I had time to run one more test to continue the 3458A guard switch saga, which was to check if the same problem also occurs if I use a different voltage reference than the Fluke 7001. So now I used the same 3458A #1 DMM and cable, but I used a Fluke 732A running off mains power (it does not have batteries yet) instead of the Fluke 7001 running of battery I used for the previous experiments. I tried both with connecting the guard lead to the F732A guard terminal, which is isolated from the 10 V HI/LO terminals. Unsurprisingly the guard switch did not make a difference there, and the difference are way down in the noise (0.05 ppm). Note that the vertical scale spans 0.2 ppm here:
Then I connected the guard to the same terminal as the low input lead to the DMM (so 10 V LO when measuring positive polarity and 10 V HI when measuring negative polarity). Here I observed a very similar effect to what I saw on the F7001, -6 to -8 ppm for this meter. So clearly the effect isn't just that the setting upsets the F7001 output stage:
Error bars again indicate standard deviation, n >= 20.
branadic and I on short notice came up with the idea of a mini (or micro) metrology meet at branadic's place in Stuttgart. The goal was to compare and transfer our best representation of 10 V and 10 kOhm between our labs over the span of a bit over 24h with a few hours of sleep. branadic was also so kind to add a thermistor to my 10 kOhm reference that was sorely missing an internal thermistor (see picture). All raw data is
available on xDevs.com and the preliminary analysis (without nice plots / summary tables and without proper uncertainty calculations) are
here on Github in a Jupyter Notebook.
A full article on xDevs.com should follow soon.
Fun it was, attached an image of the DUTs that we focused on.
-branadic-
Since I'm currently in the process of comparing the volt with Frank facing some trouble with measurements performed on our 3458A, I remembered again similar trouble we had at MM2023, that were fixed once we applied the guarding scheme as described in
HP AN-123.
So I thought it is worth mentioning here and drawing it in color, as it might be helpful to others, too.
-branadic-
Dr. branadic, did you already find the culprit?
Not yet, but I hope to spot it this weekend. All my measurements at home look good so far.
-branadic-
I finally managed to find the culprit and leave it here as an additional reminder to the above image.
I used a wall socket in the lab for powering your reference that was nearby. Obviously this wall socket is connected to a different phase, so meter and wall adapter of the reference were running on different phases, which led to large shifts during polartity reversal, that vanished when I ran the reference from batteries.
I also cleaned all posts and cable connectors again before I repeated the measurements and tada the result now make sense.
-branadic-
Hello Dr. branadic,
that's a great finding!
I have never seen that polarity change effect in my lab, as all instruments and all the references are supplied from the identical phase.
We should ask the colleagues from CERN, PTB tomorrow, how they analyse and manage the mains and power supply (CMMR), illumination (EMC) and temperature (T.C.) characteristics in their labs.
This is another very important aspect, to achieve lower uncertainty and higher stability.
In our EE company, all mains plugs are marked individually, I.e. the red ones are connected to the USV system, and the black ones are marked for the 3 different phases, if I remember right. Have not been in the company recently.
I guess that our engineers don't care, because they are not doing such high precision stuff, usually
Frank
Just the normal mains phase alone should not have much effect. The meters should have enough shields and mains hum suppression to cope with this. Especially if the guard terminals are connected there should not be much mains effect left.
The problem is likely more some higher frequencies that also live on the mains and also different ground paths. It may be worth testing if an EMI mains filter make a difference.
Interesting, if polarity reversal has a effect I'd assume there is a dc component to it.
Leakage from the power supply?
I think there are two questions here: Why does connecting to a different AC circuit (different phase or frequency?) matter, and why does this affect polarity reversal. To me phase cannot matter unless the meter is sampling an odd number of power line cycles, like NPLC 1. It also seems unlikely that the frequency is very different (probably coming from the same three phase feed). So I agree with Kleinstein that this is something else than just the phase difference. Like indeed ground currents or EMI. And then the question that The Soulman raies is why does this increase the polarity reversal deviation, and my best guess is there that the noise is getting rectified at the bridge rectifier of the reference's power supply and is leaking through the regulator.
I wonder if the same problem would happen if you connected the 3458A and reference both to the other phase. If it's EMI on that phase, it should show the same result. And then it would be interesting to see what the ref's output looks like on a scope (AC coupled and ideally differentially) when connected to either phase. Maybe the noise is so high that you don't even need a low noise amplifier
Would offer a handful of split ferrites to
@branadic for CM and DM supression games, if its possible to do the right thing (tm) with them.
Will there be a meeting in 2024?
Asking for a friend(ly Keithley 2000 and a Prema 5000.
)
Hello,
we can have MM2024. Possible dates are:
06.07.2024
13.07.2024
20.07.2024
27.07.2024
03.08.2024
10.08.2024
17.08.2024
24.08.2024
I would like to ask you to give me a feedback. Depending on that and some additional feedback from a person I'm asking for a talk we will fix the date.
-branadic-
Personally I would prefer July over August.
Local school holidays start end of July, so generally more people might be on vacation from then on.
Hiho
I would second Wollvieh‘s preference for a date in July over August
cheers, Ole
From July 8, 2024 – July 12, 2024 there is the CPEM while IMEKO World Congress is from August, 26 – August 29, 2024.
On the other hand, if I would schedule the meeting on August 17, 2024 we could have a talk by PTB.
-branadic-
August 17th is perfectly fine for me, as well July 8th to August 29th, which are the Summer holidays. André, thank you!
Frank
Hi,
i'd like to come too. I don't mind the date.
Does someone from southern-bavaria go there and could give me and my Prema-5017 a lift?
(DB / german railservice sucks big time)
Would bring some "cheap" voltage-references as well.
Best regards
I think I'll be there this year too. Last time it didn't work out for me due to the coronavirus.
I would bring my "new" 1Ohm standard resistor and my Prema5017.
Greetings until then
Save the Date for MM2024: August 17, 2024
-branadic-