About 9 hours into the lock today at 76W, we put squeezing in and can get about 4dB of FDS, with pretty clean broadband noise reduction, see spectra. Red = shot noise, taken shortly before beam diverter. Brown = FIS. Blue = FDS.
I think we can make several further sqz optimizations here, and possibly get it up to 4.5 dB again. For example, it seemed like TCS tuning to reduce laser frequency noise like the 4.5dB time (68251) was super helpful, but for now, I think we can continue to win by injecting some more squeezing, at least at high frequencies. Low frequency shot noise is still being worked on in IFO, so we have a clear look of FDS at low frequencies yet, but there is a nice broadband reduction, reaching ~3dB just above 100 Hz. The preliminary, possibly uncalibrated, range increase with FDS looks promising (bottom-right-red trend).
Since optimizing squeeze angle at the start of the lock, the squeeze level / angle has drifted over the last 3:45 hours by I think ~15-20 degrees. In the top right sqz blrms of screenshot, we show (blue) BLRMS 5 = 1.7 kHz, and (purple) BLRMS 6 = 4.7kHz.
This is unlike previous long locks at lower power, see e.g. late March (68136): then, we saw ~0.5-1dB squeezing level drifts in 1-1.5 hours that were unrecoverable by changing squeeze angle, adjusting opo temperatures, mode-matching, sqz asc alignment offsets, etc. Today at higher power (76W), we see coherent sqz level drifts of 2-3 dB; "coherent" meaning it is not loss, because we can rotate the squeeze angle back, and entirely recover at least the exact same sqz level if not more. Nutsinee has also observed this recently (that rotating the squeeze angle back has been able to recover squeezing). Dhruva also points out that, multiple sources of coherent sqz rotations from e.g. SRCL detuning or mode-matching, which can drift during thermalization (and have been measured many times now), all pile up in this 100-500 Hz frequency band for rotations around the darm pole. From kevin and dan's simulations, both of these two effects can produce sqz angle rotations of similar scale, with potentially opposite sign; plots forthcoming. Interesting to see if going forward, if this squeezing drift is definitely related to thermalization, and if it remains just a coherent sqz rotation and not sqz loss.
