Typical Calibration points for 200mV, 2V, 20V, 200V, 1000V DMM

[Denny Wa]

I have a GWInstek GDM-8255A 199999 count Multimeter.  The user manual and specification sheet list ranges of 100mV, 1V, 10V, 100, 1000V, but it is actually 200mV, 2V, 20V, 200V, 1000V.  The question is, what calibration points should be used for calibration?  I say full scale, the calibration lab said , we use 10mV, 1V, 10V, 100V, 1000V.  Any comments from professional calibration technicians?  UNI-T also has a 199999 count, 5 1/2 DMM, UT805A , they document the ranges as, 200mV, 2V, 20V, 200V, 1000V.

Update:  I asked this lab to measure the DC input resistance of the meter at 100mV, 1V, 10V, 100V, 1000V for some high accuracy voltage dividers I need.  They said they are strictly prohibited from making those measurements, or they will return the meter to me.  I said ok, you are the experts.

[bob91343]

I would check both full scale and half scale, and decide whether to compromise to get most readings correct.

[Denny Wa]

Yes, that is what I would do also.  This is the second calibration of this instrument.  I was looking at the last calibration report and noticed they were not using full scale.  So I asked them about this, they said, "we use; 100mV, 1V, 10, 100, 1000V", does not seem right to me.

[Irv1n]

You can read document "Guidelines on the Calibration of Digital Multimeters - EURAMET"

[Denny Wa]

Thank you, I will find that document.  I did read that document, this is typical industry stand procedure, worldwide.  This calibration lab I am using is only giving me one point per range at 50% of range.  I mean really, all they have to do is hookup some leads and run a test program.

[mendip_discovery]

I have a GWInstek GDM-8255A 199999 count Multimeter.  The user manual and specification sheet list ranges of 100mV, 1V, 10V, 100, 1000V, but it is actually 200mV, 2V, 20V, 200V, 1000V.  The question is, what calibration points should be used for calibration?  I say full scale, the calibration lab said , we use 10mV, 1V, 10V, 100V, 1000V.  Any comments from professional calibration technicians?  UNI-T also has a 199999 count, 5 1/2 DMM, UT805A , they document the ranges as, 200mV, 2V, 20V, 200V, 1000V.

If its a 17025 calibration then the customer gets the specify the test points. If the customer doesn't state it then I usually find the limit of the range, 1, 2, 3, etc. and then I would test at lets say 0.9V, 1.9, 2.9 etc. DC I will pick a range like 20V and do a linearity test at 5V increments both + and -

Update:  I asked this lab to measure the DC input resistance of the meter at 100mV, 1V, 10V, 100V, 1000V for some high accuracy voltage dividers I need.  They said they are strictly prohibited from making those measurements, or they will return the meter to me.  I said ok, you are the experts.

Well, it is 10MΩ, will your divider be affected by it?

I would hazard a guess that as the manufacture doesn't state any spec of the input resistance so the cal lab won't get involved in doing that. Its not a normal request.

[bdunham7]

The question is, what calibration points should be used for calibration?  I say full scale, the calibration lab said , we use 10mV, 1V, 10V, 100V, 1000V.

What does the GW Instek manual say?

I don't know that model specifically, but not all DMMs give you any choice as to what the input stimulus is for setting the calibration constants.  More accurate, better meters generally do allow for non-standard stimuli, especially on resistance ranges, but I have seen ones that do not.  And many of those that do, including other GW Instek models that I've seen, only allow a limited deviation from the required test point.  Of course you can do a performance check without adjustment using any points that you want, regardless of what is in the manual.

Update:  I asked this lab to measure the DC input resistance of the meter at 100mV, 1V, 10V, 100V, 1000V for some high accuracy voltage dividers I need.  They said they are strictly prohibited from making those measurements, or they will return the meter to me.  I said ok, you are the experts.

How accurate?  You may need to characterize the input impedance with more than one simple resistance reading.

[TheDefpom]

Depending upon the multimeter design the mid scale point could actually be the best place to calibrate it as that could be where the greatest error is, if response is perfectly linear then yes the ends of the scale are theoretically better, BUT if there is non linearity due to compensations etc. that could change things.

For instance the Datron multimeters and calibrators all default to the middle of the range 10V for example, despite the range going to 19.999999V, they actually refer to the upper half as being over range.

I have attached a Datron design note, a few pages in it mentions this kind of thing.

[David Hess]

Depending upon the multimeter design the mid scale point could actually be the best place to calibrate it as that could be where the greatest error is, if response is perfectly linear then yes the ends of the scale are theoretically better, BUT if there is non linearity due to compensations etc. that could change things.

But then calibrating at the mid scale point would maximize the error at the end.

Analog to digital converters, and other types of circuits and transducers, often have a known shape of non-linearity curve, in which case the calibration point can be chosen to minimize the peak error.  If the non-linearity was S shaped, then calibrating at midpoint or endpoint would yield the same minimum error.  But if the non-linearity was parabolic which is common, then calibration at the midpoint or endpoint would maximize error and a point at about 2/3rds would be better.

A parabolic non-linearity is common from self heating of precision resistors in a feedback network, which makes the gain monotonically change with signal level.  I think S-shaped non-linearity most commonly comes about from first order correction of parabolic non-linearity.

[alm]

Analog to digital converters, and other types of circuits and transducers, often have a known shape of non-linearity curve, in which case the calibration point can be chosen to minimize the peak error.  If the non-linearity was S shaped, then calibrating at midpoint or endpoint would yield the same minimum error.  But if the non-linearity was parabolic which is common, then calibration at the midpoint or endpoint would maximize error and a point at about 2/3rds would be better.

On the last page of the article that Scott posted, Datron says they correct for switch resistance by measuring the deviation at its maximum (50% point), and then use the quadratic deviation through the two points (null and 50%) to correct linearity. What I find a bit confusing, though, is they talk about both adjusting and verifying linearity. But if they use a quadratic equation to correct the linearity, then it's not obvious to me that the residual non-linearity after correction still has its maximum at the 50% point.

[Kleinstein]

The self heating in resistors in an amplifier feedback network, input divider and the input resistor to an current integrating ADC gives an v³ contribution. After a linear fit or with the slope to minimize the maximum difference this gives the S shape INL part that is relatively common.

Another part can be a square law part, that can come up form FET switches (e.g. in the ADC).
The correction for the quare law part seems to be somewhat common by using calibration at +10 V and -10 V. This is relatively easy (no special instrument needed - but still a good zero reading !) and doable with good accuracy. The nonlinear part from CMOS swtiches also has a relatively predictable size.  A correction for the V³ part is more tricky and the V³ part can vary quite a bit including both signs.

Using the 10 V point instead of a possible 20 V end of range could be just from practical considerations: 10 V references are relatively common.
Calibration at the center of the range keeps the maximum deviation in the presence of the S shaped (or V³) INL small. Calibration at the end would have all the error to one side.

p.s.:
The Datron text about modern calibrators is about the error of the calibrator, not the DMM error. The case there is a PWM DAC with mismatch in the switch resistance. This gives a square contribution and with calibration at 0 and the end gives maximum error in the center.