Decent benchtop multimeter for beginner?

[skander36]

...

For example, if I have 3 meters, and 2 agree but one doesn't, the 2 agreeing don't mean a lot if they're the ones out of cal. Whether he ever renews cal again is up to his needs.

...

If the two meters  that agreed and not in cal are 34465A and DMM6500 and the third is a Siglent 3065X which is in cal but disagree with the other two, you can bet that the Siglent is wrong ...    ... my case. 
The two meters were verified with pdvs2 mini in calibration period.

[nctnico]

...

For example, if I have 3 meters, and 2 agree but one doesn't, the 2 agreeing don't mean a lot if they're the ones out of cal. Whether he ever renews cal again is up to his needs.

...

If the two meters  that agreed and not in cal are 34465A and DMM6500 and the third is a Siglent 3065X which is in cal but disagree with the other two, you can bet that the Siglent is wrong ...    ... my case. 
The two meters were verified with pdvs2 mini in calibration period.

Agreed, calibration is nothing more than taking a snapshot (like a photo) from the situation at a single point in time. Equipment can and will break which is why you need to do checks using known sources before and after measuring a DUT if you need to be very sure the results are correct.

Another point I have not seen mentioned in this thread is that some handhelds can give false readings when the batteries run low. And this is not only a problem for the $5 meters but I've seen (ballpark) $200 BK 'Precission' handhelds go wrong as well.

[EEVblog]

...
For example, if I have 3 meters, and 2 agree but one doesn't, the 2 agreeing don't mean a lot if they're the ones out of cal. Whether he ever renews cal again is up to his needs.
...

If the two meters  that agreed and not in cal are 34465A and DMM6500 and the third is a Siglent 3065X which is in cal but disagree with the other two, you can bet that the Siglent is wrong ...    ... my case. 
The two meters were verified with pdvs2 mini in calibration period.

It's basic statistics. The odds of two random meters agreeing to the last digit adds value above and beyond any "calibration". Especially so if they have agreed for years and not drifed from each other, in which case, bingo, you've started to add some "tracability" and "measurement confidence" with those two meters.


For those interested:

[Martin72]

If the two meters  that agreed and not in cal are 34465A and DMM6500 and the third is a Siglent 3065X which is in cal but disagree with the other two, you can bet that the Siglent is wrong ...    ... my case. 

I don't think so.
If a calibration laboratory, which is incomparably “better” equipped than you, determines that a multimeter is within its tolerances, then the device is OK.
The point is not that it should display something that you expect.
Exaggerated example:
If 10V is fed in and you expect a “nice” 10.00000V to be displayed, but the device “only” displays 9.99999V, but that is within the tolerance, then that is OK.
You can stand on your head, but that's the way it is.
If other meters then show the hoped-for 10.00000V, then they are not “better” or more OK, then they just show 10.00000V, because these meters also have a range from...to...
Then you have to make sure that the 10.00000V is actually 10.00000V.
This also varies, although not as “strongly” as with the meters.
This fluctuates and the “true” value is not known anyway.
A snapshot.

[nctnico]

I'm pretty sure skander36 is not argueing about the 6th digit. But more like 2 meters saying the voltage is 10.00V and the third saying it is 9.90V. Then 9.90V is more likely to be wrong than right no matter whether that third meter was calibrated 5 minutes ago.

[skander36]

@Martin72, If two meters that are out of period of calibration show a similar value confirmed by a voltage reference and a third meter show a consistently different value even if is freshly calibrated, for me is wrong for a 6,5 meter. For a handheld this can be tolerable.

[Fungus]

If 10V is fed in and you expect a “nice” 10.00000V to be displayed, but the device “only” displays 9.99999V, but that is within the tolerance, then that is OK.

I hate that.

10.00001 is perfectly fine but 9.99999 is really annoying. 

[Martin72]

I can well imagine that.
But things usually work differently.
If I want a meter that definitely displays 10.00000V when 10.00000V is actually present (don't forget that), then I have to buy one that can do that.
Then you won't get far with 6.5 digits, regardless of the brand.
An SDM3065X with its 0.0015% tolerance(look at the tolerances of other 6.5 digit meters) can fluctuate between 9.99985 and 10.01500.
If it's within that range, it's not broken or bad, it's just “in there”.
If I don't want that, I have to buy a 7.5 digit or higher meter.
Incidentally, in my opinion, the cheap voltage references that are freely available are all unsuitable for estimating a 6.5 digit meter.
If you really want to know, you must at least have your meter calibrated to ISO.
It costs about 150€ here, I can't buy a suitable reference for that.

[tautech]

Well there's Calibration and there's Calibration and what I've seen and learnt in the last few days was a real eye opener. 
Clean room, all working with equipment wore antistatic clothes, hats and booties and so so so many new units were having automated calibrations performed at once, meters, scopes, AWG's PSU, analyzers and so on.

But consider this, should just a single measurement in each measurement range be performed or many within each range ?
Some instruments took several hours to have the full calibration process complete. 

So so many top of the line calibration instruments performing these tasks was a sight to behold and will be etched in my mind like a good cal for the rest of my days.

It need be done properly or you're 

[Fungus]

Incidentally, in my opinion, the cheap voltage references that are freely available are all unsuitable for estimating a 6.5 digit meter.

Obviously... you need temperature control for that many digits.

[Phil1977]

Maybe a blasphemy question, but what do you need 6-digit absolute accuracy for - except for calibrating other instruments?

All validation/qualification labs I know had basic requirements of 0.1% or in rare cases 0.05%. If your meter should be one order of magnitude better than your specification you get to 0.005% - these are 4 1/2 digits.

Of course you can use 7 or more digits for relative purposes. Seeing small changes also over a large timespan is a great feature, so the instrument should have sufficient stability. From a pragmatic point of view this stability can also be ensured by tracking an external standard instead of going through the exhaustive and expensive complete calibration process.

What are real life applications for 6 digit absolute accuracy and not only stability? I´m curious.

[uargo]

Well, you will see in precision applications (medical, industrial, military, satellite, etc.) that you need to capture 3.3v or 1.8v from a sensor with a 24-bit ADC. We are talking about precisions of up to 107 nanovolts (1.8v / 2^24 = 0.000000107v). Any noise greater than 0.0000001 will distort the measurement, and if you are designing that circuit, you need to measure with that precision to mitigate the noise as much as possible.

[Kleinstein]

It is indeed very rare that one needs the high resolution. One point would be checking the linearity of a slow DAC / ADC or maybe DC amplifier. This usually does not need an accurate absolute scale.
Another point can be checking resistors or voltage references if they are really stable. For this one may want to see changes of only a few ppm.
A point here can be that the resistance ranges may not as stable / low noise as the preferred DC volt range.

A more common point is looking at really small signals, like a thermocouple voltage. For this one may want 1 µV or maybe 0.1 µV resolution. With a 100/200 mV range this would need 5 or 6 digit resolution. A few meters offer smaller ranges like 20 mV or 2 mV, but not very many. In this respect the old but still comming up used HP3478 can be a good pick with a 30 mV range and 5.5 digits.

The number of digits is really not the only thing to look at. There can be other points that can be important, like small ranges, speed, readability, having 4 wire ohms, high impedance to a certain voltage, frequency range for AC, PC interface, calibration. It really depends on the expected use which features are important and where one can compromise. A beginner tends to not need the calibration paper-work.

[robert.rozee]

[...] If your meter should be one order of magnitude better than your specification you get to 0.005% - these are 4 1/2 digits [...]

depending on how you view things, the above (0.005% for a 4.5 digit meter) may be out by a factor of 10, or even 100...

consider my 3.5 digit meter: it can at best be used to measure an arbitrary input to a 0.5% resolution. the limiting factor is where i want to measure, for example, 2.01 volts. this is too much to measure using the 2v range, so i need to use the 20v range instead. this can yield readings of 2.00 (a little low) or 2.01 (spot on) or 2.02 (a little high), giving a 0.5% usable resolution. while i could have my 3.5 digit meter adjusted to significantly better than 0.5% accuracy, this would be relatively pointless as the meter is unable to display results to anything better than 1 part in 200 (0.5%).

my 4.5 digit meter can, by extension, resolve an arbitrary input to within 0.05%, and hence is it pointless to have it adjusted to anything significantly better. my 5.5 digit meter can do 0.005%, while my 6.5 digit meter can do 0.0005%


now lets work the numbers backwards. lets say i want to measure something that is between 1.9 and 2.1 volts using my 3.5 digit meter.

if i see the number "2.01" displayed, what can i confidently say about the measurement that i am observing? it is fair to say that the actual voltage being measured could be anywhere between (just over) 2.005 and (just under) 2.015 volts - assuming the meter is spot on (0.0% accuracy error). but then we need to factor in the 0.5% level of accuracy we already arrived at above - this gives us an expanded range of 1.994975 to 2.025075 volts for what we are trying to measure, and possible displays of "1.99", "2.00", "2.01", "2.02", or "2.03".

this rather wide range gives me little confidence in that ".01" i'm seeing! i feel that i need at least to see one more digit to be confident - or that i need to downgrade the accuracy i'm willing to ascribe to my measurements by a factor of 2. i would be far more comfortable if i was looking at a display showing "1.990", as then i can happily dismiss the last digit and say "this is within 0.5% of 1.99 volts" with an air of certainty!

hence, when making arbitrary measurements, we really would like a 'guard digit' that we can throw out. while this could be a half-digit on the right (where the rightmost digit on our meter can only be either "0" or "5"), the marketing department prefers that every displayed digit beyond the first can cover the full range of "0".."9". instead, when we see the display of "2.01" we need to take that as being something between "1.99" and "2.03" - even if the reference and resistor divider of our meter is good for 0.5% accuracy, the reading we see displayed only has a +/-1% useful accuracy (what we might decide to call 'confidence').

by extension:
our 4.5 digit meter has a maximum useful accuracy of 0.1%,
our 5.5 digit meter has a maximum useful accuracy of 0.01%,
our 6.5 digit meter has a maximum useful accuracy of 0.001%.

and in cases of there being a little noise within our measurement system (as happens with 5.5 and more digit meters), we may wish to instead take a full decimal digit as the 'guard'. then we get:
our 5.5 digit meter having a maximum useful accuracy of 0.05%,
our 6.5 digit meter having a maximum useful accuracy of 0.005%.


cheers,
rob   :-)

[Fungus]

Well, you will see in precision applications (medical, industrial, military, satellite, etc.) that you need to capture 3.3v or 1.8v from a sensor with a 24-bit ADC. We are talking about precisions of up to 107 nanovolts (1.8v / 2^24 = 0.000000107v). Any noise greater than 0.0000001 will distort the measurement, and if you are designing that circuit, you need to measure with that precision to mitigate the noise as much as possible.

Nobody's arguing against having precision. The question was about accuracy:

Maybe a blasphemy question, but what do you need 6-digit absolute accuracy for - except for calibrating other instruments?

[Fungus]

consider my 3.5 digit meter: it can at best be used to measure an arbitrary input to a 0.5% resolution. the limiting factor is where i want to measure, for example, 2.01 volts. this is too much to measure using the 2v range, so i need to use the 20v range instead. this can yield readings of 2.00 (a little low) or 2.01 (spot on) or 2.02 (a little high), giving a 0.5% usable resolution. while i could have my 3.5 digit meter adjusted to significantly better than 0.5% accuracy, this would be relatively pointless as the meter is unable to display results to anything better than 1 part in 200 (0.5%).

Benford's law: https://en.wikipedia.org/wiki/Benford%27s_law

[Martin72]

Well there's Calibration and there's Calibration and what I've seen and learnt in the last few days was a real eye opener. 
Clean room, all working with equipment wore antistatic clothes, hats and booties and so so so many new units were having automated calibrations performed at once, meters, scopes, AWG's PSU, analyzers and so on.

But consider this, should just a single measurement in each measurement range be performed or many within each range ?
Some instruments took several hours to have the full calibration process complete. 

So so many top of the line calibration instruments performing these tasks was a sight to behold and will be etched in my mind like a good cal for the rest of my days.

It need be done properly or you're 

What you saw was a mass calibration, there is no other way with the quantities.
The good thing is that Siglent does not “throw away” the data obtained, even if the (cost-effective) devices are only accompanied by a one-page declaration that they have been successfully calibrated.
This is how I got the complete protocol for my SDM3065X in a roundabout way, which was important for me, because this is the only way to see trends and where the journey is going with each further calibration.
Incidentally, the protocol for an SSG3000X is a whopping 14 pages long...
Privately, these efforts are more than sufficient, for us they are null and void.
We have different standards and every piece of measuring equipment we purchase is calibrated by an accredited measuring laboratory before it is used.
Even if they basically do nothing other than what I have already described.
Is the device within its own tolerances yes/no.
But this is documented and completely traceable.
To be able to calibrate according to ISO, for example, the references used must be calibrated according to DakkS, the higher standard.
This in turn must also be completely traceable, the laboratory must be listed, otherwise their protocols can be contested.
And that is why a calibration by the manufacturer is null and void.

Maybe a blasphemy question, but what do you need 6-digit absolute accuracy for - except for calibrating other instruments?

Well....
I would like to see the person who has a private project going on where his resolution is really necessary.
There are probably not many people who own a 6 digit or more instrument.
At work, we sometimes have projects that require voltage adjustment to 1mV, which is very precise.
Anyone can work out what resolution is sufficient for us.
As a MIL, AERO and ISO certified company, we do not have a 6.5 digit instrument in our inventory.
If you buy something like this privately, 90% of the time it will go to the dog.
Or according to the dog question.
Why does he lick his balls?
Because he can.

[tautech]

Well there's Calibration and there's Calibration and what I've seen and learnt in the last few days was a real eye opener. 
Clean room, all working with equipment wore antistatic clothes, hats and booties and so so so many new units were having automated calibrations performed at once, meters, scopes, AWG's PSU, analyzers and so on.

But consider this, should just a single measurement in each measurement range be performed or many within each range ?
Some instruments took several hours to have the full calibration process complete. 

So so many top of the line calibration instruments performing these tasks was a sight to behold and will be etched in my mind like a good cal for the rest of my days.

It need be done properly or you're 

What you saw was a mass calibration, there is no other way with the quantities.
The good thing is that Siglent does not “throw away” the data obtained, even if the (cost-effective) devices are only accompanied by a one-page declaration that they have been successfully calibrated.
This is how I got the complete protocol for my SDM3065X in a roundabout way, which was important for me, because this is the only way to see trends and where the journey is going with each further calibration.
Incidentally, the protocol for an SSG3000X is a whopping 14 pages long...
Privately, these efforts are more than sufficient, for us they are null and void.
We have different standards and every piece of measuring equipment we purchase is calibrated by an accredited measuring laboratory before it is used.
Even if they basically do nothing other than what I have already described.
Is the device within its own tolerances yes/no.
But this is documented and completely traceable.
To be able to calibrate according to ISO, for example, the references used must be calibrated according to DakkS, the higher standard.
This in turn must also be completely traceable, the laboratory must be listed, otherwise their protocols can be contested.
And that is why a calibration by the manufacturer is null and void.

Check my current sig. 

I was not about to take pics of the several multiple calibration stations but I can add I've seen the high end Fluke universal calibrators elsewhere but those I saw at HQ were next level and many had leveling heads where they were connected to the DUC.
This alone indicated calibration was done to a very high standard.

[Martin72]

I was not about to take pics of the several multiple calibration stations but I can add I've seen the high end Fluke universal calibrators elsewhere but those I saw at HQ were next level and many had leveling heads where they were connected to the DUC.
This alone indicated calibration was done to a very high standard.

That is clear, because if you also offer high-precision instruments like Siglent, the calibration references have to be higher, otherwise it won't work.
But no quality auditor will say, OK, a certificate directly from the manufacturer is enough for me.
It has to be confirmed independently.
Professionals like Siglent know that.
But that's the game in the industry, siglent says our devices meet the self-imposed standards, we have tested EVERYTHING accordingly.
And an independent measurement laboratory then says “right”.

[Fungus]

I'm still unclear how "calibration" provides a guarantee unless it's done before and after each critical measurement.

Sending these devices through the abusive transport system for calibration or saying "it was calibrated six months ago" doesn't seem enough.

I'm more of a "I have two or three decent meters and they all agree to within a digit" sort of a guy.

This is something that calibration doesn't provide, per se.

[KungFuJosh]

I'm more of a "I have two or three decent meters and they all agree to within a digit" sort of a guy.

This is something that calibration doesn't provide, per se.

The best way to deal with that logic is to have a calibrated reference.

[J-R]

At a basic level, a hobbyist can use whatever method they want to obtain a desired result, but this is not realistic in a company/manufacturing context.  You can't just give an employee three DMMs and tell them to figure things out.  That leads to extra labor cost, mistakes, poor knowledge transfer, etc.  So you go with set procedures and automation along with calibration and multiple levels of QC to provide a statistically desirable result.

Fungus, how do you think you were able to obtain three completely DMMs that agree with each other???  Calibration...

[EEVblog]

I'm more of a "I have two or three decent meters and they all agree to within a digit" sort of a guy.
This is something that calibration doesn't provide, per se.

The best way to deal with that logic is to have a calibrated reference.

Multimeters are "calibrated references" in their own right.

[EEVblog]

I'm still unclear how "calibration" provides a guarantee unless it's done before and after each critical measurement.

Technically, it doesn't.
That's what "tracability" and "measurement confidence" are all about.

[KungFuJosh]

Multimeters are "calibrated references" in their own right.

If they're calibrated. 😉