Broadband noise injections on the four FP suspensions have been performed, along longitudinal and pitch d.o.f.. The transfer functions towards hrec are the expected ones (up to 60 Hz), if the horizontal resonance of F7 crossbar is added to the simplest model (just the double pendulum marionette-mirror). The frequency of the modes has not been calculated; in the attached figures the models have been adapted to the measurement in order to match the peaks. The guess that the peaks are due to the crossbar resonance has to been confirmed, but at least it produces the correct shape of the tf.

All the peaks are present in hrec in clean condition. The most famous is 48.05 Hz (due to WE, as already known), but also the others are there. The presence of the peaks in the sensitivity is not explained by the simple noise projections of the actuation or control noise, through the measured tf: the input noise is to low. In the last attached plot the projection is forced to match the sensitivity putting an enhanced value as input of the tf.

Above 60 Hz the shape of the tf is not explained at all. It looks really to high, especially for the INPUT susp. It seems to suggest a possible direct coupling from F7 cage to the mirror, bypassing the double pendulum.

A complete and precise projection of F7 residual motion into the sensitivity cannot be easily done at the moment: some of the high frequency structures could stay above the safety margin foreseen by the requirement.

During the acoustic injections at NE building  (#47122), we notice the peak at ~ 48.45 Hz excited a bit  Figure 1 and Figure 2.

As reported in Figure 1, we observe an excitation of peak at ~ 48.1 Hz during the injection acoustic at WE building (#46708), as well.

We had a look to the evolution of rms around the 48Hz peak (47.9Hz to 48.2Hz) all along O3: it is the green colored line in the attached figures. we used Didier's script that reads from archived VIM spectrograms, see 46410.

A few times the amplitude of the peak changed up or down by a factor 2 or 3 or 4, here are the times of these occurences:

• june 16 (SUNDAY) btw the lock ending at about 7:30 UTC and the lock starting at about 9:30 UTC (line went UP by a factor ~3)
• june 23 (SUNDAY) btw lock ending at about 8:30 UTC the lock starting at about 10:30 UTC (line went DOWN by a factor ~2)
• june 25 (TUESDAY) across the maintenance (line went UP by a factor ~2)
• july 30 (TUESDAY) across the maintenance (line went DOWN by a factor ~2)
• august 20 (TUESDAY) across the maintenance (line went UP by a factor ~2)
• august 25 (SUNDAY) around 10 UTC the line excited and then reduced by a factor ~4 (this is the event noted by PAOLO in 46854)
• september 3 (TUESDAY) across the maintenance (line went UP by a factor ~3)

the line's amplitude seems more or less stable since then ... (Figure 3)