High Precision Resistance Measurement - What would be best to use?

[retroware]

The 1010 is not really built for very precise voltage dividing. Also, the 242 temperature drift adds. So the 1ppm is not a surprise. The 1010 is intended e.g for 1:100 ohms transfers and is pretty good at that, as temporary drifts are compensated out (see theory section of 1010 in user manual). For stable voltage divisions, there are better solutions (720A...). What you can do is to apply the error propagation methods of GUM, to calculate the temperature drift related uncertainty. It would go a little far to go into details how to calculate. In a nutshell, use the divider formula, calculate the weighted drift impacts of each of the contributors and combine this. You could assume a worst case scenario, all resistors on each side of the divider drifting 0.5x 5ppm up, the others down, but thats pessimistic. RSS would be more realistic.

I started down this path in hopes of calibrating my RV622 and RV722 KVDs. The ESI documentation specifically calls out the SR1010 for this purpose. But, as I'm finding out, getting an accurate read of each resistor value is proving to be a challenge.  I might try my hand at estimating the TC contributions using some of the techniques you mentioned.  Maybe things would still be good enough to calibrate a 1ppm linearity KVD such as the RV722.

After acquiring the 622 and 722 and trying to self calibrate them, I broke down and acquired a Fluke 720a which I know I can use to accurately calibrate the 622 and 722. But, I'd still like to see how far I could push this 'old school' method.

The cheapest way to get accurate readings is to buy some PT100 (100 ohms) or PT1000 (1K) Platinum temperature sensors.

Thanks for reminding me that I have such a sensor! Plus I already have a 3456 in the 'Frankenstein' stack just waiting to measure the value.  I think I'll directly mount the sensor to the SR1010 case to see how much temperature variation I'm getting over a period of time.

[Chipguy]

The PT1000 sounds better then the PT100 with the higher resistance, but the PT100 is less susceptible to noise. You can use 1V across the PT1000 (as long as it is in thermal contact) but the lower the voltage the better. If you have 6 digit meter, you could a 100K range to measure a PT1000 just to minimize the power dissipation.

I tried both and found that PT100 works better for me due to higher noise immunity.
Since a PT1000 (the ones I used) should be driven with max. 0.1mA I did not see any advantage compared to the PT100 driven at 1mA.

I have designed a programmable resistance decade that can "output" resistance values from 1 Ohm to 1.6 M Ohm with a resolution of at least 6 digits.
It was calibrated against a Datron 1271 end of 2017 and is currently doing burn in tests since it has been calibrated.
For what I can tell it's current deviation of it's originally calibrated values less than 0.001% (10 PPM).
However I can only do these checks with a DAkks calibrated 34401A. A more reliable result will be revealed once it is going back to the Lab for a meetup with the Datron 1271.

That decade can simulate all sorts of sensors, like PT100, Ni120, YSI400/700.
It can also simualte water/ salt solutions with temperature compensation (using a second programmable decade as slave to make the temperarure sensor simulation)
Currently working on customisation so you can define "My resistive something" using Excel/Calc, outputting the table to CSV, whack the USB stick into the decade box and import it.

So now you know why I am thinking about a high precision resistance meter.
I need something better than the 34401A but do not want to spend around 10K for that.

[retroware]

I'm going to second Edwin Pettis's comment regarding the 242 as being a pretty darn good device for accurate resistance measurements. While I don't have a SR104 to compare against, I did measure the resistance of my SR1010 in series mode against the RS925. The reading was 11ppm off from the expected nominal of 10K. This is for equipment that is probably 30+ years old.  Now of course, the resistors in both the RS925 and the SR1010 could have both drifted in the same direction and rate over that period of time but I'd be willing to bet that all of resistors in the system are still well within the stated 50ppm lifetime drift spec.

This thread also suggests that the RS925 ages well:

https://www.eevblog.com/forum/testgear/esi-rs925a-decade-resistance-standard-mini-teardown

In terms of cost, the two critical components are the RS925 and the 240C. If one is patient, one can probably get the pair for around $500.

[ManateeMafia]

If you are real patient, and quick, you can get a complete 242E for $300. At least that's what I heard.  At that price any one part needs to be functional to cover the cost.

[dacman]

The laboratory I work at was once NVLAP certified using the ESI 242D as an NVLAP standard (although we were also certified using a Guildline bridge and an MI bridge).  Also, several decades in the RS925D have adjustment trimmers for each step.  IET Labs sells an RS925 equivalent.  One use for the RS925D by itself is as an RTD 4-wire simulator.