A similar but much wider bump between 465 and 485 Hz (first attachment) also has coherence with some other channels, for example:

The second attachment again shows the power spectrum and coherence with some channels. You can see an increase in coherence around this same frequency range, with all channels compared showing similar behaviour. 

This being a 20 Hz wide feature, it would be interesting to try to do more studies in order to get rid of this, if possible.

Particularly since the ISS channels are coherent with the ISI, Pep and I talked about this being potentially beam motion that is then coupling in to the ISS second loop, which is then putting it onto the 1st loop.  The attached figure shows that while we're in lock, the second loop QPD is near one of the edges in pitch.  Note however that when we have just the IMC and not the full IFO that the QPD doesn't look centered.  So, I want to move the spot during an IMC-only time the amount that I think it needs for full lock - I don't want to actually center the beam while we have IMC only. 

[Pep, Jenne]

We move the picomotors in front of the ISS second loop array by +0.8 in pitch and +0.47 in yaw, such that the beam should be centered when we're in full IFO 35W lock.  On the HAM2+oplev picomotor controller, we mostly used the picomotor channel 8 (PSL ISS QPD/PD (PM5)) to do the centering, and then checked with channel 1 (PSL ISS QPD/PD (PM1)) that there was no change in the SUM if we moved a few fractions of a beam diameter.  This check helped us feel more confident that the beam is not clipping.  Pep will check if there is an improvement in this coupling, once we're relocked.

After the changes made this Tuesday, the noise floor has improved (see here https://ldas-jobs.ligo-wa.caltech.edu/~pep.covas/O3/AfterPSLPicomotor/Comparison.png), but it could be better. We have used again the picomotors (mostly in YAW) while at 35 W to try to make a better improvement.