Updating the Bob Smith Termination Technique.docx

Updating the Bob Smith Termination Technique.docx

《Updating the Bob Smith Termination Technique.docx》由会员分享，可在线阅读，更多相关《Updating the Bob Smith Termination Technique.docx（9页珍藏版）》请在冰豆网上搜索。

UpdatingtheBobSmithTerminationTechnique

UpdatingtheBobSmithTerminationTechnique

WhiletheBobSmithTerminationtechniquehaswonacclaimworldwide,ithasitsflaws.Fortunately,anewterminationtechniqueisproposedherethatcansolvemanyoftheproblemsdesignerswillencounterwhenusingtheBobSmithapproach.

ByJimSatterwhite,TeltestElectronicLaboratories

FordesignersofEthernetsystems,theBobSmithTerminationtechniqueisregardedasanimportantapproachformigratingpotentialRFemissionandsusceptibilityproblemsforEthernetinfrastructure.But,whatifthetechniquehasflaws?

Unfortunately,therearesomeflawswiththeterminationtechniquepatentedbyBobSmithINthe90s.ThisarticlepointsoutpotentialflawsandthenshowshowaproposedterminationtechniquecansolvesomeoftheproblemsencounteredusingaBobSmithterminationapproach.Thearticlealsoprovidesaphysicalverificationmethodthatdesignerscanusetoanalyzetheimpactoftheproposedterminationtechnique.

Theproblematicresultsofimproperterminationaredifficultdifficulttomeasureandverify.Theyshowupasproblemswithelectromagneticcompatibility（EMCcertificationandworseasintermittentanddifficulttoidentifyperformanceproblemsinservice.Thus,thisarticlewillalsoprovideasoundtheoreticalapproachforevaluatingtheproposedterminationscheme.

ProblemswithMr.Smith

InaU.S.patent,RobertW.（Bob）Smithdescribedamethodtoreducethelongitudinalorcommonmodecurrentonmulti-pairconductorsystemswherethepairsareinterrelatedinauniformmanner.HealludestothefactthatthepairtopairrelationshipsofaCAT5cableformtransmissionlinesinthemselves.

Smithassertsincorrectlythatsuchtransmissionlinesexhibitacharacteristicimpedanceofapproximately145ohms.Hisdescriptionisconfusingandmisleadingbecausehedoesnotdefinetheconfigurationinwhichthisissupposedlyexhibited.Hecorrectlygoesontosaythatthecablecansupportandradiatecommon-modesignals.Moreover,hestatesthatthereisamethodtoreducethecommonmoderadiationbymatchingthecommonmodeimpedanceofthecable,andthenpresentsamethodofterminatingthecommonmodetransmissionline,thatdoesnotinactualityaccomplishthatgoalwell.Histerminationtechniquedoesreducethecommonmoderadiation,butnotoptimallyandnotfromthecorrectunderstanding.

ForaCAT5cable,therearetwocommonmodesofinterest.Thefirstisthepair-wisecommonmodeandthesecondisthecable-wisecommonmode.Inthearticletofollow,we'llnoteswilladdressthepair-wisecommonmode.

Thepair-wisecommonmodehasacommonmodecharacteristicimpedancethatcanbereadilycharacterized,whilethecable-wisecharacteristicimpedanceisgrosslydependentonimplementation.Actually,thecharacteristicimpedanceofthecable-wisecommonmodeisnotameaningfulconceptexceptinsomewell-definedcircumstances.

Withthatbackgroundinmind,let'slookatthepair-wisecommonmodecharacteristicimpedancerelationshipswithinaCAT5cable.AsFigure1pointsout,thepair-wisemodecallsforfourpairswithanintrinsiccharacteristicimpedancebetweeneachpair.Duetotheapproximatesymmetryintheconstructionofthecable,eachpairhasthesamerelationwithanyotherpair.Itisprudenttodefinetherelationshipstobedescribed（Smithdoesn'tdothat）.Sincetheinherentcharacteristicimpedancescannotbemeasureddirectly,toverifythetheoryitisnecessarytomeasurethemincombinationsandresolvetheindividualelements.

Figure1:

Pair-wisecharacteristicimpedancesonaCAT5cable.

Thefirstrelationshipisbetweenanytwopairs（Z1-1）.Thisparameterissomewhatmeaninglessinapplicationbecauseallfourpairsareinterrelated,butitisusefulandcanbecharacterizednone-the-less.Thesecondandmoremeaningfulrelationshipisthecharacteristicimpedanceofonepairrelativetotheotherthreepairs（Z1-3）.Thethirdistherelationshipoftwopairstotheothertwopairs（Z2-2）.Forthecaseofonepairversusanyotherpairwiththetworemainingpairsfloating,themeasuredcharacteristicimpedanceisapproximately100ohms.

Forthecaseofonepairversustheotherthreepairstiedtogetherthemeasuredcharacteristicimpedanceisapproximately70ohms.Forthecaseoftwopairstiedtogetherversustheothertwoparstiedtogetherthemeasuredcharacteristicimpedanceisapproximately50ohms.NowheredoesthereexistarelationshipthatthecharacteristicimpedanceisanywherenearSmith'sstated145ohms.

Figure2showsthedefinitionsofsomeofthepossiblemeasurablecharacteristicimpedances.

Figure2:

Definitionsofpossiblemeasurablecharacteristicimpedances.

Table1showssomemeasureddatasamples.TheseweretakenusingtheAgilent4396BNetworkAnalyzer.

Table1:

MeasuredDataSamples.

Analysis

Analysisofthemeasuredvaluestodeterminetheinherentcharacteristicimpedancevaluesiscomplex.Sometechniquesinclude:

Z1-3（1,234）=Z14//Z12//Z13

Z2-2（{12,34）=Z14//Z23//Z13//Z24

Z1-3（1,234）=Z12//（ComplicatedcombinationofZ14,Z34,Z23,Z13,Z24）

Thisdoesnotappeartobereadilymanageable.Therefore,asimulationwasdevelopedselectingvaluesforZnntogivevaluessomethingnearthemeasureddata.Fromthis,theintrinsiccharacteristicimpedanceZnnappearstobeinthevicinityof200ohms.Fromthissimulation,itcanbesuggestedthatthevalueofZnnvariesalongthelengthofthecableaswellasfrompairtopair.Thus,toproperlyterminatethecommonmodetransmissionlines,thetechniqueinFigure3mustbeused.

Figure3:

Diagramshowingtheproposedterminationtechniqueapplied.

OnewaytovisualizeaproposedterminationmethodistoterminatetheZ1-3characteristicimpedance.First,thattransmissionlinemustbeterminatedwithavalueequaltoitscharacteristicimpedance.Asingleresistorterminationwillaccomplishthis,howevertheotherthreepairsarenottiedtogetherinnormaluse.

Amoregeneralapproach,intheory,istoformaresistornetworkcomprisedoffourresistorsofequalvalue,oneinserieswiththeotherthreeinparallel.Oncethenetworkissetup,designerscanthencomputetheresistancevaluesuchthatthenetworkmatchestheZ1-3characteristicimpedanceofthecable（Figure4）.

Figure4:

Matchingequations.

Smithclaimsthatallfourresistorsshouldbe75ohms（Figure5）.Thismeansthata75-ohmresistorinallfourterminationswouldmatchacommonmodeimpedanceof100ohmsforZ1-3.

Figure5:

DiagramshowingtheSmithterminationtechnique.

ButSmithclaimsthecharacteristicimpedanceof"some"common-modeconfigurationthatheisattemptingtomatchis145ohms.Sinceit'sdifficulttounderstandinthepatenttheconfigurationSmithisdescribingandconfigurationstypicallyarenotanywherenear145ohms,itcanonlybeconcludedthatSmith'snumbersandapproachtoderivingthemarewrong.Additionally,hismatchingtechnique,givenacommonmodecharacteristicimpedance,doesnotproduceamatch.

Thus,Smithhasn'tpresentedareallyworkablesolutiontotheproblemhehasrecognized.Forexample,forCAT5andCAT5Ecableanimprovementinreturnlossachievedbyreplacingthe75ohmresistorsby52.3ohmresistors,wouldbefrom15dBtomorethan28dB,givenavariationincommonmodeimpedanceof+/-5ohms.Thevariationofcommonmodecharacteristicimpedanceistakenintoaccountinthecalculationofreturnloss（Table2）.

PhysicalVerification

Fortheaboveconclusionstobeaccurate,it'simportanttohaveaphysicalverificationmethod.Fortunately,thereisarelativelysimplemethodfordoingverifyingtheaboveclaims.

Undertheproposedverificationmethod,designersconfigurealengthofcablewiththeterminationresistors,theninsertacommonmodepulsewhilemonitoringthatinsertionwithanoscilloscope（apoorman'sTDR）.Apulsegeneratorwithanoutputimpedanceof50ohmsandarisetimeofapproximately1nswasused（Figure6）.

Figure6:

Diagramshowingtheverificationsetup.

Intheverificationmethod,ifthecableisproperlymatched,noreflectionshouldbeseen.Ifnot,areflectionwillbeobserved.

Intheexampledescribedinthisarticle,terminationsofthreevalueswereappliedateachendofacablesevenfeetlong.Theresistorvaluesemployedweresetsof52.3,75and100ohms.Thecableemployedwas7feetlong.

TheseriesofscopepresentationsinFigure7showimplementationsofthevariousterminationsforcomparison.Thefactthataterminationinthevicinityof52.3ohmsisproperforthematchcanbeclearlyseen.

Figure7:

Diagramshowingreflectionmatchviews.

Forcalibrationpurposes,severalconditionsareshownareshowninFigure7.Thefirstscopepresentationshowstheoutputofthepulsegeneratorwithnoload.ThesecondshowsthecableconnectedtothepulsegeneratorasinFigure6andthefarendopencircuited.Thethirdisthesameconditionswiththefarendshort-circuited.Theremainingthreescopepresentationsshowtheresultsforthethreevaluesoftheterminationresistance.

Itisclearthatforthe100ohmand75ohmresistancesthereisareflection,whereas,with52.3ohmsthereisnot.Theconclusionisthatinthevicinityof52.3ohmsiscorrectandthatBobSmith's75-ohmvalueisnotcorrect.

Asanote,CAT6cablehasahighercommon-modeimpedance.Thisisexpectedduetothesignificantdifferenceinitsconstruction.Itisanticipatedthatalsoduetoitsmethodofconstructionthatthecharacteristicimpedanceswillbemuchmoreuniformbothversuseachotherandalongthelengthofthecable.

TheoreticalAnalysis

Figure8showsthemodelforcomputationofthetermination