The 1/sqrt(f) noise that is limiting the sensitivity doesn't seem to have disappeared after the change of the NE mirror. So we should look for other hypothesis than thermal noise.
One place is that B4 DC shows 1/sqrt(f) spectrum, which was not the case during O3.
Figure 1 shows data from April, and the spectrum shows a 1/sqrt(f) shape every day, with the exception of April 19. This coincides with the test installation of the RAMS servo on the 6MHz and 56MHz sidebands. There was only two successful locks with the RAMS servo, and it was removed on April 20 as the lock acquisition was not reliable.
Figure 2 shows that the test had not changed the sensitivity in the 100-200Hz band (the comb in B1 on the purple curve are very likely due to calibration injections). There is also some excess noise on B7/B8 DC, but this did not change during the RAMS test.
A few thought on this
- It does not seem that this issue is affecting the sensitivity, but it is still worthwhile understanding it more.
- It surprising that the RAMS servo can affect the DC light stored in the central interferometer
- It is also surprising that this was not an issue during O3, where the modulation configuration was the same.
- With the RAMS servo the B4 spectrum become flat up to ~500Hz and then decreasing (quickly limited by shot noise which is only a factor 2 lower).
Looking further back through the logbook, Edwige had noticed a a 1/sqrt(f) noise on B5, which was changing with the 6MHz modulation depth.
Figure 3 shows that the B5 noise 1/sqrt(f) behavior also disappeared when the RAMS servo was installed.
Figure 4 looking back at the data from April 7 with the mod depth changes (red nominal, purple 6MHz mod depth increased by 3dB, blue 56MHz mod depth decreased by 6dB). Both B5 and B4 spectrum increase with higher 6MHz mod depth, but are not changed by the 56MHz mod depth. This could also be the explanation why the problem was not apparent during O3, the 6MHz modulation was depth was practically zero at that time.
Conclusion: Installing the RAMS servo would help clearing the B4 DC spectrum, so other issues can be explored looking at the B4 spectrum.