In message CAKn+a3sNg3fE-C8=g-_UTwj3ibmeujGVPfApV5yvR7cNN8FqPw@mail.gmail.com, Mark Goldberg writes:

I researched this a bit a couple of years ago in relation to HP5065 C-field tempco

The major tempco in pots are mechanical in nature.

Plastics have horrid tempcos, most in the hundreds of
PPM and the best (nylon) barely making it under 20 PPM.

Mind you, that is usually measured on relatively large linear
extrusions, not small bits of geometrically complicated plastic,
like you would use to encapsulate a trimpot.

There are trimpots on the market which claim 5PPM ratio stability,
but the conditions under which that is measured are not very easy
to implement in practice.

If you want anything close to 1PPM trimpots, hunt eBay for "ESI dekapot"

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Mark is correct, but with a caveat:  Unless the pot slider sees a load
impedance that
is much much larger than the pot's end-to-end resistance, contact resistance
variations can also play a big role, especially when the pot gets old.  For
this reason
alone I favor sticking with the 3-terminal "ratiometric" configuration, AND
using a
high-Z buffer amplifier between the pot slider and the load whenever
appropriate.

The downsides are noise and drift contributions from the amplifier, but
these can often
be held to acceptable levels with careful selection of the amplifier type
and model.

It all depends on the specific situation.  But in any event, I cringe when
I see a design
using a pot as a 2-terminal variable resistor, especially in situations
when an open
condition could cause damage.

Dana

On Sun, Dec 24, 2017 at 2:57 AM, Mark Goldberg marklgoldberg@gmail.com
wrote:

Hi

The other issue with putting in an amp is noise. You really do not want to feed
a bunch of added “stuff” into the FM modulation port on the oscillator. You get
some thermal noise from the pot, so it is never zero. How close you need to get
to zero - who knows? When the part was tested to see if it met spec, it probably
didn’t have an amp in there …..

This of course all assumes a stand alone device like the Pulse Puppy. Once you
go to a structure like a GPSDO, things inevitably get a bit more complicated. That
is one of the many reasons testing this and that on a servo’d system is part of the
normal design process.

Bob

I'm glad that the PulsePuppy post spawned some good discussion!

The pot I'm using is a Bournes 3296W-1-103LF which is a 25-turn, 10K,
cermet pot, spec'd at 100ppm/degree, so it's not anything super fancy.
The number of turns provides decent setability, and it seems to be a
good match for the class of oscillator the PulsePuppy is designed for --
I found that I could trim the Isotemp oscillator without problems.

I'll admit up front that the PulsePuppy wasn't designed as a host for
ultra-stable oscillators.  I tried to keep the circuit board size and
cost down, as well as the number of components that users would have to
install.  And since the EFC trimmer is one of those user-installed
parts, it's possible to substitute as nice a pot as you'd like.  I have
one unit that's hooked up to an external precision pot with a
turn-counter dial (just because it was there).

John

On 12/24/2017 03:57 AM, Mark Goldberg wrote:

John,

Do you notice a backlash effect when homing in on the desired setting with
those
tripots?  I last used such things back in the 1980's, and remember often
having
enough backlash to make close trimming rather difficult.

I wonder if they have gotten better in that regard.

Dana

On Sun, Dec 24, 2017 at 9:58 AM, John Ackermann N8UR jra@febo.com wrote:

Hi

The nice thing is that the footprint on the Pulse Puppy is pretty generic.
There are a lot of OCXO’s on eBay that will fit the pattern on the board.
When you get into the more exotic parts, the footprints (and functions)
don’t tend to be as standardized.

The small package OCXO’s will always have an issue with drafts and
temperature stability. There just isn’t much room in there for a fancy
stuff in there. Putting a bunch of baffling around them is problematic.
You can quickly bump up the thermal gain and likely destabilize the control
loop ….

Lots of fun …

Bob

I didn't really notice much backlash, though when setting oscillators I try to approach (slowly) from one direction until it's "good enough" and then stop, to avoid that problem.

On Dec 24, 2017, 11:28 AM, at 11:28 AM, Dana Whitlow k8yumdoober@gmail.com wrote:

Hi

With these little OCXO’s you likely will move them 0.1 ppb simply walking
past the bench. If they have a total of 0.25 ppm of trim, a 25 turn pot will get
you 1 ppb per turn. Setting these pots to 1/10 turn is not all that crazy.

Yes, there are a bunch of assumptions made there and your setup may or
may not behave quite like the one I have described.

Bob

I am using a Bournes 3224 and I do see backlash issues. I do come from one
direction to set it and if I overshoot, I go way past and come back from
the other direction. I also see non-monotonic sections. If one of those is
where you want to set the frequency, it is pretty hard to do. I chose one
with significantly lower impedance than the input impedance of the TCXO
control port. The 3296 datasheet has Adjustability specs and the 3224 does
not. The 3269 is only 12 turns but does have an Adjustability spec on the
datasheet. Maybe I will consider that. I would have to change my board to
use a throughole part.

Thanks for the info.

Mark

On Sun, Dec 24, 2017 at 8:58 AM, John Ackermann N8UR jra@febo.com wrote: