FAMIS 31068

PMC REFL continues to rise. Since I didn't get a chance to bump up the ISS diffracted power last week and it's continued to fall, I'll adjust that opportunistically this week between lock stretches.

TITLE: 01/13 Day Shift: 1530-0030 UTC (0730-1630 PST), all times posted in UTC
STATE of H1: Lock Acquisition
OUTGOING OPERATOR: Oli
CURRENT ENVIRONMENT:
    SEI_ENV state: USEISM
    Wind: 6mph Gusts, 4mph 3min avg
    Primary useism: 0.10 μm/s
    Secondary useism: 0.47 μm/s
QUICK SUMMARY:

• We lost lock at 12:31 UTC from an earthquake (6.8 from Japan), currently relocking at DRMI
• 2ndary microseism is decreasing, wind is low

TITLE: 01/13 Eve Shift: 0030-0600 UTC (1630-2200 PST), all times posted in UTC
STATE of H1: Observing at 159Mpc
INCOMING OPERATOR: Oli
SHIFT SUMMARY:

IFO is in NLN and OBSERVING as of Jan 12 14:03 (16 hr lock!)

Very calm shift where we were locked the whole time. violins are somewhat high but don't seem to be increasing (EY1 likely needs new damping controls).

LOG:

None

At 16:37 Sun 12jan2025 VACSTAT detected a sensor glitch on PT132 BSC3.

I restarted the service at 17:30 to clear this alarm.

TITLE: 01/12 Day Shift: 1530-0030 UTC (0730-1630 PST), all times posted in UTC
STATE of H1: Observing at 160Mpc
INCOMING OPERATOR: Ibrahim
SHIFT SUMMARY: We've been locked for over 10 hours, ITMY 5/6 and ETMY1 remain elevated, I turned off ETMY1s damping after seeing it start turning around.
LOG: No log for this shift.

• 14:03 UTC Observing

TITLE: 01/13 Eve Shift: 0030-0600 UTC (1630-2200 PST), all times posted in UTC
STATE of H1: Observing at 160Mpc
OUTGOING OPERATOR: Ryan C
CURRENT ENVIRONMENT:
    SEI_ENV state: USEISM
    Wind: 4mph Gusts, 3mph 3min avg
    Primary useism: 0.06 μm/s
    Secondary useism: 0.64 μm/s
QUICK SUMMARY:

IFO is in NLN and OBSERVING as of 14:03 UTC the previous day (10 hr lock!)

Microseism is increasing and violins are elevated.

We've been locked and observing for over 7 hours, 2ndary microseism looks to have peaked.

Sun Jan 12 10:18:16 2025 INFO: Fill completed in 18min 13secs

TCmins [-106C, -99C] OAT (3C, 38F)

TITLE: 01/12 Day Shift: 1530-0030 UTC (0730-1630 PST), all times posted in UTC
STATE of H1: Observing at 160Mpc
OUTGOING OPERATOR: Oli
CURRENT ENVIRONMENT:
    SEI_ENV state: SEISMON_ALERT_USEISM
    Wind: 6mph Gusts, 4mph 3min avg
    Primary useism: 0.06 μm/s
    Secondary useism: 0.76 μm/s
QUICK SUMMARY:

• We've been locked for 1.5 hours
• 2ndary microseism is rising, wind is low

Got called at 14:15UTC due to assistance required. I *believe* the issue was that the NLN timer had run out, but by the time I logged on we were in Observing. We had been in OMC_WHITENING for about an hour presumeably damping violins. It looks like the total amount of time we were trying to relock after completing an initial alignment was almost 3.5 hours. The strange thing is that looking at the H1_MANAGER log, it looks like the NLN timer ran out at 12:42UTC, triggering H1_MANAGER to enter ASSISTANCE_REQ, but like I said, I did not get called until 14:15UTC. There was also something weird going on with the IFO_NOTIFY guardian too, when I logged on it was rapidly cycling between ALERT_ACTIVE and WAITING, although it looks like the cycling started once we got into Observing, so that might have just been due to no longer needing the alert (although I never noticed it doing that on other similar occassions but maybe that's just me missing it).

All is good and we've been Observing for 30 minutes.

Alerted by H1_MANAGER that it needed intervention. There was a large earthquake coming through that had tripped the ISIs for the ITMs and ETMs. Also, ETMY stages M0 and R0 had tripped (not yet sure if that is just due to the ISI tripping).

There was also a notification on verbals from 08:51UTC, a minute after everything tripped, that says "ETMY hardware watchdog trip imminent", but when I checked the hardware watchdog screen everything looked normal with no countdowns. Once it looked like the worst of the earthquake had passed and the ISI values were all within range, I reset R0 and M0 for ETMY and reset all four ISI watchdogs. We are still in LARGE_EQ mode so we haven't started relocking yet, but it looks like we are close to leaving and we should be good to go for relocking.

TITLE: 01/12 Eve Shift: 0030-0600 UTC (1630-2200 PST), all times posted in UTC
STATE of H1: Observing at 159Mpc
INCOMING OPERATOR: Oli
SHIFT SUMMARY:

IFO is in NLN and OBSERVING as of 02:36 UTC

Overall pretty quiet shift with one random lockloss (alog 82227). There doesn't seem to have been an EX glitch and the environment was stable (same microseism as last few days and low wind). We managed to get back up to NLN in 1 hour, with DRMI Locking and ENGAGE_ASC_FOR_FULL_IFO taking the longest to get past.

About 1.5hrs into the lock, SQZ unlocked and took us out of OBSERVING. It relocked automatically and we were observing again 4 minutes later.

LOG:

None

Unknown cause Lockloss - not environment and no EX saturation.

TITLE: 01/11 Day Shift: 1530-0030 UTC (0730-1630 PST), all times posted in UTC
STATE of H1: Observing at 163Mpc
INCOMING OPERATOR: Ibrahim
SHIFT SUMMARY: I ran the calibration sweep this afternoon, lockloss shortly after. I tried restarting the IOC for the lab dust monitors but it still couldn't connect to the lab2. "ISS diff power low" - DIAG_MAIN We've been locked for ~2 hours.
LOG: No log.

• 20:03 lockloss
  • ALSY couldn't lock  there was a "Find YARM by hand" notification, so I had to step in and I adjusted ITMY in yaw and it fixed it.
  • 20:42 I started an initial alignment
  • 22:35 UTC Observing, range was in the 140s (probably the OPO temp)
  • 23:00:04 - 23:01:36 I adjusted the OPO temp

TITLE: 01/11 Eve Shift: 0030-0600 UTC (1630-2200 PST), all times posted in UTC
STATE of H1: Observing at 156Mpc
OUTGOING OPERATOR: Ryan C
CURRENT ENVIRONMENT:
    SEI_ENV state: CALM
    Wind: 10mph Gusts, 6mph 3min avg
    Primary useism: 0.04 μm/s
    Secondary useism: 0.38 μm/s
QUICK SUMMARY:

IFO is in NLN and OBSERVING as of 22:34 UTC

Closes FAMIS 26353. Last checked in alog 82116.

All trends look well. No fans above or near the (general) 0.7 ct threshold. Screenshots attached.

We had just finished the calibration measurement 3 minutes earlier, 20:03 lockloss

Jeff K, Oli P

In alog 75947 part3, we had modeled the BBSS using what we thought were the correct Final Design parameters and then adjusted the d1 and d4 values* in order to get our model to match how the actual BBSS was responding. However, the parameter set we were using had the boolean variable stage2** set to 0, when the parameter set from the BBSS Final Design document(T2000503) had stage2 = 1. This was a change that I had made due to a misinterpretation.

Because of this mistake, the parameters that we had thought were correct were actually wrong, and so when we adjusted d1 and d4, we were unknowingly fitting to an incorrect model. These results were shared with the SUS group, where it was mentioned that stage2 was supposed to be 1.

We've now rerun the model with the actually correct Final Design(FD) parameters, and have adjusted d1 and d4 to fit our data. This adjustment has d1 = FD - 2.5mm and d4 = FD - 1.0mm, which are smaller changes than when we were using the wrong parameter set (d1 and d4 were increased by 5.0mm and 1.5mm respectively). Attached are our results in pdf form, along with pngs of the transfer functions for Pitch and Longitude, since those are the two most affected by the changes to d1 and d4. This adjusted parameter set can currently be found at $sussvn/Common/MatlabTools/TripleModel_Production/bbssopt_pendstage2on_d1minus2p5mm_d4minus1p0mm.m, but eventually will replace the bbssopt.m that is in that same directory.
Figure legend:
- Black - Model using the wrong Final Design parameters
- Dark blue - Model using the wrong Final Design parameters plus adjustments for d1 & d4 to match data
- Dark green - Model using the correct Final Design parameters
- Light Green - Model using the correct Final Design parameters plus adjustments for d1 & d4 to match data
- Gray - BBSS M1 Transfer Function data

Appendix:
* d{top, 0, 1, 2, 3, 4} is the vertical distance between the break off point of a wire and the CoM for each mass. These values are important for calculating the locations of fundamental modes for each DOF. There are two different types of d's, effective d's and physical d's.
    - The physical d's are the vertical distance between the break-off point of the wire and the CoM of the mass. These values assume an infinitely flexible wire, which isn't the case in real life.
    - The effective d's are the vertical distance between where the wire leaves the clamp and the CoM. Since the wires aren't actually infinitely flexible, they'll start bending before they reach the break off point, meaning we'll have a larger effective d as compared to the physical d.
These two relate to each other as  physical d = effective d - effective flexure point, with the effective flexture point being the extra distance between the clamps and the breakoff point.
In the building of the suspension model, both types are used at different times, so there is a need to be able to swap between them. That's what the stage2** variable is used for.

** stage2 is a boolean variable that determines whether the effective(stage2 = 0) or physical(stage2 = 1) d's are used.
    - If stage2 = 0, the d values are used as effective values, and there are no corrections made to account for the flex of the wires.
    - If stage2 = 1, the d values are taken to be the physical values, and so an effective flexture point is calculated and added to account for the flexing of the wires.
Since we entered in the physical d's, me changing stage2 to 0 meant that the model wasn't adding in the effective flexture point distances, changing our results.

I. Abouelfettouh, J. Kissel, O. Patane, B. Weaver

Executive Summary: The first article BBSS transfer functions look great. Though there is some confusion about the M1 P 2 P modeled transfer functions drastically disagreeing with the measured TFs, there is a consistent story between 
    - the adjustments to the mechanics that were made during construction and 
    - deviations from the "production" model parameter set that could re-create those construction adjustments. 
Further discussion will be had with the assembly / design team as to the future course of action.
Kissel suggests that -- even as the first article stands now -- the resulting measured transfer functions with the mechanical adjustments would/should happily meet A+ O5 requirements.

%%%%%%%%%%% Begin kLOG (You missed these...) %%%%%%%%%%%%%%

I got a debrief yesterday from Betsy, Oli, and Ibrahim of the comparison between 
    - measured transfer function results from the first article construction and
    - what had been deemed the production model parameter set for the BBSS,
i.e. what's discussed in LHO:75787.

The existing "production" model parameter set starts from Mark's update to the BBSS parameter set post-final-design after adjusting for the production wire thicknesses (see TripleLite2_mark.barton_20211212bbss_production_triplep.m, changed at rev 11625, circa Sep 2023). Oli successfully copied over to the usual matlab formating to create bbssopt.m (created at rev 11734, circa Jan 2024).

At the start of the debrief, there were (only!!) 3 outstanding issues / questions they had:
    (1) The overall magnitude scale for all DOFS for all measured transfer functions was a consistent factor of ~3.15x more than the model estimates,
    (2) After browsing through the EULER-basis drive to OSEM-basis response plots, and some of the off-diagonal EULER-basis showed little-to-no coherence, and
    (3) The measured M1 Pitch to Pitch transfer function's frequency response was significantly different than the model.

For (1), this is typically a sign of a mis-calibration of the data. We reviewed the calibration of the measured data from the processing script, plotBBSS_dtttfs_M1.m, created by Oli and Ibrahim in Nov 2023. The DTT templates that measure the transfer function use the pre-calibrated output of the sensors for response channels i.e. the channels come in units of microns and microradians, so they only need a factor of 1e-6 [meters / micrometers]. The only substantial thing that needs calibrated into physical units during post-processing is the excitation. The review of the calibration of the exciation revealed nothing suspicious in the script based on our current expected knowledge of chain
    - the test stand electronics (an 18-bit DAC = 20 / 2^18 [V/ct]), 
    - BBSS coil driver (a TTOP coil driver, coupled with a BOSEM coil = 11.9 [mA/V]), 
    - 10x10 magnet strength (1.694 [N/A]).
    - (lever arms and numbers of actuators are pre-calibrated out via the EUL2OSEM matrix, generated by make_susbbss_projections..m, and installed in EPICs)
The above factors result in an overall calibration of 1 / 1.5405 [(m/N) / (um/DAC ct) or (rad/N.m) / [urad/DAC ct]] that's displayed in the legend of each of the plots from LHO:75787.

In the end, we were more interested in understanding (2) and (3) rather than getting to the bottom of the calibration. Further, the test stand is some old, pre-aLIGO concoction whose records and modifications are unclear. So we figure we just move on, accepting that we need to fudge the data by the extra factor of 3.15x. We'll get serious about figuring it out if there's still such a discrepancy after moving the BBSS over to the production H1 system.

For (2), all concerns can we waived off with expectations. 
    (a) The first plot of concern was the P to F1F2F3 plot (page 17 of 2024-01-05_1000_X1SUSBS_M1_ALL_TFs.pdf), in that the magnitude of the F2 and F3 TFs were low and/or noisy. This is expected because F2 and F3 OSEMs are along the (center of mass / axis of pitch rotation) of the BBSS's top mass. So they see no pitch by construction (for better or worse).
    (b) The second collection of plots of concern were the off-diagonal DOFs, 
        (i) showing noise and/or 
        (ii) the opposite -- showing well-resolved cross-coupling in DOFs that we *don't* want cross-coupled. 
        We shouldn't be mad about (i) -- e.g. page 7 showing incoherence between L response to V drive and V response to L drive. What power is resolved in those transfer functions -- typically on/around resonances -- is because the TFs were taken undamped an in air. So there's just a ton of movement that an FFT might / cross-correlation might *think* is coherent with the drive, but it's really not.
        We looked closer at any of the off-diagonal TFs that *were* resolved, (ii) -- e.g. page 9 showing well-resolved cross-coupling between R response to V drive and V response to R drive. In each of these TFs, we found that the magnitude of the cross-coupling, off-diagonal TF was less, if-not-MUCH less that the on-diagonal TF, which is good. Where it was close, it sort-of "is what it is." Little attention has been typically paid to mitigating the off-diagonal transfer functions during the design phase of LIGO suspensions to-date. Further, they often are a result of the unique construction of each individual instantiation of the suspension type. There's no much we can do about it post construction, and what we *do* do if it proves problematic to the detector, is dance around the problem with fancy controls techniques if needed.

For (3), we arrive at the meat of this aLOG ::  The *model* of the M1 Pitch to Pitch transfer function looked very weird to me. Betsy mentions the during the construction of the first article they 
    (a) found a discrepancy between the fastener model vs. measured mass budget that resulted in an unclear relationship between the center of mass of each stage and their suspension points (typically called the "d" parameters)
    (b) acknowledged there would be uncertainty in the location of the suspension point for the bottom mass / dummy optic given the wire-loop + optic prism system since the final distances between masses have not been measured.

This, coupled with the fact that no *other* DOFs disagreed with the model besides P to P, led me to suspect the model parameters that only impact the pitch dynamics may be incorrect: 
    (i) each stages' separation between suspension point and center of mass, the "d" parameters, and 
    (ii) the pitch moments of inertia.

For a reminder of the physical meaning of all of the triple suspension parameters, see T040072.

As such, using the bbssopt.m "production" or "Final Design" (FD) parameter set as starting point, we tweaked these parameters by 10%-ish or factors of 2 to gather intuition of of how it would impact the response of the P to P transfer function. As a result, we have come to the conclusion that, in order to explain the data, we need to
    - increase "d1" by + 5 [mm]. This is the separation between the top (M1) mass center of mass and it's M1 to to M2 blade tip heights. In the absolute sense, this is increasing the "physical" d1 from -0.5 [mm] to +4.5 [mm], and
    - increase "d4" by 1.5 [mm]. This is the separation between the bottom (M3) mass / dummy optic center of mass and the wire/prism break-off point. In the absolute sense, this is increasing the "physical" d4 from +2.6 [mm] to +4.1 [mm].

Check out the attached plots which demonstrate this.
Citing discussion of overall scale (1) from above, all *measured* transfer functions have been scaled to the model by a factor of (1 / 3.15). This just makes comparing model to measured frequency response a lot more clear.

First attachment :: comparison between the final design model parameters and a variety of reasonable deviations of d1 between *decreased* by 2.5 [mm] and *increased by 5 [mm]. You'll notice that once d1 surpasses +1.0 [mm], the transfer function starts to look more like a standard triple suspension's transfer funtion. a d1 of FD + 5.0 [mm] lines up well with the upper two resonances of the measured data, but reduces the frequency of the lowest two L and P modes to below the data.

Second attachment :: comparison between the final design model parameters and a variety of reasonable deviations of d4. You'll notice that d4 really only have an impact on the lowest two L and P modes.

Third attachment :: comparison between the final design model parameters and a variety of reasonable deviations of the top (M1) mass' moment of inertia, the I1y parameter. 

Fourth attachment :: comparison between the final design model parameters and a variety of reasonable deviations of the middle (M2) mass' moment of inertia, the I2y parameter. 

Fifth attachment :: comparison between the final design model parameters and a variety of reasonable deviations of the bottom (M3) mass' moment of inertia, the I3y parameter. 

None of the modeled changes to the moment of inertia -- shown in the third, fourth, and fifth attachments -- show promise in reproducing the measured results.

Sixth and Final attachment :: comparison between the final design model parameters and one with only d1 increased by +5 [mm], and d4 increased by +1.5 [mm].

The modified model in this last attachment fits the data the best, so we conclude that the issues with mechanical construction (3a) and (3b) are consistent with the measured data :: the reconfigured mass budget needed from fastener issues resulted in a deviation from design value for d1, and the imprecision of the mass-to-mass distances and wire-loop / prism system resulted in a roughly ~2 [mm] slop for this assembly.

%%%%%%%%%%% End kLOG (You missed these...) %%%%%%%%%%%%%%

Big Picture Systems Level Commentary by Jeff :: If these measured transfer functions end up being the reality of the final frequency response of the BBSS -- this will be totally fine. The pitch isolation one gets above the resonances (defined mostly by the moment of inertia) is the same, the lowest L and P modes are sufficiently low, and the details of where the rest of the resonances land are totally inconsequential / amenable to a damping and global control design.