Stupid Question about Thermal EMF on connections.

[mendip_discovery]

So I understand thermal emf is a thing I need to take into account.

One thing that has me wondering is if it has to do with the difference in temperature between the two connections or the temperature they are in.

For example, if I plug in some plated 4mm banana connectors I get a voltage offset at first then after 10 minutes or so it drops significantly, and I have put that down to the difference in temp.

But at the same time, I need to take into account room temperature changes, but is that just causing a short-term issue or is it a long-term issue?

34401A <-- gold plated connector with copper wire --> 10 V Ref.

Copper + Gold + Copper + Gold + Copper
0.5 µV/°C + 0.5 µV/°C + 0.5 µV/°C + 0.5 µV/°C + 0.5 µV/°C

If I were to use copper all along I would straight away start getting, <0.3 µV/°C

[Gyro]

...
Copper + Gold + Copper + Gold + Copper
0.5 µV/°C + 0.5 µV/°C + 0.5 µV/°C + 0.5 µV/°C + 0.5 µV/°C

You are perhaps forgetting polarity. Copper-Gold is 0.5uV/'C but Gold-Copper is -0.5uV/'C so you will get a good deal of cancellation. Also, the plating is intimate contact with (and so, the same temperature as) the underlying metal of the connector, so whether this is Copper, brass, Nickel etc. has a much greater consequence than the plating. This applies both to the mated contacts and the copper wire to connector junction (solder or crimp).

As you observed, the thermal emf reading tends to decrease with time as the connector temperatures equalize. Other things being equal, accumulated thermal emf the +ve and -ve connections should cancel as their temperatures become the same.

If I were to use copper all along I would straight away start getting, <0.3 µV/°C

Yes you should (and these should mostly cancel over time, as above). Note though that Copper-Copper Oxide is >500uV/'C. The Copper needs to be freshly clean, otherwise the thermal emfs can be huge and unequal.

It might be worth getting one of the guys in your lab to give you a more detailed refresher, as this can be quite a significant factor for low level, or long scale instrument calibration uncertainties.

[TimFox]

Another thing to watch out for is that the interior of the voltmeter or whatever will usually be warmer than the ambient air, due to power dissipation of the equipment, so you may have an unexpected temperature difference somewhere in the circuit.

[dietert1]

If you have a lab multimeter that resolves uV, you can do some experiments yourself to get intuitive understanding. For example make a short across the meter input using a short piece of copper wire. Then wait a minute until you get stable readings. Then you heat one of the binding posts between two fingers and observe the EMF. If you succeed to heat by 10 K, you may see 5 uV. Now if you let it cool down for another minute and heat the other binding post, the sign will be opposite.
Binding posts, even low EMF ones are usually made from copper alloys easier to machine than the pure electro-copper of wires. That difference times a temperature difference is enough to generate EMF.
For low EMF measurements i often wrap both binding posts with a sock to protect them from air draft. This does not guarantee compensation of EMF, for example if the upper binding post gets more heat from inside the meter. Nanovoltmeters usually have connector systems with a metal shell around both contacts for better temperature equalization.
Another proven approach is soldering copper wires or crimping copper parts to each other in order to avoid the oxide layer. You will normally find that a thin layer of gold does nothing, it the copper to gold and the gold to copper temperature is the same. In general using thick wires is worse than thin wires, as thin wires cause lower temperature differences. But that's all subject to experimentation.

Regards, Dieter

[mendip_discovery]

Yes you should (and these should mostly cancel over time, as above). Note though that Copper-Copper Oxide is >500uV/'C. The Copper needs to be freshly clean, otherwise the thermal emfs can be huge and unequal.

Thanks for the clarification. I was just going in circles when reading about the subject.

It might be worth getting one of the guys in your lab to give you a more detailed refresher, as this can be quite a significant factor for low level, or long scale instrument calibration uncertainties.

erm, yeah about that. I am the HoL so it is kinda me who has to deal with the educating stuff. Sadly I don't have any wise of wizards (grey beards) to ask at work as I am the eldest and yet to be blessed with wisdom. Thankfully I don't use my 3458A for much other than resistance and the odd voltage but my uncertainies are big enough to cover this.

[mendip_discovery]

If you have a lab multimeter that resolves uV, you can do some experiments yourself to get intuitive understanding. For example make a short across the meter input using a short piece of copper wire. Then wait a minute until you get stable readings. Then you heat one of the binding posts between two fingers and observe the EMF. If you succeed to heat by 10 K, you may see 5 uV. Now if you let it cool down for another minute and heat the other binding post, the sign will be opposite.
Binding posts, even low EMF ones are usually made from copper alloys easier to machine than the pure electro-copper of wires. That difference times a temperature difference is enough to generate EMF.
For low EMF measurements i often wrap both binding posts with a sock to protect them from air draft. This does not guarantee compensation of EMF, for example if the upper binding post gets more heat from inside the meter. Nanovoltmeters usually have connector systems with a metal shell around both contacts for better temperature equalization.
Another proven approach is soldering copper wires or crimping copper parts to each other in order to avoid the oxide layer. You will normally find that a thin layer of gold does nothing, it the copper to gold and the gold to copper temperature is the same. In general using thick wires is worse than thin wires, as thin wires cause lower temperature differences. But that's all subject to experimentation.

I did some tests a while back to work out the best leads we had. Got some PTFE shielded wire and that made a big difference. I have seen some have a 3d printed cover that goes over the terminals to prevent drafts etc. I guess the best I can hope for at the moment is to keep things as equal as I can and avoid too many different metal contacts being involved.

[bastl_r]

Hi
I just got another lesson regarding EMF...
I am currently trying to build a small Hamon divider according to Conrad Hoffmann and have adjusted the lower resistor to approx. 0.5ppm to the upper three in parallel.
Applied 1V with the calibrator and... 60ppm....
What did I do wrong?
I calibrated the resistors with offset compensation. So they should be the same - if you ignore the EMF. Grrr.
So again and please only calibrate with a stable DC voltage so that the EMF is always taken into account. Also that of the internal solder joints.

PTFE, i think, do you not need with low voltages an 6.5 digits. EMF is much more important to begin with!

[Gyro]

If you have a lab multimeter that resolves uV, you can do some experiments yourself to get intuitive understanding. For example make a short across the meter input using a short piece of copper wire. Then wait a minute until you get stable readings. Then you heat one of the binding posts between two fingers and observe the EMF. If you succeed to heat by 10 K, you may see 5 uV. Now if you let it cool down for another minute and heat the other binding post, the sign will be opposite.
Binding posts, even low EMF ones are usually made from copper alloys easier to machine than the pure electro-copper of wires. That difference times a temperature difference is enough to generate EMF.
For low EMF measurements i often wrap both binding posts with a sock to protect them from air draft. This does not guarantee compensation of EMF, for example if the upper binding post gets more heat from inside the meter. Nanovoltmeters usually have connector systems with a metal shell around both contacts for better temperature equalization.
Another proven approach is soldering copper wires or crimping copper parts to each other in order to avoid the oxide layer. You will normally find that a thin layer of gold does nothing, it the copper to gold and the gold to copper temperature is the same. In general using thick wires is worse than thin wires, as thin wires cause lower temperature differences. But that's all subject to experimentation.

I did some tests a while back to work out the best leads we had. Got some PTFE shielded wire and that made a big difference. I have seen some have a 3d printed cover that goes over the terminals to prevent drafts etc. I guess the best I can hope for at the moment is to keep things as equal as I can and avoid too many different metal contacts being involved.

Yes, your best bet is allowing everything to reach thermal equilibrium, by minimizing draughts and thermal shielding where possible. Ultimately, settling time is inevitable, even though it doesn't make for good 'productivity'. I guess you just have to work around that one by good scheduling - fill in the gaps with the more  mundane handheld stuff.

[donlisms]

If you can reverse the current in your measuring circuit and get another reading, half the difference will effectively cancel the thermal emf.  So… ((sig+emf) - (-sig+emf))/2 = sig.

That’s why 100 year old bridges have a reversing button.  :-)

A “Hamon divider” as in Conrad’s implementation has the compromise (simplification) of being two-wire rather than four-wire, as was/is Hamon’s transfer standard.  Thus not only is thermal EMF a potential issue, but the resistance of the connections and switches when reconfiguring between series and parallel is an issue. Most folks seem to fight for super-low resistance contacts and such to try to get as much out of it as possible. Wenner had his approach with the NBS bridge 10:1 accessory, where he wanted low resistance, but also did a lot of math to deal with the errors. Fluke has their way of looking at it.   

Hamon’s idea was to eliminate the issue with Kelvin connections, which are great at working around unpleasant resistances. The resulting circuit is more complex - it requires tetrajunctions and more connections and low resistance busses and balanced compensation resistors.  But in the end it Works Pretty Good!

[Gyro]

If you can reverse the current in your measuring circuit and get another reading, half the difference will effectively cancel the thermal emf.  So… ((sig+emf) - (-sig+emf))/2 = sig.

That’s why 100 year old bridges have a reversing button.  :-)

Indeed, I recently picked up a Tinsley reversing switch on ebay for just this purpose (reversing one half of bridge network to zero the upper and lower legs of my Hamon divider). I wasn't expecting it to be quite that large (150 x 150mm - they went a bit overboard on the font size!) but it is beautifully built and very low thermal emf...

[Overspeed]

Hello

Nice Tinsley equipment .

A parameter to check is the contact wetting which have a direct effect on contact resistance.

Surfaces roughness and surface cleanness are very important .

Marco reps have built an interesting prototype : Reinventing Rotary Switches for Nanovolt Accuracy Scanner

Regards
OS

[Gyro]

Thank you. Yes I like his video, his stepper driven switch experiments are interesting, even with the present limitations.

I see he has a Tinsley switch at 8:15. Not sure why he taped over the label, maybe to discourage the market. It's good to have his confirmation of performance at a higher resolution than I can manage (100nV). His sample has added thermal isolation on the knob, possibly more important on a single-ended switch than a 'differential' one. The contacts and wipers are identical.

Somewhere on here there are pictures of the Keithley(?) nV meter sprung  copper stud range switch, which doesn't look as if it would be too hard to duplicate.

[Overspeed]

Thank you. Yes I like his video, his stepper driven switch experiments are interesting, even with the present limitations.

I see he has a Tinsley switch at 8:15. Not sure why he taped over the label, maybe to discourage the market. It's good to have his confirmation of performance at a higher resolution than I can manage (100nV). His sample has added thermal isolation on the knob, possibly more important on a single-ended switch than a 'differential' one. The contacts and wipers are identical.

Somewhere on here there are pictures of the Keithley(?) nV meter sprung  copper stud range switch, which doesn't look as if it would be too hard to duplicate.

Hello

Tinsley is not ''rocket science'' but that high quality machining , high quality raw material .... possible to clone with some machining equipment

Regards
OS

[Gyro]

Somewhere on here there are pictures of the Keithley(?) nV meter sprung  copper stud range switch, which doesn't look as if it would be too hard to duplicate.

Ah, here it is, the Keithley 148. Not quite as I remember - only SPST. It's a nice example of low thermal emf circuit construction though...

https://www.eevblog.com/forum/metrology/keithley-148-nanovoltmeter-(a-brief-teardown)/msg1063743/#msg1063743

[Overspeed]

Hello

Interesting crimping on the Keithley 148 and no soldering connection

Regards
OS

[lowimpedance]

If you can reverse the current in your measuring circuit and get another reading, half the difference will effectively cancel the thermal emf.  So… ((sig+emf) - (-sig+emf))/2 = sig.

That’s why 100 year old bridges have a reversing button.  :-)

Indeed, I recently picked up a Tinsley reversing switch on ebay for just this purpose (reversing one half of bridge network to zero the upper and lower legs of my Hamon divider). I wasn't expecting it to be quite that large (150 x 150mm - they went a bit overboard on the font size!) but it is beautifully built and very low thermal emf...

For those interested in some discussion and pics of another vintage reversing switch,  see link to old thread below.

(Apologies to the OP for the diversion !)


https://www.eevblog.com/forum/metrology/switch-yes-but-what-for-!/

Fixed link thanks @Gyro

[EC8010]

Got some PTFE shielded wire and that made a big difference. I have seen some have a 3d printed cover that goes over the terminals to prevent drafts etc. I guess the best I can hope for at the moment is to keep things as equal as I can and avoid too many different metal contacts being involved.

Metal biscuit and tobacco tins are your friend. They shield from drafts and electrostatic hum. And if the test instrument is on the bench, then the test box can be made to plug directly to the instrument's terminals. Hewlett-Packard/Agilent/Keysight had an excellent white paper on LCR testing where they stressed the importance (and difficulty) of moving the measurement interface from the test equipment terminals to the DUT without degradation. Good test jigs with short leads make all the difference in the world.

[Gyro]

For those interested in some discussion and pics of another vintage reversing switch,  see link to old thread below.

(Apologies to the OP for the diversion !)

https://www.eevblog.com/forum/metrology/switch-yes-but-what-for-!/

(Link corrected)

Very nice! I never saw that one. A thing of flexible low emf beauty.