[time-nuts] FE-5680A tuning vs resonant peaks

[Jeff Woolsey]

> Hmmm...
> You mention varying suply voltage by 1.5V, but from where as a starting point?

Sorry, that is a red herring.  The regulator on the support board needs
enough headroom to regulate to 15V for the FE5680A.  Thus anything less
than about 17V input will drag down the voltage into the unit, and the
frequency rises. I got down to 14V--the unit remains locked but 10MHz
doesn't.  (It also doesn't help that the voltage display on the power
supply is about half a volt off.)  On the other hand, the regulator has
a high-temp cutoff which I managed to hit at around 20V--there is no
real heat sink on the support card.  Usually it's running around 80C;
cutoff at 125C.


> It's been a while since I calibrated an FE5680A but looking back through my notes, doing it "properly" is, or was for me anyway, a non trivial exercise.I'm not familiar with the Windows software mentioned but the approach I remember was first to determine what, if any, offset was programmed into the unit as received,


The software I'm using doesn't seem to be able to read and display the
previous offset....  Sigh.


> then to measure the frequency of the unit as received, then to calibrate the tuning itself by setting positive and negative tuning extremes and measuring the frequency range before interpolating to find an initial tuning word, followed by calculating a further approximation, and so on and so fifth, and of course eventually programming the FE5680A accordingly.


Which is what I should be doing instead of winging it....


> At that time I was using a similar test setup of a 53132A referenced to a Thunderbolt, although I did use a second Thunderbolt feeding the second channel of the 53132A as a confidence check.


Somebody should push me to pulling the trigger on buying a little ublox
LEA M8F-based GPSDO (VCTCXO) (currently on that auction site).   It
would replace a wireless cellphone eval kit (with ublox LEON chip) I
picked up MAD-magazine-cheap at a flea market that I managed to short
out the TIMEPULSE on...  My tools are too big for soldering a wire to
surface-mount, let alone replacing the chip.


> Much as I love Lady Heather, hmmm just how kinky is that?:-), I don't rely too much on her reported offsets etc,

The particular figure I think I'm looking at is straight out of the
TBolt.  It's "10MHz offset" bytes 20-23 in the 0x8F-AC report packet.
All that LH does is multiply it by 1000 to report it as parts per
trillion.  The TBolt is measuring the frequency offset of the 10MHz
output relative to GPS/UTC in parts per billion.  Positive values
indicate the 10MHz clock is running slow relative to GPS/UTC.   Watching
this value in LH shows a lot of jitter.  I try to take measurements on
the counter when this value is closest to 0.  This is impossible to
predict, of course.  Next best would be to correlate this value with the
readings I take (via GPIB) from the counter.  My impression is that the
Rb may be more stable over the short term than the GPSDO, but I'd like
to be more certain.


> preferring to trust hardware measurement for that, but would suggest that if she is showing your thunderbolt as locked and tracking a reasonable number of sats then experience suggests you should be able to trust your Thunderbolt as being on frequency.


Trust but verify...


> Experience also suggests, at least with all the units I've seen, that the FE5680A generally reached the surplus market with a programmed offset of zero, presumably because that was good enough for its intended purpose.Soooo, I would suggest that if you have any doubts at all the first obvious thing to do is to reprogram the offset to zero, and start again from there.
> I'd be happy to share my programming notes, but must admit I'm having a bit of fun understanding them myself right now:-)
>
>
I have to pay more attention to whether increasing the offset I send
makes the frequency rise or fall.  And it has to be a fairly large
offset to make the output change obvious.  The manual isn't very clear
whether this offset is for the 50.5MHz oscillator or for the divided
result (to give 10MHz independently).  I'd expect the latter.

I suppose I should re-learn how linear regression works on my calculator.

-- 
Jeff Woolsey {{woolsey,jlw}@jlw,first.last@{gmail,jlw}}.com
Nature abhors straight antennas, clean lenses, and empty storage.
"Delete! Delete! OK!" -Dr. Bronner on disk space management
Card-sorting, Joel.  -Crow on solitaire