Posting the analysis for this FIS + SRCL offset data, where we can compare FIS + SRCL data to QN models, to try inferring the physical SRCL detuning (in degrees) for each offset value (in counts). 

Comparing data + models for FIS at different SRCL detunings, with kHz sqz optimized - Attachment 1. This seems like a reasonable way to estimate SRCL detuning.

• Previously at -175 cts, it was likely around +0.3 deg. Now at -290 cts, it is likely around -0.08 deg. Note these measurements were taken over ~40 minutes, but it only uses ~2 min of data at each setting, so this could be faster.
• These plots show the measured squeezing dBs after subtracting non-quantum noise based on an unsqz noise model. Data = circles. GWINC QN models = solid lines. In models, the SRCL detuning is just set to qualitatively match the sqz data (by-eye). Red dashed trace shows a reference at 0-degree detuning.
• For calculating the qn models in gwinc: at each srcl detuning, the sqz angle is fit to minimize kHz sqz. Just like how, in real life, we minimized kHz sqz at each SRCL detuning for the measurements. That is, kHz sqz is minimized in both measured data and in qn models, so the data+models can be readily compared.
  • It seems like you could get a good signal on this if optimizing bucket squeezing because kHz sqz would be a good lever arm. But based on this week, that seems potentially harder as 1) harder to optimize sqz angle in real life (noisy/slow as it requires more averaging to optimize 350 hz noise, leaves more room for meas error) and 2) potentially more complicated to model, because in the models, the sqz angle needs to match the one used in the measurement at each srcl detuning.

Comparing FIS models at various SRCL detunings + fit how SRCL offset scales from counts to degrees - Attachment 2

• Left panel shows that around 0-degree srcl detuning, the differences in squeezing are very marginal and hard to see (compare the red / yellow / green traces).
  • From this plot, we can see why we thought blue = -400 cts trace was a reasonable point to try yesterday in lho79929 -- that setting seems to minimize 100 Hz noise if we only had FIS. But we have FDS, where the anti-squeezed quantum noise below 100 Hz should be "taken care of" by the FC. So, seems reasonable to aim for zero detuning.
• Right panel shows a fit of SRCL offset from counts to degrees, fit suggests -274 counts would minimize the detuning; we set it at -290 this week in lho79929, so very close to 0! Note that the different signs of the SRCL detuning seem to impact squeezing quite differently.

So far this code is living here.

Here also reproducing the list of times I used for the analysis, all span = 120 seconds. Also including the SRCL offset in counts of the filter bank, and the corresponding estimated SRCL detuning based on FIS quantum noise models, with khz squeezing angle minimized.

nosqz: 1409511600
FIS -100: 1409509687, pink.        ~~> +0.9 deg
FIS -175: 1409509282, brown.    ~~> +0.3 deg
FIS -250: 1409510096, green.     ~~> +0.1 deg
FIS -325: 1409510392, orange.  ~~> -0.1 deg
FIS -400: 1409510673, blue.       ~~> -0.5 deg
FIS -475: 1409511323, black.     ~~> -1.1 deg