A few comments on what we can see now that the RAMS noise is nicely reduced by the new servo.

Figure 1 shows the difference in spectra with the RAMS loop closed (purple), and without it (blue). On B1s we see the RAM nicely removed, on B5, B4 and B2 it is also reduced, and reveals another noise of a different origin that has a shape of a pole at ~500Hz.

Figure 2 shows that none of the cleaner audio channels shows any very interesting coherence with B1, appart from B5 below 20Hz which would be interesting to study further. But it shows that the newly revealed noise in B5, B4 and B2 is common between these 3 output ports.

Figure 3 shows how B2 can be used to remove that common noise from B5. The result on B5 it still a factor 2 above shot noise.

Figure 4 shows how B5 can be used to remove the common noise from B2. The result is a factor 3 above shot noise, it might have a significant contribution of frequency noise, the 227.1Hz line is a factor 500 above this noise level (for 5s long FFTs), and it is also the case for the SSFS Q noise compared to the SSFS I line at 227.1Hz (see figure 7).

Figure 5 shows B1s. After removing the RAM noise, there is a factor 2-3 more noise than expected just from shot noise.

Figure 6 compares the PD spectrum with the RAM servo enabled, and a time without a laser beam (IMC unlocked). It shows that on B1s at 2-3kHz we are actualy limited by PD electronic noise, which is a factor few higher than for other photodiodes. On B1p PD1 the spectrum is limited by electronic noise above 20Hz, as there is 100 times less power on that photodiode than on B1s. And B4 PD2 6MHz I, the extra noise visible since last week, is actually a factor few below the PD electronic noise, so it may not be a very relevant issue, but it is still worth to follow this up to understand why the 6MHz I signal from the same photodiode is different on SPRB and in the SSFS process.

It would be interesting to replace B1s with a photodiode with less electronic noise, and check if the 5Hz comb on B1s is due to some LED blinking inside the electronic tank or some similar issue.

Figure 8 shows B4 after subtracting from it the coherent noise with B5. The resulting noise floor is flat, and a factor 1.5 above shot noise, it is also a factor 3 above the PD electronic noise.