[time-nuts] Oscillator questions

[Robert Watzlavick]

Hi,
I'm new to the list and I'd appreciate some advice regarding my
application.  One of my hobbies is building and testing liquid fuel
rocket engines (http://www.watzlavick.com/robert/rocket).  As part of my
instrumentation system, I have some Datum 9550 Video Time Overlay units
which stamp time from IRIG-B onto NTSC video from various cameras
positioned around the static test stand.  I also have a Datum bc635PCI
card which generates the IRIG-B signal and outputs a TTL start pulse to
synchronize the data acquisition system so I can associate a particular
video frame with the measured data.  The data system uses National
Instruments hardware (PXI-6030E, SCXI-1520) and can accept a start
trigger and a reference clock.  I discovered that the accuracy of the
built-in 10 MHz reference clock in the PXI chassis is sufficiently
different from the flywheeling bc635 card that even over a few minutes,
they can drift by up to tens of milliseconds.  That sort of negates the
goal of lining up the video with the data so I routed the 10 MHz out
from the bc635 into the PXI chassis CLKIN to lock them together.  That
seems to work fine as even after several hours, the data and video are
still synchronized.  When I did that experiment, I didn't have any
"good" 10 MHz reference sources to compare against other than the PXI
chassis which is rated at +/- 25 ppm and 5 ps RMS of jitter.  The bc635
has a VCXO which probably isn't that good either.  I looked at the 10
MHz out from the bc635 and it looked pretty clean and stable so I
assumed it was good enough.

A few months ago, I decided I needed a better 10 MHz reference source
because I was trying to calibrate some pressure transducers against a
Paroscientific DigiQuartz which has a frequency output as a function of
pressure.  On a related note, I had always wanted a GPS-based NTP server
of my own (just because!) so I picked up a Datum TymServe TS2100 w/GPS
from eBay.  As an added bonus, it has a GPS-synchronized IRIG-B output I
could use for the rocket testing and a disciplined 10 MHz I could use
with my HP5316A and the Paroscientific.  All good, or so I thought,
until I compared the 10 MHz out from the TS2100 to the OCXO in the
5316A.  When locked to GPS, the TS2100 jumps all over the place relative
to the 5316A.  I know it's the disciplining algorithm because when I
turn it off, it's actually pretty stable compared to the 5316A.  I see
the same thing when I compare the ref clock out from the bc635 to the
5316A on a scope - it seems to be correcting once a second but the phase
is all over the place, sometimes moving 2-3 cycles before it reverses. 
I'm not sure how to quantify it but I'm wondering whether the jitter
from the TS2100 or bc635 will be enough to screw up the reference clock
on the PXI chassis for the data system.  The data sheet on the PXI-1042
chassis says it uses a PLL to phase lock to an incoming clock so maybe
it dampens out the jitter and it's not a problem.

But, I'm the type that likes to look at the details so I bought a
Thunderbolt from eBay and I was able to confirm it is definitely the
TS2100 and bc635 that has the issue.  No amount of playing with the gain
and filter settings in these units cuts down on the phase shifts so I'm
guessing it's a byproduct of the cheap VCXO they use.  I was thinking of
upgrading both of those units to an OCXO to improve the stability, an
MTI-240 for the TS2100 and an MTI-210 for the bc635.  The boards appear
to be pin compatible with these units so they should just work.  I have
the schematic for the bc635 and it looks straightforward to change it
out but on both units I'll have to reprogram the OCXO gain settings each
power cycle because I can't figure out how to make them stick in the
EEPROM.  I've talked to the MTI rep and he is going to see if he can
waive the $1500 min order requirement for me.  No idea how much the
oscillators will cost, probably more than I want to pay.  Has anybody
successfully upgraded the oscillators in these units?

For now, running the entire data system locked to GPS isn't necessarily
a requirement (cool but not really needed) so I could use the 10 MHz
from the Thunderbolt to run into the bc635 and on to the PXI chassis. 
I'd just have to manually set the time in the bc635 and let it
flywheel.  Another option is to jam-sync the bc635 to GPS and then let
it flywheel during the test so it's 10 MHz output doesn't jump all over
the place.

Sorry for the long post but I'd appreciate any comments or advice you
could offer an apprentice time-nut.

Thanks
-Bob
K5RLW