Error of a measurements taken with a multimeter

[MementoMori96]

Hi, normally when we take a measure with a multimer we consider the error given by this type of tables:

http://www.transcat.com/media/pdf/mete35xp.pdf

Do you know how these errors are calculated?

 

[Asymptotic]

If you are asking how to interpret the accuracy specifications then read through this article at Design World.

https://www.designworldonline.com/how-to-determine-digital-multimeter-accuracy/

 

[MementoMori96]

No i’m asking how these errors are calculated. How does the company estimated these errors ?

 

[phinds]

No i’m asking how these errors are calculated. How does the company estimated these errors ?

Presumably the are calculated based on the known tolerances of the components of the meter

 

[berkeman]

No i’m asking how these errors are calculated. How does the company estimated these errors ?

They are by design, and subsequent design validation.

For example, during design you decide whether to use 1% tolerance resistors, or 0.1% tolerance resistors. The 0.1% resistors are more expensive, so you only use them in places that affect the overall accuracy.

You also typically would design custom ICs for such a product. And in those custom ICs there are design techniques to improve accuracy and matching of component values. The more sophisticated you make the IC, the better the accuracy specs that you can publish. But the tradeoff is that it takes extra silicon area to do all that matching (and sometimes trimming and calibrating), which makes the product more expensive.

So early in the design specification stage of a product, you figure out where you want to position such a product on the price/performance curve in the market, and then design accordingly.

Does that help?

 

[MementoMori96]

Thanks . Is possibile that these errors contain also systematic errors ?

 

[berkeman]

Thanks . Is possibile that these errors contain also systematic errors ?

As opposed to what? Random errors from thermal noise?

 

[MementoMori96]

I consider the error that i calculate thanks to these tables a random error, is a mistake?

 

[Asymptotic]

Part of the manufacturing and ongoing maintenance process is to calibrate the DMM against a known standard of superior specification, typically, one with at least 4 times less error that the meter function being calibrated. I don't know if his name has resonance today, but one of the pioneers of electrical technology was Edward Weston who developed a saturated cadmium cell that generated an extremely stable voltage, and was an NIST voltage standard up to the 1990s.

If you are interested in metrology in general, "Standard Cells: Their Construction, Maintenance, and Characteristics", Walter Hamer, 1965, has a history of how voltage standards were developed, although these days the NIST standard is based on Joesphson junctions instead of electochemical cells.

https://www.nist.gov/calibrations/voltage-measurements-calibrations

 

[berkeman]

I consider the error that i calculate thanks to these tables a random error, is a mistake?

It depends on the instrument and the thing being measured, but for a DVM I would think that the errors are more systematic rather than random. You mainly get random errors when thermal noise is significant in a measurement, like with a picoammeter. For standard ADC measurements, the errors would tend to be more systematic.

For example, for a data acquisition system that I designed recently, I had offsets and gains that we calibrate at manufacturing test time, to ensure that we stay within the error tolerance numbers we publish. After that calibration procedure, I get a fairly straight line for my ADC transfer characteristic, but the line may still have a little offset and/or a little gain error (but it stays within the allowed error bands). It is not a random function within those error bands, since thermal noise is not a problem for this ADC system.

 

[sophiecentaur]

I consider the error that i calculate thanks to these tables a random error, is a mistake?

If your DMM has not been calibrated specially (you pay more for that) then the probable errors are what is in your table and they will be due to the tolerances of all the components in the instrument and the range of temperatures it is specified to operate over. A number of different instruments will have been tested to back that up. As Berkeman says, the errors are more likely to be systematic but there will be a random element from instrument to instrument. You could calibrate your own at some chosen points and it would indicate that your meter consistently reads about 2% high on one particular range, for example, even if the table tells you +/- 5%.