TITLE: 06/05 Day Shift: 1430-2330 UTC (0730-1630 PST), all times posted in UTC
STATE of H1: Planned Engineering
OUTGOING OPERATOR: None
CURRENT ENVIRONMENT:
    SEI_ENV state: MAINTENANCE
    Wind: 6mph Gusts, 3mph 3min avg
    Primary useism: 0.05 μm/s
    Secondary useism: 0.14 μm/s 
QUICK SUMMARY:

ISS alignment final steps, BSC in-chamber of activities at BSC1+2, HAM3 SPI, viewport inspection are among activies on the docket today.

Alexandra, Shoshana, Arnaud, Fil, Dean (remote)

Today we completed the balancing of the CRS with a period of 61s (or 16.4mHz, which is our target frequency), aligned and tuned the HOQIs (with >75% fringe visibility) and connecterized the picomotor to the mighty mouse connector. Tomorrow, after verifying the balancing and alignments have not changed, we will mount the HOQIs to the CRS, which will complete this first assembly. 

In detail : 

• Alexandra fixed the miswiring of the monitor signals on the HOQI board which allows us to connect those outputs directly to the oscilloscope
• Fil wired a 18V power cable to the interlock so it can be used for the laser driver. This allowed the transition of the H2-PSL to laser hazard.
• Shoshana rebalanced the CRS to achieved the target frequency of 61s. picture 1 shows the vertical mass arrangement : x2 8-32x3/8 on the back side, and x2 8-32x3/8 on the front siide +2 washers. picture 2 shows the measured resonant frequency.
• After deciding to use the SPI spare pico channel, the picomotor was connecteriized to the might mouse connector via an extension cable (red (+) wire to pin 1, and white (-) wire to pin 2). Thanks to Jeff for offering the components and tooling. (picture 3)
• x2 HOQIs (x1 Left and x1 Right versions) were setup on the optical table with a fixed target to align the various components. We tested with the laser driver knob at 6.3 and an output monitor voltage of 9.3V, which corresponds to ~20mW of laser power. We achieved the following fringe visibility: 
  • HOQI Left : PD1: 91%, PD2: 88%, PD3: 86% (Picture 4)
  • HOQI Right: PD1: 86%, PD2: 78%, PD3: 83% (Picture 6)
• The alignment was done as follow :
  • Rotate the Fiber Coupler to maximize the power out of the first PBS (in the hoqi)
  • Align the Fiber Coupler in Yaw to center the beam on the 3 PDs
  • Align the HOQI baseplate to the target so the retro beam is centered on the 3 PDs - at this point we should see fringing when tapping on the table.
  • Continue aligning to the target to maximize fringe amplitude
  • Optimize the Lissajou figure with the 2 waveplates (see video for Lissajou figure)

Note1: We did break a set of flexures during balancing today (SN 18 and 11) 
Note2: We could not get good fringe visibility on a 3rd HOQI (max 60%) we tested. We will have to get back to this one.
Note3: The Right HOQI has a PD wire that seem to have been pinched during assembly (see last photo). This doesn't seem to be an issue, we will just have to be careful when routing that it doesn't touch the baseplate.

Belated, as usual.

Things make more sense, no railing of RM2

We continued HAM2/3/1 alignment. Things are making more sense now after realizing that RM1 and RM2 slider polarity was wrong but we're still finding things.

1st attempt: We rotated IM1 to relieve RM2 YAW (IM1 y -987->-687 i.e. +300 from Tuesday value), aligned PRM so the beam retroreflects (but didn't touch up other IMs in the interest of time). After some confusion it worked and we were able to center REFL ASC sensors without railing RM2. Happy. (See ASCREFL_centered_RMs_dont_rail.png. If you do the math using flash peaks, REFL_A PIT=0.034, YAW=0.013, REFL_B PIT=-0.0097, YAW= -0.017, so it was very good.)

2nd attempt: Encouraged, we proceeded to center the beam on the IFI output baffle using IM2, steer the beam to the nominal beam position in front of PRM using IM3, steer the beam to the nominal position in front of PR2 using IM4 (which moves the beam position on PRM by a small amount but we didn't bother to iterate), and finally aligned the PRM so the retroreflection is restored. We were able to center REFL ASC sensors without railing RM2. Happy again.  (eod.png, REFL_A P= 0.034, Y=-0.009, REFL_B P=-0.023, Y=-0.022.)

We looked at the beam position on IFO REFL baffle and it was off in YAW,  so we changed the baffle position slightly in +Y direction.

Large difference in RM2 DAC output for two attempts, 80% or DAC range used in the 2nd attempt

We didn't change IM1-IM2 line, which means that the alignment into HAM1 should not have changed assuming that the PRM was retroreflecting, but the YAW offset necessary for RM2 to center the ASC sensors in the 1st and 2nd attempt were very much different.

In the 2nd attempt, RM2 Y offset changed by negative 500urad and the DAC output of two of the RM2 coils reached ~107 million, which is about 80% of the DAC range. That's closer than I'd be comfortable with before closing down the chamber.

That's consistent with the retroreflection error 

It turns out that this is consistent with our retroreflection accuracy which I claimed to be "like +-50urad" in PRM rotation (not the beam rotation) in alog 90451.

Equivalent of positive PRM rotation of 50urad in YAW is replicated by the combination of physical negative 55urad for RM1 and physical negative 380urad for RM2 (note that the sliders for RMs as of now has the opposite sign as the physical rotation). 

If retroreflection is off in terms of PRM YAW rotation by +-50urad, when the REFL ASC sensors are centered, RM2 YAW would be off by +-380urad from what would be required to center the beam that is truly retroreflecting. See attached script. 500urad difference between two attempts is consistent with that.

On top of that, it's possible that +-50urad error estimate is too optimistic. We set the PRM angle by centering an iris (placed between IFI input and IFI HWP) to the forward going beam and centering the back propagating beam on the back of the iris. The basis for the accuracy is that we could start seeing how the circumference of the back of the iris is unevenly illuminated by the back-propagating beam when we gave the PRM 50urad offset, but the beam is always moving in YAW and sometimes rather slowly, so we have to eyeball the average beam position.

We don't know if the 2nd attempt was closer to the true retroreflection or, for that matter, if the true retroreflection is "on the other side of" the 2nd attempt relative to the 1st.

As such, it's prudent to relieve RM2 YAW offset further.

Tomorrow

Relieve RM2 further by giving IM1 a positive 300urad YAW rotation. IM2, IM3, IM4 and PRM should be readjusted accordingly.

IM1 YAW offset hasn't changed much in the past 2 years except for this week, and we'll go back to that neighborhood. Even though this means that the beam on IFI input baffle will be off-centered (probably it's been like that for years),  I'm absolutely sure that the forward-going beam won't be clipped, I'm also quite sure that the clipping-like thing of the IFO REFL beam on the IFI input baffle won't come back as far as PRM is retro-reflecting.

We might have to move the IFO refl baffle again.

We WILL move the IM4 baffle because whenever the beam is aligned to the nominal position in front of PRM the beam is too close to the +X edge of the baffle.

(Betsy, Randy, Jim, Tyler, Travis, Jordan, Gerardo, Mitchell, Corey, Ibrahim, Danny on watch)

Ibrahim and crew prepped the BBSS yesterday for flight. 

This morning Jim and I finished cable routing and stowing and made some final pre-flight preps.

This afternoon, we successfully flew the BSC2 Cartridge from the West Bay Test Stand to the BSC2 chamber and installed in onto the support tubes via the dome.

• We had a pre-flight meeting just before lunch, then met at 1pm to gear up and start.  
• Moved the West Bay cleanroom South to allow for cartridge lift off.  Cartridge was covered via the multiple C3 cover layers (ISI cover+BBSS cover+Cartridge mega cover top to bottom over the whole thing which then was rolled up as the cartridge lowered into the chamber.
• The cartridge was lifted off the test stand, checked for good level but no adjustment to rebalance the lifting fixture was needed - the setting used previously was good. Yay.
• With the load cell, the cartridge weighed 9430lbs (pretty spot on for what it weighed at de-install flight as a BSFM).  It was then set back down, load cell removed and flown the pre-planned path East over the X-Arm Beamtube manifold section and around diagonally onto BSC2. (Stays still not included, nor the HRTS for BHD, nor the HR/AR BBSS baffles, nor 1 of 4 vibration absorbers - basically the same ish payload as before with the BSFM weighing not much more than the BBSS plus or minus some bits at flight)
• The hand rails with the notch worked well (switched in by Randy earlier this week) - no handrail sections needed removal as the cartridge passed right over this time.  During the BSC chamber install the Platform tent was pulled back to allow for craning.  Particle counts near the fans on the platform were 0-100 before the tent was pulled back.
• By 2:45pm the cartridge was landed on the support tubes in the  BSC2 chamber. The tent was restored and the ISI left covered at the dome.

Tyler, Travis, Jordan, Corey, Jim in suits at the Test Stand for lift off

Randy on the Crane controls

Mitchell, Gerardo, Betsy in support

Jim and Ibrahim in the chamber for the lowering, alignment and support tube bolt attachments

Danny on Safety watch

More pictures to follow, but attaching a good one of the LVEA perspective actross the West bay and of Danny monitoring the flight.

Well done all, big job complete.  Now tune-up, align and close!  Easy button.

TITLE: 06/04 Day Shift: 1430-2330 UTC (0730-1630 PST), all times posted in UTC
STATE of H1: Planned Engineering
INCOMING OPERATOR: None
SHIFT SUMMARY:

The LVEA is LASER HAZARD

BSC2 cartridge flown from test stand and into BSC2 DONE
SPI PSL pickoff shutter installed DONE
HAM2 B+K measurement DONE

LOG:                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            

When I hammered the below table "beard" baffle on May 26 (alog90335), something wasn't correct with the way I saved the data or my configuration or something else. Either way, Jeff and went in today and hit that as well as the SPI ISIJ assembly with and without the cylindrical shroud (D2500030). Preliminary results are looking much better, so I'll clean those up and post them soon with a whole writeup.

In 89751 we found we could not trust the FC1 top mass osems for an accurate FC1 alignment, we then set an FC REFL iris on SQZT7 to help us capture correct FC1 alignment 89771, as we thought we couldn't trust any stage of the FC1 suspension. But this iris also depends on ZM1,2,3 which is confusing.

Since the FC1 issues this week, we checked that the sliders are working as expected 90466.

We have decided to not use the FC REFL iris to set FC1 as it puts FC1 in a very different position to O4.  We will align FC1 using the slider positions from O4. This is as in the December vent, between the O4 under-vacuum and the vents, doors off time, the FC1 WIT channels were within ~10-20urad, see top two rows of table below. Additionally in February when we had our FC1 trouble, we though that we would have been fine if we had never touched the FC1 sliders.

From August to November, the FC1 M3 Pitch WIT drifted between 175 and 195urad so where we are currently at 170urad is not far outside of this range. 

* drifted down to 162 over next week

We have put ZM1,2 3 back to their O4 sliders as a starting place, you can see below that the top mass osems say this is not the same location as when under vacuum but we don't think this is trustworthy. We will change these alignments as needed to get the FC beam to retroreflect off FC1.

*WIT is M3 WIT for FC1 but for ZMs there is no bottom mas WIT so we're using M1_DAMP_{P,Y}_INMON
 

R. Short, J. Oberling, J. Kissel

This afternoon we installed the shutter in the SPI pickoff path on the PSL table.  We installed it in between the 2 SPI lenses.  The cable was tied down with some dog clamps and cable ties, and we checked to make sure it was well clear of all known beams (it doesn't really look like it in the picture, but the shutter cable is ~3 inches away from the ALS beam and cannot drift into its path).  Once we ran the cable outside the enclosure, Jeff plugged it into the driver and we ran several on/off tests, with the shutter successfully actuating as expected, every time.  I'll update the as-built PSL table layout to include the shutter.

[Tom, Oli]

Summary: There is a polarity flip between the BOSEM and QOSEM coils due to the wiring of the QOSEM flexi-circuit. To maintain consistency with existing BOSEM controls a sign flip should be implemented in CDS coil drives.

When commissioning the damping loops for the BBSS on the teststand, we noticed that there was a sign flip required in the QOSEM coil drive for the loops to be stable, when copying over the existing BOSEM damping. See LHO alog 90441. After some investigation, this sign flip is due to a polarity flip in the wiring from the coil drivers to the QOSEM coils. The QOSEM coils are wound in the same direction as the BOSEM coils, however on the QOSEM flexi-circuit I accidentally flip the coil connections with respect to that of the BOSEM. All QOSEM version to do date suffer from the coil sign flip. 

In the attached screenshot, we see the BOSEM flexi-circuit has this weird flip between the coil pads (on the very left) and the uDB9 connection. I did not copy this flip over to the QOSEM flexi, which has caused this coil polarity flip.

J. Oberling, P. Thomas, R. Short

TLDR - Following the power outage(s) over the weekend, we've successfully restored the PSL.

Monday morning, Jason started by turning back on the benchtop power supply behind the PSL racks in the LVEA which powers CB2 and then restarted the Beckhoff computer in the diode room (we've seen in the past that the Beckhoff computer needs everything to be powered on before it boots so that it can "see" everything). However, even though the EPICS IOC auto-started when the computer came up, the PLC did not, so we were unable to bring the system up at first. I was able to find and launch the correct PLC (which, for our future reference is in the "Projekt_LIGO_240124" folder), but we noticed after the GUI came up that we were missing the amplifier PD WD, one of the newest additions to the software. Thinking we had an old version of the software somehow, we called in Patrick, who reminded me that instead of just starting TwinCAT in run mode, I should have also logged in and started the PLC to pull in all the updates to the code. This brought the Beckhoff computer back to full functionality, and we were able to bring the laser up without issue after clearing the site laser interlock (which wouldn't clear until the Beckhoff computer was functional). As usual, our calibration settings and operating hours for the NPRO, amps, and chiller were lost with the computer reboot, so we found the last saved screenshot of the settings table we took after the last outage (alog88368) and used it repopulate the table. The operating hours were all fixed also since we could trend back what they were before the outage. A new screenshot of the settings table is attached.

Relocking the PMC went without issue, and it's now been coming up to temperature since Monday with the transmitted and reflected powers slowly settling out. I can tell by the camera spots and the lower transmitted power that some alignment is needed into the PMC, which we can do eventually. The PMC relocked at a temperature a couple degrees cooler than before, which we've seen before, so this slightly different operating point isn't too surprising.

Trending the enclosure temperatures shows that the environmental controls seem to have come back no different than before the outage.

Before starting the cage baffle install we took a reference of the optical levers (with ISI locked) such that later when we unlock the quad we can tell if there was a significant shift. NB there were small shifts at LLO when we installed these baffles and we are still baffled as to why. Since they attach to the cage, which is connected to the ISI and the ISI gets rebalanced, it's unclear why they'd shift more than some 20-30 urad. 

I took one reference pre-vent, circa Nov 28, when we were still locking full ifo. The second reference is recent, at air, with the ISI locked.

Ibrahim, Oli, Thomas, Betsy, Jim, Mitchell

• ISI rebalancing masses
• BBSS stages locked
• General torque checks done
• BRD Measurements re-taken - these were too noisy so will re-re-take in-chamber
• Some cabling done
• Face sheilds secured for cartridge
• Earthquake stop replaced (glass on fluorel tip was chipped)
• Calibrate BOSEM vs QOSEM vs Slider Values
• BBSS covered for cartridge flight

Here's the analyzed data for the QOSEM transfer functions, as well as a comparison of transfer functions between the BOSEMs and the QOSEMs for the BBSS on the test stand. They're pretty similar.

I didn't modify the analysis script yet since we are currently only using the X of the QOSEMs, so I just exported the M1_OSEMINF_{F1,F2,F3,LF,RT,SD}_X_OUT_DQ channels in theplace of the M1_OSEMINF_{F1,F2,F3,LF,RT,SD}_OUT_DQ channels.

Data
/ligo/svncommon/SusSVN/sus/trunk/BBSS/H1/BS/SAGM1/Data/2026-06-01_1700_tfs/2026-06-01_1700_H1SUSBS_M1_WhiteNoise_R_0p02to50Hz.xml
r13024
Results
/ligo/svncommon/SusSVN/sus/trunk/BBSS/H1/BS/SAGM1/Results/2026-06-01_1700_tfs/2026-06-01_1700_H1SUSBS_M1_ALL_TFs.pdf
r13024
/ligo/svncommon/SusSVN/sus/trunk/BBSS/Common/Results/allbbss_BBSS_teststand_BOSEMsvsQOSEMs/allbbsss_BBSS_teststand_BOSEMsvsQOSEMs_ALL_TFs.pdf
/ligo/svncommon/SusSVN/sus/trunk/BBSS/Common/Results/allbbss_BBSS_teststand_BOSEMsvsQOSEMs/allbbsss_BBSS_teststand_BOSEMsvsQOSEMs_ALL_ZOOMED_TFs.pdf
r13025

Shoshana, Alexandra, Jim, Arnaud, Fil

Yesterday's activity log: 

• Shoshana and Jim suspended the first CRS with flexures and after a first round of balancing obtained a resonant frequency of 37mHz (using the Panasonic optical interferometer -HL-G105- to read the frequency, see picture). We had to add x8 8-32 screws (8) at the bottom of the CRS to compensate for a high COM. As per the SW model the balance mass type 2 in d230099 used in the mass adjuster, and the new coupler (-v4) move the COM up by 43um (compared to the MIT build). We will swap d230099 to type 1, which will help reduce the number of Ti screws we have to use.  Table below shows the COM position wrt to the flexure center point with different mass adjuster configurations:

• Alexandra confirmed the signal chain from each HOQI was working with the TIA we received from Caltech. The HOQI board D2300154 has some minor wiring issues on the monitor signals, but we can use the TP internal to the board instead for now. 
• Fil has started wiring the interlock system to the H2-PSL enclosure so we can turn on the laser.

A. Effler, R. Schofield, R. Short

All of the new SLiC baffles have been installed on both ITMX and ITMY (both the test mass and compensation plate sides), but the CP side of ITMX still will need some alignment positioning touched up. Earthquake stop brackets on the test mass faces for both quads were also swapped out to accommodate for the new baffles.

One person always remained near the open door in BSC1 while others were around the corner working in BSC3, and oxygen monitors were worn out of an abundance of caution.

Sheila, Begum, Ryosuke, Camilla

Sheila checked that FC1 is now working as expected but was not in May 90462.

Started the morning by setting FC1 to the FC REFL iris ZM1,2,3 sliders set to 90183  (when Sheila had the beam on SQZT7), we then set irises after ZM1 and ZM3. Doing this clipped the return beam on B:L1 aperture, this is expected as we could not get the return beam through the OPO when FC was aligned in Feb.  Later we realized that this doesn't make scene, so set ZM1,2,3 sliders set to when we had good FC flashes in  89694 and then FC1 was moved to get the beam centered on the  FC REFL iris, again this clipped the return beam on B:L1 aperture but the beam was considerably higher at the ZM3 iris than when we originally set it. Around or a little higher than 6.25". Tomorrow we will fully think this though and compare to O4 times. 

We locked the OPO SUS iteratively looking at the beams off the OPO. It was finally lock with the FC REFL beam getting to the second iris but not centered and when we changed the alignment to the IR SQZT7 alignment, the beam made it through both irises when closed. 

Sheila, Begum, Ryosuke then took beam profiles of the beam after ZM1 and ZM2 with the borrowed Phasics camera.

Before locking down the ITM suspensions as part of the new baffle installations, we measured the positions of each test mass and compensation plate (CP) from the outer bevel to the inner cage:

EDIT: clarifying that "left" and "right" here refer to when looking at the face of either the test mass or CP, so left and right flip when going between TM and CP on the same suspension.

ITMY

Compensation Plate -

• Left: 9mm
• Bottom: 24mm
• Right: 8mm

Test Mass -

• Left: 14mm
• Bottom: 22.5mm
• Right: 16mm

ITMX

Compensation Plate -

• Left: 8.5mm
• Bottom: 20.5mm
• Right: 8mm

Test Mass -

• Left: 13mm
• Bottom: 20mm
• Right: 16mm

See attached photots for Anamaria's sketches of nominal positions and how they relate to the baffles.

J. Kissel, S. Koehlenbeck [remote], J. Wright, M. Simmonds 

Here, we post an aLOG to cover the whole week's slog, picking up where we we left off at that end of last week (LHO:90332). 

%%%%%%%%%%%%%
Executive Summary: the remaining outstanding issue to solve before instal for the SPI pathfinder is the irregular and intermittent drift in heterodyne efficiency in the SPI's MEAS IFO (item (3) from the list of issues after finishing re-assembly up to D2400107-v5 after ECR E2600106). Importantly, we've seen this issue in the MEAS IFO even with the previous version of the assembly back in Mar 2026.

After a week of hard work, we've ruled out just about everything that it could be *except* for flaws in the alignment system of the picomotor-actuated IXM100 mirror mounts which steer the MEAS beam to the MEAS IFO (see D2400107 for the annotated layout).
%%%%%%%%%%%%%

I spend the rest of the aLOG summarizing what we've ruled out and how. Remember -- thru out all of these studies, the REF IFO remains a rock-solid high efficiency, and the power monitor PDs so the input power on the MEAS and REF is quite stable.
And a reminder for those not-so-versed in heterodyne interferometers, the primary metric is the efficiency, defined as 
    eta = [peak-to-peak beat note amplitude] / (2 * beat note mean)
(nominally 100% if the mode overlap between two interfered beams is perfect).
"Drift" of this efficiency means that -- without 

Electronics Issues with the IFO MEAS PDs.
    - Suspected cause: "maybe the bias voltage is drifting?" "Something's going on with the electronic ground because we're only looking at the positive leg?"
    - RULED OUT: we swapped the REF B PD (which had been measuring a rock-solid, high efficiency from the REF IFO) into the MEAS B PD position, and had it measure the MEAS IFO beam. The drift was measured by the REF B PD as well. This convinces us that "it's on the incoming interfered beam.
    - PLUS: The bias voltage is supplied by a single circuit for all PDs in the ISIK transceiver system, so the bias-related issues would be common to REF or MEAS PDs.

Spurious drive on the the picomotors
    - Suspected cause: "maybe the ambient electric field in the room is driving the floating pins of the picomotors, pushing the mirrors around?"
    - RULED OUT: At the start of the week, the picomotor signal chain was hooked up as designed from motor might-mouse to quadrupus to mock feedthru and out a ~6 ft D25 in-air cable that had been used for testing. 
       (a) Just like we'd already done with the PDs, we hooked up ever single picomotor pin to rack power supply ground via a breakout board and clip leads at the end of the in-air cable. Still see drifts. 
       (b) We also tried simply disconnecting all picomotors at their might-mouse connector. Still see drifts.

Clipping along the MEAS path into the MEAS IFO
    - Suspected cause: "If there's some sort of clipping anywhere, you lose efficiency. If it's just barely clipping, the clipper might be moving / breathing with the environment."
    - RULED OUT: 
       (a)Checked centering on M_M1, M_B4, M_M2, M_B3, and the two MEAS IFO PDs, and can confirm they're either
           :: On Optics -- offset in yaw by design (but still several beam diameters away from the optic barrel in REFL or TRANS), or 
           :: Well-centered.
       (b) We'd started the week with the M_M4 and M_M5 mirror mounts and adapters in place (with no dump in either). The beam comfortably goes through the transmission ports of both adapters. But, over the course of the week, as we ruled out "everything else" out, we removed them so now they're off the board. We still see drift.
       (c) Confirmed that "tight-squeezes" between beam dumps still allow for at least ~1x beam diameter of clearance (on an IR card).

Optics' Quality, dust and schmutz
    - Suspected cause: "If there's a scratch or schmutz on the optic, then maybe as the beam alignment drifts, the optical quality of the reflected beam is intermittently spoiled."
    - RULED OUT: We replaced all optics in the MEAS path with just-as-good, same production/coating run, equally class-A cleaned, high quality mirrors. Still see drifts.

Beam quality and Optical Mode-shape
    - Suspected cause: "We haven't found anything, so let's look at the beam shape on a profiler. This should be the definitive answer."
    - RULED OUT: We removed IFO_MEAS_B, and projected the beam on to a WinCam beam profiler head, and read it's data out in the WinCam software. We looked at the beam both with a very slow heterodyne beat-note of 250 mHz, and the nominal 4096 Hz (adjusting the modulation frequency at our commercial RF source's digital interface).
        (a) With the MEAS IFO's alignment just tuned up, Compared mode shape and alignment during episodes of high and low efficiency, and we see nothing but excellent beam shape. (In the slow-beat-note-configuration) the only difference in profile results is the expected lower intensity when the efficiency was low.
        (b) We deliberately misaligned the REF and MEAS beams going into the MEAS IFO such that both beams still visible on the WinCam profile, but not interfering. No movement of beams, no mode shape change, no beam diameter change, nothing that would cause the efficiency drift.

Polarization
    - Suspected cause: "The last thing that it could be if it's not mode shape / quality / alignment is polarization drift..."
    - RULED OUT: With the same IFO_MEAS_B beam, we inserted a polarizing beam splitter and monitored the s-pol power in the reflected port.  S-pol (max) power (at peaks of beat note) is stable at ~10-15 [microWatt] during periods high and low efficiency. The designed p-pol light at IFO is at the 5-10 [milliWatt] level, so the extinction ratio is awesome. And we still see drifts. This also rules out, e.g. the nightmare scenario that the optics' mount PEEK set screw -- only on the MEAS path optics -- are cranked on the optic so hard that is causes birefringence.

    - PLUS: if there was a board-level polarization drift, the REF IFO would also be seeing this same drift.
 
Optical Power fluctuations (Sanity Check)
    - Suspected cause: "I know you don't see any drift in the REF IFO, but ... is the input power drifting?"
    - RULED OUT: 
        (a) The REF IFO beam samples the same beams as the MEAS IFO, and
        (b) Yes, we've had the power monitor PDs up on the o-cope for days, and we see no obvious or reproducible correlation.

RF Modulation Modulation power fluctuations (Sanity Check)
    - Suspected cause: "I know you don't see any drift in the REF IFO, but ... did you check the RF electronics?" 
    - RULED OUT: Yup, we've had the RF power monitor outputs of the laser prep chassis on the scope most of the time, and these are rock-solid and at the expected level.

Environment (Insanity Check)
    - Suspected Cause: "The only thing that intermittently drifts on the minutes time-scales you're seeing is things like air currents and thermal effects."
    - RULED OUT: We tried SO MANY different combinations of 
        (a) Clean room Lights ON/OFF
        (b) Room Lights ON/OFF
        (c) Clean room Fans ON/OFF
        (d) People in/out of clean room
        (e) People in different positions in the clean room
      and none of these showed obvious or reproducible correlation.

Conclusion -- It's gotta be the mechanical assembly of the picomotors in the IXM100 mounts.
The picomotor-actuated assembly process was documented in LHO:87497 following the assembly procedure E2500163.
 
[1] The most accurate mechanical drawing we have the IXM mount itself D1100362
[2] The vendor drawings of the picomotor E1000197
[3] The most accurate assembly of a pico-actuated IXM mounts D1500494 and D2100433
[4] The SPI's drawings for its IXM mounts D2400144 and D2400145

Looking at the assemblies:
    - We're worried that the pico-motor "stopper nut" is pushing the movable-mounting plate (item 1 in [1]), rather than the pico-motor's ball bearing. And somehow this poor kinematic connection is drifting and slipping with the environment or something.
    - The v-groove of the item 7 in [1] carbide plate faces toward the pitch actuator, and away from the yaw actuator. While this is the right thing to do to prevent over constraint -- because the same part is used in both actuator interfaces, the ball of the pitch actuator (be it manual- or pico- driven) will sit "deeper" in longitudinal than the yaw actuator. The yaw actuators' stopper nuts all easily clear the movable mounting plate, but the pitch actuators do not.
   - The manually driven IXM100s show there's no issue, given that the cupped 8-100 alignment screw clears the carbide plate holes in the moveable optic-mounting plate.

ACTIONS FOR NEXT WEEK
We're at our wits end, and considering just ax'ing the picomotor actuation on M_B4 and M_M2. In fact, we've already replaced the mount of M_M2 with a manually driven IXM100, and may do so for M_B4 if we find enough clean IXM100 mounts.

However, we've got some ideas to fix the problem and still have remote actuation:
    - Place a washer between the carbide plate and the front plate.
    - Remove the stopper nut.
    - Mill away parts of the front plate.
    - Get a thicker carbide plate.
    - Use only one pico-motor per holder.

Pictures and further commentary to come in due time.