Cellular / Andreas
Not sure if Andreas is coupling the resistors -directly- to the LTC2400 ADC input (if that's what is in use) - but be aware of the warnings on pp22~24 on the datasheet. You'll have a source impedance around 1k, plus capacitance plus inductance + thermal flow issues on those long resistor leads waving in the wind. Not the most accurate setup. Realize when the datasheet is talking about capacitance issues that also means "impedance" (as in all LCR) so be careful of that - even a few nH can result in detectable ppm shifts depending on how the circuit is setup. The input of any of the LT switched capacitor devices is far from DC. Always keep that in mind. The datasheet is a bit optimistic when it is telling you an input impedance of 10k might be OK...then again it might not if you've got a bit more LC in your circuit than you thought.
Not to mention the variable TC of the input protection diode leakage of the ADC - which is also going to add data you weren't counting on when you think you're looking for ppm-level changes.
If Andrea's test is using a ppm-accurate buffer on his resistor checker, then it might be less of a concern.
NOW: Realize your LTZ ratio resistor is going to be very out of balance in terms of heat dissipation between the high and low resistors. Yes the power is small number, but the ratio changes you're looking for are even smaller - and so most of the resistor ratio performance is generally dictated by how it's mounted mechanically to the board and thermal flow into / out of the leads. Having a matching TC ratio on both resistors helps but that's not the whole story because of mis-matched power dissipation. You can try using something like 14ea. 1k resistors to develop the correct 13:1 ratio, but that builds in a much larger current loop area for noise injection. Etc. No perfect solution I'm afraid.
Here's a suggestion for more accurately evaluating resistor performance for LTZ heat ratio resistors - and because you're happy to spin boards
:
Build a simple LTZ circuit with a few different sets of heater ratio resistors on the -same- single board. You want to be able to jumper the sets into the same LTZ circuit for comparison of performance (without soldering). The concept here is the resistors are
under real Vbias and are mounted a realistic distance from the LTZ with a somewhat realistic
thermal flow in and out of the resistors. It won't be 100% perfect but it will be a close approximation to realistic use condition.
Jumper in you various resistor sets for a week or month at a time and watch LTZ performance.
See if you detect any difference in Vref TC or general LTZ performance as you change resistor sets. You will see a slight change in -absolute value- of Vref output when you change resistor sets, but you're looking for any difference of TC and noise performance of the LTZ.
As long as you're using somewhat decent ratio resistors you'll probably find no huge (or even detectable) difference between a couple QUALITY $6 PWW or expensive Vishay Voodoo Magicals. Even if you've got a JVS or multiple 732's on hand to really <accurately> compare ratio resistor performance on the -same- LTZ circuit (where the only thing changing is the ratio resistors) - it's hard to see any difference, really. Sometimes you can spot a really crappy ratio resistor set... But most of the better resistors we've checked makes no huge difference either way.
That's the real beauty of the LTZ circuit!
My advice is to don't over-spend on LTZ resistors - because it may not make any difference to your lab equipment and your measuring ability. Remember that even multiple 732's on hand will get you down to a few ppm measuring of absolute value of a Vref, and anything below that is just pure uncertainty - if an -accurate and traceable- absolute Vref value is important to you.