With all due respect, I don't think that you can characterize the drift and noise in a LTZ1000 -based reference using a multimeter that itself uses a LTZ1000 reference. You will need something several times better than what you are trying to measure. For a LTZ1000 ... well ... that means basically a JJA.

Hello macboy,

that's an interesting issue you critically pose here..., it's a very basic question about any kind of "standards".

Well, I think, you are not right that a JJA or something much better than an LTZ is necessary..

That's the famous problem described by: " A man with one clock knows the time, a man with two clocks is never sure, a man with three clocks..."

OK, let's get "scientific", by shifting your objection one level upward:

How is it in principle possible, to specify the uncertainty or stability of the very best standards, if there do not exist any better ones?

How can metrologists even conclude, that a JJA standard itself is uncertain to better than 1e-16, if there exists no better quantum volt standard?

How can they conclude, that the Cs fountain clock is stable to 1e-15..1e-16, if there exist no better clocks? (OK, today we already know better ones, but how do we judge these, again?)

How did the metrologists find out, that the kilogram prototype in Sèvres, France, thought to be "absolute", is in fact unstable to 2e-8 in 100 years, and that all the other, identically made ones, are consistent?

Very simple, they have always compared two, or more physically identical standards, against each other.

The idea is, that if they have an intrinsic drift or uncertainty, they will presumably drift apart, and the uncertainty is then defined as being about half, or something like sqrt(N) - statistics of the drift against each other.

The more different samples of the standard in question you have, the better the statistics is, and the lower the probability, that the whole entity does not drift in the same direction.

That's the purpose also of ring comparisons, btw.

Therefore, comparing the homemade LTZ1000 reference against the Keithly one, will give a good indication of the (combined) stability, at least.

As the Keithley reference is characterized by its stability specifications, for this case of relative measurements, its 10min and 24h stability, the Keithley instrument has a more profound weighting.

Therefore you can at least judge the DUT exactly by these Keithly specs, or even better, if the drift between both is less than the Keithley spec.

I admit, having 3 references (including the DUT), would allow to judge, which one drifts outside the entity, and 4 or more allow to do a better statistical analysis.

Also, the LTZ1000, like the LTFLU in the 732B, Fluke 57x0 instruments, are known to have tendentiously a negative drift (on the order of less than -0.7ppm/yr.) , see correspondent specifications or analysis. But this is not important in this case of short to mid term drift measurement.

Myself, I have 4 different voltage references, which I monitor for now about 5 years: 2 homemade LTZ1000, one LTZ1000A in the 3458A, and 2 stacked SZA263 in the 5442A.

This whole group did not drift apart more than +/-1 ppm over all these years.

Therefore, I have quite a good evidence, even without the aid of a JJA, that each of my 4 references are stable to a level of <1ppm/yr.

(If I would have access to a JJA, I could determine the absolute values of my references, but that's secondary.)

In the same manner, I was also able to make some short term stability (noise) measurements in relation to my 3458A, which had been specified by HP concerning this parameter.

So I could tell, that the 5442A was a little bit less noisy than the homemade LTZ1000 references, each of them on an order of <0.2ppm, and that an old Fluke 332B was 10 times noisier.

Therefore, it may sound weird, but this bootstrapping procedure, by comparing 2 or more virtually identical references, actually is a well accepted metrological practice to determine (estimate) their stability figures.

Frank