3458A differencies after calibration

[Drirr]

A few days ago, our company decided to have one of our multimeters calibrated after it was repaired.
The black 3458A is a new one, calibrated at Keysight. The white one is our older multimeter, calibrated at a Czech laboratory.
I have to say, I did not expect such a difference. I know they are probably within tolerance, but… you know.

We left our calibrator and both multimeters running for three days in a temperature-stable environment.
We also tried low-EMF cables from Pomona, but the difference was minor.

Please see attached image.

Would you consider this OK?

[picburner]

For such a comparison I would also use the "GUARD" connection (and shielded cables).
But at least I would press the "GUARD" buttons to "LO Position".
I would not leave the "GUARD" floating.
I do not see the Agilent button because it is hidden, but the Keysight button is definitely not pressed.

[Drirr]

@picburner
Thank you for your reply. Next week, I will try upgrading my low-EMF cables to shielded ones.
We experimented with the Guard, but there was almost no noticeable difference.

[scl1992]

The difference between the 2 meter readings was 9.2 ppm. It would be OK if the output from the calibrator (Datron/Wavetek 4000A or 4700A or 48xx?) was around 9.99988 V. So the 10 V output from the source should be known to determine whether any of the meters are out of specifications.

Might try these:
i) Perform ACAL prior to test. (ACAL should be performed every 24 hours).
ii) Interchange the connection cables to the meters.
iii) Swap the positions of the meters.
iv) Swap the power cables to the the meters.
v) Using Remote Sense (4-wire).

The 3458A Guard terminal should normally be connected either internally to LO or connected to the Calibrator LO terminal.

[aronake]

Was it calibrated or adjusted? what does the calibration certificate say it should show with 10V input?

[SHF]

I would place the devices next to each other so temperature transfer is excluded

[ElizatronicWarfare]

I also second the advice to run ACAL and move cables around to see if the error follows them. At this precision, you are fighting your connections and environment. Make sure your 3458As and setup is configured per the caveats in the specifications - many 3458A specs are only valid under certain configurations.

[IanJ]

Probably not going to be an issue but my own OCD would be looking to have the bottom 3458A fitted with feet......just to make sure both meters experience the exact same airflows/contact etc.

As it is, can you look at TEMP? on both meters and report back.

Ian.

[Oldtestgear]

 I would not be happy with this discrepancy. I get twitchy when comparing my 3458A with my 34470A after both come back from Keysight. The error between them (at 10V) is 2ppm.  But I am just a hobbyist with OCD.

Phil

[CalibrationGuy]

Would I be ok with this? No, but a few things must be considered. Here's some suggestions:

1) Same cables, Pomona or equivalent.
2) Set units to NDIG 8
3) ACAL ALL, wait 30 min to stabilize.
4) NPLC 100, or higher.
5) Look at calibration sheets. 4ppm/year is opt02, 8ppm/year drift is standard.
6) Two standard units can be up to 16ppm apart, without considering calibration uncertainties.
7) The units will drift closer and farther apart, several ppm over a day. Watch them and take readings every 30 min or so and see.

You need the absolute value of the reference, and the values from each unit's calibration to really assess the situation. On the face of it, yes, those 2 meters can be within their one year spec with those values.

PS - Most calibration houses will not adjust a meter that is in spec unless specifically requested to do so by the customer. If you want the numbers to match up, you have to request adjustment to 24 hour spec. When I send our master 3458a to Keysight, I ask for calibration to 24 hour spec only if the meter has drifted more than 2ppm or so above their reference values.

TomG.

[mendip_discovery]

As @CalibrationGuy points out just the spec of the meters may cover this if they are the 8 µV/V (reading) + 0.05 µV/V (range).

KS will have a nice low uncertainty, and the Czech lab may not.

Highly recommend calculating the actual uncertainty for both meters and then doing an EN ratio calculation. (Reading A - Reading B) / (Square Root (Uncertainty A * Uncertainty A + Uncertainty B * Uncertainty B)). You are looking for <1 potentially <0.5.

[bdunham7]

Would you consider this OK?

No, I don't think it is OK.

Which one was repaired and recently calibrated--I presume the white Agilent one? 

When was the calibration done on the other one? 

Do you have the calibration certificates and data from each calibration?

If you let them stablize for some time (>4h) and then perform an ACAL on each one simultaneously, what results do you get immediately after that?  And after that, do they drift?  Try doing the ACALs, recording the measurements and then reading them again 24H later.

[aronake]

To answer the question if this is OK, letting people here know what the calibration certificates say would be of much help.

[IanJ]

As it is, can you look at TEMP? on both meters and report back.

As I mentioned above this would be a helpful test to identify any possible fan/temperature related problems which are giving you your apparent discrepancy.

Ian

[Rax]

Most relevant points have been made, I think, but I'd second a couple:

• In my experience with a 3458A, running ACAL right before any critical measurements is critical.
• NPLC as high as possible.
• (Did I mention this before?) run ACAL right before any critical measurement.

I'd also recommend low-emf cables my AB-Precision (a resident here), which I think are excellent and very fairly priced.

[CalibrationGuy]

Most relevant points have been made, I think, but I'd second a couple:

• In my experience with a 3458A, running ACAL right before any critical measurements is critical.
• NPLC as high as possible.
• (Did I mention this before?) run ACAL right before any critical measurement.

I'd also recommend low-emf cables my AB-Precision (a resident here), which I think are excellent and very fairly priced.

I've seen this mentioned twice now, but Keysight recommends waiting 30 minutes after ACAL ALL to let the thermal emf from the switching relays die down. It's in their precision measurements white paper.

TomG.

[Rax]

I've seen this mentioned twice now, but Keysight recommends waiting 30 minutes after ACAL ALL to let the thermal emf from the switching relays die down. It's in their precision measurements white paper.

TomG.

The two are not contradictory - running ACAL before a critical measurement, and letting the thermal emf die down (per your reference). The point is not to expect high accuracy measurements without a fresh ACAL preceding it. What "right before" means should be defined by taking into consideration the literature.

[CalibrationGuy]

Just clarifying that right before doesn't mean immediately before because there are many here who will take that advice as written. That's all.

TomG.