Keep in mind that potentiometer R6 is one of the key components of the
C-field circuit.
Together with R8 and R12, it defines the reference voltage to Q6A/B.
The tempco of that wire-wound potentiometer can certainly not be neglected.

I would suggest to check the C-field temperature dependency with an
external precision current source feeding the C-coil directly.

Adrian

Poul-Henning Kamp schrieb:

In message 55E1ADA8.2000102@arcor.de, Adrian writes:

Absolutely not, it's clearly part of the picture, but something has
10x the tempco of that one right now.

That's what I'm doing right now: C-field driven by Fluke 732A
through 2.5k resistor.

--
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.

Hi

Most potentiometers have very good TC when used in a “ratio mode”.
You can quickly mess that up by putting resistance in series with
the pot. It’s a bit surprising how may times you see this inherent stability
messed up that way.

Bob

In message 8BE4F2B1-171B-4117-88A5-A6DC64D9890D@n1k.org, Bob Camp writes:

Yes, but they do have a good reason for it here:  They want the
C-field pot to be "calibrated" in frequency so they need the
resulting voltage to follow the quadratic C-field sensitivity
decently well.

I still havn't figured out what component is responsible for the
huge tempco, but it is not the C-field pot.  There's a few PPM
there but not that many.

--
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.

Bob,

agreed. But in the 5065A C-field circuit, the pot is wired as a variable
resistor.
The Bourns data sheets specifies +/- 50 ppm / deg C max.
Together with the two series resistors, the effective TC of the pot
drops to about 20 ppm.
That is indeed just 2 times more than the effective temperature
coefficient of the reference diode (1N938).

Btw. the 1N938 is running at about 7.5 mA +/- 1 mA, not at 12 mA as
mentioned earlier.
Some 3.5 to 5.5 mA are going through R8, R6, R12.
To get the correct diode current, this current has to be subtractrd from
the total current through R7 (Q5, Q6 base currents neglected).

Adrian

Bob Camp schrieb:

Hi

Another counter argument against the “ratio mode” is that with zero current
through the wiper, odd things can happen. There is always something
else to consider in any real design.

Since they needed to calibrate the voltage swing, there aren’t a lot of options
with the technology they were using at the time. Today there are a lot more
choices of how to get this sort of job done.

Bob

In message 1698D85B-EBB6-45E3-9CB0-CBF780CE5550@n1k.org, Bob Camp writes:

It's not just the swing, it's also the shape of the curve:

http://phk.freebsd.dk/hacks/HP5065A/20150828_c_pot/index.html

If it were just the range things would be a lot simpler.

--
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.

Poul-Henning,

On 08/30/2015 09:54 PM, Poul-Henning Kamp wrote:

Ehm, eh... that transistor pair you have there. How tight together is
really the transistors thermal connection? I bet not all that good.

The reason I got involved with counters and atomic references was
originally my interest in analog synthesizers, and there we use a
transistor pair for exponential conversion, which has scale and offset
issues and thermal issues. The use of a Q81 +3300 ppm/C resistor in the
division network helped to compensate the thermal properties of that
transistor pair, and you wanted stuff like MAT-01 where the two
transistors is thermally ties to each other and the put your tempco
resistor to that for good performance. All this requires good
measurement and good reference, so that what motivated me towards that
step. Anyway, the take-away is that you should look at that discrete
op-amp and see if it is not causing you the thermal dependence you are
trying to locate. Maybe replace it with a more modern op-amp like the
741 or something (irony may have been used).

Cheers,
Magnus

Hi

As we go joyously bashing the poor guys that designed this beast, it’s worth noting
just how old the design is. 741 op amps were indeed “modern” when they did much
of this and quite possibly to modern to be trusted. Most of the design would have been
right at home in the late 1960’s at a conservative design house. As time has shown, in
a lot of cases that mistrust of the early linear stuff was well founded ….The 741 only
was designed in 1968….The 5065 design dates to roughly that time.

Bob