Sheila, Nutsinee
Today we took our time to carefully map out phase vs. amount of squeezing we get. Turned out the best squeezing can't simply be judged by looking at QMON (although I believe this works for asqz). In the end we were able to get 2.02 dB of sqz and 4.5 dB of asqz at a nonlinear gain of 2.39. This measurement still indicates a significant amount of loss. Optimizing the CFL phase is probably what helped the phase noise here. We think we could be limited by some technical noise (to be calculated). But attached below a usual sqz/asqz fit. It seems like our loss is approximately the same as last time (30-40%, alog46814) with better phase noise. We also got a little more RF3 today (-14.7dBm compared to -15.5 dBm last time). We used the same amount of CLF power as last time.
When the beam diverter was opened to inject squeezing this evening we noticed that the 20-29Hz DARM BLRMS got noisier. I made a quick spectrogram and it looks like there are lines at roughly 24 Hz, 26 Hz, 30 Hz, and 32 Hz that show up when the beam diverter is opened. In the first attachment the beam diverter is opened at ~t=500s. Note that the squeezing angle was unlocked for the duration of this spectrogram.
The second attachment shows the spectrogram up to 1kHz when the squeezing angle was locked, showing satisfying broadband noise improvement upon locking.
Since apparently you gain one dB at each attempt, you just need to do squeezing injection one more time! :-D
Great progress, well done!
Very cool! Take that, infinite sea of noisy energy.
Do you see a time dependence or fluctuation in the RF3-Q? That might indicate that some of the loss is time-dependent (and will be removable by controls).
One pedantic note about your plot. It looks like your X-axis is really the nonlinear gain you quote above. Since the seed field you use to measure NLG is injected through a different mirror, it is not measuring the parametric gain that the vacuum is squeezed by, so I wouldn't use that label. The vacuum is a reflection from the cavity M1, not transmission between M2 and M1 like the diagnostics fields. If you do use that label, then you should pass the NLG through the NLG->dbs formula and plot vs. "generated squeezing".
Attached frequency dependent sqz level plot. I divided squeezed spectrum by the reference, turned that into dB and then plot the average (using smooth() function in matlab).
We were moving CLF phase pretty much the whole time we were squeezing so I can't say there is or there isn't a time dependent Qmon fluctuation.
