Instrument Accuracy and Calibration

Hi there,

You will have to get your multimeter calibrated. You should get a certifacte back with it showing the %error that it was calibrated to to. You will need some calibration equipment for example Beamex. In the case of a pressure transmitter for example, you will use as well as the beamex with its pump the multimeter to calibrate the transmitter by pumping air into the transmitter and setting the zero and span. The bemex will note the mA reading at various points and can print out a calibration certificate with all the necessary data for %error, repeatability and all that good stuff you need. These certificates can be used for ISO ratings as well as custody of transfer. - www.beamex.com

They also used to have a portable calibrator for temperature transmitters - but I have not seen one in years.

Cheers

Craig

I have a calibrated Meter calibrator, Decade box, as well as, a Volt meter and DMM.

I am calibrating, in house, our other DMM's not in calibration lab. Alot of the models give different accuracies as to their readings. That's where the problem lies for me.

If a specific model states it's accuracies as +/- .02% of reading + 2 digits, how do you figure this in to TUR? ( if you may know).

Without long explanations just for practical using:

I'd conduct 30 test readings. All thirty should be in within an accuracy span -0.02%...+0.02%. If some of them is out. Ok there's a chance here (to add or subtract 1 digit related to multimeter discrimination performance). It gives some room to improve overall result. It might help or it might not. For latter, alas, the multimeter does not fit had been declared accuracy class (3Sigma rule).

The accuracy of an instrument is often stated as a percentage. For example, a voltmeter may be specified as 1% accurate. An important question is: 1% of what? Is it 1% of the reading or 1% of the "full-scale"? Suppose you have a meter which reads voltages in the range of 0 to 100 volts. Then the full-scale value is 100 volts. Now suppose you use that meter to measure an unknown voltage, Vx, and it reads 50 volts. If the accuracy of your meter is ±2% of the reading, then the actual voltage is somewhere between 49 volts and 51 volts since 2% of 50 volts is 1 volt. On the other hand, if the accuracy is ±2% of full-scale (or f.s.) then the actual voltage is somewhere between 48 volts and 52 volts since 2% of 100 volts is 2 volts. Accuracy is often given as percentage of full-scale, which means you should use the lowest scale you can to make the measurement. Suppose a 5% voltmeter has two ranges, 0-10 volts and 0 to 20 volts. If you want to measure a 9-volt battery then you should use the 10-volt scale since 5% of 10 volts is 0.5 volts while 5% of 20 volts is 1.0 volts.

Digital Meters: That Last Digit
Digital meters are often compared by the number of digits they can display. For example, a 2-digit meter can display values from 00 to 99 while a 3-digit meter can display values from 000 to 999. Suppose you have a 2-digit voltmeter that reads 0 to 99 volts. Effectively it has a full-scale capability of 100 volts. Suppose you use it to measure a voltage with a value of 50.5 volts. What will the meter read? The only choices are 50 or 51, so either way there will be an error. That fact about digital meters is expressed by saying that all readings are plus or minus a count of one.

Digital Meters: Accuracy vs. Resolution
The fact that the reading on a digital meter is always uncertain by a count of 1, either up or down, defines the resolution of the meter. Resolution is the smallest change an instrument can measure (or "resolve"), so in a digital instrument it is the last bit: +/- a count of 1. Like accuracy, resolution can be expressed as a percentage. A 2-digit meter has 1% resolution (1 count out of 99) while a 3-digit meter has 0.1% resolution (1 count out of 999). However, resolution is not accuracy. A 3-digit meter has 0.1% resolution buy may only have 0.5% f.s. accuracy. Read the specifications of the meter carefully: it is usually the case that the resolution is better than the accuracy.

Which is the most accurate - a sundial or a watch? The differences in tolerance v accuracy may be considered with reference to time. A sundial may have dial marks at every half hour - its resolution is strictly half an hour (although you could estimate more closely) However, its accuracy, is absolute i.e. = zero. (if properly positioned of course - and it's sunny!)

Whereas a watch may have a second hand and therefore a resolution of 1 second, and yet the actual time, and therefore accuracy, may be totally wrong. As the saying goes "A stopped watch is absolutely accurate twice a day!"

Hi bettsd44,

"Wanting some clarity on accuracy information on equipment that states, for example, +/- .02% + 1 digit."

If you have a voltmeter that reads 10.0000 volts, one digit is .001%, so the spec is .021% for that resolution. TUR is test uncertainty ratio, it's the spec divided by your lab's uncertainty for that voltage. If the spec it 0.1%, and your uncertainty is 0.025%, then the TUR is 4 for that point.

The following site can give you information about training in calibration:

http://www.ncsli.org/

Regards,

S

..understanding of specification (from the inside of instrument)

In measurement there are two main problems: You have to have reference, and You have to compare reference with measured value. Both, reference and comparison suffer from imperfections and involve errors. Let's take voltmeter as example.

Inside voltmeter, in the beginning may be Voltage Divider, then is voltage follower/amplifier. Then something that compares measured voltage with reference voltage, and converts to digital form (A/D converter)

Input divider can be not accurate (temperature, humidity, time dependent) and involves scale error (for example 1:10 divider giving at output 1.0001V instead of 1.0000 when on input is 10.000V). Amplifier/follower usually has drift of null. That means, that even if at input is applied 0V, output voltage drifts around null (again influence of temperature,....time..) and voltmeter does not always displays 0V. This gives zero error. And from here there are those +- digits. Amplifier also introduces its scale error. Then Reference voltage has its scale error(Zener diode drifts with temperature and time). Comparison circuit involves another null error. Amplifier is usually nonlinear (higher voltage ==> more power ==> higher temperature ==> change of resistance). But lets leave going more deeply in details. So..

In general there are two basic errors: error of null and error of scale.

Error of scale (slope) can be described as % of reading.

Error of null can be described as % of FS (Full Scale ) or in digits (if digital meter).

For example If we have voltmeter with 5 1/2 digits : with max display on 200V range 199.999 with specification: 0.01% of reading +-3 digits. That means: 0.01% is scale error, and +- 3 digits (equivalent to +-0.003V) is zero error + resolution error (see AussieBob)

Resulting final error depends on measured voltage (on selecter range): When You will measure 190V, You will get 190*0.01% =,019V scale error + .003V zero error = 0,022V, being 0.0116% of measured Voltage.

Measuring 100V will give 0.013% , .. measuring 21V will give 0.024%. Measuring 10V on 200V range would give 0.04%, ...but it is better to measure it on 20V range.

The same zero error could be specified as 0.0015% of FS (Full Scale) 200V * 0.015% = 0,003V

But if specification is: 0.01% of FS (Full scale) +- 3 digits, then final errors will be for the same measured voltages as follows; 190V ==> 0.012%, 100v ==> 0.023%, 21V ==> 0.115%

Above errors could be not all. If above mentioned Voltmeter has input impedance 10 MOhm and You will connect it to Voltage source with output resistance 1 MOhm, then You will get extra 9.1% error. This kind of error can be calculated and corrected (if realised! and taken into consideration).

In general, if You think about TUR/calibration - I recommend You to see >Guide to the Expression of Uncertainty in Measurement<

Thanks for the help.

Along these lines, here is an example. Trying to write a procedure for a conductivity meter. We have 1500ppm solutions as 442 Std. here with a stated NIST accuracy of +/- 1%. The meter is accurate to +/- 1% full scale.

I am trying to do the TUR, so I need to know how to work the accuracies correctly to do so. I am not sure if the solution is more accurate to being 1%, period, or if at full scale % is more accurate?

±0.02% + 2 digits

0.02% should be defined as 0.02% of what. Is it 0.02% of full scale or of reading. this is easy to calculate. Then add the 2 least significant digits to this value. both must be in the same units.