Getting the best out of biccmicall antennas for emission measurements and test site evahatiom
Martin J Alexander Marth H Lopez Martin J Salter NationalPhysical Laboratory CENAM National PhysicalLaboratory Teddington km 4.5 can-.8 Los cuts Teddington TWll OLW,UK. El Marqu&, Cl?76900 QuerCtraro,M&&o TWll OLW, UK.
Abstract:
Thebiconical antemlahas been adopted for EMC testingbecause of its broadband,frequency coverage, 20 MI-J?to 300 MH?z,and compactsize. Potentiallythe antennaworks very well, but some models have poor reproducibility resulting in perceptionsof excessivelylarge uncertaintiesfor emissionmeasurements and predictingpessimistic performances of test sites.The antelmacan Side View End View be accurately modelled, which enables normalised site attenuation @ISA) values to be based on more rigorous Fig. lb Biconical antennaelements with cross-bars theoreticalprinciples, andchallenges the validity of the currently acceptedpractice of calibrating antennasat 3 m and 1 m The tuned dipole antelmais the industry standardand its use is distances.Antemla factor can be measuredwith uncertaintiesof enshrinedin CISPR Publication 16-l’. The problem with using f 0.2 dl3, which togetherwith calculability,make the biconical this antennais that it has to be tuned at every measurement antenna a worthy reference antenna. The use of properly frequency.These days modern receivers and spectrum analysers constructedcollapsible elements is recommendedfor portability. are able to make the equivalent of swept frequency measurements,which is highly desirable in ensuring that all 1 Introduction unwantedEMC responsesare identit?ed.It is customaryto use broadbandantennas, the biconical being the acceptedtype LIPto The tuned“ half-wave” dipole antennahas the statusof being the 300MHz, andthe log-periodicantenna taking over to 1 GI-Tzand acceptedreference antenna for measurementof E-field strength. beyond. CISPR 16-1 requires that the antenna factor of This is becausethe theory of the radiation propertiesof such broadbandantennas be referredto the tuneddipole, including the antennasis well establishedand the antemlafactor is accurately effect of radiationpatterns. It is advocatedin this paperthat it is calculable, and the antenna is relatively easy to construct. possible to be weanedoff dependenceon the hwd dipole and Around the 1950sthere was a requirementfor a more compact that broadbandantennas should be establishedas reference antennafor making emissionmeasurements in screenedrooms. antennasin their own right. This is feasible on two accounts. The size of a typical room ruled out the useof a dipole tunedto Firstly there is a handful of laboratoriesin the world that offer 30 MHz since it was 4.8 m long. The biconical antennawith cost effective and accuratecalibration of broadbandantennas. “wire cage” eleme& was introduced,whose overall length is Secondly it is possible to model these antelmasso that they approximately1.4 m in the dimensionsspecified in MIL-STD- becomecalculable reference antennas. 461, still usedto this day, seeFig. la. Over tile yearsthe baluns havebeen improved,enabling a frequencyrange of 20 MHz to The performanceof broadbandantennas can be modelledwith 300 MI-h to be covered,with an improvedreflection coeficient. great precision using NEC2*3 . This paves the way to perfonn Schwazbeclcintroduced the singlecross-bar in eachelement, see experimentswith antennason the computer,which hasbeen done Pig.lb, which suppresseda high-Q resonanceat 287 MHz. for this paper to investigatethe methodsof biconical antemia calibration for 1 m distanceaccording to ARP958”, and NSA Balun , Metal Supports measurementsat a distanceof 3 m accordingto ANSI CG3.4$. The NEC calculatedproperties of antennashave been validated by measurementto uncertaintiesas low as& 0.1 dB on a National Standardground planeG.
The applicationof the tuneddipole antemlato openarea test site validation and EMC testing has its own problems, For many x years only free-spaceantenna factors were considered.More Side View End View recent technology has enabled overall EMC measurement uncertainties to be reduced, which has shown up antelma Fig. 1a Biconical antenna“wire cage”elements propertiesthat were originally ignored,see clause 4.3 of ANSI
0-7803-4140-6/97/$10.00 84 C63.57.Antenna factor, AF, is affectedby mutualcoupling ofthe 10 antennawith its image in the ground plane, such that AF can changeby 6 dB when a dipole tuned to 30 MHz is scannedin Sleightfrom 1 m to 4 m. Variation of Al! with height for a 16 biconical antennais dealt with in Section 2. When it bec So far in this introduction intrinsic electromagneticproperties of IO antennas have been considered. Ways of measuring these propertieswill be given in the following sections,together with appropriate ways of using the measureddata. These will be _.-.-.- 1.5 m - 1.0m comparedwith cunent published n@hods, with shortcomings ----- Frees/~ace 6 highlighted.But now anotheraspect of the use of antennasmust be introduced. Some antenna models have exhibited poor 4 reproducibility of results. With hindsight one can say that these 30 50 70 90 110 130 150 170 190 210 230 250 models had design deficiencies, but this has wrongly led to exaggeratedmeasurement uncertainties for EMC activities.It has Frequcncy(MHz) led to speculationabout OATS defectsand aboutthe criticality Fig. 3 Antennafactor calculatedat 7 heightsand in free-space of equipment layout. Two common problems are bahm imbalance in biconical antennas and breakdown of RF Themeasured AFs are about 1 dB higherwhich is partly because connection on log-periodic array elements. Both of these the balun is not in the model, thereforeno balun loss, and also problems are eliminated by the choice of well designedmodels the model may not be complete,for examplethe solid conical of antenna. sectionswhich hold the wires togetherwere not in the model. The measurementswere of a 1:1 or 50 ohm balun. 2 Antenna factor clmwteristics of biconical antennas Note that for frequenciesbelow 55 MHz the antennafactor changesvery little with height. This is becausethe antemlais The antennafactor of a horizontally polarisedbiconical antemla behaving like a short dipole with a large input reactance.If the varies by up to 1.8 dB when the antennais scannedin height higher antennafactor can be tolerated,an accuratelycalibrated from 1 m to 4 m over a conductingground plane. This has been biconical antennawill generallymake it easierto obtain lower accurately measured”Or height increments of 0.5 m against measurementuncertainties than will a tuneddipole antennaused calculablestandard dipoles on a national standardGO m gound below 55 MHz. Becauseof the antenna’shigh self-impedanceat plane, Fig.2. In contrastantenna factor changesmuch less for a these lower frequencies,the effect of mutual coupling to its vertically polarised antenna. The change can be ignored for sunoundingsis negligible. heights above 1.5 m, where the antennafactor is effectivelythe free-spacevalue. This large variation of antenna factor with height causes a - problem with the accuratecalibration of biconical antennasby the ANSI method7.One of the three ‘\,- The insertion loss between two bicones was measured and ,\, compredwith prediclion. The agrm~nent wt.~s better 111~1 1 dI3 asshown in Fig. 4. The biconeelements were on detachableNPL \ balls whoseS-parameters could be measured.The biconeswere -- horizontally polarisedat a height of 2 m and separatedby 10 m. 30 50 70 90 110 130 150 170 190 210 230 250 The shapesof the curvesare very similar, andthe model is being refined to give better agreementwith measurement,by altering Frequency(MHz) the segmentationand including the solid metal endsof the cages Fig. 2 Antennafactor measuredat 7 heightsand in free-space in the model. 85 Biconical antennaswere originally intendedfor use over the frequencyrange 20 MHz to 200 MHz. Thefirst designshad a poor return loss, typically less than 4 dB over most of the frequencyrange. This figure was improvedto 6 dl3 by a new balun designwith a 1:1 transformer,which also gavea good performanceup to 300 MHz. Howeverthe AF was typically 19 dB/m at 30 MHz. Thenext ma jorimprovement was to usea 4:l balun,otherwise known as 200 ohm,which gavean AF of 13dB/m at 30MHz andimproved return loss. The poor return loss below 40 MI-Iz is causedby the high reactancetypical of a shortdipole antenna. In orderto makethe antennaa more eficient transmitterfor immunity testing, the antennaelements can be replaced by longerelements. The propertiesof thebilog antennaare equally calculable using 20 40 60 60 100 120 140 160 100 200 220 240 260 260 300 320 NEX, which can be used to explore radiationpattems and Frequency (MHz) uncertaintiescaused by the variation of phasecentre position Fig. 4 comparisonbetween NEC andmeasured SA with tiequency.This is the subjectfor L?esignimprovements Collapsiblebiconical elements One problem with the popular designof biconical element, Fig. 1a, is t&e occurrenceof a 10 dl3resonance at 287MHz, see Anotherdevelopment is the use of collapsibleelements which Fig.5 are useful for compact storage of the antenna and for transportation.The cost of transportby air can be more thran halvedif the antennacan travel by weightinstead of by volume, asfor the rigid elements.The collapsible design that is a hinged version of the rigid designis not recommendedbecause the hingescan give intennit&entW contact Fig. 5 SimulatedAFs of biconeusing NIX This is causedby a cavity resonancein thewire cageand can be removedby placingone cross-bar within eachcage, F ig,lb, or by usingan open structure as in Fig.6,below. One bicone mode lhas three cross-barsin eachcage, but this is a retrogradedesign featureas it throws up a sh At 1 m height there is a feature of 1 & in magnitudeat * 116 MHz. This “resonancewas” measuredat 120MHz of this gaveexactly the samemagn itudeas predicted.In practicethis will rarely affect measurementsbecause the horizontally Side View End View 86 polarised bicone is usually higher than 1 m at this frequency The computedH-plane pattern at 300 MHz is shown in Fig. 10. when recording results. The antennafactors of the collapsible Whereasthe pattern for using collapsible elementsis circularly and rigid bicones are comparedin Fig.8, which showsthat the symmetrical, the one for cage elements is not. There is an A.F of !lie collapsible bicone is lower at the top end of the asymmetiy of 0.6 dB. When the cross-barwas removed in the frequencyrange and is better behaved. model, the symmetrywas restored. This explains a difference betweenthe measuredand NSA and the ANSI prediction above 260 MHz. 20 40 60 80 100 120 140 160 100 200 220 240 260 260 300 320 Frequency (MHz) Fig. 8 Comparisonof AFs of cageand collapsible bicones -.----- Ordinary Bicone ------Collapsible Bicone Cross-polar performance The computedcross-polar signal of a biconical antennaover the Fig. 10 I-I-planeradiation pattern at 300 MHz frequencyrange 20 MHz to 300 MHz was more that 1000 dB lessthan the co-polar signal.Providing a physicalantenna is built Balun imbalance symmetrically, the measuredcross-polar level should be well below the - 20 dB required in CISPR 16-l) clause15.4.2. A lot of EMC engineerswill be familiar with the following scenario.They designand build an OATS andthen evaluateit by Radiation patterns measuringsite attenuation.Despite adherenceto bestpractice for OATS designthey fmd that the NSA fails to confom~with the The good agreementbetween calculatediu?cl meastlred antenna CISPR 16-1 criterion, and can daer by as much as + 10 dB for factor adds coltidence to the alreadyknown ability of NW to verLica1polarisation. So a lot of money and effort is spent in accurately predict radiation patterns. It is not so easy to repeating the measurements,covering other locations on the accuratelymeasure radiation patternsof dipole antennasbecause groundplane, inspecting the integrity of the welds,raising doubts of their lack of directivity. At 300 MHz the bicone length is 1.4 about edge effects and even enlarging the ground plane. All wavelengths, at which a single wire dipole would exhibit along the culprit was the antennaswhich had unrecognised multilobing. This is not app‘arentwith the bicone, as shown in faults, and the OATS was perfectly satisfactory.These faults Fig 9 below, but the E-plane patternis slightly more directive at have led to all sort of speculation,which hasbeen perpetuated in ,300 MHz than at 250 MHz. I-Iowever the pattern of the the literature, aboutthe best way to designsites, the best way to collapsible elementsat 300 MHz is similar to that of the cage lay out the antennacable, the use of ferrite clamps on the cable, elementsat 250 MHz, again showing the superiorperformance all of which are secondaryor insignificant issuescompared to of the collapsibledesign. proper antennadesign. The fault is balun imbalancewhich occurs on some models 01 biconical antenna. This causesunwanted current to flow on the outer conductorof the input cable. The radiationfrom the cable interferes with the deld from the EUT being measured.In the worst casesthe signal is nearly cancelled.The effect is most pronouncedwhen the input cable is laid out parallel to and in proximily to the antenna elements. Fortunately antenna -__-__-__- 250 MHz rn:&‘~lurers tie becomjngaware &this problemand recti&ing ------300 MHz a7 it where needed.Imbalance is less unacceptablein high power the ANSI formula for R = 3 m andfixed antennaheights of 2 m. models which are used to set up field for immunity testing, becausethe field is calibratedusing a field probe. The effectsof The defmition of NSA usedin ANSI 63.4 doesnot take account imbalancecan be reduced by extending the cable by several of near-fieldterms. metresbehind the antemlabefore allowing it to becomeparallel to the antemlaelements, and alsoby placingfenite clampson the However the usethe ANSI NSA valuesin Table 1 of Ref. 5 was cable,though this latter cannoteliminate the effect. found to be consistentwith the methodof calibrating antennasat R = 3 m in ANSI C63.5, which likewise doesnot use near-field terms. This was done by running a computer simulation of the ANSI method. In other words the ANSI method can give an accurate evaluation of site perfonnauce, but it requires that antennasbe calibratedaccording to C63.5. It hasbeen proposed to CISPR/A that free-spaceantelma factor is adopted as the _. primsuy parameter&aracterising EMC antennagain. This has -. the advantage9that a unique and fundamentalproperty of the antenna can be used with the minimum of measurement uncertainty for different antemla separations, heights and _.J’ polarisations. The C63.5 method yields AF with an inherent uncertainty, as alluded to in the 3rd paragraph of Section 2 _. above, and thereforemay be no more accuratethan the method proposedas a replacementin the following section. Proposedclerivrction of NSA using AF,, Frequency (Ml-k) VP ANSI C63.4 usesthe formula, Ref. 11, in Smith et al’s classic paper to calculate NSA. Whilst the far-field approximation serveswell for distancesgreater than R = 10 m, it doesnot hold Fig. 1I Balun imbalanceof high power biconical antenua for 3 m. The requirementby C63.4for a 3 m calibrationfor NSA purposeshas causedmuch confusionin the minds of test house A typical plot of imbalance is shown in Fig.11, showing operators as to the need for an Al? value which is related to uncertaintiesof up to f 4 dB. This is measuredby settingup the distance. It has also generateda demand for the elToneous receive antenna vertically polarised with the cable hanging concept of distancerelated Al%. Whilst 3 m Al? is neededfor vertically behind it as in nornlal use, but without fenite clamps. C63.5 NSA, it is the cause of increasederror when used for A transmit Retimmces 1 CISPR publication 16-1 1993. Spec$cation ,fir radio disturbance and i77~7nunity7neasuri77g apparatus and 777ethods.EC, Geneva. 2 LOGAN, J.C. and BURKE, A.J., Nz,mericul L?lcctro7nagnetic Code, 1981, Naval OceansSystems Cenlre,CA, USA. 3 Maml, S.M. and Marvin, AC., Characteristicsof the 89