Hi Markus,
You've found my problem I think. The 32 bit frequency word was correct
for 137.777 but one bit low for 137.477 kHz. Today the DOXCO was
restored to its original calibration and the bit error corrected.
Now I've further humiliated myself by blaming oscillator drift
for what was my round down instead of round to nearest integer mistake!
Both freqs should now be within about 250 uHz. Your OPDS tonight
showed 137.477000 kHz and so the 50s phase shift experiments on
137.777 have resumed... but I expect to be humbled yet again
unless 20 mHz sidebands appear tonight!
Thanks for keeping me on track.
73
Joe
On Fri, 13 Nov 2015, Markus Vester wrote:
Joe,
well maybe the "crystal drift" is not the whole story... I measured both your
Op-32 and QRSS transmissions (nominally on 137477 and 137777 Hz respectively), before and
after the OCXO tweak. Opera has been -0.7 mHz before and is now closer to nominal (~ -0.1
mHz). However on Nov 9 the QRSS has been only slightly below nominal (-0.17 mHz), and the
tweak has moved it up to +0.44 mHz. So the offset seems to depend on the selected
frequency.
I wonder if you are using a binary DDS, which would be expected to show a
frequency-dependent error due to the finite stepsize (e.g. 10MHz/2^32)? Anyway
this won't be a problem for EbNaut, as long as the offset is known and constant.
| 