When I am looking at a datasheet for a DMM (Digital Multimeter) to calculate the accuracy of a measurement it specifies two error sources: of-range and of-value (also called of-reading). For example this datasheet for the Keysight 34470A: https://www.keysight.com/us/en/assets/7018-03846/data-sheets/5991-1983.pdf

I understand that the of-range error is mainly the offset error and the of-value error is the gain error.

But I don't get two things:

  1. Is the quantization error included in these of-range specifications or do I need to account for it separately? The examples always seem to suggest that the quantization error is in fact included.
  2. Is the random error included in the of-range error? If not, then how can I account for it, because I have never seen a DMM datasheet specify the noise (except for AC measurement)? If yes, than what is the confidence interval (2 sigma -> 95%, 3 sigma -> 99%, 4 sigma -> 99.99)?
  • \$\begingroup\$ Read the manual. In the one you've linked to, this is at the head of the specifications page, These specification are compliant to ISO/IEC 17025 for K = 2.. Get ISO 17025 and read its fine print. Generally, K=2 means 2 sigma, means 95% of readings fall within the limits. For (almost) all instruments, Q noise is insignificant compared to systematic error. \$\endgroup\$ – Neil_UK Apr 11 '20 at 3:31
  • \$\begingroup\$ "K=2 means 2 sigma": thanks, that is useful information! However, I obviously don't want to buy the standard for 142 USD. Especially, that in my general experience with these things, that standard would probably reference a dozen other standards, which would all reference a dozen other, and so on. So I suspect at the end it would cost me over a 1000 bucks. Plust, in my experience at least, the text of these standards are hard to read and for the general practitioner not clear what they mean at all... So that's why I'm asking the good people of the internet. \$\endgroup\$ – Cerike Apr 11 '20 at 11:12
  • \$\begingroup\$ At least you've bought a decent meter, not some dubious make never heard of before (I've got one made by Sino!). Keysight was HP, which means they should be fairly honest about the specifications. Once you've got your head round any temperature variations, and the number of averages they specify, and the cal interval, etc, etc, the total error should be within spec at least 95% of the time. For ADCs, getting small quantisation noise is cheap and easy, getting accuracy, stability, low additive noise etc is far harder, you can ignore Q noise as an issue. \$\endgroup\$ – Neil_UK Apr 11 '20 at 13:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.