.. _hp5065_tempco:

HP5065A Temperature Sensitivity
===============================

I found a HP5065A on eBay recently, and have started playing with it.

It is evidently quite sensitive to temperature, so I set out to find
out why that might be.

My first idea was the C-field, if the current driving that isn't incredibly
stable, the frequency will obviously vary.

My first experiment was to measure the voltage over the current sensing
resistor (A15R10 & A15R11) with my HP3458A, and bingo!  It was not stable.

What's more, the long term MVAR floor I had measured could be
entirely explained by the C-field current noise I had measured.

I decided to try to get all the relevant data in one go and hooked my
HP34972A up with four 10k NTC termistors and six shielded cables:

.. image:: DSC_0002.JPG

The termistors (twisted pair wire) measure the temperature of the OCXO,
the Rb physics package, the C-field pot and the A15CR5 zener diode.

.. image:: DSC_0001.JPG

The six shielded wires probe the A15 assembly as follows:

.. image:: schem.png

* The +20V supply voltage (top of R7)
* The Zener voltage (CR5)
* The high and low connections to the C-field coil ("1" and "9")
* The high and low connections to the C-field pot (two wires at bottom right)

Result:  The zener is temperature sensitive
-------------------------------------------

The A15CR5 ener diode is temperature sensitive at the 20 PPM/C level:

.. image:: plot1.svg

This is not good, that zener is supposed to keep everything rock-stable.

It is a 1N938, which was pretty much state of the art back when the HP5065A
was designed, but the component list does not indicate if it was a selected
part or just run-of-the-mill.

The 1N938 is driven with almost 12mA ((20-9V) / 925R) in the schematic
above, and that may not be a good idea.

Later data sheets for 1N938 list 7.5 mA as the optimal drive current for
zero temperature coefficient:

.. image:: tempco.png

An obvious experiment would be to put a 2K trimmer in series with R7.

This also explains why it takes about an hour for my HP5065A to stabilize:
the zener needs to reach a stable(-ish) temperature.

Result:  The current regulator isn't very good
----------------------------------------------

When the HP5065A is switch to "LOCK" mode, the +20V supply drops a
millivolt or so, and disconcertingly the voltage over the R10||R11
C-field current sensing resistors do as well:

.. image:: plot2.svg

I switched in and out of "LOCK" mode a number of times, and finally
left it locked from 1000 seconds onwards

I am not entirely sure what's going on here, but my theory is that
Q7 simply isn't up to the job.

The voltage across the C-field coil is only 56 mV which rides on
top of 3.2 V across R10||R11.

This leaves 16+ Volts for Q7 to dispose of, and it looks like the
noise from the +20V supply pretty much leaks straight through.

I have not done the math on this one yet, it may be below the
radar, but I think not.

Possible remedies would be to put a pre-regulator in front of the
current driver.

More radically one could design an entirely new C-field driver
using a dedicated voltage reference.  A good low noise op-amp
can easily drive the C-field coil directly.

Conclusion - so far
-------------------

At least for my HP5065A, it looks like the MVAR floor comes from
temperature sensitivity in A15.

I'll leave the data collection running overnight and as time permits
I will mine for further insights, and add them to this page.

*phk*