Hello Radu,
4W Kelvin connection is definitely necessary for a 10kOhm reference standard (as your thread title implies).
Please see any other standard resistor like the FLUKE 742A-10k, or ESI SR104. Latter has an oil bath, a thermometer and a R(T) table built in, so the Fluke is not a really good standard resistor.
How could you otherwise precisely measure this standard resistance with your DMM, if the cables from the DMM to the resistor, and the internal cabling are on the order of several 100 mOhm, or ~10ppm error?
Contact resistances come on top, and these might vary by several 10mOhm, or ~ 1ppm. If you study the schematic of the FLUKE 5450A standard resistors box, providing 2W or 4W resistances from 0 Ohm to 100 MOhm, all resistors from 0 to 10MOhm are 4W connected; only 19MOhm and 100MOhm are 2W.
If you want to use any of these as standards on 1ppm "accuracy" level, you need to be able to measure their resistances with ~ 0.1ppm resolution and repeatability.
With several additional tricks, my 3458A is able to make measurements on resistors with 0.01ppm resolution, but 0.2ppm standard deviation, which is a measure of stability of this measurement.
Both parameters are not specified in the 3458A data sheet. These parameters are also important, if you want to precisely characterize the R(T) curve for later compensation of T.C.
Those binding posts cost on the order of 10$ each, so that makes no sense to try to save money here. The VHP101 might cost 10 times of that.
You also might buy a cheaper type, with higher T.C., due to the mathematical compensation of R(T). But please chose a hermetical sealed, oil filled type for best timely stability.
The VHP202Z was promoted to be stable as 2ppm/6years, which turned out to be a false promise by Vishay Precision, again. I think, they made a mistake during production.
VHP101 might perform much better, what I heard from others. The VHP101 / 40k inside my 3458A is stable to < 2ppm over many years, see diagram.
Yes, I drilled another hole into the aluminum block, so that the head of the NTC (epcos type S863 or eq.) is a few mm near the body of the resistor. I did not drill all the way through to the cavity for the resistor.
I filled the hole with a bit of thermal compound, and I as well applied that between alu block and case.
You might use a PT100 instead, but this requires another 4W connection, using a LEMOSA connector, maybe.
This way, any external temperature variation is instantaneously transferred to thermometer and resistors (because latter is oil filled), so you get a very tight thermal coupling between NTC and resistor, and a certain thermal damping, due to the thermal mass. I always place the box on top of my 3458A, which gives sort of thermal ovenizing effect at ~ 25¡ãC, see picture.
So that looks like a very simple assembly, but it's got plenty of physics tricks included.?
I made a mistake to shorten the legs? of the VHP202Z. You should leave them long, briefly soldering a wire at 10mm distance from the body, and soldering the ends of both wires to each binding post. Wire wrapping is a good idea to make a very stable solder connection.?
Always use a heat tweezer, to avoid inducing any heat hysteresis into the resistor element. The resistor element is of the strain gauge type.
I got the uncertainty (on the order of < 1ppm) of my resistor group from Metrology Meetings or transport to other volt-nuts or the PTB at a Maker Fair, who had an uncertainty of < 0.5ppm.
I'm monitoring the drift of my group for over 10 years, meanwhile, so I can detect all kind of instabilities.
That' also what I recommend, to build at least 3 of these standard resistors to be able to compare them against each other.
You know, a man with one clock knows the time, a man with two clocks is not that sure any more..
Frank
