Inspired by Valera's alog demonstrating the improvement in the LLO sensitivity from O3b to O4a (LLO:66948), I have made a similar plot.

I chose a time in February 2020 for the O3b reference. I'm not aware of a good time without calibration lines, so I used the CALIB_STRAIN traces from both times. Our range today just got a new bump up, either due to the commissioning work today or the EY chiller work (72414). Note: I am quoting the GDS calib strain range!

I am adding a second plot showing O3a to O4 a as well, using a reference time from May 2019.

There is a significant amount of work that gave us this improvement. I will try to list what I can recall:

• ITMY replacement in 2021 to remove large point absorber
• Removal of septum window, new output faraday isolator, new OM2 TSAMS
• Addition of filter cavity for frequency dependent squeezing- squeezing levels around 3.7 dB (or more)
• Increase in circulating arm power ~200kW to 375kW
• Commissioning of camera servos- no more 20 Hz dither lines
• Reduction of ASC loop bandwidths, commissioning of HAM1 feedforward
• Reduction of LSC loop bandwidths, improvement of LSC feedforward
• Improvement of HAUX, HSTS and HLTS damping loops, reduced injection of BOSEM noise
• Damping of various baffles
• HAM1 table work to clamp RM blade springs, power attenuation on POP and REFL to allow power up
• Optimized test mass ESD biases to minimize electromagnetic interference coupling, improved grounding at end stations and corner to reduce ground loop noise
• Ring heater and CO2 laser commissioning to reduce frequency noise coupling and contrast defect

Closes FAMIS 26434, last completed in alog 71777

All BRS channels are within their nominal regions.

TITLE: 08/25 Day Shift: 15:00-23:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Observing at 150Mpc
INCOMING OPERATOR: Austin
SHIFT SUMMARY:

H1's been locked 53.5hrs (1hr from our O4 longest lock *knock on wood*).

Main news of today was the H1 Chiller Pump#1 going down.  Tyler as well as a contractor spent most of the last 6hrs dealing with this and we are currently set and running with Chiller Pump #2.

There was 2hrs of commissioning by Gabriele & Elenna as well.

We appear to have a ~5Mpc increase in range....waiting to hear if this is due to the commissioning or working with different chiller pump!

LOG:

• 1707-2240 EY visit to address HVAC chilled water issues with incursion to Mechanical room (tyler)
  • 1814-1826 quick run to EX to compare chiller.
  • Richard at EY
  • Contractor contacted to come out to work on chiller
  • 1838 Richard walking into electronics bay to check on temperature
  • 1839 Propping door between E-Bay & Garb Room (Richard)
  • 1850 "Move & Cool" Portable air cooler being forklifted into the EY E-bay (the area you enter (tyler, chris, randy)
  • Ryan autoscaled the temperature scope on nuc32
  • 2011 Mac Miller (hvac contractor) starts driving down to EY with tyler
  • 2041 Heading out to try to turn on more AC units (i.e. EY Lab room) [Richard]
  • 2240 Done for the weekend.  Tyler said we are looking good and returning to a normal state with only Chiller Pump2.  (Chiller Pump1 still needs to be looked at.  In n
• 2018 COMMISSIONING---LSC & MICH Feed Forward measurements by Gabriele & Elenna

Recently we've been receiving groups of SubGRBs or long grbs from our igwn-alert subscription system that are often old events or repeated. Since we don't normally react to these and will ignore them, we don't need to even have them update in epics. I've added more filtering for these types of events in igwn_response.py and also in the tests.py for VerbalAlarms. The new code is running for both of these systems.

Another related issue I'm looking into is why we haven't received any superevents from out igwn-alert listener in maybe a month. I don't think the issue is on our end, but I have a few more checks. We still get phone calls in the control room, so operators are still informed.

TITLE: 08/25 Day Shift: 15:00-23:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Observing at 159Mpc
OUTGOING OPERATOR: Tony
CURRENT ENVIRONMENT:
    SEI_ENV state: CALM
    Wind: 13mph Gusts, 9mph 5min avg
    Primary useism: 0.02 μm/s
    Secondary useism: 0.06 μm/s
QUICK SUMMARY:

- H1 has been locked for 53.5 hours, all systems appear stable

Jim installed and updated the HAM1 HEPI feedforward, which he describes in alog 72393. The HAM1 to ASC feedforward has not been updated in some time, and Jim's results showed some difference in CHARD P from his HEPI FF update. Jenne and Jim turned off the HAM1 ASC FF a few days ago (72395) to gather data for retraining.

I used data from that time to retrain the HAM1 to CHARD P, PRC2 P, and INP1 P feedforward. All new filters are labeled with today's date "0825". I turned on these new filters and saw small improvement in the CHARD P and INP1 P error signals. There was no evident improvement in DARM. We are not limited by CHARD P in DARM at this time so I am not that surprised (see third attachment in 72245).

The attached plot shows all ASC REFL loop error signals. The blue trace is the previous HAM1 feedforward in use, and the red live trace shows the new feedforward on.

The filters are not guardian controlled. I updated the SDF observe file in the SEIPROC model. I then asked Corey to take us to the "SDF to SAFE" guardian state so I could also update the SEIPROC safe file. However, that guardian state does not change the SEIPROC model to safe, so I had to change to the safe table by hand. Here are the steps:

"SDF restore screen">"! select request file">choose "safe.snap">open>"load table". Once the diffs are accepted and confirmed, follow those steps to load the "observe.snap" file. I took a screenshot of the safe SDFs I accepted before hitting "confirm".

[Elenna, Dan, Gabriele]

We tested a filter (FM8 in DARM2) that increased the DARM gain below 3-4 Hz, where most of the RMS is accumulated.

The DARM RMS is reduced by a factor of 3. There is no immediate evident effect on the DARM noise aboev 10 Hz. However, it would be useful to test this new filter for a longer time in the future.

Weekly FAMIS Task

1. Alerts from running the script:

ITMX_ST2_CPSINF_H1 high freq noise is high!

2. All other "floors" from the spectra look normal.

[Elenna, Gabriele]

We tried increasing the SRCL gain by a factor of 2. As expected SRCL_IN got better, SRCL_OUT did not change. No effect on DARM RMS.

Gain is back to nominal

As the title says, we retuned the MICH feedforward, and the new filter performs better at all relevant frequencies.

Guardian has been updated to engage FM9 instead of FM8.

Quoting Elenna: "It's been 0 days since we retuned the LSC FF"

J. Kissel, for T. Guidry, R. McCarthy

Just wanted to get a clear separate aLOG in regarding what Corey mentioned in passing in his mid-shift status LHO:72423:

The EY HVAC Air Handler's chilled water pump 1 of 2 failed this morning 2023-08-25 at 9:45a PDT, and thus the EY HVAC system has been shut down for repair at 17:35 UTC (10:35 PDT). The YVEA temperature is therefore rising as it equilibrates with the outdoor temperature; thus far from 64 deg F to 67 deg F.

Tyler, Richard, and an HVAC contractor are on it, actively repairing the system, and I'm sure we'll get a full debrief later.

Note -- we did not stop our OBSERVATION INTENT until 2h 40m hours later 2023-08-25 20:18 UTC (13:18 PDT), when we've gone out to do some commissioning.

J. Kissel, L. Dartez

I'm attaching a plot of the DARM UGF that compares two measurements taken ~3.5 months apart. The blue trace is taken from the calibration sweep that Ryan C. took a few days ago (LHO:72392) and the orange trace is from a similar sweep taken back in May (20230506T170817Z). 

The DARM loop UGF has moved from ~58Hz to ~66.4Hz and the phase margin has increased by about a degree since May. 

There is no immediate need to adjust the DARM open loop gain (DRIVEALIGN_L2L gain, as mentioned in LHO:72416).


The fact that we don't see the loop dip near 20Hz is a rough indicator that the [actuation stage] crossovers are stable. I'll be following up with a more in-depth look at that.


The script used to generate the attached plot lives at: /ligo/home/louis.dartez/projects/20230825/plot_olg/plot_olg_meas.py

H1 is currently at almost 50.5hrs for a lock (current record is ~54hrs).

We are approaching 1pm with the start of 2hrs of commissioning.

We are also dealing with increasing temperatures at EY due to HVAC chilled water issues.  Contractor is on their way.

J. Betzwieser, L. Dartez, J. Kissel

Ryan Short recently updated the control room FOM for the BNS range (LHO:72415) which now shows -- with a clear legend -- the range computed using CAL-DELTAL_EXTERNAL_DQ vs. GDS-CALIB_STRAIN_CLEAN for both H1 and L1 observatories -- see example attached. 

This makes it dreadfully obvious that "L1's DELTAL_EXTERNAL range is right on top of the CALIB_STRAIN range -- but H1's is not, and DELTAL_EXTERNAL is *higher*." -- see First Attachment from the control room FOM screenshots.

The natural questions to ask then are "why?" "is something wrong with H1's calibration?"

No, there's nothing wrong.***
The discrepancy between DELTAL_EXTERNAL and CALIB-STRAIN at H1 is because the static test-mass stage actuation strength hasn't been updated since 2023-05-04 -- before the observing run started -- and it has slowly drifted due to test mass ESD charge accumulation -- and it's now at 8% larger than the May 04 2023 value. See the current value for the past 24 hours and a trend of the whole run thus far. L1's ESD strength has *not* drifted as much (see similar L1 trend), and they also regularly "fudge" their DELTAL_EXTERNAL actuator strength gains in order to get DELTAL_EXTERNAL more accurate (and they do so in a way that doesn't impact GDS-CALIB_STRAIN). H1 has chosen not to, to date.

This drift is tracked and accounted for in our "time dependent correction factor" or TDCF system for that test-mass stage actuation strength, \kappa_T -- and GDS-CALIB_STRAIN (and STRAIN_NOLINES, and STRAIN_CLEANED) all have this correction in place. Check out the Second attachment from the same day's "CAL" > "h(t) generation" summary page, and walk with me:
This plot is showing the ASD ratio (and thus roughly analogous to the magnitude of the transfer function) between all of the various stages of the calibration pipeline.
    - GDS-CALIB_STRAIN, GDS-CALIB_STRAIN_NOLINES, and GDS-CALIB_STRAIN_CLEANED are all this same from this perspective. Thus the ratio between these three channels with DELTAL_EXTERNAL in the denominator is highlighting the DELTAL_EXTERNAL is a preliminary product, and NOT corrected for TDCFs and thus there's a huge ~16% systematic difference between the two "stages" of product.
    - Recall that *all* of the four paths of the calibraion -- UIM, PUM, TST, and Sensing -- are being summed, and the cross-over frequency for these sums are all culminating around 50-200 Hz -- and in that region there's a factors of 2x to 3x gain peaking (see e.g. Figure 4 of P1900245) -- and thus the 8% drift in the TST stage strength means 16% systematic error in the DELTAL_EXTERNAL calibration.
    - However, the front-end version of the preliminary product that is corrected for TDCFs is also shown in the plot -- CFTD-DELTAL_EXTERNAL. The ASD ratio between this channel has MUCH less systematic discrepancy -- indicating that correcting for time-dependence (get it? CFTD!) does a LOT of the heavy lifting of accounting for this 8% TST drift. 

Of course, these ratios of different portions of the calibration pipeline don't *really* tell you if you've *really* done the right thing in an absolute sense. They only tell you what changes from step to step. (And indeed, the CFTD-DELTAL_EXTERNAL to GDS-CALIB_CLEANED ratio still shows *some* discrepancy.)

The fact that fifth attachment, from the archive showing the constant *direct measurement* of the systematic error in the calibration -- from the absolute reference, the PCALs -- is nice and low (i.e. the transfer function is close to unity magnitude and zero phase) indicates that all of the correction for time-dependence is doing the right thing.

*** Yet. In O3, L1 suffered a lot from TST strength drift. Joe has shown repeatedly that if you let an actuator TDCF drift too far beyond 10%, then the approximation we use to calculate these TDCFs breaks down (and see Aaron's work discussing it as a motivation for P2100107). In addition, since the real ESD strength is changing -- :: -- which is corroborated by the in-lock charge measurements -- I think -- see highlighted red region of sixth attachment from LHO:72310 -- :: -- that means the DARM open loop gain TF gain is also changing. 

This may impact the DARM loop stability (see e.g. LLO aLOGs 50900 and 50639). So, *eventually* we should resurrect the two things we've done in O3:
    (1) Reset the model of the static actuation strength for the TST stage to a more current value. (And thus start a new calibration epoch)
    (2) Potentially change the actual DARM loop by adjusting the DRIVEALIGN_L2L gain
    (3) Work up a solution to mitigate the drift -- perhaps doing something similar to what was done in O3, and play gains with turning on the ESD Bias voltage with the opposite sign when we're not in observing.

Following up on TJ's alog from 8/17 (72293), it was noted that a garbage truck at LSB was quite loud. The seismometers in the LVEA clearly picked up the noise, and it seems to coincide with noise in DARM. It's hard to be certain with only one signal, but it probably warrants further investigation.

Spectrogams attached of one seismometer and DARM for 2 minute time span around when noise was reported:
-Of signal
-Zoomed in with boxes on noise regions
-Stacked with some boxes around correlated noise