[GPS_Standard] VE2ZAZ

[Dave Platt]

> The system performance you are describing (1/30th to 1/300th of a 
> second) appears to fall short of what most see. you should get better 
> than 1x10e-9, with the mid to low 10e-10's a more common place. I don't 
> know if you meant that your S sampling period is between 8 and 16 
> minutes, but this is likely too short. An hour to a couple of hours is 
> more common on this type of system. Of course, the stability of your 
> OCXO may force you to go with shorter sampling period...

Yes, I agree, it's too short for best accuracy (assuming that the
oscillator is stable).  I'm still in the early phases of getting
the controller settings "tuned in", and figuring out the factors
which limit this particular device's accuracy.  The current settings
favor a relatively rapid ability to track oscillator speed changes,
but very probably result in a system whose accuracy is dominated
by sampling error and GPS jitter.  A significantly longer averaging
period will help reduce this, once I figure out how much the oscillator
itself actually drifts (e.g. with ambient temperature) and how fast
I'll need to allow the controller to adjust.

A few details on my particular implementation.  I'm using an Efratom
oscillator, which has a relatively large adjustment range (around
10 Hz).  I've actually "padded down" the controller's adjustment range,
by using a voltage divider circuit with a separate, manual "coarse
adjust" cermet pot fed from a separate +5VDC regulated reference... I can
center the oscillator's rate with the coarse adjust, leaving between
1 and 2 Hz of pull-range for the controller.  The oscillator is fed via
three separate voltage regulators (two 7812s for the oven and the
oscillator itself, and a 78L05 which regulates down from the oscillator
7812 and feeds the cermet coarse-adjust pot).  The controller board has
its own on-board 7805, of course, and I added a separate 7805 sitting
on top of the output-driver chip (I call it the "mad hatter") to remove
the effect of output-load power variations from the controller's main
7805 (which provides the DAC reference voltage).  The GPS and dual-port
Maxim serial-port chip are fed from another 7805... and all of the
7805s are fed from a 7809 pre-regulator mounted at the power-entry
panel.  Lots of regulators :-)

Over a period of a couple of weeks, I recorded the controller's
output reports, and ran 'em through some PERL scripts and then
into GNUPlot.  You can see the result at

  http://www.radagast.org/~dplatt/hamradio/fll.pdf

The interval from 3/31 through 4/4 was the initial turn-on
and lock-acquisition period.  I had set the coarse-adjust pot
using a decent HP counter as reference, and with a short averaging
period (a minute or so) the controller had indicated a lock, but
it doesn't seem to have actually approached stability until
around 4/2 or so (where the curve briefly flattens out).  On
4/4 I manually tuned the controller to its mid-range,
tweaked the coarse-adjust pot until the controller re-established
a short-period lock, and then let the controller run (with an
8-minute averaging period).

The results are interesting.  For about a week thereafter, there's
a very pronounced 24-hour tracking cycle visible... which then
largely disappears.  The box was sitting in my unheated garage,
which goes through fairly pronounced temperature changes in good
weather (probably close to 30 degrees F on those days).  Rainy
weather arrived during the second week of April, with much more
constant temperatures, and you can see the effect on the
system's frequency stability!

My next step (after finishing up the box bezel) is going to be
to try to reduce the standard's temperature sensitivity.  Although
the oscillator itself is in an oven, the other circuitry is
probably temperature-sensitive to some degree - in particular I
suspect that the various voltage regulators may be drifting a bit.

I'll probably try sealing up the vent holes in the cabinet I've
used, perhaps adding some insulation or thermal mass, and perhaps
adding a sort of auxiliary-oven temperature regulator which would
keep the entire interior of the cabinet at a more constant
temperature.  Simply moving the standard into the interior of
the house would help, of course, but it's more convenient to keep
it in the garage/shop.

If I can get the temperature sensitivity problem licked, one way
or another, I can then use a longer averaging period and gain
improved accuracy.