I've found that after yesterday's maintenance, REFL_A_RF9 RF monitor got significantly smaller and more stable at the same time.
Attached left is a trend before the change. In 20W lock the RF level (CH5) was going all over the place between -18 and -8 dBm.
In the middle is a trend after the change (now) in 20W lock. After an initial hike associated with powerin up, it quenched down to about -17dBm on average and the fluctuation is much smaller than before.
To the right is the spectrum comparison of RF9_I and Q, red is now and the green is before. A small part of the RF level change could be attributed to the smaller micro seismic motion, but RMS is only a factor of 1.5 or so smaller now than before.
Interestingly, a broad bump (10-70Hz or so) that was present in REFL_A_RF9_Q as well as POP_A_RF9_I is gone now. But that isn't good enough to explain the RF level change either, something should have been going on out of band.
Could this be somehow related to the change of CM-MC board connection cable from TNC-coax to 2-pin lemo twisted pair and/or the limiter (alog 45046)?
Craig is heading to the floor to redo the RF measurement (old measurement: alog 44811).
I've attached scope prints of the current RF levels. Remember the factor of 14 for the (Actual RF to demod/RF monitor pickoff) compensation.
Actual RF to demod
--------------------
REFL9pp = 210 mVpp
POP9pp = 98 mVpp
REFL45pp = 700 mVpp
POP45pp = 28 mVpp
Overall, RF levels seem reduced from before. Also attached are the RF slider settings at the time of the measurement.
Old Pre-O3 RF levels
EDIT: The RF slider now for 9 MHz is the same as it was during the old measurement, but the 45 MHz slider is 3 dB lower.
A comment on the limiter: There is a limiter in the fast path of the common mode board. It 'soft' limits the output voltage to ~3V. It is intended to limit the transients during lock acquisition. This at some time was important for the IMC to keep the FSS well behaved. It should never be engaged, once lock is acquired. For the common mode boards other than the IMC, it was never required and should be off at all times. The signal of the ifo CM board was extremely small at the fast output, so it couldn't have been limiting.
Update: alog 45132
In my post above, the 9 MHz are in the first plot, and 45 MHz in the second. REFL is green, and POP is blue. The names are in the png name itself.