Fluke 5440A calibration

[Rax]

You might monitor the environmental temperature as well.
Frank.

Today I got a glimpse of what excessive temperatures look like in the readings. Prior, I almost developed a sense of "invincibility" from the P6048/F5440A pair. But the temperature at the bench having exceeded 28C by quite a bit (SoCal heat wave), I saw an almost 1ppm divergence between the two, which is pretty unsettling, having seen such rock solid agreement this far.

Hopefully, there's no other, more permanent, reason for this divergence.

[bdunham7]

Today I got a glimpse of what excessive temperatures look like in the readings. Prior, I almost developed a sense of "invincibility" from the P6048/F5440A pair. But the temperature at the bench having exceeded 28C by quite a bit (SoCal heat wave), I saw an almost 1ppm divergence between the two, which is pretty unsettling, having seen such rock solid agreement this far.

Hopefully, there's no other, more permanent, reason for this divergence.

Tempco within listed limits is going to be a big part of the specified tolerances.  What you won't know is whether the tempco of your two units is complementary or oppositional.  If the latter, you're in good shape, but if the former the actual values could be varying more than you'd like.  The 8846A has very good tempco AFAIK (the old PCBA did for sure) but it certainly can't be the arbiter of a 1ppm discrepancy. 

For your data acquisition issue, how about just using time lapse video?

[Rax]

Tempco within listed limits is going to be a big part of the specified tolerances.  What you won't know is whether the tempco of your two units is complementary or oppositional.  If the latter, you're in good shape, but if the former the actual values could be varying more than you'd like.  The 8846A has very good tempco AFAIK (the old PCBA did for sure) but it certainly can't be the arbiter of a 1ppm discrepancy. 

I think my observations this far point out oppositional tempco between the F5440A and the P6048 - by which I mean they seem to essentially compensate their T gradient - but they do that at slightly different rates. So over 28C this slight divergence starts becoming apparent. But throughout my usual "summer bench" environmentals (say, 24-28C, sunny with a slight breeze at the beach), they stick together very well. Data collection and "armchair processing" still ongoing.

For your data acquisition issue, how about just using time lapse video?

I'm not quite sure what you mean by this, but if you could elaborate?...

[Rax]

Now regularly, say every week or every month, do the same. And maybe after a number of months you'll be able to say something about the relative long term drift

Fine. What I'm not sure I have a handle on, is how to subtract (or maybe "abstract?...") the drift you're talking about while, obviously, everything I could throw at this data collection has its/theirs own drift(s)?... Differently put, how can I guarantee I see the F5440A drift, vs. the (relative) P6048 drift?

[bdunham7]

I'm not quite sure what you mean by this, but if you could elaborate?...

Just take a long video of the instruments and then fast forward through it later and manually record the readings every xxx seconds or minutes.

[dietert1]

When you go sub-ppm thermal EMF can get into the way, typical at a time scale of minutes.
As the Prema doesn't come with binding posts but with copper safety sockets, you need copper banana plugs. And you want to wrap the connectors against air draft. Others made a small box to cover them.
How is the Connection on the Fluke calibrator? Probably one needs copper parts, too.

Regards, Dieter

[Rax]

When you go sub-ppm thermal EMF can get into the way, typical at a time scale of minutes.
As the Prema doesn't come with binding posts but with copper safety sockets, you need copper banana plugs. And you want to wrap the connectors against air draft. Others made a small box to cover them.
How is the Connection on the Fluke calibrator? Probably one needs copper parts, too.

Regards, Dieter

Dieter - definitely so; if I open the garage door, I see the draft and temperature gradient pretty quickly up to a level of a ppm or so I think - I mentioned this here before. My better half knows the car gets out first thing in the morning, and gets back into the garage last thing in the evening. When needed, I run the AC in the garage and I can also see those cooler drafts in the readings pretty distinctly.

But to answer your question - I am using low-emf banana plugs fitted 1m cables by AB, so definitely up for this service. I think those are some really neat cables. I always leave enough time for temperature settling after maneuvering them, and the 20s integration I always use helps with that too.

[Rax]

But the temperature at the bench having exceeded 28C by quite a bit (SoCal heat wave), I saw an almost 1ppm divergence between the two, which is pretty unsettling, having seen such rock solid agreement this far.

Hopefully, there's no other, more permanent, reason for this divergence.

Or, maybe I just ran out of luck. I've had these on since, ran zeroing on the P6048 a few times, INT CAL on the F5440A, also internal soft and hard checks, and I seem to see a permanent slide up of about 1.2ppm that's not going away. May try a cold boot or other cures. Pretty disheartening, to be honest.

My other bench reference is a 931B 731B - and please recall it's a verified extremely stable reference relative to what's expected from a 931B 731B - is still in "under .5ppm" agreement with the P6048. Maybe the F5440A just experienced a slight reference jump?

[Rax]

I should add the perfect agreement between the F5440A and the P6048 has been unshaken for almost two months of essentially daily checks. I probably should have never restarted the instruments.

[bdunham7]

My other bench reference is a 931B 731B- and please recall it's a verified extremely stable reference relative to what's expected from a 931B - is still in "under .5ppm" agreement with the P6048. Maybe the F5440A just experienced a slight reference jump?

I don't think the SZA263/LT-FLU reference is all that prone to power cycle jumps like that because it isn't heated, at least not internally or to the same temperatures as the LM399 and LTZ1000 types.  But there are probably good reasons that they assign the tolerances they specify.  What would be your best guess as to your absolute uncertainty of 10 volts in your garage lab? 

[Dr. Frank]

Hey Rax,
definitely you should acquire a cheap GPIB adapter, a monitoring program and a third reference source.
You're turning in circles, not exactly knowing what your both instruments are doing.
Frank

[alm]

I should add the perfect agreement between the F5440A and the P6048 has been unshaken for almost two months of essentially daily checks. I probably should have never restarted the instruments.

Yet the only actual data you presented is in disagreement with this. This "almost two months of measurements" appears to have produced zero useful data. Could you have written a college lab report with "The mass of all the liquids I measured was in agreement with their density and volume"? That might be a conclusion, but that's not a report.

I don't know the Prema 6048, but some Prema meters have fairly advanced math features, so it wouldn't surprise me if you could configure the P6048 to acquire say 32 samples and show the mean and sample standard deviation or variance. While not as good as acquiring individual samples, it could make manual registration more convenient until you set up automation.

[Rax]

You're turning in circles, not exactly knowing what your both instruments are doing.
Frank

Hi Frank,
There's three instruments, and as I presented here before:

My other bench reference is a 931B 731B - and please recall it's a verified extremely stable reference relative to what's expected from a 931B 731B - is still in "under .5ppm" agreement with the P6048.

They've been in very close agreement until just now. Three instruments excludes a vicious circle fallacy.

[Dr. Frank]

You're turning in circles, not exactly knowing what your both instruments are doing.
Frank

Hi Frank,
There's three instruments, and as I presented here before:

My other bench reference is a 931B 731B - and please recall it's a verified extremely stable reference relative to what's expected from a 931B 731B - is still in "under .5ppm" agreement with the P6048.

They've been in very close agreement until just now. Three instruments excludes a vicious circle fallacy.

Hi Rax,
sorry, no, this 731B is a transfer standard at best, but not a reference standard.
It's too much depending on temperature variations.
And you know: A man with one clock knows the time, but a man with only two clocks is not so sure any more..
Frank

[Rax]

I don't think the SZA263/LT-FLU reference is all that prone to power cycle jumps like that because it isn't heated, at least not internally or to the same temperatures as the LM399 and LTZ1000 types. 

But, as you know, this is ovenized inside the unit. I have two hypotheses to go by:

• The reference jumped by the observed 1.5ppm. I should have kept it running, minimizing the risk of this occurring. Sometimes learning to trust yourself over external input comes with tolls.
• The environmental temperature (28C at the time) triggered some sort of failure. I'll probably open the unit and see if there's any signs of trouble. I suspect something like power carbon comp resistors, one of the very few original wet caps leaking due to heat, etc.

Thank you for your thoughtful points.

[Rax]

Yet the only actual data you presented is in disagreement with this.

It's not, but I'm sure you'll tell me that it is (hopefully, also "how" it is, or "why"). If you could make that less of a "teaching down at you" moment, that'll drive your point a little better, and I mean this both respectfully and hopefully (if you're a tenth as prominent in what you do, as you seem to be, you'd know this. I am in what I do, and I do know it). It's a disservice to your knowledge to have this demeanor.

But I haven't seen that this far (with all of your generous posting with resources, articles, and links, etc., which I appreciate every single one of), nor have I seen a close read of what I'm posting here. You're glazing over my discrete data or qualitative points concerning hours of observations I've made (none of which I can really substantiate, as they're "qualitative," and you'd just have to trust me that what I'm saying is the result of hundreds of hours hard work of "memorized" data accrual) to drive points you want to make regardless of any of it. That, in particular, is not helpful. It's disrespectful of my work.

The ball is in your court, and I'm hopeful.