---

..

---

INT:~RCOMPARISON OF NIST, NPL, PTB, AND VSLTHERKA:' VOLTAGE CONVERTERS FROM 100 kHZ TO 1 MHZ

J.R. KinardNIST*, Gaithersburg

MD, USA

R.B.D. Knight , P. MartinNPL, Teddington

En~'land, UK

Abstract

Coaxial, thermal voltage converters (TVC's)have been hand-carried between NIST, NPL, PTB,and VSL for intercomparison of ac-dcdifferencefrom 100 kHz to I MH::. This paperbriefly describes the methods ..ndunderlyingprinciples on which ac-de differencedeterminations are based in ecch laboratoryand gives the results of the inturcomparisons.

Introduction

Thermal voltage converters form the primaryand working standards at the fourparticipating laboratories for ac-dcdifference, and hence ac voltage, in the100 kHz to 1 MHz frequency range. Anintercomparison of the ac-dc cifferences ofTVCs in this range is particularly informativebecause the methods and hardwale used by thefour laboratories differ significantly. Theagreement between methods adds considerableadditional confidence to the pr-ocess. NISTacted as the pilot laboratcry for thisintercomparison.

Methods of Individual Laboratories

National Institute of Standards lnd Techno1oav(NIST). A group of multijun=tionthermalconverters(MJTC's) has been esta~lished as theNIST primary standards for ac-ec difference.For these MJTC's constructed with bifilarheaters, thermoelectric and othur errors havebeen shown theoretically and exp'~rimentallytobe below the 0.5 ppm level in the range 30 Hzto 10 kHz from 2 V to 10 V [1]. The fact thatthese MJTC's were the resul1. of variousfabrication techniques over seme period ofyears, improves the confidence :.nthis group.

The extension of the frequency range from10 kHz up to 1 MHz was achie\ed using twotechniquesemployingseveralTVC:'scontainingcylindrical, deposited-carboll resistorsmounted coaxially with 5 or 10 mAthermoelements [2]. Firstly, 5- to 25-V TVC'swere made with geometries alld electricalcharacteristics expected to have nearly zeroresidual reactance and to be neilrlyfrequencyindependent up to 1 MHz. Interc:omparisonsofnearly identical structures wLth differentrange resistors demonstrated frequencyflatness of :t5 ppm out to 1C0 kHz and of:t10 ppm out to 1 MHz.

Secondly, the major ac-dc differencecontributions from each structul"alelement orregion of TVC's similar to th:lse describedabove were theoretically and Hxperimentallyanalyzed up to 100 MHz. These contributionswere: voltage standing-wave :.n the inputconnector, transimpedance of the seriesresistor, current standing-wave in thethermoelement, and skin effect. Examinationof both range dependent and rancreindependentcontributions confirmed ac-dc cifferences at1 MHz to be <10 ppm. The methods ofintercomparison are described ill [2].

.Electronfcand Elee. Eng. lab., u.s. Dept. of Corm., Technology Acinin.U.s. Goverrmentwork not protected by u.s. copyrfgh t.

M. KlonzPTB, BraUnschweig

Germany

J. de Vreede , J. DessensVSL, Delft

the Netherlands

+

National Phvsica1 Laboratorv (NPL). Thefoundation of all ac-dc transfer measurementsat NPL is a collection of MJTC'S[3] obtainedfrom a variety of sources. Single-junctionthermal converters (SJTC's) are selected andcalibrated in the current mode by comparisonwith an appropriate MJTC in the mid-frequencyrange where the errors in MJTC's are know tobe small. The errors introduced byextrapolating the response of the SJTC up to1 MHz are estimated and form a majorcontribution to the uncertainty of themeasurements. Calculations of the parasiticinductive and capacitive properties of 1 kOmetal film resistors mounted in relativelylarge cylindrical enclosures show that theseeffects produce negligible errors, even at1 MHz. Various versions of suitablestructures have been described, e.g.[4], andare widely used for transfer measurements.The combination of such a resistor with a 5 mAor 10 mA SJTC provides a transfer standardrated at 5 or 10 V.

Comparison of standard and test items iscarried out using a highly automateddigitalsystemwhich modelsthe operationof an analogbridge [5]. The two signal outputs areamplified and digitized by high resolutionDMM's. Sequences of 13 measurements arecollected with the signal being switchedbetween ac and dc sourcesin accordancewitha specificpattern. Arithmeticalmanipulationof the data allows the resultsto be largelyindependent of drifts in the devices andsupplies.

Phvsika1isch-Technische Bundesanstalt (PTB).At the PTB multijunction thermal convertershave been developed as basic standards forac-dc voltage transfer over the entirefrequency range from 10 Hz to 1 MHz [6,7].Transfer differences caused by thermoelectriceffects, which are present to some degree innearly all thermal converters, as well asfrequency-dependent transfer differences inthe aUdio-frequency range, are negligible inthe PTB MJTC. At higher frequenciesthefrequency-dependent change of the inputimpedance of the heater circuit is responsiblefor ac-dc differences.

The bifilar heater, the intermediate leads,and the coaxial connector up to the referenceplane of the tee connector have been takeninto consideration. The heater resistance withits residual inductance, capacitance, anddielectric losses; the intermediate leads withresidual impedance, skin effect, and proximityeffect; and the connector circuit withresistance, skin effect, and dielectric losseshave been measured or estimated as far aspossible, and the transfer differencescalculated for different heater resistances.As an experimental check, the differencesamong the various MJTC's have been measured.Quite good agreement between calculated andmeasured differences was obtained which givesconsiderable confidence to the correctness ofthe method.

For frequencies up to 100 kHz, a 190-0 MJTC(2.3 V) has been chosen as a referencestandard. For this resistance, the effect ofresidual capacitances and dielectric losses

are just compensated by inductance, skin andproximityeffectswithin the MJTC. only theconnector contributionis left, and it hasbeen measured using a MJTC as a referenceplaced directly at the pins without aconnector.

As the skin effect increases with frequencysquared, such a MJTC has a transfer differenceof 92 ppm at 1 HHz. Therefore a MJTC withhigher heater resistance is preferable, and a700-0 MJTC (5.6 V) is taken as a referencestandard. For higher voltages, seriesresistorsand 190-0 MJTC 1S are combined andcharacterized in a step-up procedure.

Van Swinden Laboratorv (VSL). In the VSL theac-dc difference standards between 100 kHz and1 MHz have been developed on the basis ofcalculable HF ac-dc standards which are usefulup to 30 MHz [8]. Recently a new set ofstandards was constructed with smaller ac-dcdifferences above 1 MHz [9].

Up to 100 kHz a set of multijunction thermalconverters of PTB design and mounted inVSL-built housings are used as references [6].These converters are considered to be nearlyindependent of frequency up to 100 kHz. Thecalculable converters and the PTB MJTC's havebeen compared to confirm the validity of anumber of assumptions in the calculation forlow frequencies. within the combineduncertainties, both types of converters givethe same results up to 100 kHz.

As standards for this intercomparison,oneMJTC, two old style HF TVC's, and one new HFTVC were used. All of the standards wereprovidedwith type-N male connectorsand theplane of reference for all the VSLmeasurements was the symmetry plane of atype-N tee connector. The measurementsystemwas an updated version of the VSL systemdescribedin [10]. In the updatedsystem theindividual thermal converter outputs areamplified by a factor of 100 and then thedifference is amplifiedagain by 500. Theoverall VSL uncertaintyis estimatedon thebasis of the above mentioned analyses andinternal comparisons of 2- to 4-V TVC's in therange 100 kHz to 1 MHz.

Transfer Standards and Results

The transfer standards used in thisintercomparisonwere TVC's of three basictypes. The first type consisted ofcommercially manufactured, coaxial, 5- & 10-VTVC's containing UHF-pattern thermoelements,internal shield structures, and conventionalseries resistors. The partial results givenin Fig. 1 for 200 kHz, 500 kHz, and 1 MHzare for these commercial TVC's. The secondtransfer standard was a VSL constructedcoaxial, 5-V TVC containing a thin, straightwire for the series resistor. The third typewas a PTB constructed MJTC with a 700-0,bifilar heater. The commercial TVCs and thePTB MJTC were mounted with type-874connectors. The VSL TVC had a type-Nconnector. The 100 kHz results include valuesfor all three types of TVC's. Allmeasurements were referenced to the center ofa type-874 tee structure.

The results from each laboratory were averagedto produce a representative value for thatlaboratory at each frequency. Therepresentative values, or deviations, for alllaboratories at each frequency were averaged,and these values are plotted in Fig. 1 asrelative to the overall average which isindicated by zero ac-dc difference.

Fig.1.Partialresults of intercorrpedson wfth the overall averageindicated as zero BC"'dedf1ference.see Results Section for details.\.'here given, error bars res resent t2a.

References

[1] F.L. Hermach, J.R. Kinard, and J.R.Hastings, "Hultij met ion Thermal Converters asthe NBS Primary ,.C-DCTransfer Standards forAC CUrrent and /01tage Measurements," I.Illi1itTrans. Instrum. Heas., vol. IM-36, pp.300-306, June 1987.

[2] J .R. Kinard .md T.X. Cai, "Determinationof AC-DC Difference in the 0.1-100 MHzFrequency Range," IEEE Trans. Instrum. Meas.,vol. 38, pp. 360-367, April 1989.

[3] F.J. Wilkins, T.A. Deacon and R.S. Becker,"Multijunction Tharmal Converters," Proc. lEE,vol. 112, pp. 794-805, April 1965.

[4] F.L. Hermach and E.S. Williams, "ThermalVoltage Convert'lrs for Accurate VoltageMeasurements to 30 Megacycles per Second,"Trans. AlEE (Comm. and Elec.), vol. 79, pt. I,pp. 200-206, Jul} 1960.

[5] R.B.D. Knigh:, D.J. Legg and P. Hartin,"Digital 'Bridge 1 for comparison of AC-DCTransfer InstrumEnts," Proc. lEE A, vol. 138,pp. 101-104, March 1991.

[6] M. Klonz, "IC-DC Transfer Difference ofthe PTB-Multijunction Thermal Converter in theFrequency Range :'rom 10 Hz to 100 kHz," ~Trans. Instrum. Iofeas.,vol. IM-36, pp. 360-367, June 1987.

[7] M. Klonz, "AC-DC Voltage TransferDifference of tte PTB-Multijunction ThermalConverter at 1 HIz," Digest Conf. Prec. Elec.Meas. 1990 (CPEM'30), ottawa (Canada), pp. 60-61.

[8] M. Nomair (ndAccuracy CalculaJ 'Iefor the LF-30 offizTrans. Instrum. ;illd345, April 1989.

K.J .P.M. Harmans, "HighAC-DC Transfer StandardsFrequency Range," lEEE

Meas., vol. 38, pp. 342-

[9] M. Nomair ard K.J.P.M. Harmans, "A HighAccuracyCalculalleAC-DC TransferStandard,"VSL Report ES-88-D4 (implementation in annex),1989.

[10] K.J.P.M. H,rmans, J.J. Schmitt and R.Kaarls, "A High Precision Automated system for

AC-DC Transfer Using Thermal Converters,"Digest Conf. pre=ision Elec. Heas. 1982, pp.D5-D7, 1982.

100 kHz II 500 kHz

: ipilIiis

i

.."' ... on ... ----J I

M.ST "'.." ..

1 MH.200 kHz

L*jIII

PT. "." II MI8T ...." "T8 vaL