Two multimeters, two different voltages. Now what?

[ComfyChair]

I am measuring the output voltage of a power supply set to 1V with two multimeters:

• Mastech MS8264: 0.998 V
• Aneng AN870: 1.0066 V

Which one is lying? What should I do now?

For reference, here are the (nominal) specs of the two multimeters:
MS8264(https://mastech-group.com/as/en/MS8264)
DC Voltage - 200mV/2V/20V/200V - 0.1mV/1mV/10mV/100mV - ±(0.5%+1)

AN870(https://www.anengmultimeterstore.com/product-aneng-an870-multimeter)
DC Voltage - 1.9999V - 1000.0V - ±(0.05%+3)

[Kleinstein]

The difference is less than 0.3%. That is well within the specs of the Mastech meter and chances are also that this meter is older.
The meters are not lying, they just don't know better and don't even claim to know that accurate.

Chances are the AN870 is the more accurate one, as it has the better specs.

One could do another test with reversed polarities. This could reveal at least the even oder error from the meters (e.g. offest and U² part), though the modern DMMs are usually relatively good at this.
A voltage closer to 2 V could give a little more resolution, but should give a similar ratio.

[forrestc]

Which one is lying? What should I do now?

They are so close that this is basically the same reading:

MS8264 is +- 0.5%+1  = which means that .998V is approximately anything from: .997*0.995=.992V to .999*1.005 = 1.004V. 

AN870 is +- 0.05%+3 = which means that 1.0066 can be anything from 1.0063*.9995 = 1.0058 to 1.0069*1.0005 = 1.0074V

(Note the numbers above are approximate since technically you should probably do the math based on percentage from the actual voltage, not the reading - but close enough for this discussion).

Since the range .992-1.004V is pretty close to the 1.0058V to 1.0074V range I might be inclined to say that both are as accurate as one can expect, especially since those specs are for a certain time after  they leave the factory, and tend to drift as time goes on.   

Plus, a few thousandths of a volt can be created through a whole bunch of other effects.  Like as in you need special test leads to get to that accuracy.  (See Seebeck or thermoelectric effect among other things).  So there is every possibility that those readings are accurate at the input of the DMM.

Are you doing things where a thousandth of a volt matters?

[TheDefpom]

The trick here is to either:
Only own ONE multimeter.
Own about 10 multimeters and take the average.

[forrestc]

The trick here is to either:
Only own ONE multimeter.
Own about 10 multimeters and take the average.

This is the EE equivalent of the old saying "A man with a watch knows what time it is. A man with two watches is never sure."

[Doctorandus_P]

This is how volt nuttery starts 

Maybe it's the meter on your power supply that is off by the biggest margin.

After a while you have a stack of 8 1/2 digit DMM's and a bench full of LTZ1000 references and an automated measurement system that compares them to each other all year long.

I am just starting on this path. I thought about buying one of those voltage reference and resistance / capacitance test boards, Then thought a bit more about the cost of a decent one, and bought a BM869s instead. It still compares quite well against my other meters. A 30+ year old Dynatek 112, and a Wavetek Meterman 35XP.

I also bought 2 Aneng 8009 meters when they were all they hype, and I don't like them at all. They beep at you annoingly all the time and they are not reliable. I once bought a strip of 1k 1% resistors and wanted to test their tolerance, so I used my new AN8009, and I almost gave the aliexpress shop a negative review because the resistors were out of tolerance (by 2% or so), and this is a usual scam. Just to be sure, I verified it again with another meter, and the resistors were in spec. Then I fiddled a bit with the dial on the AN8009, and then it measured OK again. That was the last time I used those Aneng meters for anything. I had to wipe off a thick layer of dust to veryfy the type number of the aneng meters just now.
(The other 5 or so meters I have are not worth mentioning).

If you are serious / curious about calibration, then get some "decent" parts (10V voltage reference, some resistors, capacitors etc) and mount them on a board so they are easy to measure. And then once a year measure all the devices on that board with all your DMM's and log all the data. When one of your DMM's start deviating, there is something "wrong" with it. If all your DMM's suddenly start deviating, it is your reference board that is somehow wrong.

If you have higher quality meters, you can also have them calibrated every now and then, but this is to much to bother with and too costly for most DIY'ers.

[mendip_discovery]

You have a meter that isn't performing to specification.



Normally I would expect the real result to sit where the error bars overlap but this time it isn't as I suspect the Aneng has overestimated its accuracy.

Did you perform the measurement at the same time? Was the power supply drifting up and down?

[J-R]

Plus, a few thousandths of a volt can be created through a whole bunch of other effects.  Like as in you need special test leads to get to that accuracy.  (See Seebeck or thermoelectric effect among other things).  So there is every possibility that those readings are accurate at the input of the DMM.

I'm confident the Seebeck effect and test leads won't be an issue in this case.  You'll need a lot more digits, a mV range and a higher temperature differential than holding the test leads would provide.

I ran some what-ifs on this effect when I picked up a 7.5 digit DMM and had a hard time producing an impact with standard banana jack leads on the 1V range.  My primary thought on the reason is that there is a lot of contact area between the plug and jack, so the temperature differential remains small, even when holding only one plug.  The easiest way to see it is with DMM test probes, where you can heat up the end with your fingers quite easily, and the temperature differential is better maintained due to the small contact area between the probe tip and the DMM input jack.  In that case, I could make the digits wander at will.

Lastly, it is frequently the case that you have an even number of connections, so the effect is cancelled out, such as with a hot, fanless bench DMM and ambient test leads.

[J-R]

The trick here is to either:
Only own ONE multimeter.
Own about 10 multimeters and take the average.

Instead of a blind average, I think a better method would be to remove the obvious outliers, then contemplate the average of the remaining readings.  If the spread is large, the average may not be particularly reliable.  But if there is a tight grouping of most of the DMMs, then an average of those could be a reasonable result.

For a hobbyist headed into metrology, I think three DMMs and one externally calibrated reference is probably the minimum setup, assuming the DMMs are not calibrated but agree otherwise.  Or a minimum of four regularly calibrated DMMs.  Without that, you'll be chasing your tail when there is a disagreement.

[Kim Christensen]

It's not an issue in this case, (Connecting directly to power supply) but if you're measuring higher impedance circuits, then the input impedance of the meter becomes an issue also. Some of these cheap meters can have an input impedance as low as 1M which could become a problem for you.
Your Mastech is spec'd at 10M. I can't find that info on the AN870 so who knows?

[J-R]

The input impedance of the AN870 is 10.11M Ohms, per the review here: https://lygte-info.dk/review/DMMAnengAN870%20UK.html

The OP provided a single photo showing both DMMs at the same time, so it's likely they are both connected to the same supply simultaneously.  So that is good.  But the quality of the supply might have a minor impact to the readings, as there could be a significant AC/ripple component that each DMM filters differently while in DC mode.

So I think a battery is a more reliable source to use for this test.  I'd suggest a 1.5V nominal Alkaline AA/C/D cell since the Mastech is 2,000 counts and the Aneng is 20,000 counts.  It's absolutely critical in this case to connect both DMMs to the battery simultaneously.  Give the DMMs a little time to settle, then post another photo with the result.

[bastl_r]

Hi
I think you can trust the 870.
Ihave one here wich i bought by interest.
Ists values are in all voltageranges between -0.02% (2V)and -0.04% (200V) the mesured values of my Prema5017.
Only th mV range shows -0.07% of the Prema.

[J-R]

The OP's Aneng may be correct, but "trust" is taking things too far, especially for the Metrology Forum.  He would need to check his specific DMM as well.  So either get another DMM to send things in a certain direction, or live with the discrepancy.

[bastl_r]

OK, let's say it in other words:
I would trust the Aneng more than the Mastech.

My Aneng870, compared to my Prema which I had freshly calibrated at Metrologyday, only showed an insignificant difference of 0.02%(2V; 20V) to 0.07%(in the mV range).
For a 4.5-digit meter, the comparison should be sufficient for an assessment.
For metrological requirements, the measuring device itself would have to be calibrated and documented. But is this necessary for the thread creator?

[ComfyChair]

Thanks

To summarize:
0. I can consider the measurements equal
1. Get a third (better) multimeter
2. Get / build a reference to compare the multimeters against.

I have a few ideas for the third point. I'll open another post for more info regarding the voltage reference design.

[tggzzz]

Thanks

To summarize:
0. I can consider the measurements equal
1. Get a third (better) multimeter
2. Get / build a reference to compare the multimeters against.

I have a few ideas for the third point. I'll open another post for more info regarding the voltage reference design.

Then get a better voltmeter to calibrate the reference.
If it is a new reference, leave it powered up for at least 1000 hours before using it in anger, then check how much it drifts over time.
Welcome to the rathole.

[tooki]

I recently discovered a 25-year-old Fluke 87 at work that we had somehow inherited and didn’t know we had. Hadn’t been used in years. I gave it a new battery, cleaned the zebra strips and their contacts to fix the weak display, and gave it a thorough cleaning. Then I tested it against the less than 1 year old DMM6500 in my lab, and the old 87 is still in spec, just one or two counts out from the DMM6500.

Mastechs don’t keep that good a calibration within a decade, never mind 25 years. Aneng… we shall see?

[Geoff-AU]

This is the EE equivalent of the old saying "A man with a watch knows what time it is. A man with two watches is never sure."

The version I've heard is:

Someone with one clock knows exactly what time it is.  Someone with two clocks has no idea what time it is.  Someone with three clocks has a slightly better idea than the person with two clocks.

[moerm]

Thanks

To summarize:
0. I can consider the measurements equal
1. Get a third (better) multimeter
2. Get / build a reference to compare the multimeters against.

I have a few ideas for the third point. I'll open another post for more info regarding the voltage reference design.

Then get a better voltmeter to calibrate the reference.
If it is a new reference, leave it powered up for at least 1000 hours before using it in anger, then check how much it drifts over time.
Welcome to the rathole.

That - or: think about the precision&accuracy one really needs.

That's why I tend to smirk when I see someone measuring a power supply, (ab)using it as a kind of reference. Never mind that the PS's specification almost certainly is far worse than even that of a decent 3.5 digit DMM ...

@ComfyChair

I have simple advice for you: Do not look at, let alone be irritated by, the difference of the two measurements but rather focus on the 2 digits right after the decimal point ... and you'll see 0.99xx vs. 1.00xx.

The real message of what you see is that (a) you seem to have a decent power supply, and (b) both your DMMs are OK. Do not worry about 8 mV difference when measuring a (normal) PS!

Or, alternatively pick the route @tggzzz hinted at after thinking long, deep, and hard ...

(Edit: small corrections)

[tggzzz]

Thanks

To summarize:
0. I can consider the measurements equal
1. Get a third (better) multimeter
2. Get / build a reference to compare the multimeters against.

I have a few ideas for the third point. I'll open another post for more info regarding the voltage reference design.

Then get a better voltmeter to calibrate the reference.
If it is a new reference, leave it powered up for at least 1000 hours before using it in anger, then check how much it drifts over time.
Welcome to the rathole.

That - or: think about the precision&accuracy one really needs.

Booooring! But usefully pragmatic

That's why I tend to smirk when I see someone measuring a power supply, (ab)using it as a kind of reference. Never mind that the PS's specification almost certainly is far worse than even that of a decent 3.5 digit DMM ...

My ancient Power Designs 2020 supplies 2A with 1mV resolution, and tweakable by a pot to ~20µV with maybe 40µV short term drift. It helps that its reference is an LM399

I have simple advice for you: Do not look at, let alone be irritated by, the difference of the two measurements but rather focus on the 2 digits right after the decimal point ... and you'll see 0.99xx vs. 1.00xx.

The real message of what you see is that (a) you seem to have a decent power supply, and (b) both your DMMs are OK. Do not worry about 8 mV difference when measuring a (normal) PS!

There you go again, i.e....

Booooring! But usefully pragmatic

[ComfyChair]

Oh well it is a power supply I'm designing and wanted to check how good the calibration was.

I immediately realized that, even considering the error bands (see also @mendip_discovery's comment) the two multimeters where off. And BTW both are attached via banana clips, in parallel to the PS.

I guess I can consider the calibration OK then

[Wallace Gasiewicz]

Comfy:   

You Need:
At least a 6.5 digit meter, maybe two   
A Volt Standard     
Very good wire connectors     
A climate controlled room where the temp is always within 1 degree C.  Also Humidity control     
You need temp and humidity monitors to verify your climate control.     
A budget for having the equipment calibrated once a year, count on about $100 bucks for each piece of equipment     
Eventually you will need Resistance Standards and have them Calibrated alsoYou need to age the equipment, turned on in the climate controlled room         
You need another piece of equipment to record and preserve the measured values compared to temp and humidity  over time       
The most expensive part will be when you join Test Equipment Anonymous for treatment, they will talk you into buying more equipment.     
Can you take out another mortgage on your home?     
You can see where this leads.........     HA!!!     

Please note that I will deny any of the above if anyone infers that the statements apply to me personally   

Or just accept that your meters are within specification.....the numbers actually are pretty good.

[moerm]

My ancient Power Designs 2020 supplies 2A with 1mV resolution, and tweakable by a pot to ~20µV with maybe 40µV short term drift. It helps that its reference is an LM399

Kindly note that I put the adjective 'normal' next to 'power supply'. And then note that you used the adjective 'ancient' next to yours ...

I think we can agree that an ancient PS (read: quite likely a well designed and built one) with 399 as ref. ... and ... what most hobbyists with Mastech and Aneng guesstimeters have for a PS are, uhm, not really comparable ... *g
(congrats though. Very nice PS you got there)

There you go again, i.e....

Booooring! But usefully pragmatic

OK, you got me. Full confession: I did not get a high-end hand-held DMM but thought (and still think *blush) that a Brymen 235 (blue, of course!) is good enough for > 95% of all tasks I address with a guesstimator, you know, things like "are all the rails OK?".

Oh well it is a power supply I'm designing and wanted to check how good the calibration was.

Got it. But still, then either your PS design specs are low enough to be sensibly checked by a, uhm, not exactly high-end DMM like the ones you mentioned - or - you tried, to put it like this, to measure micrometers with a cheap supermarket plastic-calipher.

Being at it, not that I'm immune to numbers porn, but do you really need a 1 mV precise PS? I mean, I don't know about yours but my Op-Amps don't care a whole lot whether they get a few millivolts more or less into their supply pins. If anything, they do (and sometimes you should) care about stability.

Maybe a Fluke 8842 in good condition is something you may want to look for, in case you want a decent starting point for your precision quest/adventure

[tggzzz]

My ancient Power Designs 2020 supplies 2A with 1mV resolution, and tweakable by a pot to ~20µV with maybe 40µV short term drift. It helps that its reference is an LM399

Kindly note that I put the adjective 'normal' next to 'power supply'. And then note that you used the adjective 'ancient' next to yours ...

I think we can agree that an ancient PS (read: quite likely a well designed and built one) with 399 as ref. ... and ... what most hobbyists with Mastech and Aneng guesstimeters have for a PS are, uhm, not really comparable ... *g
(congrats though. Very nice PS you got there)

Yes, it is nice The using the four voltage setting knobs has a pleasing tactile feel, almost pr0nographic 

I can say the same about my ESI DekaPot and DekaStat.

There you go again, i.e....

Booooring! But usefully pragmatic

OK, you got me. Full confession: I did not get a high-end hand-held DMM but thought (and still think *blush) that a Brymen 235 (blue, of course!) is good enough for > 95% of all tasks I address with a guesstimator, you know, things like "are all the rails OK?".

Well..... yes.

At school most of our meters were 2% accurate, but some physics experiments required a <1% accuracy. We achieved 0.1% with a NiFe cell, 1m of resistance wire, a 1m rule, any old crap meter, and a Weston standard cell.

And I designed and implemented my first computer (6800 128bytes RAM) with LEDs switches and a 2% multimeter.

My .sig hints at my attitude

Oh well it is a power supply I'm designing and wanted to check how good the calibration was.

Got it. But still, then either your PS design specs are low enough to be sensibly checked by a, uhm, not exactly high-end DMM like the ones you mentioned - or - you tried, to put it like this, to measure micrometers with a cheap supermarket plastic-calipher.

Being at it, not that I'm immune to numbers porn, but do you really need a 1 mV precise PS? I mean, I don't know about yours but my Op-Amps don't care a whole lot whether they get a few millivolts more or less into their supply pins. If anything, they do (and sometimes you should) care about stability.

Need? Go and was your mouth out with soap!

Yes, stability is usually more important than precision.

Realistically I rarely use the 2020 as a PSU. I use it more often as an easy way to set a voltage, change it in steps, and tweak under 1mV. There are many other things I could use for that purpose.

Maybe a Fluke 8842 in good condition is something you may want to look for, in case you want a decent starting point for your precision quest/adventure

Not me; I don't ned that. I have an Agilent 34410A, a Keithley 2015 THD, several Panaplex/neon meters, and a Solartron 7081.

The latter is very nice, but after not being used for a while, is reading 50ppm higher than before. Most irritating; pondering how to proceed.

[ivo]

I had bought an AN870, annoyingly it reads 5.005 on my 5V reference(s). Such a reading relies on the +digits to be within spec, it's outside of the % tolerance.