DMM # of digits versus ADC bits

Hi guys,
I was wondering how to compare DMM # of digits versus ADC bits. Let me explain. Suppose I have a 5.5 digit DMM. On the 10V range the last digit would then indicate 00.001V is 1mV. If I have a 16bit ADC that is almost 100dB dynamic range, lets take 100dB for the discussion. That's a factor 10^5, so if I input 10V to the ADC the LSB would be 0.1mV.

Is that reasoning correct?

Background: I have a 16bit USB scope and the software can also display the signal in a voltmeter format. That voltmeter shows 6 digits, with the last one fluctuating all the time. I can set it to 5 digits and then it is steady. Just trying to get my head around the numbers and accuracy.

Jan
 

BSST

Member
2018-07-07 2:21 am
The 1/2 digit means that the most significant digit will be either 0 or 1. So the extremes of the DMM readings will range from 000,000 to 19,9999; therefore there are 20,000 possible codes. This would correspond to a little more than 14 binary bits. log(20000)/log(2) = 14.29 bits.

I hope I've done this right. ;)
 

1audio

Member
Paid Member
2004-03-24 5:16 am
SF Bay Area
Jan- you are opening Pandora's box as you know. First the basic reasoning is correct. The bits line up. However for a meter (and for an ADC for audio) its not that useful to just have bits. How many are meaningful? And how do you define that. For audio and other ADC applications there are various measures of accuracy and the issue of noise pops up so you need really clear definitions.

If you look at the spec sheet for a 5 1/2 digit DVM its very carefully drawn to show what the manufacturer defined as accurate. That us usually a DC spec with a number of averages under specific conditions. The averaging is a sort of noise filtering which is necessary to get a useable reading. My Keithley 2015 can be almost useless on DC if there is any noise on the DC. Its just a blur of numbers.

I would be cautious about even 5 digit DC accuracy with a scope. The probes are not that precise a division. however the magnitude of change may be reasonable accurate. I.E. 1.0015V not really. But a 15 millivolt change under load may be.

Audio ADC's usually digitally remove DC making those comparisons impossible.

I have 6 1/2 digit and 7 1/2 digit DVMs and an interesting device from QuantAsylum; the QA350, which is a 24 bit ADC with DC precision. Its similar to the DVM's but the lowest bits don't line up and its clear the steps are binary not decimal. However its a great null meter with isolated USB power and data. It also incorporates an interesting RMS function similar to the Keysight 3458.
 
Thanks Demian. I also own a 2015 and I noticed that you can select how many digits you want to see. As well as the averaging.

But my new USB scope has a selection of vertical bits versus BW. At the lowest BW, 3MHz, it has 16 bits resolution (250MHz at 8 or 10 bits).

So in the 16 bits resolution setting it should perform at least as good as a 5.5digit DMM, no? No just DC, also AC. And with all the caveats of filtering and averaging, which I believe goes for both cases.

(This is just a demo setting).

Jan
 

Attachments

  • WS6.PNG
    WS6.PNG
    47.6 KB · Views: 142
The 1/2 digit means that the most significant digit will be either 0 or 1. So the extremes of the DMM readings will range from 000,000 to 19,9999; therefore there are 20,000 possible codes. This would correspond to a little more than 14 binary bits. log(20000)/log(2) = 14.29 bits.

I hope I've done this right. ;)

I think it is 200,000 counts, no? 17.x bits?

Jan
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
Minor quibbles:

The conversion from binary bits to decimal digits seems to be 3.322, so 16 bits ought to be 4.8 digits.

A 5-digit DC meter uses all those bits for positive signal. (AD-DC conversion is often done before the ADC.) The 'scope has to handle + and - signal, so only 15 bits on each side.

But when you get to this many bits, a lot of it is in the details.
 
Last edited:

1audio

Member
Paid Member
2004-03-24 5:16 am
SF Bay Area
The scope uses an averaging algorithm similar to noise shaping to get the additional bits. Useful for some things but not necessarily accurate. It the 15 mV vs 10.015V example above. If its at the hardware level it may be more accurate and useful for looking at noise and artifacts. With FFT you could validate its real accuracy with some low level tones.