Setting up a small cal-lab. Am I on track?

[Fortran]

So we have a bunch of handheld multimeters at work that hasn't been calibrated lately. (ever)
Mostly 3½ and 4½ digits. I think we may have one or two 5½.

Plan is to buy a calibrated 6½ digit bench meter (Keysight 34460A) and use that to calibrate all the other meters.
That way we only need to send one meter to cal.

Then what?
Get a couple precision resistors, some voltage references etc.
Since I'll be comparing each DMM to the Keysight, I'm guessing accuracy isn't that important as long as there is stability?
How do you get stable ACV for instance? 

Any thoughts, tips, experiences?

[ap]

Since you buy this for work and so labour hours is an issue, and since you will need a calibrator anyhow, it may be more efficient to buy a calibrator (cheapest is probably a fluke 5100 or a 9000) and get that calibrated, or calibrate it with your keysight meter. That way you only need to do one measurement rather than a comprison (i.e. two measurements) each time.

[e61_phil]

I think that doesn't make sense. I'm from germany, but I think it is everywere the same: If you're accredited to ISO9001 your quality polices forces you to use a meter which is traceable to your NMI for specification verification measurements. To have a real traceable chain of uncertainties you need to read and understand the GUM, at least. And it take some time to do a proper calibration. In most cases a proper calibration from a cal lab isn't more expensive than doing it by your own.

Isn't it enough to "verify" the normal meters from time to time and use the proper calibrated one for the crucial things?

[Fortran]

Good point.
Maybe verification is a better word for it.
If up to spec, keep using. Else discard.

We "calibrate" a ton of equipment simply by comparing to known references. Figured this would be no different.
All we need is to be able to, by measuring, put a sticker on a meter saying it's within spec.
Knowing the exact amount it's off by isn't that important. Pass/fail is.

[mzzj]

Even a some sort of  verification can take some time and lots of fiddling with cables if you intend to do it with only resistor boxes and voltage references.
For example Fluke 175/179 manual calls for about 100 different measurements
https://dam-assets.fluke.com/s3fs-public/17x_____cieng0600.pdf

If you do only a very limited test like only 10 VDC test  the 1000V range could be dead, or current ranges, or AC ranges...

[GigaJoe]

something similar, do the same at home. consider 2 major obstacles: 1000V and AC. - if it need.   I haven't reach 1000V but made stable 400V,  1000 would-be my next step
secondly you may consider an older DMM for a better stability. some anchor box for example , or few of them. (i have 3 HP3456A)
building reference resistors need to match with upper level DMM for calibration,  as an example I have 1K,  2K ,3K ohm, limits on a 3 different units. same for everything. another thing some DMM allow to use adjusted reference like 1.0014K , so you define this numbers during cal-proc ,  some req.  precise 1.00000K ....

are you 3 and 4 digits adjustable?

[twbranch]

You really need a Fluke calibrator for calibrating meters. Like this: https://us.flukecal.com/products/electrical-calibration/electrical-calibrators/5730a-high-performance-multifunction-calibrat

Otherwise you don't have a clue what your meter is reading. Send this to Fluke every two years (I think not sure how often we send ours) and you are good to go with traceability.

[Fortran]

A Fluke calibrator does looks nice.
Indeed that would be ideal, although a simpler one is probably sufficient. Most of our meters are Fluke 179's.
I'm betting they're not cheap though, so I'll need to count the number of DMM's we have and calculate how long (if ever) it will take to get return on investment.

Still.  Can't you calibrate one meter with a better one?
I figured that's basically how it works.

[e61_phil]

Still.  Can't you calibrate one meter with a better one?

This is often done with "golden" 3458A for example.


A Fluke 5730A starts at >60k€. A calibration of a Fluke 179 is available for less than 50€ (which includes the work of the people in the cal lab and it is traceable).
I don't know how much a Fluke 5500A or something like that is, but you have to do really much calibrations to save money with your own calibrator. And calibrating such a calibrator isn't cheap either.

[AlfBaz]

Still.  Can't you calibrate one meter with a better one?
I figured that's basically how it works.

That's exactly how we complied with the ISO accreditation in a previous company I worked for.
We had a set of master instruments for this sole purpose, traceable back to our national standards body.
I made up a panel that contained all the various things that needed measuring such as DC and AC volts, DC and AC amps including larger currents for tong testers and a switchable set of resistors that covered the various ranges on our meters.

Compared measurements between master and the DUT writing down values on a pre-printed sheet. If the meters where within our required tolerance they got a sticker

[Kleinstein]

For the calibration with a second (better) meter one still need the signal sources. This is relatively easy for DC <= 100 V, but starts to get tricky with high and low resistance and especially AC if a frequencies different from normal mains is needed.  Even the 50 Hz AC can be a little tricky.
The resistors an related switches may need to be good quality as the readings are not simultaneous.

It also usually takes more time switching the sources by hand - the second reading from the better meter may not be such a problem, if it is connected to a computer and thus read automatic.

An reduced check to the ranges actually needed (e.g. no 1000 V), AC only at 50 Hz  and limited resistors should be easy and fast. A full calibration might take quite some time.

[mzzj]

Still.  Can't you calibrate one meter with a better one?
I figured that's basically how it works.

Sure you can, I'm just trying to say that it's not probably cost-effective on company time. For hobby use and maybe only checking some of the most interesting ranges its totally doable.
AlfBaz's solution of panel with enough switches and relays could make it reasonably efficient as then you wouldn't need to swap around billion combination of wires.

And you certainly need a boatload of meters to justify calibrator too. Fluke 5520 is something like 20-30keur and calibration is abouts 2-3keur every year or two.

For relatively stable and clean ACV you could use function generator or PC sound card output feeding a power amplifier (audio amp), voltage stepped up by a transformer.
And step-down transformer for AC current.
In my experience PC soundcard gives better results, less noise on the measurement.

[ap]

Sure, a 5730 or even 5520 is overkill for this application. But an old 5100 or newer wavetek 9000 are much lower cost and will do the job easily. His approach, under these conditions, absolutely makes sense and is efficient.

[e61_phil]

But what is the benefit of a calibration with a 5100? It might be ok for the 3.5 digit meters, but for the 4.5 and 5.5 digit meters you can't be sure that they are in spec. And what specifications do you apply to these measurements (done with the "calibrated meters")? Do you want to create a new specsheet with higher tolerances? How do you calculate this lower specs to be sure it fits?
In my opinion that doesn't make sense at all.

I think if you provide such an internal service you must guarantee that the measurements are reliable. And if you use some gear which isn't able to deliver the needed TUR, than you have to create a new datasheet for these meters.

On my desk at work I have a couple of uncalibrated gear and it is on me to be sure what I'm doing. I think that is a better way, because the engineer itself ensures the reliability of his measurements.

[dl1640]

FYI:
at work, we used to use 9100, now we substitute it with 5502A.
it can calibrate bunch of workload (dc/ac voltage, dc/ac current, dc/ac power, resistance, thermocouple/pt100 simulating & measuring, frequency, capacitance etc.), price is relatively nice, compares to 5730 beast...
also it has scope option which we have it and good for calibrating oscilloscope up to 600MHz bandwidth.
we have it calibrated at local fluke, price is not sky-high, and is traceable to SI.

[e61_phil]

@dl1640: Do you know how much time goes into a handheld DMM for the whole calibration?

[dl1640]

@phil

if we need *full* cal of hand-held dmm, I see 2 options to reduce time if dmm has no com port.

1. semi-automate the cal procedure to control calibrator (usually has com port) and give instruction for operator to manually switch to a desired function and/or range under test, record the dmm reading in the program or a spreadsheet.

2. make dedicated jig to switch dmm rotary knob by automated mechanical hand, read back dmm reading by some *visual sensor* to recognize the numerical number...this to fully automate the process.

I saw a local NMI use MET/CAL with 55xx calibrator to calibrate hand-held dmm in first way...