Summarizing some thoughts regarding the sqz levels at LHO going into O4. Best-case, we've observed up to 4.5dB DARM noise reduction with squeezing at 50W and 60W. Ignoring phase noise, this 4.5dB sqz is about 30-35% total sqz losses (~19% known, ~15% mystery). At 76W at start of O4, we've typically seen 3-3.5 dB at high frequencies, and almost up to 4dB once (we can hopefully recover this after fixing on-table alignments). Some of the major questions that I'm thinking about, towards more squeezing at LHO:

1. IFO output losses – is shot-noise-limited displacement sensitivity consistent w/circulating arm power? (LHO:67610)
   • ~13% known SQZ output losses (sqz budget)
   • Model suggests ~20-30% total IFO output losses (i.e., reconcile quantum noise calculation for no-sqz DARM with arm power). How do IFO+SQZ output losses relate?
   • LHO:69707, measures only 82% ham6 carrier throughput. Seems relatively consistent w/previous measurements ~20% output losses.
     - new thoughts on the losses this time-- a lossy OMC (92% throughput?)    
     → updates known total sqz losses to ~19%.  if only these known losses + good phase noise, we could see up to 6dB sqz (nearly what L1 sees).
     - still leaves ~7% excess ham6 ifo loss    (could this be ifo-omc mode-matching? TSAMS?)
     - in principle, SQZ-OMC could have different mode-matching than IFO-OMC, so output losses like this could be different between sqz / ifo (unlike omc losses, which would be common).
   • To-do: precisely measure OMC finesse? Is it worse now? Could tsams help? Is this ifo power dependent?
      
2. IFO technical noise – how shot-noise-limited is DARM? 
   • SQZ can reduce quantum noise, which only helps up to the next limiting noise source. That is, if there's no quantum noise (or infinite squeezing of it), DARM will simply be limited by the next thing.
   • 76W H1 noise budget suggests only shot-noise-limited by about 3-6x @ 1kHz?  
     • 69767, this level of classical noise can look like an effective 5-10% sqz loss.
     • Comparison: L1 noise budget shot-noise-limited by >10-fold @ 2kHz, with sqz (a lot more clearance)
        
3. SQZ losses – what are excess sqzer losses, and does sqz introduce its own technical noise?
   • ~7% known sqz injection losses (w/o mode-matching)  (sqz budget)
     • In addition, homodyne sees < 10% excess mystery sqz loss 67219 (so total in-chamber losses < 17% (from e.g. clipping), much of this loss is likely related to HD and not ham7)
   • CLF can introduce sqz technical noise, projected at ~5-10x below DARM, 68991
   • Phase noise? can re-measure loops
   • To-do: can we directly measure sqz losses (vary gen sqz, more NLG sweeps on darm)? use absolute calibration of the ADF line to measure sqz losses? optimize sqz alignment? Psams? sqz loops/phase noise? Fc backscatter (could contribute to #4)? Recover homodyne to check sqz? Can we even observe changes in sqz level given the large losses + drifty technical noise? Helps to have stable ifo for sqz commissioning. And there's probably more thinking to be done re: what are good measurements to make.
      
4. Is there residual noise ~50-200 Hz related to the squeezer? Is it worse at higher power? Related to RPN or backscatter (e.g. LLO:64968, 65120)? Can we make it better? How? This noise is not obviously quantum noise (started analysis previously, w/some evidence of it), but need to compare it no-sqz vs. sqz with the quantum subbudget to confirm
   • To-do: more modeling of high + low power scenarios to learn what various knobs do, more commissioning, etc, but not sure yet..