Application Note

From the Fluke Digital Library @ www.fluke.com/library

Introduction

Until the late 1980’s electricalcalibration systems used tocompare primary and secondaryvoltages and resistance standardsconsisted of several differentcomponents. Systems like theFluke 7105A and the Datron4900 were the backbone of themajority of electrical calibrationlaboratories the world over.These systems were specificallycombined to provide a traceablesource, according to a set ofmeasurement parameters. Forexample, the Fluke 7105Asystem comprised the following instruments:

• Fluke 720A Kelvin Varley Divider

• Fluke 750A Reference Divider

• Fluke 335A DC Voltage Standard

• Fluke 721A Lead Compensator

• 845AR High Impedance Null Detectors

Similarly, a comparable systemfrom Datron (later acquired byFluke in January 2000) wasalso available. Much like theFluke 7105A, the Datron 4900system included:

• 4901 Calibration Bridge/Lead Compensator

• 4902 DC Voltage Divider

• 4903 DC Calibration Unit

• 4904 Standard Cell Buffer

However, as new innovativetechnology and techniqueswere introduced, both the7105A and 4900 calibrationsystems were soon replaced. Sowhat caused their extinction?

Fig. 1 Fluke 7105A calibration system Fig. 2 Datron 4900 calibration system

Migrating from dc voltage dividers tomodern reference multimeters

Calibrators evolve withpulse width modulation(PWM)

Pulse width modulation topologycan be found in a variety ofapplications, including varioustelecommunications applications,power generation and signalprocessing. Because of itsexceptional linearity benefits,most calibrators today nowinclude this technique in theirown internal ratio divider. Sucha circuit is typically made upfrom a two stage switching FETdesign with synchronouscontrol clock. This circuitpasses the dc reference voltagethrough the switching FETarray and then filters theirsummed outputs to provide anaverage output voltage that isdetermined by the resultingwaveform’s duty cycle. (see figures 3 and 4).

2 Fluke Corporation Migrating from dc voltage dividers to modern reference multimeters

Evolution caused thesemore mature calibrationsystems to begin theroad to obsolescence

There have been several contributing factors to thedemise of the old 7105A and4900 calibration systems. First,the development of artifact calibration has not only consolidated the system into asingle device, but has also fullyautomated the process.Second, the design of moderncalibrators incorporates pulsewidth modulation (PWM) techniques to maintain a ‘rightby design philosophy’ that provides extremely repeatablesource linearity. Furthermore,zener reference technologyimproved, and, when incorporated within calibrationequipment, subsequentlyimproved stability reducinguncertainties. Finally, high resolution DMMs like theWavetek 1281 then managedto combine these features intoa highly accurate electricalmeasurement instrument.

More recently, the introductionof the Fluke 8508A ReferenceMultimeter has taken all ofthese philosophies a stepfurther to improve accuracy,linearity and stability, and hascombined them into a functionally versatile, easy touse solution. This has enabledmetrologists to perform highlyaccurate and automated measurement tasks within asingle instrument, replacing theneed for Kelvin-Varley dividers,null detectors, resistance bridgesand even PRT (PlatinumResistance Thermometer) calibrators. This ultimatelymeans faster calibrations,reduced support costs, greaterthroughput and minimalmanual operations.

Fig. 3 Two stage PWM circuit

The output is then passedthrough a multi-stage, low passfilter network, capable of elimi-nating all ripple and noisecontent, and thus providing ahighly stable and linear outputvoltage. The output voltagecan be expressed using theformula: VO = VIN x X/nThe ratio divider criterion in acalibrator is consequently setby the frequency of the controlclock driving the two FETs.

As with any ratio divider,the PWM technique operateson the basis of ‘dimensionless’ratio. That is, there are noabsolute quantities involvedthat are subject to change overtime offering very repeatablelinearity dependent only uponan extremely reliable digitalclocking waveform. To furthermaintain high confidence, linearity is subsequently verifiedduring artifact calibration. Thisapproach compares the sametwo fixed voltages, V1 and V2,on different ranges. Figure 5illustrates this comparison. Ifthe PWM is perfectly linear,then N4/N3 = N2/N1.

Traditionally these comparisonswere performed using anassortment of ratio measuringequipment to achieve this.However, over the last twentyyears, instruments with artifactcalibration capability have allbut eliminated many of theselabor-intensive measurementtasks. Many traditional manualoperations to establish voltage,resistance or current ratios cannow be accomplished auto-matically within the instrument,consequently providing consis-tent and efficient calibrations,as well as significantly reducingthe costs previously associatedwith higher labor intensity anda larger equipment inventory.

Fundamentally, an instrumentwith artifact calibration capabilities will first transferand then reference to a set ofexternal artifact voltage andresistance standards. Havingbeen transferred, this internalvoltage reference can then beconfigured to appear as if ithad been applied to an internalarray of comparable instrumentslike a Kelvin-Varley divider, anull detector or even a decadedivider, though in practice thisis not really the case.

Fluke Corporation Migrating from dc voltage dividers to modern reference multimeters 3

Fig. 4 A representation of the output voltage prior to filtering

Fig. 5 Converter linearity verification

The development of artifact calibration

Artifact calibration is a processwhere calibrators automaticallyperform internal ratiometriccomparisons and store the correction data relative to afew precise external artifactstandards.

Technology advances haveshrunk the null detector onto aminiature hybrid integratedcircuit, the ratio system is nowa single PWM printed circuitboard, and improved lower costthin film resistor networks havereplaced bulky wire-woundresistor ratios. This kind ofadvance in technology nowmeans that designers canproduce highly comprehensiveinstruments with artifact calibration capability. So now,having eliminated these extraexternal devices and mimickedthe same capability from withinthe instrument, we can nowtransfer the accuracy of theartifact to the various ranges ofthe instrument with minimaluncertainty, with greater accuracy and stability, andwith complete traceability.

The dawning of the highresolution DMMs

As discussed earlier, artifactcalibration is a particularly efficient and easy method ofcarrying out a multitude of calibrations. However, whilebeing an acceptable method ofcalibration, it does come at aprice. Therefore, artifact technology is normally foundon calibrators at the premiumend of the range.

While lower performance, lessaccurate calibrators forego artifact design, adopting more traditional direct function-to-function, range-to-range verification.

Even with artifact calibration,it has been generally recom-mended by all manufacturers tofully verify each range usingexternal methods at least twicein its first year, and then subsequently every two years.It is this reason that many laboratories would, and in somecases still do, resort to moretraditional calibration systems,like the Fluke 7105A or Datron4900, to accomplish this.

Today, most laboratorieswith a large installed base ofcalibrators carry out theprocess of their verificationsusing high resolution digital multimeters like the Fluke8508A Reference multimeter.These meters can be used toperform all of the functionspreviously associated with theolder calibration systems withlittle or no degradation ofuncertainties. Another realbenefit with using these meterscomes from the considerablereduction with inter-connectionleads and the immense timesaved to re-configure the setupfor a different measurement.

Add to this the ability to fullyautomate the calibrationprocess, and the reference multimeter becomes very easyto justify over most of the traditional systems.

4 Fluke Corporation Migrating from dc voltage dividers to modern reference multimeters

Fig. 6 An example of a Fluke 7105A or datron 4900 voltage calibration system

Calibrating the calibratorbefore the days of long-scale DMMs

The advantages of using asingle high resolution DMMover the traditional multi-instrument calibration systemsare probably best demonstratedby firstly describing how atypical calibration would havebeen carried out. Figure 6illustrates the source calibrator(UUT), an external divider, nulldetector and a 10 V dc referencein a conventional voltage calibration setup.

Here the 100 V dc source(UUT) is being verified againsta 10 V reference, using a ratiodivider of 10:1. In reality theratio divider would haveseveral ‘taps’ calibrated to a setof definitive ratios i.e. 100:1,10:1, 1:1 V and 0.1:1 V. Beforeany verification of voltage couldtake place, the individual ratioswould have first been calibratedseparately so that the givenratios exactly represented thesource instrument’s outputvoltage at each voltage step.The possibility of drift withtime, due to the temperature

coefficient of the ratio resistors,meant that this process wouldhave required a skilled metrologist who knew how toperform this operation bothcompetently and promptly.Having calibrated the divider,the UUT calibrator could nowbe connected as described infigure 6. With a null detectorbetween the ratio divider and10 V dc reference, the sourcecalibrator would now beadjusted until the null detectorindicator displayed zero. (Anyresidual error would contributeto the expanded uncertainty.)

From this brief description,you can probably begin tocomprehend how complex thisparticular measurement processis. In addition, the lack of anykind of remote capability, thetime consuming makeup of theprocedure and, above all, theoverall cost of the system, onlyserves to further compound thesituation. Nonetheless,advances in technology, coupledwith the need to make themethodology simpler, faster,cheaper and more efficient,help set a new precedentwithin the industry.

Early precision high resolutionDMMs like the Fluke 8505/8506consolidated the methods usedby all of the test devices illustrated in figure 6 into asingle instrument, so eliminatingmost interconnecting leaderrors, greatly reducing theoverall cost of the calibrationsystem, but moreover, allowingfull automation of virtually allmeasurement tasks. This inturn liberated the seniormetrologist from this task andallowed him/her to concentrateon other important laboratoryresponsibilities.

Reference multimeterwith reference standardaccuracy and stability

High resolution precision DMMshave been available for almostthirteen years, but since theirlaunch in the late 1980s theproducts have remained comparable in both performanceand application. Since Fluke’sacquisition of precision instrument manufacturerWavetek-Datron in 2000,design teams in the US and UKhave worked together andpooled their expertise toproduce the best in precisionand long-scale DMM design -The Fluke 8508A ReferenceMultimeter. The 8508A hastaken many of the leadingFluke and Wavetek-Datronpatented multimeter designsand then improved themfurther, using the latest state ofthe art technology and newelectronic measurement designtechniques.

For the example given inFigure 6, the Fluke 8508Aeliminates every instrumentother than the 10 V reference.In essence, the function of boththe ratio divider network andnull detector has now beenreplicated into the 8508A.

Fluke Corporation Migrating from dc voltage dividers to modern reference multimeters 5

Fig. 7 An example of the reference multimeter being used to accurately verify the output of the artifact calibrator to a known voltage reference

Furthermore, all inter-connecting leads that existedbetween these two instrumentshave also been eradicated,removing the probability oflead errors and consequentlythe need to compensate forthese errors. Having connectedthe 10 V reference to the Fluke8508A’s second input channel(rear input), the 100 V dc fromthe UUT being standardizedcan then be applied to the8508A’s front input channel.This is typically done as shownin figure 7. The 8508AReference Multimeter has twoinput channels that can beautomatically switched toperform a ratio measurement.The 10 V reference would beconnected to the 8508A’s rearinput (channel B), with theUUT’s 100 V dc voltage connected to the 8508A’s frontinput (channel A). In Ratiomode, the 8508A displays theratio of the inputs in the formF-R (front minus rear), or F/R(front as a percentage of therear), or (F-R)/R (the differenceas a percentage of the rear). Inthe example given, the F/R (i.e.the front as a percentage of therear) mode would be used. Inthis mode, with the 10 V dc reference connected to the rearchannel and 100 V connectedto the front channel, the displaywould show +10.000 000 %.This is the ratio of the unknown100 V to the known 10 V reference. Note that the referencemultimeter is measuring thewhole voltage for each channeland is configured to a single dcvoltage range (200 V).Consequently, the only significanterror contributions to thismeasurement are the uncertaintyof the 10 V reference standard,the noise and differential linearity of the reference multimeter and the noise of theUUT 100 V standard.

Typical noise of the referencemultimeter is less than 50 nVpk/pk (7½ Normal & 8½ FastADC modes) with the differen-tial linearity in 8½ digit modebeing better than 0.1 ppm ofrange over a value rangingfrom 10 V to 1 V (halve thetypical linearity spec for valuesspanning the entire DMM scalefrom 0 to 19.99999 V).Note: the above procedureassumes that the 10 V dc reference standard being usedis calibrated and has anassigned value. The assignedvalue is keyed into the 8508Amath memory subsequentlycorrecting any residual gainerror on the multimeters 20 Vrange.Typical procedure sequence:

1) Select DCV 20 V range

2) Connect low thermal cablesto the 8508A front and rearterminals (8508A-LEAD).

3) Short the Front A and Rear B inputs at the cable ends of the 8508A and perform zero range function.

4) Remove shorts and connectcalibrator to 8508A front input.

5) Connect 10 V reference standard (e.g. Fluke 732B) to 8508A rear input.

6) Configure 8508A, rear input,rolling average, 16 samples,8.5 digits, and record referencereading after initiating a two-minute delay.

7) Select math mode, deselect rolling average, enter the reference standard value (as recorded in step 6) into the 8508A “m” variable.

8) Select scan, F/R and allow two minutes for stabilization.

9) Select Math *m.

10) The 8508A will now perform the ratio reading, multiply it by the traceable reference standard value and display the UUT normalized value on the front display.

11) It is recommended that after approximately 10 averaged readings have been recorded, refer back to step #4 to regain a new reference point.

The uncertainty associatedwith this measurement issimilar to that which might beobtained by a skilled metrologistwith a newly calibrated voltagedivider and a null detector. Inaddition, the reference multi-meter can make this measure-ment for prolonged periods, asits linearity does not changesignificantly over time.

Uncertainty components.1) DMM short term stability

when measuring 100 V.

2) DMM short term stability when measuring 10 V.

3) DMM linearity.

The 8508A specifications detailthe uncertainty for the DMM as:10 V measured on the 200 Vrange to eliminate rangeswitching using 20 minutetransfer specification = 0.4 ppmof Reading + 0.1 ppm of range

RSS above result with 100 Vmeasured on the 200 V rangeusing similar transfer specificationand the total measurementuncertainty is approximately2.5 ppm of Ratio.

The 0.4 ppm of readingaccounts for noise and 0.1 ppmof range accounts for a conservative linearity specification.Therefore as two individualmeasurements are being performed during ratio modethe transfer specification foreach measurement is RSS’dtogether yield a measurementuncertainty of 2.5 ppm of Ratioreading. Some additionaluncertainty maybe consideredfor the source.

6 Fluke Corporation Migrating from dc voltage dividers to modern reference multimeters

A conservative linearityspecification is assigned tomultimeters during developmentas part of type testing. Thespecification supports a worstcase linearity measurementwhich is often at the twoextremes of a range. In practicethis spec is often very conservative,an example would be tocompare a known (traceable)10 V with unknown 10 V standard on a fixed range -what is the linearity spec? Inthis example there is negligiblecontribution to linearity as eachmeasurement is being made atthe same point in a givenrange. Linearity measurementscould be performed on everymultimeter to yield betterspecs. However, who shoulddetermine how many points aremeasured to gain confidence?This exercise is time consumingand requires extremely lowmeasurement uncertainty oftenonly achievable using JJ Arraystandards.

Figure 8 shows an example oftype testing linearity on the8508A multimeter. The graphindicates deviation from idealin ppm of range at variouspoints on the multimeters +20 Vrange. A published spec of 0.2ppm of range is assigned, yettype testing results in betterthan ±0.035 ppm of range asindicated by the two extremesof deviation from the ideal lin-earity.

Summary

In summary, a comparison ofmeasurement uncertaintybetween divider system and8508A will often favor the traditional divider systems.However the improvements inmeasurement uncertainty willbe at a cost!

Uncertainty may be compromised using moderninstrumentation, but furtherconsideration must be given tothe intended application andhow it may have changed.

Precision calibrators oncerequired divider type measurement uncertainty now implement artifact calibration.Where linearity verification isinvolved divider technology isnow internal to the calibratordesign. This is not true of allcalibrators where linearitymeasurement using multimeterperformance remains adequatefor lower priced, less accuratecalibrators.

Fluke Corporation Migrating from dc voltage dividers to modern reference multimeters 7

Fig. 8 A typical example of type testing linearity on the Fluke 8508A

Fluke Corporation Migrating from dc voltage dividers to modern reference multimeters

Fluke. Keeping your worldup and running.

Fluke CorporationPO Box 9090, Everett, WA USA 98206

Fluke Europe B.V.PO Box 1186, 5602 BD Eindhoven, The Netherlands

For more information call:In the U.S.A. (800) 443-5853 or Fax (425) 446-5116In Europe/M-East/Africa (31 40) 2 675 200 or Fax (31 40) 2 675 222Canada (800)-36-FLUKE or Fax (905) 890-6866From other countries +1 (425) 446-5500 or Fax +1 (425) 446-5116Web access: http://www.fluke.com/

©2003 Fluke Corporation. All rights reserved. Trademarks are the property of their respective owners.Printed in UK 10/2003 2114953 D-ENG-N Rev A, DS271

Reference standard accuracy and stability, in onefunctionally versatile, easy to use solution

Fluke 8508A Reference Multimeter

Model8508A 8.5 digit Reference Multimeter, Certificate of Calibration and User Manual8508A/01 8.5 digit Reference Multimeter with front & rear input binding posts,

Certificate of Calibration and User Manual

AccessoriesNVLAP NVLAP Accredited CalibrationUKAS UKAS Accredited Calibration8508A-SPRT Standard Platinum Resistance Thermometer8508A-PRT 100 Ω PRT8508A-LEAD Comprehensive Measurement Lead KitY8508 Rack Mount KitY8508S Rack Mount Kit Slides8508-7000K Calibration Kit

Ordering information

Other precision instrumentsin the range

5720A MultifunctionCalibrator

9500B OscilloscopeCalibrator

The lowest uncertainties of anymultifunction calibrator

High accuracy calibration ofanalog and digital-storage oscilloscopes up to 3.2 and 6 GHz

Superior accuracy and functionality in an economicalbenchtop package

525A Temperature/PressureCalibrator