[time-nuts] Coupling between oscillators -- an example
John Ackermann N8UR
jra at febo.com
Fri Mar 25 19:41:46 UTC 2022
One of the things that needs attention as we measure oscillators more
carefully is ensuring that there isn't mixing or injection locking
between multiple sources. I just encountered a doozy of a case where
that happened.
I have 3 FTS 9110 5 MHz OCXOs that came out of (dead) FTS military
cesium units. They run on 24V and have 4 SMA outputs. They are pretty
loud: +17 dBm.
I powered up all three off a common 28V battery supply and let them warm
up for a day or so (I connected only DC power and not any of the EFC
connections). Then I hooked them to the three DUT inputs of the
PhaseStation using well-shielded cables and started logging. The
attached "initial_adev" plot attached threw me for a loop. That's quite
a roller coaster ride, particularly on #3019.
I thought that maybe the three +17 dBm signals going into the test set
were causing some overload, or at least allowed for cross-talk. So I
put a 9dB attenuator right at the output of each unit. That didn't
really help things -- in fact things look worse on the first run after
doing that (see the "worst_ripple_adev" plot).
The frequency counter plots show that #3019 is about 0.168 Hz high,
#3028 is about .012 Hz high, and #9025 is about 0.588 Hz high. Those
differences seem about right to create beat notes in the 1 to 10
second-ish range. And then there are beat notes beating against each
other... it gets complicated! The "worst_ripple_freq" plot is really
interesting; you can easily see the beat notes.
My next step was to put each oscillator on a separate power supply.
That helped a lot, but there were still numerous but smaller ripples
left. (And I saw that these oscillators are very sensitive to noise or
variations on their power input; there may not be much regulation inside
the box.)
Finally, I put an isolation transformer at each oscillator output (just
after the 9 dB attenuator) so the signal cables had no DC path except at
the measurement inputs. That cleared things up almost completely. The
"isolated_ripple_adev" plot shows the results -- #3019 still shows a
possible ripple at around 10 seconds, but that's a small dip on a line
that's in the low 13s. I'll take that!
This exercise shows that sometimes you need to use some tricks to
eliminate coupling, but it can be done.
John
PS -- an interesting feature of these oscillators is that they are an
"S12" variant that has a little accelerometer bolted on and hooked to
the EFC circuit to compensate for G forces. These were ship-board Cs
units, so I wonder if the accelerometers were intended to compensate for
ordinary ship vibrations, or maybe the firing of the big guns. :-)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fts9110_initial_adev.png
Type: image/png
Size: 70235 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220325/575eeb2c/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fts9110_worst_ripple_adev.png
Type: image/png
Size: 37163 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220325/575eeb2c/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fts9110_worst_ripple_freq.png
Type: image/png
Size: 49920 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220325/575eeb2c/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fts9110_isolated_ripple_raw_freq.png
Type: image/png
Size: 100227 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220325/575eeb2c/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fts9110_isolated_ripple_adev.png
Type: image/png
Size: 64833 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220325/575eeb2c/attachment-0004.png>
More information about the Time-nuts_lists.febo.com
mailing list