Why is measuring a single voltage considered useful for checking multimeters?

[niconiconi]

I've noticed on this forum that getting a voltage standard is the first step of becoming a voltnut. For example, LT1021-based DMMCheck's 5 V standard is popular. Some people also use this as a benchmark for multimeters to post measurements results of all types of meters. If the screen reads "4.99 V", then the meter is said to be accurate.

But it only tests a single point and a single range, so I don't understand why is it considered an useful test.

1. How can you be sure that the error is consistent for all input voltages? Isn't it possible that the error can be higher or lower for a different voltage even in the same range?

2. It only checks a single range. How can you be sure that other ranges are accurate? As far as I know the range switch is a voltage divider, it's possible to have a drifted or even defective range without being noticed.

Thus I think the single voltage reference test, at best, tells you the analog-to-digital converter is working, its internal reference is accurate, and the divider used in a single range is accurate, not more, for example, it doesn't say anything about linearity.

For a more complete test, I think one also needs something like a Kelvin-Varley divider and a high-voltage standard as well. Or just use a multimeter calibrator. But since those are very exotic and out of reach from most, they are not commonly used, thus a single reference voltage became the most popular test.

[TiN]

But since those are very exotic and out of reach from most, they are not commonly used, thus a single reference voltage became the most popular test.

Well, you just replied your question here . Many volt-nuts do this for hobby, and its easy to do single point check. In most DMMs you have gain errors and voltage reference drift as biggest contributor to uncertainty, so doing even single point check is giving some useful information. More serious voltnuts indeed use dividers and test linearity of meters to confirm the whole range. Also older calibrators such as Datron 4000, Fluke 5440, Datron 470x are reasonably accessible for more senior voltnuts. Those sources are well capable of checking any 6.5digit meter. After all, such checks and results are not pretending to be a full characterization of DMM under test or any commercial application purpose here (I hope).

I'd say getting first DIY LM399 or LTZ1000 is a first step for a volt-nut. DMMCheck stuff is for mortal engineers who don't want to worry about ppms too much 

[Kleinstein]

A single voltage testpoint is no way good enough to test a DMM. They just sell those cheap reference board as DMM-check, but this is just name not saying the parts are suitable for a DMM check. The single voltage test mainly looks for drift of the reference. Even with a single external reference one can also do the test with the negative voltage. This so called turn over test is a common simple test point for linearity. This is by no way a complete test, but a simple point to check without much instruments.

The usual calibration procedures require at least 2 test points (e.g. +-90% FS) per range.
Many of the moder integrating ADCs have quite good linearity and it is not very common for the linearity go bad over time. So one may get away without a full linearity test. Usually one still does a test for both signs and sometimes also points around half the full scale.
With other ADCs (e.g. the Fluke recirculating reminder type) there is a chance for linearity to get worse and there should thus be more points in the cal proceduce.

A point I am somewhat missing is a test for the actual input current. Quite some damages to a DMM can cause a higher than normal input current - a test with a low impedance test source may not show such a damage.

[Berni]

Well you can get a 3 point calibration out of it. The +5V then -5V and 0V by shorting the inputs together. So it does check both for reference accuracy and slope of all the circuitry.

Yes a proper way to do it is to get a calibrator but you probably know how much one of those costs.

Thing is that voltage references tend to be the most drifty component so a single test at 5V will catch that. The divider networks in DMMs tend to be stable once calibrated. People care the most about there 10 or 20 V ranges on DMMs since they are the most used and this works fine. You can also spot check a 100V range with 5V to make sure its reasonably close. You can also make a reasonable 100mV reference out of that using divider resistors.

To be honest i don't care exactly how accurate the 1000V range on my DMM is as long as its reasonably close. Just find some high voltage DC source and measure it with 2 multimeters simultaneously, if they agree its good enough for me. I test my DMMs using a Fluke 410B high voltage supply. Can be found pretty cheap used and will generate anything from 1V to 9999V with enough accuracy to calibrate most handheld DMMs. For higher spec DMMs i just use it as a voltage source and hook up multiple DMMs simultaneously to compare.

It's a different story for the voltnuts that chase the last fractions of a ppm on there fancy 8.5 digit DMMs

[niconiconi]

Well you can get a 3 point calibration out of it. The +5V then -5V and 0V by shorting the inputs together. So it does check both for reference accuracy and slope of all the circuitry.

Ah! Thanks for the trick. 

[mendip_discovery]

Personally I want one as a something I can send to other labs as a interlab comparisons.

So if you measure it and its 10.001V, you lend it to another person and they get 10.050 V, then they pass it to another and they get 10.001 V. So you can pick on the one person who has odd readings, if one of your mates has a 8.5digit meter and a recent calibration then you can trace your kit to national standards even if its limited.

I also like the idea of them as they are something you can measure yourself regularly and get better at measuring as you go down the rabbit hole. Also once you have them figured you can then use it to check you meter hasn't ran off the the old lady down the road and doing strange things with its probes, you know you go to test a dc voltage and the meter says 4 rather than 8, and you start to wonder if you are going mad but it's just that your meters batteries are dying.

[IanB]

But it only tests a single point and a single range, so I don't understand why is it considered an useful test.

It's just meant to be a sanity check. If you have a 5 V reference and your meter shows 4.94 V, you will know your meter is a bit off in its reading.

Some information is always better than no information. Such DMM check devices are simply intended to provide more information than you would have otherwise. The more elaborate ones can check DC volts, AC volts, resistance, current, and maybe some other things. You can at least have an idea that your meter is not broken.

[David Hess]

1. How can you be sure that the error is consistent for all input voltages? Isn't it possible that the error can be higher or lower for a different voltage even in the same range?

2. It only checks a single range. How can you be sure that other ranges are accurate? As far as I know the range switch is a voltage divider, it's possible to have a drifted or even defective range without being noticed.

Resistor network decade voltage dividers used in most designs are ratio matched so the other ranges will be accurate to the ratio matching specifications unless something is broken, which is pretty unlikely.

Some multimeters use a decade divider and then attenuate the reference by 10 for the most sensitive range.  These require two voltage tests to adjust two reference voltages.

For a more complete test, I think one also needs something like a Kelvin-Varley divider and a high-voltage standard as well. Or just use a multimeter calibrator. But since those are very exotic and out of reach from most, they are not commonly used, thus a single reference voltage became the most popular test.

A Kelvin-Varley divider would be ideal but is really only needed for diagnosing rare disorders or designing.  They are especially useful to testing linearity but there are other ways to do this also.

Another good but simple test is to short the inputs and then check different ranges.  If there is an offset and it increases at lower ranges, then the input buffer or protection circuits have high leakage.  The displayed count changes because the input buffer sees a different series resistance through the decade divider as the ranges are switched.

[TimFox]

"Useful" does not imply "sufficient".

[bob91343]

A voltage reference is not intended to be used for calibration.  As said above, it's not enough for that.

What it is used for is the master source in a voltmeter.  All ranges are set up with precision resistors and the reference is the standard of comparison.

Of course your readings are only as accurate as the resistors.  But if you use 0.01% resistors and the reference is also 0.01% you now can have a meter good to 0.02%.  Statistically it's better than that, as you can assume that one or the other is better than spec and so the combined result also is.

[TiN]

A voltage reference is not intended to be used for calibration.  As said above, it's not enough for that.

What it is used for is the master source in a voltmeter.  All ranges are set up with precision resistors and the reference is the standard of comparison.

Of course your readings are only as accurate as the resistors.  But if you use 0.01% resistors and the reference is also 0.01% you now can have a meter good to 0.02%.  Statistically it's better than that, as you can assume that one or the other is better than spec and so the combined result also is.

Essentially every sentence in this post is not correct. Voltage references are used for calibrations, that's what they are designed for. Sure, reference to calibrate handheld 3.5 digit DMM and reference to calibrate 8.5-digit benchtop meter are very different, but concept is very same. Reference whole job is to provide best stability possible, ideally regardless on what is it connected to. DMMs are much more complex devices with lot of compromises, so they don't have best references available because if these compromises.

Accuracy apply little to stability as well, it is the circuit, physical implementation and function that matters most. You can have only 5% resistors used with LM399 or LTZ1000 voltage reference, and still have 0.0002% accuracy in the end once you calibrated that reference. And shockers, LTZ1000 reference is also few % at best from "nominal", not even 0.1% 

Statistics is useless too when you have one particular reference, and one particular DMM. It may be okay to discuss guesswork or accuracy hopes in project section or test equipment, but this is metrology section here.     

[Shock]

The words are explained in names - check, calibration, adjustment, report, certificate.

A comparative check to a reference just adds some measurement confidence. In laboratories they may do a bunch of things prior to a measurement.


 

[Berni]

The 8.5 digit boatanchors and little handheld DMMs are two different ballparks all together.

A serious volt nut who is into his 8.5 digit zen will also have a proper set of calibration standards and perhaps a friend in the local cal lab. This is where you get into triple ovenised battery backed LTZ1000 references.

These simple little voltage references are mostly for the more average joe who just wants to know if there multimeter is reading right. Building a reference to get your multimeter to show 5.000V is not that hard and can be done on a budget too. Certainly not a 100% fool proof test of a multimeter but still tests most of the circuitry. Just because a DMM was cheap does not mean you can't enchant it with some measurement confidence by giving it a good testing every so often and seeing it stay bang on.

I have a few cheep chinese Aneing DMMs that have some decent measurement confidence for me since i tested a few of them a few times and they kept showing bang on in comparison to a 8.5 digit Keithey. Sure it might not be a Fluke but its nice to know the cheep little disposable multimeter rolling around the car toolbox is reading spot on. Most of the time the 3.5 digits is all you need.

[Kleinstein]

The simple reference box is intended mainly for checking handheld meters. It is only a spot test, but still better than nothing.
With many hand held meters a proper professional calibration can be more expensive than a new meter. So there is some sense in a simplified affordable test for those (hobby use) who don't need a proper calibration.
 

[David Hess]

The simple reference box is intended mainly for checking handheld meters. It is only a spot test, but still better than nothing.
With many hand held meters a proper professional calibration can be more expensive than a new meter. So there is some sense in a simplified affordable test for those (hobby use) who don't need a proper calibration.

I do not really agree.  Let's say I was designing and fabricating a meter.  I have a high impedance decade divider, high impedance buffer, analog-to-digital converter, and reference.  If my decade divider is ratio matched to 0.05% which is readily available and I accept that level of accuracy, then I only need one test voltage.  Multimeters using this type of design are intended to be calibrated in this way.

Now if I want to calibrate every range independently as many meters do now, then I need multiple test voltages or some way to turn a test reference into multiple accurate test voltages.  Some advanced multimeters can do this internally and perform a complete or almost complete calibration with a single external reference voltage.

[WizardOfBoz]

The OP is correct:
1) To be able to trust a meter, a multipoint calibration is necessary.
2) These require expensive (>$30K US) and exotic equipment to be done in an economic an reportable manner

Others have pointed out that
1) Single voltage references are a sanity check
2) Simple, multiple point references similar to DMMChecker are available (well, not now, they're out of stock) and fairly cheap
3) Most DMMs use a resistor network to divide voltages so if the 5vdc is precise and the 117VAC is about 117VAC chances are that the resistors are good (note that some voltage references like the AD587 use a resistor network themselves to provide 10, 7.5, 5, and 2.5 volts).

So checking a few points gives us an idea if the meter is within shouting distance of being accurate.  Gross errors are detected.  And if you make or buy a cheap voltage reference, you can do a sanity check every few months, or after you drop your Fluke DMM from 30 feet onto the concrete pad (which happened to me: it still worked!  But I didn't check calibration).    And last, it was noted that those in the population (most people on this forum) probably enjoy creating voltage references, so there's that.

You use a single (or two or three) point calibration to do a sanity check to tell you if you have to send your DMM in, pay $60-$280, and get the meter back with a nice note saying "Yeah, your meter was 0.1% off so we adjusted it and ists now 0.05% off.."