunisdoingagoodjobofmatchingtherabbitear
antennatoa50

system.This˝gurealsoshowsthattheelementlengthcane˙ectthemismatch
errorbyupto

0.5dBatmultiplefrequencies.
Figures5.10,5.11,and5.12showtheradiatedemissionsmeasuredbytherabbitearantennaat
eachofthethreeelementlengthscomparedtotheradiatedemissionsmeasuredbytheBicoLOG
30100XEMCantenna.Ineachofthese˝gures,theBicoLOG30100Xresultsareshownatthetop
andtheotherthreeplotsshowtheresultsfromusingtherabbitearantenna(withtheelementlength
ˆ!
increasingfromplot(b)toplot(d)).Thesourceoftheradiatedemissionsbeingmeasuredwas
thecustomEUTthatwaspresentedinSection4.1.2.
Figure5.9:TheamountofpotentialmismatcherrorwiththeEMCanalyzer
"ˆ
ˆ
forallthreeof
therabbirearantennaelementlengths
ˆ!
thatweretested.
140
Figure5.10showstheresultsfrommeasuringthecustomEUTat1mwithabasefrequency
settingof40MHz.OnecanseethattheresultsfromtheBicoLOG30100XshowninFigure5.10a
indicatethatitissigni˝cantlymoresensitivethantherabbitearantenna.Figures5.10b,5.10c,
and5.10dshowthattheelementlengthoftherabbitearantennadoesnotappeartosigni˝cantly
a˙ectwhichfrequenciestheantennaissensitiveto.TheoneexceptionisinFigure5.10b,where
theradiatedemissionsabove200MHzaredetectedmuchbetterbytheshortestantennathanbythe
othertwoelementlengths.
Figure5.11showstheresultsfrommeasuringthecustomEUTat1mwithabasefrequency
settingof50MHz.Figures5.11b,5.11c,and5.11dshowthattheantennawiththelongestelement
lengthismuchmoresensitivetothefundamentalfrequencyof50MHzthantheotherlength
antennas.However,allthreeantennaswereabletomeasurethestrongestharmonicat150MHz
withamarginoflessthan10dB.
Figure5.12showstheresultsfrommeasuringthecustomEUTat1mwithabasefrequency
settingof120MHz.Figures5.12b,5.12c,and5.12dshowthattheshortestelementlengthwasmuch
moresensitivetofrequenciesabove400MHz.However,thisfrequencyrangeisnotrecommended
becauseofthepreviouslydiscussedVSWRbehaviorabove200MHz.Theplotsinthis˝gurealso
illustratehowtheBicoLOG30100XEMCantennaissigni˝cantlymoresensitivethantherabbit
earantennaatallofthefrequenciestested.
141
Figure5.10:RadiatedemissionsfromthecustomEUTsetto40MHzmeasuredbyusing(a)the
BicoLOG30100XEMCantennaandbyusingtherabbitearantennawith(b)
ˆ!
=0.375m,(c)
ˆ!
=0.667m,and(d)
ˆ!
=0.991m.
142
Figure5.11:RadiatedemissionsfromthecustomEUTsetto50MHzmeasuredbyusing(a)the
BicoLOG30100XEMCantennaandbyusingtherabbitearantennawith(b)
ˆ!
=0.375m,(c)
ˆ!
=0.667m,and(d)
ˆ!
=0.991m.
143
Figure5.12:RadiatedemissionsfromthecustomEUTsetto120MHzmeasuredbyusing(a)the
BicoLOG30100XEMCantennaandbyusingtherabbitearantennawith(b)
ˆ!
=0.375m,(c)
ˆ!
=0.667m,and(d)
ˆ!
=0.991m.
144
5.3UHF"BowTie"AntennaPerformance
Figure5.13showsthemeasuredVSWRforboththehorizontalorientation(Figure5.13a)and
theverticalorientation(Figure5.13b)ofthebowtieantenna.ThemeasuredVSWRwaslessthan
sixintheregionfromapproximately350MHzto1000MHz.Thisresultisveryclosetotheusable
rangeof300MHzto800MHzdescribedbyAndréandWyatt[25].
UsingEquations3.10and3.11,themeasuredVSWRofeachelementlengthshowninFigure
5.13canbeusedto˝ndtheamountofmismatcherrorfromusingeachorientationofthebowtie
antennawiththeEMCanalyzer(
"ˆ
ˆ
).Figure5.14showsthevalueof
"ˆ
ˆ
forbothorientations
ofthebowtieantennaacrosstheentireCISPR-22ClassBfrequencyband.Aswiththerabbitear
antenna,onecanseefromthis˝gurethatthemismatcherrorislessthan2.00dBatallfrequencies.
ThisisduetotheimprovedimpedancematchbetweentheantennaandtheSMAcableusingthe
modi˝edbalun.Mostnotableisthesharpnullthatoccursjustpast500MHz.Here,thevalueof
"ˆ
ˆ
isnearlyzeroindicatinganalmostperfectimpedancematchwiththeEMCanalyzer.Since
theresultsfromeachorientationaresosimilar,onlythehorizontalorientationwasusedforradiated
emissionmeasurements.
Figures5.15and5.16showtheradiatedemissionsmeasuredbythebowtieantennaineach
orientationcomparedtotheradiatedemissionsmeasuredbytheBicoLOG30100XEMCantenna.
Ineachofthese˝gures,theBicoLOG30100Xresultsareshownatthetopandtheotherplotsshow
theresultsfromusingthebowtieantenna.Thesourceoftheradiatedemissionsbeingmeasured
wasthecustomEUTthatwaspresentedinSection4.1.2.
Figure5.15showstheresultsfrommeasuringthecustomEUTat1mwithabasefrequency
settingof120MHz.OnecanseethatthebowtieantennaismuchlesssensitivethantheBicoLOG
30100XEMCantenna.However,thebowtieantennawasabletomeasuretheradiatedemissions
causedbytheharmonicsofthe120MHzsignal.Onaverage,eachmeasuredharmonicwas
approximately40dBlowerthanwhatwasmeasuredbytheBicoLOG30100XEMCantenna.Also
notableishowthebowtieantennawasabletomeasurethefundamentalfrequencyat120MHz,well
outsidethesuggestedusablefrequencyrangeof300MHzto1000MHz.Thisismoreindicative
145
(a)
(b)
Figure5.13:MeasuredVSWRoftheUHFbowtieantennain(a)thehorizontalorientationand(b)
theverticalorientation.Eachtraceistheaverageof˝vetrials.
146
Figure5.14:TheamountofpotentialmismatcherrorwiththeEMCanalyzer
"ˆ
ˆ
forthehorizontal
andverticalorientationsofthebowtieantenna.
ofthestrongemissionscreatedbythecustomEUTandlessindicativeofthebowtieantennabeing
sensitivebelow300MHz.
Figure5.16showstheresultsfrommeasuringthecustomEUTat1mwithabasefrequency
settingof160MHz.Aswasobservedwiththe120MHzresults,thebowtieantennawasable
tomeasurethemajorharmonicscreatedbythe160MHzsignaldespiteitsveryweaksensitivity
relativetotheBicoLOG30100XEMCantenna.BothFigures5.15and5.16showthatthebowtie
isunabletoreceivethestrongFMradiosignalsaround100MHzthatarereceivedbytheEMC
antenna.
147
Figure5.15:RadiatedemissionsfromthecustomEUTsetto120MHzmeasuredbyusing(a)the
BicoLOG30100XEMCantennaand(b)thebowtieantennainthehorizontalorientation.
148
Figure5.16:RadiatedemissionsfromthecustomEUTsetto160MHzmeasuredbyusing(a)the
BicoLOG30100XEMCantennaand(b)thebowtieantennainthehorizontalorientation.
149
5.4ComparisontoaMoreTraditionalTestSetup
Thelastsectionofthisthesisdirectlycomparesamoretraditionalprecompliancetestsetup
withthelow-costprecompliancetestsetup.Figure5.17ashowsthemoretraditionaltestsetup
consistingoftheEMCanalyzerandtheBicoLOG30100XEMCantennawhileFigure5.17bshows
thecompletelow-costsetupconsistingoftheRSP2proandtherabbitearantennawithanelement
lengthof0.991m.Thiselementlengthwaschosenbasedupontheradiatedemissionsfromthe
EUT.Figure5.19showstheRFsignalchainsusedbyeachtestsetup.
TheEUTthatwasmeasuredfortheresultspresentedinthissectionwastheTPS57161-Q1
evaluationmodulefromTexasInstruments.Thismoduleisbasedaroundahigh-frequencybuck
converterthatisdesignedtoconverta12V
ˇ˘
inputtoa5V
ˇ˘
output[61].Theradiatedemissions
forthisevaluationmoduleweremeasuredbyTexasInstruments[61]inacontrolledRFenvironment
withabiconnicalEMCantennaliketheoneshowninFigure3.2a.Theresultsfrom[61]canbeused
todirectlyaccessthee˙ectivenessofbothtestsetupsshowninFigure5.17.Radiatedemissionsat
twofrequencies,approximately41and50MHz,weremeasuredforcomparisontotheresultsfrom
[61].
Figures5.20aand5.20bshowtheRFspectrumasmeasuredbythemoretraditionaltestsetup
near41MHzwhentheEUTispoweredo˙andthenturnedon,respectively.Theresultshereare
(a)
(b)
Figure5.17:(a)Thetraditionalprecompliancetestsetupand(b)thelow-costtestsetup.
150
Figure5.18:DetailedviewoftheEUTshowingthegold5

resistiveload.
(a)
(b)
Figure5.19:DiagramshowingtheRFsignalchainsof(a)thetraditionaltestsetupand(b)the
low-costtestsetup.
shownindBmsothattheycanbedirectlycomparedtothesamemeasurementinSDRuno(which
canonlydisplaymagnitudeintermsofdBm).Onecanseethattheradiatedemissionsfromthe
EUTwereapproximately-55.00dBmatafrequencynear49.950MHz.Figures5.21aand5.21b
showsthesameradiatedemissionsasmeasuredbyRSP2pro1.Onecanseefromthebackground
measurementinFigure5.21athatthenoise˛oorobservedbytheRSP2proisalmost15dBlower
thanthenoise˛oormeasuredbytheEMCanalyzerinFigure5.20a.Thisislikelydueinpartto
thethelowsensitivityoftherabbitearantennawhencomparedtotheBicoLOG30100XEMC
151
(a)
(b)
Figure5.20:(a)ThebackgroundmeasurementwiththeEUTo˙and(b)theradiatedemissions
fromtheEUTnear41MHz.
antenna.Figure5.21bshowstheradiatedemissionsfromtheEUT.ThelevelmeasuredbySDRuno
ismuchlowerthanwhatwasmeasuredbythetraditionaltestsetup,withapeakvalueofonly-78.8
dBm.However,thewaterfallplotinthelowerhalfofthis˝gureclearlyshowshowtheemissions
fromtheEUTarecontinuousinnature.Thewaterfallplotalsobetterillustratesthebandwidthof
theradiatedemission.
Figures5.22aand5.22bshowtheRFspectrumasmeasuredbythemoretraditionaltestsetup
near50MHzwhentheEUTispoweredo˙andthenturnedon,respectively.Theresultshereare
shownindBmsothattheycanbedirectlycomparedtothesamemeasurementinSDRuno(which
canonlydisplaymagnitudeintermsofdBm).Comparedtotheemissionsnear41MHz,onecan
seethattheradiatedemissionsmeasuredbythemoretraditionaltestsetuparemuchweaker,having
amagnitudecloserto-70.00dBm.Figures5.23aand5.23bshowsthesameradiatedemissionsas
measuredbyRSP2pro1.OnecanseefromthebackgroundmeasurementinFigure5.23athatthe
noise˛oorobservedbytheRSP2proismorethan15dBlowerthanthenoise˛oormeasuredby
theEMCanalyzerinFigure5.22a.However,Figure5.23bshowsthattheradiatedemissionsfrom
theEUTasmeasuredbySDRunoaremuchlowerthanwhatwasmeasuredbythetraditionaltest
152
(a)
(b)
Figure5.21:(a)ThebackgroundmeasurementwiththeEUTo˙and(b)theradiatedemissions
fromtheEUTnear41MHz.
153
(a)
(b)
Figure5.22:(a)ThebackgroundmeasurementwiththeEUTo˙and(b)theradiatedemissions
fromtheEUTnear50MHz.
setup,withapeakvalueofonly-92.1dBm.
Theresultsfromtheradiatedemissionmeasurementsinthissectionhavebeensummarizedin
Tables5.1and5.2.Thesetablesalsocomparethemeasuredradiatedemissionswiththeemissions
measuredbyTexasInstrumentsfrom[61].Theradiatedemissionsmeasuredbythetraditional
testsetupwereconvertedto˝eldstrengthusingthesameprocessthatwasusedinSection4.3.
Sincetheantennafactoroftherabbitearantennaisunknown,thesamecouldnotbedoneforthe
measurementsrecordedwiththelow-costsetup.
Table5.1showshowthebackgroundRFsignalsateachdatapointcompared.Themeasurement
from[61](shownincolumnthree)wassigni˝cantlylowerthanthelevelmeasuredbythetraditional
testsetup.ThisisduetotheuncontrolledRFenvironmentusedinthisthesis.Forbothdatapoints,
thebackgroundlevelmeasuredbythelow-costsetupwassigni˝cantlylowerthantheothertwo
testsetups.Thisisbecauseofthelowsensitivityoftherabbitearantenna.Again,notethatthe
low-costtestsetupmeasurementisnota˝eldstrengthbuttheequivalentmagnitudethatresults
fromconvertingSDRuno'sdBmmeasurementtodB
`
VusingEquation3.9.
Table5.2comparestheradiatedemissionmeasurementsateachdatapoint.Aswasobserved
154
(a)
(b)
Figure5.23:(a)ThebackgroundmeasurementwiththeEUTo˙and(b)theradiatedemissions
fromtheEUTnear50MHz.
155
Table5.1:ComparisonoftheradiatedemissionsmeasuredfromtheEUTshowninFigure5.18
withseveraldi˙erenttestsetups.
Table5.2:ComparisonoftheradiatedemissionsmeasuredfromtheEUTshowninFigure5.18
withseveraldi˙erenttestsetups.
inTable5.1,themeasurementfrom[61](shownincolumnthree)wassigni˝cantlylowerthanthe
levelmeasuredbythetraditionaltestsetup.Forbothdatapoints,thebackgroundlevelmeasured
bythelow-costsetupwassigni˝cantlylowerthantheothertwotestsetups.Thisisbecauseof
thelowsensitivityoftherabbitearantenna.Again,notethatthelow-costtestsetupmeasurement
isnota˝eldstrengthbuttheequivalentmagnitudethatresultsfromconvertingSDRuno'sdBm
measurementtodB
`
VusingEquation3.9.
Theresultsfromthissectionshowthatthelow-costtestsetupconsistingofthecommon
televisionantennaandtheRSP2proiscapableofidentifyingtheradiatedemissionsfromaproduct
whenusedproperly.However,withoutknowingtheantennafactorofthetelevisionantennas,
theconversionfromtheRSP2promeasurementtotheradiatedemission˝eldstrengthcannotbe
performed.Thismakesthelow-costtestsetupsuitableforperformingmultiplebench-topradiated
emissionteststhroughouttheelectronicproductdesignprocesstocheckforproblemfrequencies.
Oncetheproblemfrequencieshavebeenidenti˝ed,thelow-costtestsetupcanbeusedtoassess
156
thee˙ectivenessofchangesthataremadetotheproduct.Themainbene˝tsfromthelow-costtest
setupcomefromitslowcostandcompactsizerelativetothetraditionaltestsetup.
TheRSP2prounitsusedinthisthesiswerepurchasedforapproximately$192.00each.The
televisionantennas,balun,3D-printedmountinghardware,andcameratripodaccessoriesallcost
lessthan$50.00.Bycomparison,theBicolog30100XEMCantennathatwasusedinthisthesis
costjustunder$3000.00.Whilethespeci˝cEMCanalyzerusedinthisthesiswasprovidedbythe
MSUECEdepartment,aneBaysearchatthetimeofwritingfor"E7401AEMCanalyzer"brings
upseverallistingsforusedunitsthatrangeinpricefrom$5500.00to$6999.99.Averagingthese
pricestogetanestimatedvalueof$6250.00,thetotalcostofthemoretraditionalprecompliance
testsetuppicturedinFigure5.17aisnearly$9250.00whilethelow-costsetupshowninFigure
5.17bcostlessthan$250.00.Notethatthispricedoesnotincludethecostofthenecessaryhost
computertouseSDRuno.Thisauthorassumesthatmostelectronicproductdesignershaveaccess
toacomputer,andtheSDRUnosoftwareisfree(evenforcommercialuse).
JustcomparingtheE7401AEMCanalyzerandtheRSP2prorevealsseveralotheradvantages
ofthelow-costsetup.TheEMCanalyzerhasarectangularfootprintof409mmby373mmand
aweightof12.6kg[43].Bycomparison,theRSP2prohasarectangularfootprintofonly99mm
by87mmandaweightof0.296kg[42].ThismakestheRSP2proextremelyportableandvery
convenientforperformingquickbench-topmeasurements.TheRSP2proisalsofullypoweredvia
USBandonlyconsumesabout850mW[42]whiletheEMCanalyzerrequires120VACandcan
consumeupto300Wduringoperation[43].
ThemajordisadvantageswiththeRSP2proaremostlyrelatedtoitsrelianceonSDRuno.
Performingafullspectrumsweepwithaspectrumanalyzerismuchmoreconvenientthanmanually
tuningtoeachfrequencythatmustbetested.Additionally,only10MHzoftheRFspectrumcanbe
observedatatimewithinSDRuno.Viewing10MHzofthespectrumatatimewithSDRunorequires
theuseofthezero-IFmode,whichintroducesspuriousmixingandintermodulationproductsthat
canbemistakenforradiatedemissions.Thedi˙erencebetweeneachofthesenoisesignalscanbe
determinedbychangingthecenterfrequencyinSDRuno.Thespuriousmixingproductswillalso
157
changeinfrequency,butintermodulationproducts(andmoreimportantly,theradiatedemissions
beingmeasured)willnot[62].Todeterminethedi˙erencebetweenintermodulationproductsand
theradiatedemissionsthatoneistryingtomeasure,a10dBattenuatorshouldbeaddedtothe
RFsignalchain.Intermodulationproductswillbereducedbymuchmorethan10dBwhilethe
actualRFsignalsbeingmeasured(andthespuriousmixingproducts)willbereducedby10dB
[62].Thesetypesoffalsesignalsarenotanissuewhenusingahighqualityspectrumanalyzerlike
theE7401A.
158
CHAPTER6
CONCLUSIONS
ThisthesisexaminesthepotentialforusingtheRSP2prosoftware-de˝nedradio(SDR)forthe
precompliancetestingofradiatedemissions.Amarginoferrorforthesignallevelmeasuredby
theRSP2prowassuccessfullydevelopedbydirectlycomparingtheperformanceoftheRSP2proto
atraditionalspectrumanalyzer.Thisthesisalsoexaminestheuseoflow-costtelevisionantennas
foridentifyingradiatedemissionsacrosstheCISPR-22ClassBfrequencybandfrom30MHzto
1000MHz.BycombiningtheRSP2prowiththeseantennas,acompletelow-costprecompliance
testsetupwasdemonstrated.Thelow-costsetupintroducedinthisthesiscanbeusedbyelectronic
productdesignersfromallbackgroundstoquicklyandeasilytestaproductforradiatedemissions
throughoutthedesignprocess.
Thisworkpresentedinthisthesiscanbegreatlyexpandedupon,particularlywithregardstothe
characterizationofthetelevisionantennas.Developingamethodforeithersimulatingormeasuring
theantennafactorsoftheseantennaswouldallowthemeasurementstakenwiththelow-costtest
setuptobedirectlyconvertedto˝eldstrengthandcomparedtotheradiatedemissionlimits.Further
characterizationoftheRSP2prowouldallowforabetterunderstandingoftherelationshipbetween
themeasuredSNRinSDRunoandtheaccuracyofthemeasuredsignalpower.Thegreatestavenue
formoreresearchwouldbe˝ndingasoftwarereplacementforSDRunothathasauserinterface
moresimilartoaspectrumanalyzer.Suchasoftwarecouldimplementthefunctionalityofa
trackinggeneratorsothattheRSP2pro's10MHzmeasurementbandwidthcouldbeusedtospan
muchlargerfrequencyranges.
Atthetimeofwriting,SDRplayisworkingonasoftwareinterfacefortheirSDRsthathasmany
ofthecapabilitiesusuallyfoundonaspectrumanalyzer.WhiletheSDRso˙eredbySDRplayare
currentlytheonlysoftware-centricSDRswithabsolutepowermeasurementcapabilities,perhaps
otherswillbecomeavailableasdesignersrealizethefullpotentialoftheSDR-basedspectrum
analyzerconcept.
159
APPENDICES
160
APPENDIXA
EUTCIRCUITDIAGRAMSANDBILLOFMATERIALS
FiguresA.1andA.2showthecircuitdiagramsforeachofthePCBboardsthattogethercreatethe
customEUTfromSection4.1.2.Theschematics,billofmaterials(BOM),andthegerber˝lesfor
eachPCBboardcanbefoundintheDigitalAppendix(refertoAppendixB).
ThecustomEUTPCBsweredesignedusingKiCad,anopen-sourceelectricdesignautomation
softwareavailableformultipleoperatingsystems.Learnmoreaboutitat:https://www.kicad-
pcb.org/.
FigureA.1:ElectricalcircuitforthesourcesideofthecustomEUTfromSection4.1.2.
161
FigureA.2:ElectricalcircuitfortheloadsideofthecustomEUTfromSection4.1.2.
162
APPENDIXB
DIGITALAPPENDIXCONTENTSLIST
ThedigitalversionofthisthesisincludesaDigitalAppendixintheformofasinglecompressed
archive.TheDigitalAppendixincludesthefollowingsubfoldersand˝les:
‹
Section3.2InputDataSets(fromTable3.2)

agilent_11955A.csv

agilent_11955A_minus10.csv

agilent_11955A_plus10.csv

agilent_11956A.csv

agilent_11956A_minus10.csv

agilent_11956A_plus10.csv

bicolog30100X.csv

bicolog30100X_minus10.csv

bicolog30100X_plus10.csv
‹
Section4.1.2DesignFiles

custom_eut_BOM.xlsx

custom_eut_load_gerbers.zip

custom_eut_load_schematic.pdf

custom_eut_source_gerbers.zip

custom_eut_source_schematic.pdf
163
‹
Section5.1DesignFiles

antenna_clip_body1.STL

antenna_clip_body2.STL

antenna_holder_bowtie.STL

antenna_holder_rabbit.STL
164
BIBLIOGRAPHY
165
BIBLIOGRAPHY
[1]C.R.Paul,
IntroductiontoElectromagneticCompatibility
,Second.Hoboken,N.J:Wiley-
Interscience,2006.
[2]H.W.Ott,
ElectromagneticCompatibilityEngineering
.Hoboken,N.J:JohnWiley&Sons,
2009.
[3]K.Wyatt,RadiatedEmissionsUsingLow-CostBench-TopMethods,
InterferenceTechnology2011EMCDirectory&DesignGuide
,pp.Mar.2011.
[Online].Available:https://interferencetechnology.com(visitedon02/21/2019).
[4]Various,9:RadiatedEmissions,ECE407ClassNotes,MichiganStateUniversity.
[5]PCBDesignGuidelinesForReducedEMI,TexasInstrumentsInc,Dallas,TX,
Tech.Rep.,1999.[Online].Available:http://www.ti.com(visitedon02/20/2019).
[6]Various,8:Regulations,ECE407ClassNotes,MichiganStateUniversity.
[7]C.R.PaulandD.R.Bush,EmissionsfromCommon-ModeCurrents,in
1987
IEEEInternationalSymposiumonElectromagneticCompatibility
,Aug.1987,pp.
doi
:
10.1109/ISEMC.1987.7570770.
[8]K.Wyatt,HfCurrentProbe:TheoryAndApplication,
InterferenceTechnology2017
EuropeEMCGuide
,pp.2017.[Online].Available:https://interferencetechnology.
com(visitedon11/06/2019).
[9]R.Cadena,
VideoMatters.RFInterferencefromLEDScreensisaBIGProblem!
May2012.
[Online].Available:https://www.juliusmedia.com(visitedon11/07/2019).
[10]
FCC'sOETClari˝esEmissionsComplianceTestingforRFLEDLightingDevices
,Jun.
2016.[Online].Available:http://www.arrl.org(visitedon11/07/2019).
[11]D.Friskney,
HowtokeepyourdigitalsignagefrombeingshutdownbytheFCC
,May2017.
[Online].Available:https://www.digitalsignagetoday.com(visitedon11/07/2019).
[12]FederalCommunicationsCommission,
DOC-355270A1
,Nov.2018.[Online].Available:
https://www.fcc.gov(visitedon02/21/2019).
[13]nerdlord20,
LEDSigns:AnEMINightmare
,Jun.2017.[Online].Available:https://www.
youtube.com(visitedon11/07/2019).
[14]
AmateurRadioSleuthingPinsDownSourceofStrangeRFInterference
,Sep.2016.[Online].
Available:http://www.arrl.org(visitedon11/07/2019).
166
[15]U.S.FoodandDrugAdministration,
AbbottRecallsCentriMagCirculatorySupportSystem
MotorDuetoPumpandMotorIssues
,Mar.2019.[Online].Available:http://www.fda.gov
(visitedon11/07/2019).
[16]Engineer'sGuideToGlobalEMCRequirements:2007Edition,Intertek,Boxborough,
MA:IntertekTestingServicesNA,Inc,Tech.Rep.,2007.[Online].Available:https://www.
ieee.li(visitedon11/05/2019).
[17]KeithArmstrong,PracticalGuideforEN55022andEN5011:MeasuringRadiatedEmis-
sions,REO(UK)Ltd,Shropshire,SY79QA,UK,Tech.Rep.[Online].Available:https:
//www.emcstandards.co.uk(visitedon02/19/2019).
[18]TimothyHegarty,overviewofconductedEMIspeci˝cationsforpowersupplies,Texas
InstrumentsInc,Dallas,TX,Tech.Rep.[Online].Available:https://www.ti.com(visitedon
01/20/2020).
[19]ofAntennaFactorsinEMIMeasurementsAN-106AF,Com-PowerCorporation,
Silverado,CA,Tech.Rep.[Online].Available:http://k3hpa.com(visitedon10/18/2017).
[20]T.Karaca,B.Deutschmann,andG.Winkler,ersimulationmodelwithquasi-
peakdetector,in
2015IEEEInternationalSymposiumonElectromagneticCompatibility
(EMC)
,Dresden,Germany:IEEE,Aug.2015,pp.
doi
:10.1109/ISEMC.2015.
7256283.
[21]MakingConductedandRadiatedEmissionsMeasurements,KeysightTech-
nologies,SantaRosa,CA,Tech.Rep.,Apr.2017.[Online].Available:https://www.keysight.
com(visitedon11/02/2017).
[22]DarylGerke,P.E.andWilliamKimmel,P.E.,
EDNDesigner'sGuidetoElectromagnetic
Compatability
.Mesa,AZ:KimmelGerkeAssociates,Ltd.,2005.
[23]KeithArmstrongandTimWilliams,TestingPart1-RadiatedEmissions,Cherry
CloughConsultantsLtd,Brocton,ST170TF,UK,Tech.Rep.[Online].Available:http:
//www.compliance-club.com(visitedon01/09/2018).
[24]w-costEMIPre-complianceTestingUsingaSpectrumAnalyzer,Tektronix,Tech.Rep.,
2016.[Online].Available:https://www.tek.com(visitedon11/06/2019).
[25]P.G.AndréandK.Wyatt,
EMITroubleshootingCookbookforProductDesigners
.Edison.
NewJersey:SciTechPublishing,2014.
[26]SpectrumAnalyzer:SpectrumAnalysisBasics150,KeysightTechnologies,
SantaRosa,CA,Tech.Rep.,Feb.2016.[Online].Available:https://www.keysight.com
(visitedon02/25/2020).
[27]
UnderstandingRF&MicrowaveSpeci˝cations-PartI
.[Online].Available:http://www.ni.
com(visitedon11/08/2017).
167
[28]
WhatisSoftwareDe˝nedRadio?
[Online].Available:https://www.wirelessinnovation.org
(visitedon02/27/2020).
[29]E.Grayver,
ImplementingSoftwareDe˝nedRadio
.NewYork:Springer,2013.
[30]RodgerH.Hosking,
SoftwareDe˝nedRadioHandbook
,Twelfth.UpperSaddleRiver,NJ,
2016.
[31]
SoftwareDe˝nedRadio-RateofAdoption
.[Online].Available:https://www.
wirelessinnovation.org(visitedon02/27/2020).
[32]V.Giannini,J.Craninckx,andA.Baschirotto,
BasebandAnalogCircuitsforSoftware
De˝nedRadio
,ser.AnalogCircuitsandSignalProcessingSeries.Dordrecht:Springer,
2008.
[33]
USRPB200mini
.[Online].Available:https://www.ettus.com(visitedon02/27/2020).
[34]
Review:Airspyvs.SDRplayRSPvs.HackRF
,Feb.2016.[Online].Available:https://www.rtl-
sdr.com(visitedon10/11/2017).
[35]RTL-SDR.com,TL-SDRBlogV3Datasheet,Tech.Rep.[Online].Available:https:
//www.rtl-sdr.com(visitedon03/10/2020).
[36]
TheUltimateVHF/UHFRX"Brick"
,2020.[Online].Available:https://airspy.com/airspy-r2/
(visitedon03/10/2020).
[37]
SDRplay
.[Online].Available:https://www.sdrplay.com/(visitedon09/20/2018).
[38]mossmann,
HackRFOne
,2017.[Online].Available:https://github.com/mossmann/hackrf/
wiki/HackRF-One(visitedon03/10/2020).
[39]yRSP2,
RadCom
,vol.93,no.4,pp.Apr.2017.[Online].Available:
https://www.sdrplay.com(visitedon01/10/2019).
[40]SDRplayLtd,GainandAGCSettingsinSDRuno,Hampshire,GU341HG,UK,
Tech.Rep.,Jun.2018.[Online].Available:https://www.sdrplay.com(visitedon03/02/2020).
[41]SteveFord,yRadioSpectrumProcessor2Pro,
QST
,vol.101,no.10,Oct.2017.
[Online].Available:https://www.sdrplay.com(visitedon10/19/2018).
[42]SDRplayLtd,SpectrumProcessor2andRadioSpectrumProcessor2prov2.1,
Hampshire,GU341HG,UK,Tech.Rep.,Nov.2017.[Online].Available:https://www.
sdrplay.com(visitedon03/01/2020).
[43]E7400A-seriesEMCAnalyzersDataSheet,AgilentTechnologies,SantaClara,
CA,Tech.Rep.,Oct.2001.[Online].Available:https://www.atecorp.com(visitedon
11/14/2019).
168
[44]EMCTransducerFactors,AgilentTechnologies,SantaClara,CA,Tech.Rep.
[Online].Available:https://www.keysight.com(visitedon01/23/2019).
[45]JamesMcLean,RobertSutton,andRobHo˙man,pretingAntennaPerformancePa-
rametersforEMCApplications:Part3:AntennaFactor,TDKRFSolutionsInc.,Cedar
Park,TX,Tech.Rep.[Online].Available:https://www.ttiinc.com(visitedon01/17/2019).
[46]AaroniaAG,Strickscheid,Germany,ExcelSpreadsheet,Nov.
2017.
[47]KennethWyatt,
Review:TheAaroniaBicoLOG30100XandHyperLOG7060EMIantennas
,
Jul.2015.[Online].Available:https://www.edn.com(visitedon11/02/2017).
[48]UBBV-SeriesManual,AaroniaAG,Strickscheid,Germany,Tech.Rep.,
Mar.2018.
[49]SynthesizedRFSignalGeneratorOperatingandCalibratingGuide,Agilent
Technologies,SantaClara,CA,Tech.Rep.,Oct.1998.[Online].Available:https://www.
keysight.com(visitedon06/01/2019).
[50]
PyVISA:ControlyourinstrumentswithPython
.[Online].Available:https://pyvisa.
readthedocs.io/en/latest/(visitedon01/20/2019).
[51]AutoItConsultingLtd,
AutoIt
,2019.[Online].Available:https://www.autoitscript.com/site/
(visitedon01/20/2019).
[52]SDRplayLtd,areDe˝nedRadioAPIR3.01,Hampshire,GU341HG,UK,Tech.
Rep.,Aug.2018.[Online].Available:https://www.sdrplay.com(visitedon05/07/2019).
[53]unoUserManualv1.33,SDRplayLtd,Hampshire,GU341HG,UK,Tech.Rep.,
2019.[Online].Available:https://www.sdrplay.com(visitedon03/01/2020).
[54]P.Horowitz,
TheArtofElectronics
,Thirdedition.NewYork,NY:CambridgeUniversity
Press,2015.
[55]
UnitedStatesFrequencyAllocations
,Oct.2003.[Online].Available:https://www.ntia.doc.
gov(visitedon04/23/2020).
[56]Various,3:SignalsandSpectra,ECE407ClassNotes,MichiganStateUniversity.
[57]
NAGOYAUT-72
,2020.[Online].Available:https://baofengtech.com(visitedon04/20/2020).
[58]H.W.SilverandAmericanRadioRelayLeague,
TheARRLAntennaBook:ForRadio
Communications
,Twenty-fourth.Newington,CT:ARRL,2019.
[59]ClockMultiplier,PericomSemiconductorCorporation,SanJose,CA,
Tech.Rep.,Aug.2014.[Online].Available:https://www.diodes.com(visitedon11/06/2017).
169
[60]J.Sevick,
Understanding,Building,andUsingBalunsandUnuns:TheoryandPractical
DesignsfortheExperimenter
.Hicksville,NY:CQCommunications,2003.
[61]KrishnamurthyHegde,ScottMonroe,andHaydenHast,andTPS57xx0-Q1
DevicesEMCPerformanceComparisonStudy,TexasInstrumentsInc,Dallas,TX,Tech.
Rep.,Dec.2015.[Online].Available:http://www.ti.com(visitedon04/01/2020).
[62]Tech_Support,
VFO&LOfordummies
,Reply,Feb.2018.[Online].Available:http://www.
sdrplay.com(visitedon02/27/2020).
170