Hello DiligentMind, nice thoughts!
Just turning off the 3458A, and waiting 1 day, then turning it back on (and waiting 4 hours), then doing an auto-cal, I noticed that the reading at 10V can be different by as much as +/-0.2ppm.
I see the same fluctuations of 0.2ppm in the monthly measurements of 3458A vs, 5442A, as I also switch them off. (Energy is very expensive here in Germany).
That's perhaps not a pure hysteresis effect, as the difference drifts forth and back.
Pure hysteresis would add up or stick to fixed output values, as both oven see two different temperatures only, i.e. RT and 65°C only.
It is much more difficult to get the last 2 digits to behave well, but the next to the last digit can be made to be very stable (long term and through power cycles) if the room temperature is kept at 23C +/-0.5C, and there is at least a 4-hour warm-up period before I auto-cal. So, with care, 7.5 stable digits are possible-- the last digit is always bobbling around-- and I have an idea that it is DC-10Hz noise from the 732B (combined with the DC-10Hz noise of the DMM).
My measurements (10min stability) of the 5442A and of the DYI LTZ1000 also show jitter on the last digit only, i.e. on the order of 0.01ppm, @NPLC100, const. temp.
I also relate that to the Zener noise only, after I improved the circuitry by better PSU and shielding. The 3458A itself should be more stable / less noisy than that.
That in itself makes me confident that if the DMM manufacturers put a carefully designed LTZ1000A based reference in their 6.5 digit DMM's, then they could be made stable to 1ppm per year (ie., the last digit would change no more than +/-1 per year) on their best DC accuracy figures. I would pay a lot of extra money (up to $500 extra) to have that in a 6.5 digit DMM.
Well, I prefer the LTZ1000 @ 45°C, and this reference can be built much better than HP by yourself for about 100$.
Indeed, using highly selected components together with the special hand tuned wire-wound resistors in the boost circuit of the 732B are key to the unit's stability. I'm not certain that these techniques could be applied by the home experimenter-- you need a vast inventory just for the selection process, and this can become very expensive very quickly. ...
To my opinion and experience, instead of hand made / selected wire wound resistors, those hermetical oil filled VHP201Z will do the same job, perhaps even more stable.
Note that resistors will have double the drift rate for each 10C rise in temperature-- so keeping them cool (but at a constant temperature) is best. How cool? I don't know-- but I imagine that there is a point of diminishing returns-- so there is probably an optimum temperature to keep all of the circuitry at.
Room temperature is sufficient for 2ppm/6yrs for the VHP201Z resistors.
And that level of stability is over sufficient for the LTZ1000 circuitry.
Any heating or cooling will lead to more problems.
Another problem is temperature hysteresis-- and this occurs in the LTZ1000, as well as any op-amps that you are using-- and as Dr. Frank has proven-- in foil resistors too. (I'm not certain what hysteresis effects there are in wire-wound resistors, but I am willing to bet a premium beer that they at least have some hysteresis effects).
The hysteresis effect of the metal foil resistors is relatively low (5ppm after 125°C), and due to the 100:1 attenuation effect, does not play a big role.
On +/- 15K, there is no appreciable hysteresis, so if you take care, there's no problem at all.
Afaik, the wire wound types have neglect-able hysteresis effect, only the tightness of the winding has to be relaxed first.
Some time, we all should meet for a beer, anyhow.
The only way I can see around this is to run the LTZ1000 at cryogenic temperatures-- like in liquid nitrogen (LN2).
Will cause problems due to the thermal shock, and will stress the tightness of the package.
Also, as the internal gas will dilute and create a vaccum, the package might implode, or will get a leak and suck LN2 inside.
This would result in an absolutely stable temperature, because the LN2 always boils at a specific temperature--
No, that's not really the case. (I worked with cryogenic liquids for years, and was also responsible for air and nitrogen liquefaction in our institute for some time).
LN2 will quickly soak oxygen from the surrounding air, so the boiling point will rise from 77.2K (or so) to above 80K within a few hours, depending on the amount of liquid you use.
I used LN2 as a reference bath for a low temp thermo couple, and it was important to always use 'fresh' LN2 from the Philips liquefier.
1 l of LN2 used to cost around 50 Cents, or so, LHe4 was 10-20 times more expensive.
A usual 50l cryostate (Dewar) will cost 500-1k $/€ only, but a complete JJ is how much? 100k, I think.
A lot of first class calibrations can be paid from that.
I think, the stability of the LTZ1000 circuitry at RT is fully sufficient, and putting it completely in an oven or a crogenic bath will only cause problems and complications
Chose those ultrastable external components instead. Compared to that LN2 steadily boiling off, it is a one time investment only.
Frank
Home
Help
Search
Login
Register
Topic: Ultra Precision Reference LTZ1000 (Read 1345571 times)