Fluke 5440A calibration

[Rax]

Hi all,
Having just repaired my 5440A, I'd like to get it up to snuff.

There seem to be a few stages to this:

• Power Supplies (3-16, p.3-7) - pretty straight forward. This unit's levels were quite off when I checked them (before the repair). The unit passes all internal tests and measured exceptionally well before its first Int Cal (in my possession)
• Square Wave Generator (3-19, p.3-9) Not quite sure what the impact of this would be. Whether it absolutely needs to be completed.
• A bunch of others I may not focus on, at least in planning stages.
• External Calibration (3-27, p.3-18) Major ticket item, of course. A couple of basic questions here is that it's not able take a specific value close to 10V (such as, for instance, the Cal Club could provide). My solution to this - which I'm not sure how feasible, so I invite input - is to dial in my 731B to 10,0000000V, as indicated by my Prema 6048 (which I trust, don't REALLY ask me why ) and use that for the 10V step. Still thinking through the procedure, but do I absolutely need a null detector?
     - For the higher ranges, I can likely use my DP 8200's 100V output, though I don't think it'd be any improvement over just relying on the 5440A's own scaling linearity, as Dr Frank suggested on a different thread. I don't have anything for 1000V.
     - For the lower ranges, I'd use by KVD to divide down the 10V from the 731B.

Thanks for sharing your input.

[bdunham7]

• External Calibration (3-27, p.3-18) Major ticket item, of course. A couple of basic questions here is that it's not able take a specific value close to 10V (such as, for instance, the Cal Club could provide).
  

I'd be reluctant to actually make a calibration adjustment to either of those two sources until I had a decent history on the unit and knew its behavior (stability and tempco) over some time.  The 5440A is probably your most stable and accurate device by OEM specifications.  It is using the same reference as the 731B but with an oven.  And at this level, you start to really need a temp-controlled environment.

Are you absolutely sure about the inability to take voltages that deviate from nominal as a reference?  This would seem to be a huge oversight and this ability appears as step 9 of every 5440-series manual I could find, although I'm not at my normal computer at the moment and don't have an actual 5440A service manual handy.  After all, every other Fluke microcontroller operated device that has software calibration entry has this ability, even the old 8500-series DMMs. 

[Rax]

I'd be reluctant to actually make a calibration adjustment to either of those two sources until I had a decent history on the unit and knew its behavior (stability and tempco) over some time.  The 5440A is probably your most stable and accurate device by OEM specifications.  It is using the same reference as the 731B but with an oven.  And at this level, you start to really need a temp-controlled environment.

Which is why the way I envisage doing this is relatively consequently (within a day or, say, a weekend; even, if possible, in immediate sequence). I think I'd be able to complete the procedure with temperature variations of less than 1C.

The 5440A is, from my observations this far, in a completely different league than what I've seen before. I think the 731B is possibly not too far, but essentially, I'd use it purely as a transfer standard. I think as such it's intended to travel, be shipped maybe, experience some temperature variations, etc., but in my case it'd experience none of that.

Are you absolutely sure about the inability to take voltages that deviate from nominal as a reference?  This would seem to be a huge oversight and this ability appears as step 9 of every 5440-series manual I could find, although I'm not at my normal computer at the moment and don't have an actual 5440A service manual handy.  After all, every other Fluke microcontroller operated device that has software calibration entry has this ability, even the old 8500-series DMMs.

I'm not, this is from a relatively cursory read of the SM. Currently traveling for work, so limited capacity for this, but was hoping for some input from those with experience on this unit.

But the 5440A SM I have doesn't have a step 9 at the External Calibration procedure... ?... (mind you, the 5440A SM is only available as an Artek PDF, and is definitely different than the Bs or the 5442). Only seven steps in that one. May be a difference in software?

[Rax]

Here are the pages on the "External Calibration" from the 5440A SM.

[bdunham7]

It's in the second sentence of step 6.c. 

I think the USA Cal Club FX standard is the best possible answer for 10V, something recently transferred from that might be next-best.  But on those lower ranges are your substitutes for the 752A really adequate?  Personally I would only be adjusting it's calibration if I was reasonably confident I knew both how stable it was and approximately how much it was off--after comparing with multiple independent references.  Your 5440A may well turn out to be the most accurate instrument of your entire lot, especially after a 30-day power on bake out.

[Rax]

Interesting stuff here: "calibrate the calibrator"

[Dr. Frank]

At first, you only should calibrate your 5440A externally (@ 10V range) if you have an external standard which is uncertain relative to NIST lesser than 2ppm.
If you assume that your PREMA is uncertain to that level, you can use it directly, by measuring the 10V output from the 5440A during external calibration, and correct the 10V cal constant of the 5440A accordingly.
0.2 and 2V ranges always need such an external calibration as well, i.e. they also could be directly measured by a calibrated DMM. The 3458A would be suited best, as it auto calibrates its 100mV and 1V ranges, and therefore transfers its 10V uncertainty to these ranges. Its 1V range should be used only, for both ranges of the 5440A. Removal of offsets greatly enhances the uncertainty of the 5440.

As said, the 250V and 1kV ranges of the 3458A have too high an uncertainty, see its specifications vs. the required ones of the 5440.
The 20V range of the 5440 is additionally quite difficult to calibrate with a 3458A, as this would be done on the 100V range, and the 100:1 divider in ANY long scale DMM deteriorates its uncertainty, due to the high level of internal offsets, and impact from internal bias currents, because all applied voltages are measured in its internal 1V range only.

Anyhow, you only need to calibrate 0.2, 2, 10V ranges externally, and the 20, 250 and 1000V ranges are sort of auto-calibrated by INT CAL, so usually need no additional external calibration any more, if they are once properly calibrated at FLUKE. That's called the abbreviated external calibration, as can be found in the calibration chapter of the operation manual:

'The principal function of External Calibration is to correct for any shift in the Calibrator's internal Voltage reference.
A second function is to correct for long-term drift in the internal voltage divider resistors used for the 2.0V and 0.2V divided output ranges.
All other time and temperature dependent changes in the Calibrator are corrected by the Internal Calibration procedure.'

The differential method of full calibration originates from the old days, when there were no long scale DMMs, i.e. Fluke 335D, 720A, 752A, 732A and 845A made up a complete 7.5 digit calibration and measuring system from 100mV to 1kV. See: 'Calibration: Philosophy in Practice', 1^st and 2^nd Edition, FLUKE.

I have fully calibrated my 5442A after 13 years, for the first time ever, because HighVoltage kindly lent me his 752A, and I had a quite reliable 10V source available at that time, FLUKE 7000, which was calibrated by TiN to < 1ppm.
I have to say, that the deviation in all ranges were very minute. If I remember correctly, the shift of its internal 13V reference (i.e. its 10V range) was below 0.5ppm. The deviation of the higher ranges was about 1.. 1.5ppm, I guess. I have to search, but I think I published the calibration setup and results.

In summary, I would recommend to not re-calibrate your 5440A, as its uncertainty in its 10V .. 1kV ranges is probably much better than your PREMA.
Please get a calibrated 10V reference from your US CAL Club, to import The Volt into your lab, as bdunham7 already proposed.

Frank

[Rax]

I have to search, but I think I published the calibration setup and results.
Frank

Would it be difficult to find a link to those results, please? I've dug pretty hard, and I found different references to that thread (?), but I don't think I quite located the actual one. I'd appreciate some assistance with that.
Thank you.

[Dr. Frank]

I have to search, but I think I published the calibration setup and results.
Frank

Would it be difficult to find a link to those results, please? I've dug pretty hard, and I found different references to that thread (?), but I don't think I quite located the actual one. I'd appreciate some assistance with that.
Thank you.

I did not publish a dedicated thread about the calibration of my 5442A, so it's hard to find. Sorry that I sent you astray.
It's contained in the discussion about the Hamon Divider: https://www.eevblog.com/forum/metrology/influence-of-switch-resistance-in-hamon-dividers/msg4207093/#msg4207093

Therefore, the basic 10V calibration probably changed around 1ppm over 13 years, and the ratios of the upper ranges 1.6ppm at most, maybe caused by the assumed leakage or linearity fault of my box.
Anyhow, all ranges would still have been inside the 5440B, 30d Uncertainty specification, even before the full external calibration.
The ratio change would only violate the 30d Stability specification (0.5 .. 1ppm).
The 1kV linearity specification (1ppm) is not fulfilled, so that's another hint for a possible fault inside the box.

But that's all extremely stable, so my advice holds, to rely more on the 5440A, than on your PREMA.

Frank

[Rax]

I've looked a bit closer at my initial "post INT CAL" measurements, which were concerning with an almost 12ppm delta (just earlier tonight, 9.9998785V as read on the Prema for a dead-on 10.000000V output by the 5440A). Mind you - I've, in a less than ideal discipline of methodology, done both a re-zeroing of my Prema and the INT CAL on the 5440A at the same time, between the "pre INT CAL" and the "post INT CAL" measurements collection.

Tonight I tackled this again, as I got a bit more and better time to do this, re-zeroed the Prema on specifically the 20V range over 20s integration, left the low-emf AB cables with my weighty short contraption (= thermal inertia) rest to temp-equalize for few minutes every time I touched them, and benefitted from a more temperate and steady evening temperature in Southern California (= my garage @ a pretty steady 26C), and I seem to be within less than 5ppm of 10V.

Given the .5ppm of reading spec of the Prema for stability on the 10V range, the 5ppm on the 1yr accuracy, the 3.5ppm + 5uV 1year uncertainty of the 5440A/B, I think what I'm seeing qualifies.

I think in practical situations with a known instrument that's been performing well for a solid time and has experienced no major faults that may throw it off, one may see less than the above, but for the intents and purposes of a repair pursuit this I think tells me this is not out of the ballpark. If I get the chance of an objective reference at hand for comparison, I'll leverage that. Meaning I'd compare the reading of that and the 5440A 10V output on the Prema and draw some comparative conclusions.

But for now I'll be seeking feedback on my thoughts above. Thank you much in advance.

[branadic]

As far as I remember you neither know anything about the accuracy and/or linearity and/or t.c. of your Prema 6048 nor have you changed the opamps as suggested by dieter1 to improve its INL.
So I totally agree with Dr. Frank, don't trust your meter unless you have it verified against a well known programmable source to answer the open questions, as it is probably way off from being trustworth and rely on the calibrator instead.

-branadic-

[Rax]

As far as I remember you neither know anything about the accuracy and/or linearity and/or t.c. of your Prema 6048 nor have you changed the opamps as suggested by dieter1 to improve its INL.
So I totally agree with Dr. Frank, don't trust your meter unless you have it verified against a well known programmable source to answer the open questions, as it is probably way off from being trustworth and rely on the calibrator instead.

-branadic-

branadic - yes, all that's [sort of] true. I do have corroborating evidence the Prema is in the ballpark (I'm actually looking up records, both written up and in memory, to reassess that), but no hard evidence of exactly where it is with traceability and uncertainties and all that attached to it. As we all know, getting a Prema calibrated is actually a challenge today, and very few places would even touch it. I only have access to verified (17025 data attached) "up to 6.5 digit" meters, so that's pretty severely limiting my options.

It's entirely possible that, after I get the 5440A fully assessed, it could serve as the reference to calibrate/adjust the Prema. At that juncture, I'll be happy to do the mod nailed down by dieter1. A bit of a chicken and egg situation, for sure.

[Rax]

the accuracy and/or linearity and/or t.c. of your Prema 6048
-branadic-

I've been going through some records I have from earlier in the year, when it was my turn to have the FX and the rest of the kit (I have not been very good at submitting these to the Cal Club database, but maybe this is a good opportunity). To be noted, I got the Prema 6048 at the very end of the custody, so very little quality time was spent then with both overlapping, and this was a new to me meter, etc. So grain of salt below.

One 04/08, at 22.8C, after a fresh zeroing and using the low-emf AB cables I have (where Guard was floated as the FX obviously is a two wire deal), the FX was saying 9.9999712V on the Prema. Given the FX was presented as 9.99996194V, that's a hair below 1ppm off of 10V. I have other measurements that are not so close as this, but they are less trustworthy as they were done the day before with less experience on the 6048 etc.

This above is one data point I have that can be brought up to establish the credentials of the Prema (all its issues set aside for a moment).

[Rax]

Here is another reference point from the overlap time of my FX custody and very early ownership of my Prema. I posted these pics on the Cal Club USA #3 thread at the time.

I think this other reading shows the Prema - (can someone please confirm my math?... I apparently can never get ppm calcs to be second nature) - being .6ppm? off of the FX, but the cables are not low-emf (just decent quality ones).

Also, this obviously does 80s integration, while the example in the post just above I believe was - maybe? - 20s integration. IN my experience with the Prema, everything above 20s (providing all digits) is very consistent all the way to 80s. I'm only using 80s these days as a final vetting step when a critical reading is registered.

[leighcorrigall]

Congratulations on obtaining such nice equipment. I definitely enjoy my 5440B; it is excellent.

These are just random thoughts ...

Make certain to reapply the thermal paste to all components that require it before worrying about performance. The original thermal paste dries up and no longer effectively dissipates heat. You want to minimize heat buildup on the PCBs before considering calibration.

I heard from a former NRC meteorologist that these Fluke calibrators, in particular, should be internally calibrated every couple of weeks if they are left continuously in operation. I'm unsure of the reasoning because I had to move along in the tour, but I thought it interesting. Something about how they were designed, I guess. This meteorologist used the 5440B for voltage comparisons with some fancy equipment, which caught my eye. They are still very applicable in today's labs and outperform 5700A DCV specifications. The trouble is performing external calibrations if you don't have the right equipment (i.e., a null detector, voltage reference, and a voltage divider).

Assuming you do not have the appropriate equipment, perhaps the best choice is to verify the output voltages for each range instead of performing an external calibration. For the most part, if the instrument has not undergone serious refurbishment and one has confidence in the internal references, the calibrator should be well within the specifications after an internal calibration. As someone noted above, two SZA263/LTFLU references make the calibrator highly stable and with excellent linearity.

You will not be able to verify an unknown calibrator with an unknown meter.

[Rax]

these Fluke calibrators, in particular, should be internally calibrated every couple of weeks if they are left continuously in operation.

In my understanding, the INT CAL mostly addresses (= measures) miscellaneous drifts (internal reference resistors, for instance) such that zero and gain shift errors are factored into the calibration constants (which the procedure updates in memory). The manual actually requires it as often as daily (!) after warmup and before use. Thank you for confirming the importance of this requirement with feedback from the field team. The SM says do it daily, but I think that's under the assumption it's off every night - it sounds like the team you talked to said every two weeks if running constantly, which makes perfect sense to me.

What I haven't quite figured out is, in case the internal references are themselves subject to drifting, whether that throws off the INT CAL enough to essentially require the EXT CAL be performed to realign the instrument. I believe the INT CAL relies on internal voltages to calculate the revised cal constants. Sort of a chicken and egg situation, unless there's some other consideration here I'm not factoring in.

In this unit's case, running the INT CAL had a dramatic effect (as high as about 12ppm!), but that could be explained by decades of being stored in miscellaneous and uncontrolled conditions. Before INT CAL, the agreement between this and my Prema 6048 - as I explained previously - was nothing short of breathtaking. But I don't put much more currency on that, other than some sort of unlikely, but not impossible occurrence/coincidence.

[Rax]

Make certain to reapply the thermal paste to all components that require it before worrying about performance. The original thermal paste dries up and no longer effectively dissipates heat. You want to minimize heat buildup on the PCBs before considering calibration.

Good point. Along similar lines, I've seen some carelessly placed electrolytic capacitors (in terms of PCB design) - too close to heat sinks for long life, and therefore looking pretty toasty - and I've had to mitigate that somewhat (use 105C caps - though I use that by default for recapping - and tilt them to keep them as far away from heat as possible). As we all know, a lot of equipment from the time rather optimistically sized heatsinks, underrated tantalum caps, and made other similar design decisions.

Assuming you do not have the appropriate equipment

I have a verified 731B (by means of Keysight calibrated 6.5 digit meter), the Prema 6048 (which we're trying to assess the trustworthiness of), and few other standards to go by (obviously, I need to keep an eye on any circular reliance as I do this). Unless one has access to a metrology lab, I don't think doing this kind of thing (in a hobbyist lab) is ever easy.

You will not be able to verify an unknown calibrator with an unknown meter.

Of course!

[Rax]

Aside from the FX comps, my main bench reference point (DCV) is a 731B. I got this a few months ago, and before handover it was adjusted by a good friend and former metrology lab engineer/tech. The adjustment was with respect to a Keysight calibrated 34410A, and accounted for known deviations from the normative 10V (per Keysight's almost precisely one year old long form cal cert).

To gain more insight into my available DCV standards, I hooked it to the Prema 6048 today, and got readings between about 9.9999825V and 9.9999831V, which is about 1.75ppm to 1.69ppm off of 10V. This is at about 26C (one downside of the 731B is its less than stellar tempco compensation hardware provisions... though I think it's a pretty incredible performer particularly given that). For instance, a few hours later in the evening and a couple of degrees lower in the lab I see no discerning variations in the output of the 731B.

Over my entire ownership of the 731B (must be about a year now?), I've mostly been seeing te 731B within less than .5ppm off of 10V, though sometimes that may show a larger spread. But that was typically triggered by a change in contingent conditions - re-zeroing the meter, disconnecting (for a different project, etc.) and reconnecting the 731B, environmental conditions variations, etc. - so by itself the 731B is very consistent, at least as observed by using the Prema 6048.

To summarize, I'm taking this as another confirmation the Prema is reasonably within accuracy.

I am editing that, as it may sound conclusive, which is certainly not my intention. In my prior two or three posts, including this one, I am simply accruing as much data as I can to establish trustworthiness of my bench references, and the 5440A alongside with that.

I am looking forward to any thoughts and interpretations of the data presented this far.

[Rax]

I hope I'll be forgiven, but I've been editing the last couple of posts for form and clarity, and also data included. May not be the most conducive way, but as I solidify some of this data I feel revising the original post accordingly is the way to go. I think it'd worth taking a second look for those interested.

[bdunham7]

I am looking forward to any thoughts and interpretations of the data presented this far.

I haven't the time or energy to read the 5440A manual (actually I don't think I have it...) but I'd like to know exactly--in minute detail--how the INT CAL process works and what it depends on, etc.  I have some idea what it does for the 100V and 1000V ranges, where I think it is just using the linearity of the DAC to to a 1:10 transfer and comparison, but I don't grasp the details of the rest of it.  I'd certainly want to understand all of that in detail before doing any external calibrations.  Also, perhaps try the INT CAL again and see if it moves again or holds steady.  If it keeps changing, perhaps there is trouble somewhere.

[Rax]

Also, perhaps try the INT CAL again and see if it moves again or holds steady.  If it keeps changing, perhaps there is trouble somewhere.

This is why I'm saying this is in a different league in stability than anything I've seen before (and probably puts a good word for the Prema, as I wouldn't see it unless the DMM could read it).

Once I connect the 5440A which has been running for a few days now, it goes rock solid to the 9.9999500V I know it registers on the Prema. I am seeing variations of less than .1ppm as it cycles through 20s readings for a while as I'm doing other things around the bench (I've been watching it for maybe 15 minutes now). This is on a hot summer night in Southern California, (27.7C in the garage). I've seen the exact same value, within those variations, since its first INT CAL from a few days ago.

I am doing this to be sure what to expect after another INT CAL if I run it right now.

[Rax]

After the INT CAL, the 10V out of the 5440A seems to be pretty solidly at 3.1ppm off of a perfect 10V (in Prema's opinion) at 9.9999687V. So a bit closer to spec than the 5ppm I saw prior. Current variations are of less than .04ppm (...684 to ...692). We'll see...

I'll let run it until the morning and see what it thinks then. But maybe it just needs a good "30 day bake," as you put it, bdunham7. And maybe a few more INT CALs to really hone into the drift as the unit expels any moisture and crud and bad spirits by running for a while.

[Rax]

I'll let run it until the morning and see what it thinks then.

After running all night, at currently 25.6C, I'm seeing it somewhere near 9.9999648V (between ...44 and ...50). That's about 3.5ppm from 10V.

[Rax]

Here are the corresponding pages describing in detail the internal calibration to facilitate the analysis.

[Rax]

you neither know anything [....] linearity and/or t.c. of your Prema 6048
-branadic-

Getting more time to be thinking on this, branadic, that's not quite accurate. I am one of the contributors to the Prema 6048 with linearity tests based on my KVD measurements. I was able to replicate the "zero crossing" non-linearities observed with the Prema. I think only two or maybe three of us undertook that test.

I think I've also presented some TC data on that thread. Besides, if you have access to the Cal Club #3 dataset, you'll see some relative TC measurements that compare the Prema with a DP 8200 over a pretty wide range of temperatures.

Overall, I know far more on my Prema than I do on the 5440A. I do understand the 5440A has a following, and that some issues were determined to effect the Prema, but I don't think those really intersect this exploration here.

But as long as we're talking about the INL issue the Prema has, if I get a trustworthy 10V reference at my bench (which is at least partly why I'm doing all this), I'll then be able to tackle that upgrade to the Prema dieter1 figured out. I am not at all convinced the INL issue with the Prema impairs its ability to measure 10V accurately.