732A hysteresis

[retroware]

I few months back I purchased a Fluke 732A with unknown cal history for some at home volt-nuttery. I've had it powered on for several months and it appears to be stable. I've monitored its voltage with two different 3458s. One that was brand new 3 years ago and one that was within 1ppm of the new unit three years ago (both are option 002). Neither have been calibrated since. The 732A measures .8 ppm different between the two units today. I'm half tempted to be happy with that, but I'd really like to have the 732A properly calibrated.

I've spent some time looking for a lab within 8 hours of the Washington DC area that would calibrate this for a reasonable price and so far the closest is in New York which would require shipping and probably result in the batteries going dead in transit.

Has anyone done any tests to see what sort of hysteresis the 732A exhibits if the battery fails? At this time of year the temperature changes from shipping wouldn't be too dramatic - maybe in the 10 degrees C range.

Or does anyone know of a lab (not including NIST) that is capable of calibrating a 732A within driving distance of DC?

[Bill158]

I have collected 6 ea. 732A references over the last 18 years.  I started with one unit and added each until 2009.  During the period I have had those I had had at least 4 power outages where all lost battery power and went completely cold.  The most recent was in October 2019 when we in Northern California had a infamous Pacific Gas & Electric company "public safety power shutoff" because of high winds and PG&E power lines causing wildfires.  I am sure you have heard of this.
I had collected a great deal of data over the years on this units using the NBS TN 430 method of "round robin" inter-comparisons.  After each cold start I have waited several days and then made the usual comparison measurements and found that all but one 732A had recovered to with in 0.2 ppm, usually less, of it's original inter-comparison value (in uVolts).  I did receive one unit that had some previous data with it from a last calibration in 2002 (hand written) 10.0000281 and then in 2005 from "DELPHI" 10.000062.  That was the last 732A I got in 2009.  As best as I could determine it was 2.8 ppm high with what data I had at the time as to the exact value of the "volt" which I had gotten from where I was working 5 years before.  But that "volt" was 2.3 ppm high when I sent a 732B I had recently obtained in 2015 and then into FLUKE for a 540Z calibration.  So my basic guess is that this 732A was at least 6 ppm or so high.  So the DELPHI lab may have gotten it correct at that time.  This one could have been cold for a long time before I got it.
So my experience seems to indicate that these 732As recover VERY nicely after power down cold starts.  I have several friends in the area, one of who has experienced the same thing in cold start recovery. 
Using a 3458A has never proven to be as accurate as you would want and expect.  But if that is as close as you can get for the time being, then so be it.
Bill

[retroware]

Thanks Bill158 for the very helpful information. Given that my unit is currently uncalibrated I might run it through one or two restart from cold conditions and monitor the unit as it stabilizes.

[wn1fju]

Dave:  You might consider sending the lab's 3458A back to Keysight if they offer a metrology-grade precision calibration.  I'm not convinced the original precision calibration was ever done at the time of purchase even though we paid for it. 

[leighcorrigall]

Hello, retroware!

I recently purchased a Fluke 732A from a fellow EEVBlog member a month ago. I am the third owner and the unit has only been powered off twice that I am aware of. Labels on the unit indicate that it might have been manufactured in the early '90s. I have not bothered to open the lid because I do not have to for the time being.

The last time it was calibrated was in 2015:
2015-11-03, 10.00010198 V, 1.0180223 V, 1.0000035 V @ 21 °C
2015-12-01, 10.00010440 V, 1.0179920 V, 1.0000608 V @ 21 °C (I do not know why it was measured twice)

With a freshly calibrated DMM (September), I measured the following as of today:
2021-10-05, 10.0001116 V, 1.0180225 V, 1.0000059 V @ 25 °C (nulled before measurement on the 10 V range)

For the most important range, this equates to 0.14 ppm/a over the past 6 years without considering temperature coefficients. I got lucky with this unit because it has been well taken care of. The other ranges are not as stable, but they turned out well anyway. 2.55 ppm/a for the 1.018 V range and -4.43 ppm/a for the 1 V range. Looking at the specifications from the manual tells me that these values are well within the anticipated drift. These Flukes age like good wine but they need to be well maintained.

I would suggest finding a reliable DMM and having it calibrated, then use it to measure the Fluke. This is a cheaper alternative and gives you more flexibility. Replace the Fluke batteries. Do not tamper with the ovenized reference chamber unless you know what you are doing.

Good luck! 

[leighcorrigall]

I'm not convinced the original precision calibration was ever done at the time of purchase even though we paid for it.

That does sound like a classic business strategy. I had that happen to a few unrelated products.

[RYcal]

Can I just highlight one issue around using these DMM's to test these volt standards.

Which is the more accurate device?....Off the top of my head unless you have characterized the crap out of the DMM to lower its specifications to anything reasonable then you will find that the whole exercise when you take into account some uncertainty analysis is a mockery. And to be fair you would either need some serious long term calibration history on the DMM or you could do it using a volt standard with good calibration history. Or you could use the calibration data and apply corrections and add the uncertainty from the report into the mix. I'm not sure if these 3458A's get tested at 10V but I sure would hope they do.

@10V the 3458A option 002 has a spec of 0.0000405V (Taken from a quick google search of the specs)
@10V the 732A has a spec of 0.00003V (Also taken from a quick google search of the specs)

This gives you a T.U.R of 0.74. Based on the 1 yr specs. Not Ideal.

Food for thought.

[retroware]

Yes, that's my conundrum. Should I get the 3458 calibrated or should I get the 732A calibrated and then use it to cal my 3458 (or at least measure it at 10V).  The price for either cal is about the same - ~$850 for the 3458 + as found data, ~$650 for the 732A. As far as uncertainty, the 732A cal seems to make more sense but I'm struggling to find a local cal lab to keep the power on long enough after the cal to get it back to me.

Based on the previous posters observations, I'm guessing  even with a power down that the 732A would end up giving me better uncertainties at 10V then a 3458 cal. But, without hard data it is only a guess.

My current plan is to do some several day long stability experiments with my 732A against 2 different 3458s and also against a Fluke 5440 + keithley 181. I'll then power it down for several days and repeat the experiment to see if there is any shift. This of course does not say what really will happen with a round trip to a cal lab 600 miles away.

[leighcorrigall]

The price for either cal is about the same - ~$850 for the 3458 + as found data, ~$650 for the 732A.

These prices are amazing. The US has a great industry to be this competitive in pricing. I pay about double here in Canada for the same services and they ship them to the US anyways. Just send in your best DMM and the Fluke standard for that cost. Also, be careful about what calibration laboratory you go with; not all of them have the same capabilities or expertise.

I'm guessing even with a power-down that the 732A would end up giving me better uncertainties at 10V than a 3458 cal.

Both the Fluke standard and the 3458 DMM use the same Zener diode technology. However, the Fluke is relatively simple and less prone to uncertainty than the DMM. However, this may not be the case for shipping the devices.

My current plan is to do some several day-long stability experiments with my 732A against 2 different 3458s and also against a Fluke 5440 + Keithley 181. I'll then power it down for several days and repeat the experiment to see if there is any shift. This of course does not say what really will happen with a round trip to a cal lab 600 miles away.

Keep in mind that several days isn't going to prove much. The Fluke standard should be much more stable than any metering device that you have assuming that it is in good condition. With that said, a Fluke 732A is bulky and non-ideal as a transfer standard in my opinion. It also uses trimmers to set the three output voltages. There is a possibility that the trimmers will be disturbed enough during transit that there are noticeable voltage differences. You may also want to consider shipping the device during the summer when the temperatures are moderate. Shipping a standard in Canada during the winter holidays will be a bad decision. The annual peak-to-peak temperature swing in Ottawa is -30 °C to 40 °C. I can't speak for your situation, but I am guessing that you may need to take this into consideration somewhat.

[RYcal]

Yes, that's my conundrum. Should I get the 3458 calibrated or should I get the 732A calibrated and then use it to cal my 3458 (or at least measure it at 10V).  The price for either cal is about the same - ~$850 for the 3458 + as found data, ~$650 for the 732A. As far as uncertainty, the 732A cal seems to make more sense but I'm struggling to find a local cal lab to keep the power on long enough after the cal to get it back to me.

Based on the previous posters observations, I'm guessing  even with a power down that the 732A would end up giving me better uncertainties at 10V then a 3458 cal. But, without hard data it is only a guess.

My current plan is to do some several day long stability experiments with my 732A against 2 different 3458s and also against a Fluke 5440 + keithley 181. I'll then power it down for several days and repeat the experiment to see if there is any shift. This of course does not say what really will happen with a round trip to a cal lab 600 miles away.

For a non accredited calibration facility I'd go with the DMM calibration and make sure it gets tested at the points you want to compare against the volt standard. That way you could apply corrections and add uncertainties into the mix for some reasonable data. Of course you want to be chasing the lowest possible uncertainties for this so study your calibration providers scope well.

The reason I say do the DMM is you will then have good trusted data for other functions like resistance etc. Bang for buck scenario.

It's good to finally speak to people that understand this.

[bdunham7]

These prices are amazing.....not all of them have the same capabilities or expertise.

I'd want to see what capabilities the lab has before sending them either of these instruments.  I'm not sure I see the point of sending something like the 732A anywhere except a primary standards lab or something very close to it.

Both the Fluke standard and the 3458 DMM use the same Zener diode technology.

Not really, the Fluke version runs much cooler.  The unheated versions of the Fluke reference are pretty good about drift over many years, but perhaps not at the 0.1ppm level.  Your cold-weather observation is spot-on--I had a (much less precise) reference take quite a dive when it apparently got very cold in the mail.  That's how I learned about VR thermal hysteresis.

[martinr33]

Bottom line: you cannot reliably measure a 732A with a 3458A. You can reliably perform artifact calibration on a 3458A with a good 732A.
You can also reliably measure drift on a 3458A using a 732A. You cannot measure drift on a 732A with a 3458A unless your 3458A is super stable, and even then that is questionable.
The 732A als tends to way exceed specification.

So get the 732A sorted, and that lets you perform precision measurements with the 3458A.

There are folks in reasonable distance from you that maintain good voltage standards - better than a lower-grade cal lab.


My experience is that 732As are very, very stable against time and temperature - small fractions of a PPM per year (that said, not all are that good as per Bill's comment above).

They do drift a little, up and down - maybe a couple of microvolts.

If left unpowered for long periods, I have seen one come back in a converging fashion - very quickly settled on a voltage in a few days. The other unit was still wandering around after almost a month.

[czgut]

My 732A, after few days break in power supply and about 24h warm up, returned to -0.21ppm. After 1 year break in power supply and 1 month warm up returned to -0.14ppm. Long term drift between two 732A units seems to be within 0.4ppm/8years.
With 1V or 1.018V outputs I would not expect that accuracy. Resistors in 1V and 1.018V dividers, measured separately, had TCR in 2..3ppm/deg range, and exhibited thermal hysteresis of up to few ppm, when cooled / heated in 20..50'C range. This may be the reason of applying 'IN CAL' circuit (detecting break in thermostating) in 732A.
 I did not measure accuracy of recovery on 1V level. In my case long term drift of two 1.018V outputs and one 1V level is below 3 ppm/8years, compared to 10V output by 3R parallel / 3R serial divider. One 1V output went out of correction span.

I would rely more on 732A then on 3458, but... You may keep Your 3458A switched ON for the whole time when Your 732A will be in Cal Cab. Reference diodes are much more stable, when are continuously supplied with constant current.  However .. it may last... Primary Laboratories usually measure drift of 732A under test for a few weeks prior to final calibration.

[leighcorrigall]

...

For the most important range, this equates to 0.14 ppm/a over the past 6 years without considering temperature coefficients. I got lucky with this unit because it has been well taken care of.

...

After about a month worth of measuring the 10 V range, the drift seems to be more like < 0.5 ppm/a over a 6-year interval. An R6581T and a 3458A confirms this observation for my Fluke 732A.