Understanding multimeter accuracy ratings

[FenderBender]

Hey guys,

I was doing some comparison shopping for multimeters a few days ago and I realized I didn't completely understand what the datasheets were telling me.

Here is my current go-to multimeter's datasheet: http://content.amprobe.com/DataSheets/AM270.pdf

If you look at say "Ohms", it says 0.2% + 6digits for the 50Ohm range, 0.1% + 3digits for the 500Ohm range, etc.

I understand what the 0.2 or 0.1% means, but what exactly does the + x-digits mean? Could someone give an example of what the difference between 0.1% + 3digits and 0.1% + 6digits would mean in actuality?

This is what I would deduce. If I had a resistor that I knew was exactly 10R, according to this specification, 0.2%, the reading could be worst case from 9.998 to 10.002. It also says + 6 digits, which I'm assuming is the furthest right digit. So instead of 10.002 it could be 10.008? Then that's technically not 0.2% accuracy.

I don't understand it.

Thanks.

[kg4arn]

I recall Dave explaining this in the blog.
I understand it the way you do. 
The error is actually larger than the initial percent would lead you to believe, but by a fixed amount of least significant digits.

http://www.eevblog.com/2009/08/21/eevblog-26-multimeter-counts-accuracy-resolution-calibration/

[FenderBender]

I recall Dave explaining this in the blog.

http://www.eevblog.com/2009/08/21/eevblog-26-multimeter-counts-accuracy-resolution-calibration/

Thanks. This video was before my time!

[FenderBender]

Very well explained indeed. But I don't understand why they don't just spec a lower accuracy. I guess it must make sense. I don't feel like doing the math. 

[kg4arn]

I don't know either.  Hopefully someone will explain it to us. 

Maybe it has something to do with the resolution or operation of the A/D converters?
And that's the way you have to calculate the absolute error correctly?

[robrenz]

I always do an excell spreadsheet that calculates and plots percent of reading for all ranges of  instruments I am comparing.  Some companies show a lower percent of reading but have a higher number of digits that make them actually less accurate than the meter with the higher banner percentage.  Some are % of reading + % of range.   An example is from this post of mine https://www.eevblog.com/forum/product-reviews-photos-and-discussion/ie/msg143382/#msg143382 After you do this calculation on a few examples you will never judge by the banner numbers again.

[FenderBender]

I never really judged in the first place, but yeah I'll definitely use Excel next time and see if I can comparison shop a little better.

[FenderBender]

I don't know either.  Hopefully someone will explain it to us. 

Maybe it has something to do with the resolution or operation of the A/D converters?
And that's the way you have to calculate the absolute error correctly?

I almost think it's just a business technique...."Oh wow look! 0.0005% accuracy! fine print (+ 300digits)" Though I don't think they would be that cruel to unsuspecting buyers, who knows...

I almost thought it had something to do with the ADC limitations also but I couldn't tell you how or why...

[alm]

You can consider them offset and gain errors. Say you are measuring resistance with a 3.5 digit meter. Lead/contact resistance might be 0.5 ohm, plus an additional 0.1% uncertainty in the current source. If we want to specify the 100 ohm range from 10 to 100 ohm, the maximum error would be 5.1% (10 ohm might be measured as 10.51 ohm). This figure would greatly overestimate the error for 100 ohm resistors, where the actual error would be only 0.6% (100.6 ohm instead of 100 ohm). It would also be wrong for 1 ohm, where the actual error would be > 50%.

If we specify the 100 ohm range as 0.1% + 5 digits, it would give correct results all the way from 1 ohm to 100 ohm. Of course specifying it as a linear function is again an approximation, you may find even more detailed specs for the $$$$ high-precision meters. The offset error can be a combination of various factors, like offset in the buffer amplifier or non-linearity of the ADC near 0.

[FenderBender]

Alright! That's a lovely explanation. So they're not just being evil. Too bad the specifications can't be standardized onto some sort of scale or unit. Guess that wouldn't work. But it would sure make comparison easier.

[robrenz]

Also explains why you want to use the top of a range if possible for greatest accuracy. That percent of range really hurts at the low end of a range.