Considering that the ITF conditions were not stable since some minutes before the start of the test, my impression is that no definitive conclusion can be drawn from this test and it has to be repeated (maybe, as suggested, following a different on/off sequence).

Nevertheless, here some plots showing the behaviour of the sensitivity curve during the test:

• Figure 1 is a spectrogram with indicated the selected times (all ON, all OFF, only HALL on, only DET on)
• Figure 2 is a zoom, showing that around 13:36 a strange spectral noise onset in Hrec (see also Figure ...). Apparently this noise engaged AFTER the DET HVAC was switched OFF
• Figure 3 compares Hrec sensitivity right Before and when ALL CEB devices were Off: there is a clear reduction up to 100Hz  (these are the 3Mpc step)
• Figure 4 is a similar comparison of Hrec Before (purple) and when CEB HALL is ON (blue): no change with respect to the all off (brown)
• Figure 5 is similar, but now the blue curve is with DET airconditioner ON: noise seems to go back to the level BEFORE
• Figure 6 shows the strange noise that onsets right after DET air conditioner was turned off

Indications are that DET air conditioner is the most impacting (scattering?). The Hall air conditioner seems not contributing. The effect of the INJ air conditioner could not be tested.

It seems important to repeat the test, focussing on DET (and INJ air conditioners).

The strange noise onset seems not directly linked to our actions... it has to be further investigated.

The switch off produced a relevant noise reduction on seismometers and accelerometers in the frequency range where the excavator noise had been particularly effective. It is remarkable the disappearing of the 7 Hz bump on NI tower (fig 1). Nevertheless, the sensitivity did not improve in the 1/f^4 low frequency floor (fig 2). The guess of a possible residual noise similar to the one injected by the escavator seems to be discarded. There is only one small region, between 10 and 15 Hz, where the switch off did not change the seismic spectrum. An eventual noise injection by the shaker could focus on that frequency.

At least, the switch off had an impact at 20 Hz, where probably there is still something from MICH (fig 3). If the noise enters the ITF through MICH control, there is still margin to reduce the correction acting on the control filter. But in any case the best would be to remove the bump from MICH error signal. Maybe the data are useful to identify the source of the bump.

Comparing with switch off times: 

• the disappearing/reappering of 6.8Hz seismic noise coincides with HALL air conditioning switch OFF/ON (figure 1)
• the 20Hz bump in LCS_MICH seems consistent with the INJ air conditioning: see in Figure 2 the arrow indicates the INJ switch on time corresponding with the increase of 20Hz laser bench vibration and bump noise in LSC_ERR. However, ths is not SO clear because just a few seconds after the INJ was switched on the ITF unlocked.

the plot shows an attempted transfer function computation between the accelerometer on NI tower flange (EW oriented) and the corresponding (W) channel of the Guralp on ground. The mechanical TF seems that of a resonant mode at 6.8Hz. Might it be something similar to what was excited when shaking NE tower?

Further investigation to see the effect on CEB magnetometers linked to magnetic Noise projection in CEB (picture 1)

The lines at 47.6 and 52.4 disappear when all systems were off and reappear when the HALL system was switched on. (Figure 2)

The 43.7 line (which is irregular) does not disappear and has no counterpart above 50 Hz (Figure 2)

If we zoom in there is a small line around 43.58 Hz (which does have a side-band image above 50Hz). It disappears when all systems are OFF. But does not reappear when HALL or DET are switched on. Only when all systems are switched on again the line appears. (Figure 3)

Furthermore we studied if  the broadband noise predicted by the coherence projection is linked to any HVAC system. No prove is found. There is no significance change in the overall coherence between the CEB magnetometers and Hrec if we compare times where the system is active or not.

(Figure 4-5)