[time-nuts] HP5065A status so far

[Poul-Henning Kamp]

I have had to move the HP5065A aside for now, my work needs the
table-space, and this may be some weeks or months before I can
play seriously with it again.

Here is a summary:

A15 (PSU + C-field) is a significant source of temperature and
supply voltage sensitivity.

One simple way to improve on this, is to put a Traco TEN60-WIN
24V output DC/DC converter on A18, so that the two diodes feed
the DC/DC converter and the DC/DC converter feeds a stable 24V
to the rest of the instrument.

It would be trivial to lay out a PCB for this, but it can also be
done by drilling holes in the existing A18 and birds-nest'ing the
DC/DC converter.  (This only works on the non-battery option units,
where A18 just contains two diodes.)

Amongs the benefits:

* It reduces power consumption by 5-10Watts, because the the +20V
  supply just needs to shave 4V down from 24V, rather then 10V down
  from 30V.

* It isolates the DC supply to the ovens and the +/- 20V
  supplies from grid/EXTDC voltage changes.
  
* And thus it lowers and stabilizes the temperature on/under A15,
  which improves stability.

Improving the stability of A15 further is worthwhile, but more
involved.  The various parts of the board seem admirably balanced
and improving just one part will not help much, full a board redesign
is called for.

If you have the older model where the AC supply rectifier bridge
is on A15, then replace that with a bridge on the chassis somewhere.
(Or ditch the entire AC supply entirely, once you have the
24V DC/DC converter in.)

My suggestion is to start with a REF102CP 10V reference, but pick
your poison.

10V is a sweetspot voltage, because the voltage dividers from 20V
become 1:1

I'd produce the +20V supply with a LT317 and a LT1007 opamp using
a 10K:10K voltage divider.

Finding a good place to put the LM317 is tricky, it needs to dump
some watts of heat.  Removing the TO4 transistor to make space for
the LM317 may be a nasty bit of surgery.

The -20V needs only ~30mA.  I'd produce it with a 1W 24V DC/DC
converter, a LM337 and a LT1007.

The C-field circuit is more involved.

Keeping the C-field pot will cost on temperature sensitivity,
if nothing else because the pot itself is sensitive.

Replacing the discrete current driver with a good low-offset
opamp is worthwhile if you want to keep the C-pot.

Personally I'm tempted to go digital: A 20bit ADC over a
really good current sense resistor, a 16 DAC + a good
opamp to drive the current and a microcontroller to close
the loop.  This would also pave the way for a GPSRB solution.

I'll probably attempt to design a new A15 PCB based on that
once I resume the project.

Both the AC amplifier and the Integrator can be improved,
but I have no measurement results there yet.

The blow by blow can be found here:

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

Poul-Henning

-- 
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.