[time-nuts] HP5065A C-field current is temperature sensitive

[Poul-Henning Kamp]

I have been measuring the C-field current of my HP5065 for a couple
of days using my HP3458A.

To be more precise, I have measured the voltage across the two
parallel current sensing resistors A15R10 and A15R11.

I have not set up a precise temperature measurement for this
experiment, but eyeballing room monitoring, I find a close correlation
of approx 0.4 millivolt/C.

I also find that the phase-offset relative to a GPSDO correlates
significantely with the voltage measured.

If we assume that R10||R11 are not the cause of this, it corresponds
to 0.5 microamp/C sensitivity for the C-field current, and thus
C-field.

The total adjustment range of the C-field pot is 2e-9 which changes
the C-field current from 2.5 to 6 mA (ref: pg 8-48).

That means the C-field sensitivity is 5.7e-7/Ampere.

My 0.5 microamp/C therefore corresponds to 3e-13/C

...which is pretty much the MVAR floor in my measurements.

It seems plausible that a better C-field driver could improve the
stability.

Pretty much all the components in circuit could be causing this.

The ultimate reference for the C-field current is A15CR5, a 1N938
temperature compensated zener which is also the reference for the
+20V supply.

In the HP5065 the 1N938 is driven at approx 12mA, where later
datasheets indicates that 7.5mA is optimal for TempCo purposes.

Once the present measurement run is over, I'll do a run where I
measure the A15CR5 voltage directly, and if I can arrange it, also
with a spot temperature measurement.

If A15CR5 is the culprit, the next obvious step is to change A15R7
to 1500 Ohm, and see if that improves stability.

Another obvious experiment is to drive the C-field with a very
stable external supply, and see what that does for the MVAR.


-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at 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.