[time-nuts] Re: Temperature accuracy / repeatability / drift of double ovens

[ghf at hoffmann-hochfrequenz.de]

Am 2023-03-31 20:13, schrieb Ed Marciniak via time-nuts:
> Linear technology made a thermoelectric full H bridge temperature
> controller. At the time, I’d like to say it came as a lot of two
> components. The second was an op-amp that may have been selected or
> trimmed unless there was a good reason to include it with samples
> (maybe less commonly available).
> 
> It was supposed to be able to achieve 10 milliKelvin stability easily,
> and with more care in design, 1 milliKelvin stability was supposed to
> be not too difficult to achieve.

That really gives me some hope, I'll take the care :-)
The question is now how to prove it!


> It was a moderately expensive part…about $20 for the two components. A
> handful of passives and a a pair of N-channel and a pair of P-channel
> MOSFETs were required.
> 
> If memory serves correctly it was also switched mode rather than 
> linear.

Thanks for the pointer, also to Jim Lux.

As a first try, I used the Analog Devices ADN 8834, also from AD, now 
that
they have bought LT. But the Linear Technology LTC1923 data sheet is in
a different league, they even give absolute numbers and strip charts.
I'd bet that this went over the desk of Jim Williams, RIP. Simply top 
notch.
The ADN8843 has the H-Bridge built-in and it worked for someone I know.

I'm not too excited by the cascaded TECs. They can only provide a
temperature difference, and absolute temperature only with a regulation
loop with more time constants between the TEC and the NTC. The low 
thermal
impedance path through the TEC is a real drawback, and probably the
culprit for the low efficiency. No wonder that no one uses them in 
OCXOs.
But 87°C like in my Morion OCXOs is no solution, either, even if we 
could
dim down the absolute temperature.


> From: Bob Camp via time-nuts <time-nuts at lists.febo.com>
> Sent: Thursday, March 30, 2023 6:01:55 PM

> Thermal gain on a single oven is likely in the 300 to 600 range.
> 
> Double oven boosts that by 10 to 30X
> 
> Thermal gain in this case: take the ambient change divide it by the 
> thermal gain
> and you get the oven temperature change.
> 
> Long term stability:
> 
> low enough that it does not impact aging. ( = that’s how you test it).
> Since that gets back
> to the specific thermistor being used, you are into the ‘likely not
> much help” range.
> 
> Best guess info: The crystal + circuit is < 1x10^-9 / C in the
> vicinity of the turn. The operating
> point is offset from turn to minimize the combination of crystal +
> circuit. Just what the net
> is ….

No crystal today, just VCSELs...

> Still, if you can “see” aging at the parts in the (low?) 10^-11 range,
> it sort of kind of gets
> you to a number. Maybe 0.01C / day.


>> Cheers, Gerhard