I've been doing some modeling of the squeezer losses to try and pinpoint possible sources of missing squeezing. One thing that I have noticed is anything related to the interferometer (rather than SQZ or output-optics) will cause frequency dependent losses, which might emulate the frequency-dependent squeezing of SRCL offsets.

Georgia mentioned Shiela's posts 50104, 50473 which indicate a possibility that the budget at LHO currently requires some form of frequency dependence to explain the degradation of squeezing above the cavity-pole, and the conclusion the SRCL tunings aren't affecting it.

The impressive thing about the SQZ budget spectrum is the amount of loss required to degrade squeezing ~2db by 1kHz. This requires enough loss that one might dismiss the possibility that they are in the IFO, but IFO interactions can enhance apparent loss though coherent effects. For mode-matching to the cavity, I go though this a bit in LLO45493 to show that IFO mode matching can enhanc loss 4x in some cases, but there do not show the frequency dependence.

Another possibility is loss in the SRC cavity round-trip. I'll mention a disturbing feature of the SRC. This cavity is anti-resonant to lower the cavity pole, which should lower the effects of losses, as they are cavity-suppressed. However, the anti-resonance is established through the phase-flip of the over-coupled arms. Above the arm-pole, the arm reflectivity flips phase back and by 1kHz, the SRC is resonantly-enhancing its losses. GWINC does properly model this and increasing Optics.BSloss will show the frequency dependent degradation.

Internal mode-matching between SRC<->ARMs is also a possibility, but since it is power-conserving it is perhaps less strong. I have started modelling all of these in a simple 2-transverse-mode coupled-cavity model (no radiation pressure included). This model isolates the loss from the reflectivity in a simpler form to study (than GWINC, at least). The matrix equations could be expanded to include more modes or 2-photon formalism for RPN, but it is complicated enough as-is. The internal arm<->src mode-matching has not been studied fully yet because I have yet to automate the cavity tuning to compensate for 02-mode gouy phase that enters with the mode-mismatch. I also likely have the exact Gouy phases wrong, but the plots below show the qualitative effects to SRC reflectivity and signal cavity pole for the different levels of loss. The SRC transmissivity used is .325.

cavity_pole_SRCloss.png shows the effect of internal SRC losses.

cavity_pole_SRCmm.png shows the effect of mode-matching losses with the interferometer (assuming perfect matching to SQZ->OMC).

coupled_cavity.py has the matrix equations the notebook generating the figures are in coupled_cavity_LHO_ipynb.tgz (can't upload python notebooks without zipping). Should run in python3 for users with IIRrational.