This is a follow up with the data from 50473.  Because we had once seen some interesting frequency dependence from the squeezing 50104, I made some measurements last week of the squeezing with different SRCL offsets and phases. The braod bump centered just above 2kHz may be consistent, this could be noise that the squeezer is adding to DARM but it is not present in the correlated noise with no squeezing. There also seems to be a higer level of squeezing at low frequencies for positive SRCL offsets, although I would like to revisit that measurement.  It is interesting to compare these measurements with Lee's model of frequency dependent losses in 50589

The first attachment shows the loop corrected DCPD cross correlation for a 6 minute stretch without squeezing in the same lock as the following measurements (almost a day later, but in the same lock).  I used the pyDARM model with parameters from April 16th, but scaled the overall loop gain by 0.97 to account for the drop in optical gain. 

The second attachment shows the loop corrected DCPD spectra for the series of measurements with different SRCL offsets and squeezing phases.  You can see that in some of these measurements I may have caught a glitch (SRCL offset = -100 antisqueezing phaes 182 and srcl offset = 200 squeezing with a phase of 192 are two that seem suspicous for having a glitch.)

The third attachment shows the squeezing level for each of these measurements with the cross correlated noise subtracted. One warning about this, for each SRCL offset I did remeasure the no squeezing spectrum (not expecting them to be differenton the DCPDs), but I did not repeat the cross correlation for each of these measurements.  So if the correlated noise is changing with SRCL offset, this plot will be misleading. 

Comparing this frequency dependence to what we saw in 50104, there does seem to be a lump around 2.5 kHz which is not in the cross correlation on both sets of data.  This could be noise that we are adding with the squeezer, which seems to be at a pretty consistent level for different SRCL offsets and squeezing angles. 

The most recent calibration of the SRCL offset that I can find right now is from 2014: 14550  Kiwamu did this calibration during a DRMI lock, although it should be roughly applicable in full lock since we scale the input matrix to keep the ugf constant when we transition to POP.  So a SRCL offset of 100 counts is very roughly 0.5 nm offset. I did loose lock once with a SRCL offset of -200, which could have been a coincidence, or due to the offset. 

It would be interesting to repeat the measurements at SRCL offset = 200, and to try to measure the cross correlation without squeezing there. I've attached the scripts used, the first script attached contains the times of all of these measurements, the second is for doing the cross correlation.  The way these are written they probably only work from the control room, but the only things they are relying on are pyDARM and gwpy.