Thank you very much for all the feedback and the input.
I thought a lot about the comments and that causes one question, please help me with that:
What is the real advantage using binding posts (Pomona, whatever...) compared to massive contacts made from pure copper (like in "my" design) and using small 4 mm connectors (like Multi Contact LS4). I think the contact resistance could be neglected when using four-wire measurement, therefore I see the only improvement in fresh metal surfaces which cause less errors due to thermal voltages. But as far as I understand these errors could be minimized by cleaning the contacts and by minimizing the temperature differences between the contacts?
The idea to use home-made copper contact-feedthroughs is to achieve a semi-hemetic casing. I assume that standard binding posts will not be that gas-tight. Nevertheless it is not complicated to design another type of PTFE-copper-feedtrough-connector which allows 4 mm-plugs and 6 mm spade lugs and make a new casing or exchange the connectors of an existing casing.
Regarding machinability of copper: I found no problems during machining when using sharp tools (good fresh drillbits, fresh polished carbide inserts for turning). Therefore I assume that the usage of other copper(-tellurium) alloys will be interesting for real production scale but I think it is not necessary for home-built devices.
And I would like to add a remark regarding the temperature sensor: I decided not to incorporate a temperature sensor into the chamber in order to avoid any additional feedthroughs into the casing. Therefore I made the boring into the casing which ends directly under the resistor. By this I can measure the temperature in the neighbourhood of the resistor in a distance of approximately 2 to 3 mm, separated by aluminium. And by this I can use the same (at-home-calibrated) temperature sensor for different measurement setups and different resistors.
In the meantime I observed some miraculous multiplication... Now there are three additional casings (being only slightly modified):
#2: the "old" prototype (equipped with Dale RH25-10k)
#7: equipped with Vishay (Bedek) S102C 10k 0,01% with a small bag of desiccant (aprox. 1 gram. With this setup I should get a feeling for the performance of the sealing and the feedtroughs as the resistor is very sensitive towards humidity. Perhaps somewhat higher stability vs. time compared to the RH-resistors?)
#8: equipped with Dale RH25-10k 1% (as a comparison to #2)
#6: equipped with Dale RH50-1k 1% (as a comparison with another size and value and to "make" 1 mA or 10 mA more easily)
I did not ignore your comments and I thought about the comments carefully. I keep them in mind for the future but now I was too impatient to have a look on the performance of this design. I am especially interested in the behaviour of the Dale-resistors and the performance of the feedthroughs and the O-ring-sealing. While observing this I will think about the "improved next generation" of resistor casings. And: it is no big effort to substitute a RH25 by a better resistor if I find one.
Temperature coefficient:
For characterization I measured the temperature coefficient of the resistors (3456A, 4W, OCOMP, 100NPLC, temperature measurement using R&S UDL-45 with a home-calibrated Pt-100) as described above. And I like the results:
#2: Dale RH25-10k t.c. = +3,85 ppm/K
#7: Vishay S102C 10k t.c. = -1,91 ppm/K
#8: Dale RH25-10k 1% t.c. = +3,87 ppm/K
#6: Dale RH50-1k 1% t.c. = -0,73 ppm/K
Thermal voltage:
As a rough estimation (not a real measurement!) to study the influence of temperature towards thermal voltage I made a small quick-and-dirty test setup:
- setting the temperature with the peltier-device,
- isolation of the resistor using a towel wrapped around the resistor,
- measuring the temperature in the temperature-sensor-boring (as described above) with a home-calibrated Pt-100 with the R&S UDL-45,
- control of temperature stability with a thermistor connected to the 3456A obtaining a resolution of 1 mK,
- measurement of thermal voltage at the sense connectors of the resistors using a Keithley 2182 nanovoltmeter connected using a home-made cable (home-made connector, Cordial CMTOP222 low-noise microphone cable, Multi Contact LS4 plugs, soldered with Sn-Ag)
The results: The values are the values as indicated at the display of the meter. They are not corrected in any way (e. g. influence of the input bias current (spec: < 50 pA)).
All resistors show a "large" thermal voltage of up to 0,5 to 0,6 µV during heating. When reaching a "thermal equilibrium" (observed reading the display of the 3456A, being "constant" for some seconds) the thermal voltage reaches a "constant" value very quickly (within approx. 20 s).
#8 shows large noise in thermal voltage (and #8 shows a large standard deviation during resistance measurement, so this is a good candidate to be substituted by a better resistor).
But all resistors show a surprisingly low thermal voltage effect (what went wrong?). Therefore I assume that the design of the connectors seems to be not that bad.
The next steps:
I think I can (re)start now to monitor the resistors as good as I can. At the moment I have no better equipment than some 6,5 digit multimeters and no stable current/voltage source unfortunately. So I have to improve my equipment and I have to learn how to measure and compare these resistors. But this will take some time.
Best regards
Marcus