Our range has increased back to around 160 Mpc on the CALIB CLEAN channel the past few days. I ran the DARM integral compare plots using an observing time on Sept 28 (before OPO crystal and PRCL FF changes) and Oct 5 (after those changes). It appears the largest improvement has occured at low frequency. Some of that can be attributed to the PRCL feedforward, but not all. Based on the previous noise budget measurements, and the change in the coherence of PRCL from Sept 28 to Oct 5, I think the improvement in DARM from 10-30 Hz is likely due to the PRCL improvement. Above 30 Hz, I am not sure what could have caused that improvement. It doesn't appear there is much improvement above 100 Hz, which is where I would expect to see changes from the squeezing, if it improved from the OPO changes.
Sheila pointed out two things to me: first, that if we are not using median averaging, these plots might be misleading if there is a glitch, and second, that some of the improvement at low frequency could be squeezing related.
I went through the noise budget code and found that these plots were made without median averaging. However, changing the code to use median averaging is a simple matter of uncommenting one line of code in /ligo/gitcommon/NoiseBudget/aligoNB/aligoNB/common/utils.py that governs how the PSD is calculated for the noise budget.
I reran the darm_integral_compare code using median averaging. The result shows much less difference in the noise at low frequency between these two times. The range is still improved from 10-50 Hz, but there is a small drop in the range between 50-60 Hz. I still think the change from 10-30 Hz is likely due to PRCL.
As a further confirmation of the necessity of median averaging here, I made a spectrogram of the data span on Sept 28, and a few glitches, especially around low frequency, are evident. I didn't see these glitches in the sensitivity channel that I used to choose the data spans (I just trend the sensmon CLEAN range and look for regions without big dips). However, the Oct 5 data span appears fairly stationary.
