DMM Resolution/Accuracy, how many counts are enough?

[Trader]

There are DMMs from 2000 counts to 8 1/2 digits (millions of counts).

In what situations, experiments, projects, etc it is Really Necessary a very high Resolution/Accuracy, like 5,6,7,8 1/2 digits?

Thank you.

[bob91343]

Measuring voltage references, such as for other instruments or for A/D converters.  Sorting zener diodes.  Measuring output impedance and drift of power supplies.  Thermocouple calibration.  Voltmeter calibration.

For me, it's emotional.  I like to feel comfortable that my measurements are correct.

[daqq]

I've used high resolution current measurements for debugging various things where the current consumption could jump from a few uA to 1A. Very useful if you don't need to switch ranges. In general, you can avoid switching ranges all together when you have enough digits.
Obviously there's metrology, to calibrate/adjust/check stuff, whether they are lesser meters or your designs. And voltnuttery.
Measurement - there are a few areas where the useful data is hidden in the last digits (say, large offset voltage) or when you want to detect even the smallest change in the value, ideally without switching ranges.

[wizard69]

this is a perfect place to use the phrase "it depends".   Some might see that as a cop out but it is reality.   Further for decades engineers and technicians go by with analog meters so that highlights that digits are not even required.

There are DMMs from 2000 counts to 8 1/2 digits (millions of counts).

In what situations, experiments, projects, etc it is Really Necessary a very high Resolution/Accuracy, like 5,6,7,8 1/2 digits?

Thank you.

Given that it depends; lots of digits come in handy if you are doing calibration work.   However in industry, at least the one I work in, most of the calibration work seldom requires all the digits.   In fact one of the supplier specifically recommended an HP meter for us years ago because it did much better with true RMS measurements.   Often the better performance isn't in the digits if that makes sense.

To be honest there does seem to be an obsession with more digits when in many cases they simply are not needed.   For years I got by with a very simple 3.5 digit meter, for most field work that was good enough.  Now it is reasonable to get a 4.5 digit handheld as the cost isn't a big delta. 

Perhaps the bigger reason to buy meters with higher resolutions is to keep them in a metrology program.  When your SOP's require calibrations you grab a meter that is tracked calibration wise.   Generally we don't even bother to calibrate handhelds any more.

[tggzzz]

In what situations, experiments, projects, etc it is Really Necessary a very high Resolution/Accuracy, like 5,6,7,8 1/2 digits?

You have to understand what is important and unimportant in your specific measurements; only thereafter can you determine what is necessary.

For example, I used to work for a company that made a very expensive attenuation tester. The accuracy was 0.1dB, but the resolution and stability was 0.001dB. Guess which the customer didn't care about

[Siwastaja]

High accuracy (and resolution) is only needed when you are designing and building high-accuracy measurement devices; to verify the accuracy of your simulations and calculations with real-life components, and possibly calibrate them. Obviously, you'll know when you need it!

I'm mostly happy with 2000 count meters. Every once in a while I have a higher accuracy measurement device under design, then I borrow or go somewhere with an expensive meter available to do the verification/calibration step.

[mvno_subscriber]

For example, I used to work for a company that made a very expensive attenuation tester. The accuracy was 0.1dB, but the resolution and stability was 0.001dB. Guess which the customer didn't care about

I think this is a very good point - depending on the price range of the equipment, the accuracy can be a lot less than the number of digits. No point in having 5 decimals in a +/- 1% device 

[tggzzz]

For example, I used to work for a company that made a very expensive attenuation tester. The accuracy was 0.1dB, but the resolution and stability was 0.001dB. Guess which the customer didn't care about

I think this is a very good point - depending on the price range of the equipment, the accuracy can be a lot less than the number of digits. No point in having 5 decimals in a +/- 1% device 

Actually that is precisely the point. Sometimes there is significant value in having 5 decimals in a 1% device.

In the example I gave, all that mattered was the small change in attenuation as the cable drum's temperature was cycled over the course of a week. Yes, it was necessarily a large drum of cable, e.g. 61.8ohms coax cable.

[BravoV]

There are DMMs from 2000 counts to 8 1/2 digits (millions of counts).

In what situations, experiments, projects, etc it is Really Necessary a very high Resolution/Accuracy, like 5,6,7,8 1/2 digits?

Thank you.

It depends, imagine this various rifles capabilities ...



And with above choices ... also it depends different size of targets, want to instant kill by putting the bullet at the brain ? Which one is good enough ? Or just to hurt but no guarantee kill, and etc.



Hope this helps.

[Kleinstein]

There are a few cases when high resolution is really needed or really helping, but not that many. Quite often it is about getting enough resolution to measure also small signals.  A 3.5 digit meter would need a 0.2 mV range to get the same sensitivity as the usual 6.5 digit meter. One usually only finds something like a 200 mV or seldom 20 mV as the lowest range.

Especially at the lower end (e.g. 3.5 to 5 digits) it may today even be cheaper to add another digit resolution than another lower range for voltage and especially current. The number of digits alone no longer a way to judge accuracy / quality of a meter. The gets obvious with the cheap 5.5 digit current displays from China - they are often amazingly good for the price, but still no comparison to a real 5.5 digit meter.

[Trader]

I think an 8 1/2 digits DMM will be useful in a chemistry or physics lab to determine some electrical constant values (current, voltage, resistance, etc) in the experiments.

But for Electronics or Microelectronics, I think 6 1/2 or even 5 1/2 is much more than enough.

Seems that nobody is caring so much for handheld DMMs with more than 60000 counts.  Gossen decreased the counts of the new DMMs comparing with the previous generation and all other hi-end brands are stuck in the 50k or 60k counts.

https://www.eevblog.com/forum/testgear/multimeter-spreadsheet/

[Kleinstein]

Even is physics lab one very seldom would need 5 digits. It is more often that one wants sensitivity down to 0.1 µV or 1 pA. Usually such high sensitivity also comes together with high resolution.
Another point that sets apart most of the high resolution meters (5 or more digit resolution) from the simpler ones is usually a high input impedance (e.g. > 1 or 10 G) for the lower voltage ranges.
Another important feature for the bench meters is the PC interface - nobody likes to write down 6 digit numbers by hand. They tend to changes faster than one can write them down or even fully read them if noisy.

Even if for most tasks a lower resolution meter is good enough, one has to get a meter for the most demanding task one occasionally needs. For the "normal" electronic lab such demanding tasks are thinks like matching resistors or observing things like effect of loading, temperature or similar on a stabilized voltage / current. This can be very demanding, if done the brute force way. However there are often alternative ways to get away with less resolution. Still a high resolution meter can make it easier.

With real world voltages the noise / drift of the voltage source can also often be limiting. Even a lower noise meter can hardly (there can still be some difference in the effective bandwidth) improve on the noise / drift of the actual source.
It really helps if one is sure the noise or drift one sees is real and not an artifact of the meter. Quite some time can be lost hunting after unexpected artifact of the instruments. So some overkill can help.