Dark offsets for ISCEX and ISCEY

ALS X and ALS Y dark offset SDFs.

Accepted at H1:SQZ-ADF_OMC_TRANS_PHASE = 120. This seems to give us the best range but not high frequency SQZ. Plot attached.

We optimized the sqz angle after high CLF test and readjusted the AADF demod phase.

We should check the ADF offset with a more themalized IFO as the SQZ angle servo seems to have pulled even the yellow BLRM to a worse location, see attached.

Here is my plot which is just binning calibrated CAL-DELTAL_EXTERNAL_DQ and then excluding all the ASD values where the noise now is better than the noise in 04a (these are at low frequency and this is where our noise has improved due to ASC control noise improvements and new DARM configuration).

Then I use the same maths as Artaem to get:

excess noise with no squeezing = sqrt( (ASD with no squeezing now)^2 + (ASD with no squeezing in 04a)^2)

excess noise with no squeezing = sqrt( (ASD with squeezing now)^2 + (ASD with squeezing in 04a)^2)

These also seem to be the same in my plot, which implies that the excess noise is indeed due to some correlated noise (i.e. non-quantum).

See evolution and reduction with ZM alignment: 76757

The new phase works well in NLN (injected line at 11 Hz shows up only in I and not in Q). But the 4 Hz bump is still there. And there is also a 4 Hz line in the PZT1 signal (the one used for the 4.1 kHz dither). So it's more complicated that simply amplitude modulation of the 4.1 kHz line.

We offloaded IFO ASC and used "Save ZMs IFO" script to save the ZM settings that we found.

We moved the ADF back to 1.3kHZ as think the 322Hz ADF we used eariler is impacting the range.

Attachment 1 - looking for freq-dep SQZ loss/rotations. Here we fit a common model of frequency-dependent losses and rotations to all squeeze angle spectra simultaneously. FIS data would probably clean this up at low-frequencies, maybe removing the ~20 Hz anti-sqz bump.

• Top plot - SQZ dBs after subtraction of non-quantum noise. Note: after subtraction, we see worse squeezing now at 100 Hz vs. kHz, unlike before in O4a/O3.
• Middle plot  - fitted frequency-dependent losses (1 - \eta(\Omega)). Note: now estimating ~10-15% more loss at 100 Hz than kHz; only few% loss difference between 1-3 kHz. This was the opposite in O3 e.g. Fig. 3 mid-right plot of the quantum response paper (P2100050) shows ~20% less loss at 100 Hz.
• Bottom plot - fitted sqz angle rotation (\theta(\Omega)), in addition to the sqz angle used for each trace (\phi_0). That is, each frequency bin corresponds to a sqz angle at \phi_0 + \theta(\Omega).

Dots + thick lines = subtracted sqz data, with a moving average for clarity. Thin line = common fit model. Equations in the plot title. For each frequency bin \Omega, we fit the loss(\Omega) and the sqz angle offset theta(\Omega) given the \phi_0 for the dataset. The fit to all sqz angle spectra is done independently for each frequency bin.

This dataset suggest higher freq-dep losses at low frequencies in-band, but before we typically we had lower freq-dep losses below darm pole, e.g. Fig. 3 of the O3 quantum response paper (P2100050). I'm not really sure yet how to interpret this, and don't think there's a clear expectation for one way or another. As a basic sanity check, I compared another time with anti-sqz from March 17 LHO:76434 (which had different PSAMS settings) - there, evidence for frequency-dependent losses at lower frequencies is weaker, but there is still some evidence for it.

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Attachment 2 - squeezing-related DARM comparison, O4a vs. pre-O4b. Blue/yellow = O4a. Purple/pink = pre-O4b. Interesting things (from a sqz perspective):

• Above the DARM pole, SQZ is mostly loss-limited, notice purple better than blue -- DARM is slightly better now than O4a. For NO SQZ, optical gain looks slightly higher now. For SQZ, there's now more high freq SQZ, aka reduced SQZ losses.
  
  For reference, in O4a we saw up to [-4.5, -4.8] dB with/without subtracting non-quantum noise. Now pre-O4b, we've seen up to -5.4 dB without subtraction at the starts of locks when technical noise is lower, and can see 4.5 dB thermalized. Tracking losses in the gsheet. 
  
  From the OMC replacement plan G2302156 - the old OMC003 was estimated ~90% throughput, the new OMC001 was estimated ~96% prior to install.
  (O4a OMC loss inferences in e.g. 73873 and 74022 suggested 89-91% omc003 tem00 efficiency. See T1500060 for OMC specs.)
   
  • pre-O4b: Inferring losses from 15.8 dB anti-sqz && -5.4 dB sqz w/o subtraction {NLG = 17.3, 25mrad phase noise, in this thread 76553}
    --> pre-O4b: inferred losses ~ 22-25%, of which we expect ~16% (if new OMC001 ~3.6% loss), leaving ~12-13% excess loss. Mode-matching status unknown.
     
  • O4a: NLG sweep on DARM after crystal move, Oct 2023, see LHO:74318. This is consistent with O4a SQZ angle datasets: 74935 - Dec 2023 - cold OM2, and 73621 - Oct 2023 - hot OM2.
    --> O4a: inferred losses ~30-32%, of which we expected ~22% (if old OMC003 was 10% loss), leaving ~12-14% excess loss. Mode-matching attempted, railed PSAMS actuators, ran out of range to continue.
     
  • The higher squeezing pre-O4b (-5.4 dB vs. -4.5 dB) seems consistent with a 5-6% reduction of SQZ losses. This improvement is plausibly consistent with the expected OMC loss reduction.
     
  • On the other hand - while we see some squeezing improvement consistent with expected loss reductions, this actually suggests that the 10-13% excess sqz losses from O4a remain untouched. In O4a, based on 8dB homodyne SQZ (73562), these 10% excess losses were not "injection losses" i.e. they were after HAM7. 
     
  • May be worth considering more closely the conversion from NLG to generated squeezing level, since that factors into these loss estimates. These loss estimates are based on a model for a linear OPO. Dhruva's ADF paper P2200041 does some more math on the conversion from NLG to generated squeezing level for a bowtie OPO, like as the observatories have. I believe the same NLG corresponds to a slightly lower generated squeezing level than expected for a linear OPO.
  • Mode-matching gains from the break TBD. The sqz alignment scripts will be helpful to fully explore PSAMS optimizations. Naoki has started this in LHO:76757, bringing in Masayuki's work from e.g. LLO:64903, LLO:64458.
     
• Below the DARM pole, notice blue better than purple -- it's interesting that both no-sqz and sqz are worse, especially considering the above evidence for higher freq-dep losses below the darm pole. Attachment 1 top plot - we see worse squeezing now at 100 Hz than we used to in O4a, even after subtracting non-quantum noises.

A comment regarding the excess noise - it seems clear that the excess mid-band DARM noise is not caused by / related to squeezing, because it's there even without squeezing injected. That said, squeezing seems to be having its own issues at these lower frequencies, below the DARM pole. Not clear how the worse low-frequency squeezing (after subtraction) could be a consequence of whatever causes the excess noise without squeezing. Likely different issues / things to be optimized happening at the same time/frequencies.

After PSAMS optimization with alignment scripts, it could be interesting to try a similar SQZ dataset with FIS.

Louis and I modified the aL1L3 filter (FM6) in ETMX_L1_LOCK_L by replacing the single pole at 50 Hz with a complex pair at 30 Hz, so that this offloading filter rolls off like 1/f above 30 Hz. Since the output of L1_LOCK_L is offloaded to HEPI, we suspect that this is what was causing the kick at EX. No DARM control is directly applied to EY, so that would explain why EY does not trip. The most recent lockloss did not trip anything.

It didn't trip the ISI, but we still are spiked up to about a third of the actuators drive range on the last lockloss. I'm less worried about tripping the ISI than I am about damaging something on the suspension or HEPI. attached trend shows the ISC signal to ETMX HEPI and the drives on the horizontal actuators at the last lockloss. 

Evan added a 0.1Hz low pass filter into FM3 of H1:LSC-X_ERR_TIDAL_ERR, and for symmetry into H1:LSC-Y_ERR_TIDAL_ERR. Turned on filters and sdfs.  We expect this to stop the kicks and not effect the slower tidal offloading: titdal to UIM is ~0.03Hz, to HEPI would be even slower  68749.

Since the 0.1Hz LP has been added, the two locklosses have not kicked the HEPI. See attached zoomed out and zoomed in plots. 

Naoki Nutsinee

We changed "fcgs_trans_lock_threshold" to 120 today so FC wouldn't lock on higher order mode. The filter cavity was having trouble getting pass GR VCO lock so we have now reduced this power back to 80.

Accepted the new ASC_INMATRIX settings in sdf, see attached. Aslo accepted ASC_POS_Y's new minus sign and AS_A_RF42_YAW_OFFSET.

Accpeted the SQZ_ASC WFS as OFF. As decided, we will start the ER with SQZ ASC IFO off.