USA Cal Club Round 3

[Conrad Hoffman]

I can't prove it was whiskers. Another characteristic was that the problem would come and go. Sometimes it wouldn't show up  with my Fluke 73 DVM, but the minute I made a 10 volt (or higher) resistance measurement with the higher voltage meter, the problem would instantly come back and stay back for a good while, easily measured with the 73. Everything is clean and dry inside, plus the case is reasonably well sealed. Tin whiskers seemed as plausible as anything, but once the switch area was brushed out, the meter was as good (or bad) as ever and passes a 500 volt resistance check. "UCO" Unidentified Conductive Object.

I've been recording data morning and night on weekdays and will do a noon measurement over the weekend. I'm comparing three 731 references and two Mini-Metrology references. I don't have any kind of scanner so manual measurements are the best I can do. The lab is extremely stable at 68F all the time. The barometer hasn't moved much, nor has the relative humidity. Here's a screen shot from the Excel file. I do dates as Julian dates because they graph better. Every 10 microvolts is 1 ppm and an interesting point is the date 335.861 because the only thing that changed there was a 0.2" drop in barometric pressure, though I find it hard to believe that would be detectable. It's also interesting how everything moved in the first hours after tweaking all the pots and gently rapping things.

[Conrad Hoffman]

The kit has arrived back at Randall's so hopefully will be moving on to the next in line soon! Here's some photos of my primitive setup and a spreadsheet of results.

[Conrad Hoffman]

One more note on the GPS unit. I borrowed a small digital scope from work and put the two units upstairs where the signal is better. Over a day they stayed within about 1E-9 of each other, but either one could be running faster than the other, so they probably average to near nothing. That kind of makes sense for what I've read about GPSDO stability over short durations. At any rate (no pun intended) it's a great addition to the kit.

[RandallMcRee]

Yes, I have the kit. The next two folks in line have failed to respond to PMs sent last week, so...

next person to PM me can have it sent to them! (I mean if you are already in the list, of course)

Thanks,
Randall

Thanks--we have a winner.  

[chuckb]

The DC voltage kit arrived today in good condition. I am letting it settle down overnight before I start calibrating things. So far my DC Voltage calibration was only off 4ppm. That was based on the calibration of two Datron 4911s I picked up several years ago. I have over twenty 10V references around the place. It will be good to get a good voltage calibration assigned to each of them. The next week will be busy!

[RandallMcRee]

If you would like the kit after chuckb please PM me.

In the meantime, TiN has prepared a report concerning the recent FX incident.

https://xdevs.com/article/usac_fix23/

Kudos to TiN for his great support.

[CatalinaWOW]

Solid report and analysis, and I will add my thanks for all of the time and material that XDev has put into our club.

I have a question which could either be pertinent, a facepalm oversight on my part, or a reveal of more ignorance than I like to admit.  In any case I hope to improve my understanding.



The snippet above, taken from the first graph in the reports a snapshot of the references performance at two points in time.  Error bar is about the same on both, in the 0.3 to 0.4 ppm range.  As I think pf ppm a single digit change in the fifth digit to the right of the decimal point on this approximately 10 volt value represents approximately one ppm.  The numbers are reported in the SEP.2023 observation for three more decimal places beyond this, which is approximately parts per billion.  If the value is only valid to +/- 0.4 ppm those last three digits would seem to have no meaning.  The same reasoning would imply that the Jan and Sep measurement are not provably distinguishable from each other, with their separation being within the combined error bars on the two measurements.

What am I missing?

[bdunham7]

What am I missing?

I can't speak for the details of this particular example, but in general those tolerances are given as expanded uncertainties, usually k=2.  It's all a matter of probabilities and you need to look at is as a distribution rather than sharply defined error bars.  So if the +/-0.4ppm is an expanded uncertainty with k=2, then that implies that there is a 50% chance that the actual error will be 0.14ppm or less.  Maybe the last 1 or 2 digits aren't very useful, but the sixth decimal place one is.

[bdunham7]

Kudos to TiN for his great support.

I'm still amazed that the FX reference can not only achieve what it does, but that it does so even over many power cycles and despite enduring all of the environmental and physical punishment that the USPS can dole out. 

[CatalinaWOW]

What am I missing?

I can't speak for the details of this particular example, but in general those tolerances are given as expanded uncertainties, usually k=2.  It's all a matter of probabilities and you need to look at is as a distribution rather than sharply defined error bars.  So if the +/-0.4ppm is an expanded uncertainty with k=2, then that implies that there is a 50% chance that the actual error will be 0.14ppm or less.  Maybe the last 1 or 2 digits aren't very useful, but the sixth decimal place one is.

I will quote XDev.  No amount of statistical sugar coating beats reality.  Yes, there is a likelihood of some meaning to the sixth digit to the right of the decimal point, and it is more likely than not that there was drift over the nine months interval.  And the most likely amount of that drift is the apparent difference of the two values.

I guess another way to pose my question is this.  XDev strikes me as both careful and knowledgeable.  My first inclination is that his posting of eight digits beyond the decimal is not accidental, but don't understand how that can be consistent with the uncertainty.

[SilverSolder]

If you would like the kit after chuckb please PM me.

In the meantime, TiN has prepared a report concerning the recent FX incident.

https://xdevs.com/article/usac_fix23/

Kudos to TiN for his great support.

Amazing work, getting to this level of precision is truly astounding.

[chuckb]

The DC voltage cal kit is headed back to Randall. I was able to calibrate 15 10V sources and three DVM. I also checked my equipment at a cooler temperature to get a rough temperature coefficient for each item. It was a very busy / productive week.

More details to follow.

[RandallMcRee]

This is my last post here (as cal club moderator)!

I'm turning the reins over to bdunham7! 

I am sorry that I could not complete round3 but I have suggested to bdunham7 that he continue the round, if interest is shown.

Thanks for all the fish!  Bring your towel!

[bdunham7]

Thank you Randall for all the work so far.  As soon as I receive the kit I will take an inventory, consult with the remaning Round 3 participants and start a new thread, USA Cal Club Round 3.1.  I'm open to suggestions as to how to improve the process, such as streamlining or repackaging the kit contents.  Let's consider enrollment in this round to be closed for now.  I'll use the remaining Round 3.1 to come up with some ideas for improvements for Round 4 which will hopefully happen soon.

[Conrad Hoffman]

Randall, thanks for all the work and doing a great job!

[vindoline]

Randall, thanks for all the work and doing a great job!

I agree! Randall thank you for taking over from me and making round 3 a success!