Very interesting measurements!

There is one thing to keep in mind. The BPC is measuring the beam jitter at the BPC quadrants on EIB. The beam jitter at the IMC input might be larger since the beam is travelling in air from the BPC quadrants to the IB tower input.

Maybe a reshaping of the IMC AA loops filters can reduce the noise at 100Hz.

This said, there is still one thing that I do not understand. The IMC AA WFS is measuring the misalignement between the noisy in air input beam and the silent suspended under vacuum IMC. Three of the four IMC AA WFS signals are correctly used to act on the BPC i.e. in the noisy beam. But the fourth acts on the IMC payload. Wouldn't be possible to use also the fourth to act on the last BPC degree of freedom i.e. BPC_Y? 

Given that the two MC mirror angular degrees of freedom can be controlled using a dithering like system that keeps the beam in the center of the MC mirror, we would then be left with IB_TX. Can this remain under local control?

If not, can a blending of the fourth IMC AA WFS between IB_TX (at the IB suspension resonant frequencies) and BPC_Y (at DC and at high frequency) be the solution?

Fig.1 shows the comparison between the RIN transfer function (input beam -> B1_DC) measured yesterday (2023,04,07,19,26,00) and that measured last week (023,03,29,20,46,50).

The input RIN is computed as PSTAB_PDd_AC_MONIT/ (11*12*PSTAB_PDd_DC), while RIN_OUT is simply B1_DC/mean(B1_DC);

The TF has been compensated for the DARM closed loop (/virgoDev/ISC/matlab/DARM_loop_TF.m).

The coupling seems reduced by about a factor 2.

Fig. 2 shows the CMRF (DARM_SSFS_LINE_mag_100Hz) comparison in the two moments. The fluctuation is quite large but the mean values are respectively 2.7e-9 (old measurement) and 1.3e-9 (new measurement). This seems in agreement with the amplitude noise coupling reduction.

The measured shape of the RIN transfer function is compatible with an offset compared to dark fringe of the BS.

Figure 1 shows a simple simulation in optickle of the transfer function from laser amplitude noise to B1 RIN (so there is a factor 2 compared to laser RIN to B1 RIN transfer function), with in red a perfect interferometer, and in blue an interferometer with a large BS offset (which make the power on B1 have roughly equal contribution from the DARM offset and from the MICH offset). The shape with the MICH offset above 100Hz looks very similar to what is measured. Compared to a perfect interferometer the impact of a MICH offset on RIN coupling is mainly above 100Hz.

FIgure 2 shows from the same simulation the transfer function from the laser phase (in radians) to B1 RIN. The impact of a MICH offset is mainly below 100Hz.

This suggest that one should use the coupling of PSTAB to DARM at high frequency (the 1.5kHz PSTAB line for example), to adjust the MICH offset. Given the large fluctuation we see at each lock, it would be good to have this a signal in loop adding a MICH offset once we are in LN1. The other tunings (TCS differential, Etalon, ...), could help the working point so that this position also corresponds to a minimum in SSFS coupling and a good 56MHz balancing on B1p, which currently is not the case.

These are nice BPC injections, and I have updated the times in the noise budget to include them.

Figure 1 shows the result of the noise projection at a calm time across the frequency band, taking the transfer function where during the measurement the coherence with DARM is at least 30%.

Figure 2 is a zoom at low frequency. The bumps that are slighly below 20Hz, and which I had noticed also in MICH and SRCL, seem to be due to BPC TY. Note that it doesn't look present in h(t), presumably Hrec sees the coherence between DARM and MICH, and subtracts this way the BPC TY noise. It would be interesting to undertand better where this bump at ~18Hz is coming from.

The bump mentioned by Michal is highly coherent with the ENV accelerometers on both LB and EIB. It's not completely clear whether the dominant contribution comes from. The reason is that both LB and EIB have structures around 17Hz; what i mean is that EIB is way quieter than LB along X but not so much along Z and it hosts the BPC sensors. As input for further investigation the second plot shows the EIB displacement noise in m/sqrt[Hz] units obtained by double integrating the ENV accelerometers signals.

This afternoon, we investigated the bump around 18 Hz visible in DARM and its coherence with LB and EIB accelerometers. We have installed a temporary microphone ENV_LR_MIC in the laser room savaging the connection of the accelerometer ENV_LB_ACC_Z, Figure 1. We found the PCB accelerometer and microphone removed from the laser bench and left on the top of the plexiglass, Figure 2. We reinstalled the microphone, Figure 3. The accelerometer installation will be done on Tuesday morning (maintenance).
We switched off the air conditioning for a few minutes, from 15:56 to 16:02 UTC. We did not observe significant temperature variation inside the laser laboratory.
Figure 4 and Figure 5 (zoom) show the effect of air conditioning switched-off (blue curves) on the microphones (first row), accelerometers on the benches (second and third rows), Hrec and DARM (fifth row).  A broadband noise reduction is well visible on the environmental sensors while it seems to be a small reduction of Hrec and DARM around 18 Hz.
In Figure 6, we observe a high coherence between ENV_LR_MIC and Hrec, DARM when the air conditioning is on but it decreases abruptly during the switched-off (blue curves).
Regarding the coherence of the microphone in the laser room and the accelerometers on the benches, even in this case, we observe a reduction but some structures of coherence are still present when the air conditioning is switched off.