Attached are the transfer function measurements of the FSS taken yesterday afternoon.
The first two plots are taken with the common gain at 16 dB and fast gain of 22 dB.
The second measurement (C16F22-2) I think was affected by the modecleaner trying to
acquire lock, which might explain the notch-like feature just past 200 kHz.  The
UGF in both measurements is about 195 kHz.

    The third measurement, taken with a common gain of 20 dB and a fast gain of
22 dB is, I suspect, also affected by the modecleaner trying to acquire lock.



Jason/Peter

This is as far as I have been able to get the IFO tonight. I tried to set the intent bit, but it will not engage. I have set the Ops Observatory Mode to Observing.

Ross and I are the only ones in the control room. The last lock was lost at ENGAGE_ISS_2ND_LOOP from a pringle mode on ETMY.

[Everyone]

We have determined (a few hours ago) that flipping the ESD bias sign on ETMX does not allow us to lock ALS DIFF.  Jeff has looked through the settings, and everything is flipped to match the bias sign as appropriate, but we're still not able to lock.  It looks like a crossover is unstable, or something like that.  For now, we have reverted the ETMX ESD bias to its pre-Tuesday state, which has facilitated locking.  Team SUS will look into this later.

The ISS is much better behaved now that the alignment work was done.  However the FSS gain change was making the IMC loop very nearly unstable, and we lost lock during the power-up twice due to this instability.  Sheila and Keita will post their measurements that discovered and fixed this issue.

We have so far been able to power up to 40W once, but it looks like a PI rang up pretty quickly.  Carl will post details, but this is a new mode that hasn't been a problem previously, so it does not yet have damping settings.

At this time, it looks like there is no problem in getting to 40W, and other than PI damping we should be able to get all the way to low noise.

If Carl is unable to find damping settings relatively quickly, we may choose to instead only go to 20W for tonight, so that we can get to low noise and set the intent bit.  Stay tuned for more updates...

Jenne, Sheila, Keita, Peter and Jason are here. The work in the PSL enclosure has completed for now. The IFO has made it to DC readout twice, but lost lock on increase power. Keita and Sheila are taking measurements related to the FSS.

After a lot of alarms going off including a message from Jim, I discovered that most of the DMT processes had skidded to a dead standstill. The problem seems to have been that the broadcast frames had increased their length so that the data buffer lengths had to be increased also. The buffer lengths are now updated and the SenseMonitor outputs should now be working again. 

Yet again, Please! Let me know, preferably in advance, any time the broadcast frame list is changed.

Dan, Jim, Dave:

we tried several things today to try to make h1fw0 more stable. These are:

reintroduction of h1ldasgw2 to take NDS traffic away from h1ldasgw0 (leaving it only used by h1fw0)

upgrade the network link between h1fw0 and h1ldasgw0 from a cat5e 1GE to a fiber opitcs 10GE using borrowed intel 10GE cards from LDAS

reconfigure fw0 to not write commissioning frames

power cycle fw0 and ldasgw0

these changes have not made fw0 any more stable. fw1 continues to be more stable, some of its restarts were coincident with fw0 restarts (within several minutes)

The measurements of HARD loops with different laser powers were posted before (alog 27864, 27885, 27920).

We found that the big discrepancies between the measurements and modeling for Pitch are caused by a mistake in the simulation. I fixed it and replotted the measurements at 10w and 40w (37w actually) as below.

Now the modeling curves fit the measurements much better. In the 10w plots, the black dots are measurements without the optical levers servos on for the ETMs, and the red ones are measurements with OL on for all the TMs. (The blue curves are the modeling results, which assume only OL on for ITMs.)

State of H1: Peter and Jason are out of the PSL and locking will resume shortly

Activities today:

• Robert in LVEA setting up injections
• Travis to EX and EY for p-cal
• Karen and Christina to MX and MY
• Fil to ECR to remove a Robert electronic box

Still in play:

• JimW has new blend filters in HAM4 and will coordinate with relocking crew
• Robert heading back out to the LVEA by the rollup door

The first plot shows IM2 pitch changing by as much as 0.4urad at lock loss (lock loss after IMC power oscillations).

Subsequent lock losses on the plot show it moved by 0.3urad, 0.2urad, 0.2urad,and 0.2urad, for an average change in alignment of 0.26urad at lock loss from a 40W H1 lock.

This result supports my theory that the beam on IM2 is no longer centered, and is off center enough to explain the oscillations in it's pitch and yaw signals that I aloged here, and in an attachment here (where I showed that IM4 Trans sees the oscillation), while the IMC was oscillating in power from 22W to 57W.

The second plot shows that when H1 was no longer locking, the jumps in IM2 alignment stopped.

BS beam is not centered and it might be moving with power.

Last Friday, I placed a digital camera on a tripod in front of the CRT for the analog BS video looking at BS from BSHR side, and took some pictures. I conclude the following:

1. BS beam position is systematically shifted to the left by some millimeters when viewing BS HR surface.
2. When increasing the power from 10W to 40W, the beam moves to the left (i.e. more clipping loss) and down.
3. BS beam position is also systematically lower by some millimeters.

It's only digital camera image of CRT screen of analog video that saturates at 40W (but not really badly). Take it with a grain of salt, but if you believe that the pictures are not unreasonable representation of the reality, the following shows the BS beam positions at each power.

Note that YAW offset is seen from the -X+Y direction. You need to divide YAW (but not PIT) by sqrt(2) to convert it to the lateral beam shift.

I also took  2W picture but it's dubious because the background (dark) image subtraction didn't work well (see below).

I don't know for now how significant this is as far as the recycling gain is concerned. I'll ask simulators.

In the mean time, it might be useful to explore CSOFT offsets.

Fundamental assumptions.

Video lens, CRT screen and the digital camera are all reasonably rectlinear.

Bright thing is the beam on BSHR, not on BSAR.

BS X and Y elliptical baffle are as in D1200704, 703, 750, and the position of BS relative to baffles is exactly as in D1500239.

Mapping from picture to the physical position.

The first attachment is an annotation, so to speak, of one of my pictures (40W full lock) and the screen shot of CAD drawing  D1500239 looking at the BS from roughly the same POV.

In the picture, red line is the edge of the BS Y elliptical baffle hole, yellow is X, green is the back side (AR) edge of the BS.

There are two black rectangular things between the red and the green line. Larger one is the side of the wire protection baffle structure. Smaller one is probably the ear on BS.

Baffle hole edge comprise two identical ellipses. The narrowest part is defined by two apexes (apices?) or points where two ellipses intersect. For mapping purpose I ignore all edges and only use these two points per baffle. Red and yellow vertical lines and dots on the picture are the center lines and the center point of Y and X baffle hole, respectively.

Since I know the physical position of these four points (two per baffle), distance from Y baffle to the BSHR surface, and from BSHR surface to X baffle, I can map e.g. the physical center of BSHR surface to this image shown as green dot in the first attachment.  I can also map (X, Y) coordinate of any point on the BSHR surface in this picture to the physical coordinate.

Note that you need to account for the fact that we're looking through 60mm-thick BS to look at X baffle (but not Y and BSHR).

The position of two apexes for the Y baffle is quite well defined, but X is somewhat dubious because none of the pictures shows both sides of the apexes on X baffle (see e.g. second attachment). I assume that the edge on the picture ends at one of the points. This means that the yellow line could be to the right than in the picture, which subsequently push the BSHR center to the right, which subsequently push the YAW offsets in the table above more negative (i.e. larger YAW offset).

The length calibration on this image at BSHR surface is about 0.15mm/pixel. This is mostly determined by the distance between the Y baffle apexes and only weakly dependent on the X baffle because BSHR is much closer to the Y baffle than X.

How the images were taken and processed.

The third attachment shows the 40W image with and without dark image subtraction. I calculate the image COG inside the blue circle to find the center of the beam (yellow cross). The green circle is the center of BSHR.

The baffle looks odd after "dark" subtraction because the "dark" image is not dark for baffle scattering. That's because I used DRMI as a "dark" image (4th attachment). No true "dark" image was taken because I didn't want to interrupt locking activities. You can see in the dark image that the scattering from the baffle is already significant but not at the beam on BSHR, and that there's a reflection of the camera and the tripod on the CRT screen. Anyway, this is subtracted from the left of the third attachment to produce the right after taking into account the exposure time difference.

The beam position doesn't change much with or without dark subtraction.

Plots under the pictures show the red, green and blue channels across red horizontal lines in the pictures to see if the video image or the camera image itself was saturating, and it seems the video was, but not really badly.

The 5th and 6th attachments show 10W and 2W, respectively.

At 10W, maybe it's hard to see the benefit of dark image subtraction on the image itself, but if you look at the beam center cross section plot, removing the shadow of the camera and the tripod actually made the beam shape nicer.

At 2W, dark subtraction didn't work well. If you look at the frame of the CRT it's apparent that the dark subtraction failed. If you look at the edge of "SANYO" or the left edge of orange-ish sticky note, it's apparent that the digital camera was off in YAW significantly. But the bright image clearly shows the shadow of the camera and the tripod, and I don't know if the bright image is useful for guessing the beam position either.

After finding that the WFS on IOT2L had driffted offcenter to values around 0.7-0.8, an allert was put into the IMC_LOCK guardian.  

The alert was set to be active at the limit of 0.3 for any WFS DOF.  

WFSA yaw has drifted from 0.25 up to 0.28, causing the alert to flash on/off all the time, as the max WFSA yaw values crosses 0.3.

Since we are focused on getting ER9 started, I consulted Jenne and we came up with the plan to adjust the alert to 0.5, and recenter WFSA yaw next Tuesday.

Overnight fw0 became unstable again. We have two things to try: [1] offloading NDS QFS-NFS data using the new h1ldasgw2 machine, and [2] upgrading a frame writer's NFS network from 1GE to 10GE.

We just did the item [1]. We restarted the h1ldasgw2 solaris machine and hooked it back up to both QLogic switches. This solaris server sees both file systems (h1-ldas-frames and cds-h1-frames), and exports them as read-only NFS mounts. We reconfigured both h1nds0 and h1nds1 to only NFS mount h1ldasgw2, and set up their /frames symbolic links to both see /cds-h1-frames (i.e. the frames written by h1fw1).

So now on the h1fw0 system, h1ldasgw0 is only serving h1fw0 and not h1nds0, and no one is reading its frames except for LDAS disk-to-disk which is archiving the science frames (remember commissioning frames are archived from h1fw1).

J. Kissel, E. Goetz, K. Izumi, D. Tuyenbayev

Evan will post the details of the work we've had to do to get the model running, but in the interest of time, I've taken what we needed from the Matlab model to update the CAL-CS front-end filters. Since early results indicate that the only low-frequency (sub-Nyquist) things that have changed from the O1 model are in the sensing function (see LHO aLOG 28171):
   - Lower optical gain = 9.071e5 [ct/m],
   - Lower frequency DARM coupled cavity pole frequency, f_c = 328.7 Hz, and
   - New SRC-detuned optical spring frequency, f_s = 9.831 Hz.
I only needed to update the H1:CAL-CS_DARM_ERR inverse sensing function filter (see further discussion below). 

The new settings have been captured in the SDF system.

Details of the design:
Foton Design String --
       zpk([9.831;-9.831;328.7],[0.1; 0.1;7000],1,"n")gain(9574.81)*gain(1.102e-6)
The gain of 1.102e-6 is 1 / 9.071e5 [ct/m], this lives separately in FM4, called "ER9gain." In FM3, in the filter called "SRC D-2N" for "Signal Recycling Cavity De-Two-Ne" there lies:
- The pair of real poles at 9.831 Hz, one of which is in the right-half-plane, reflect the detuning dynamics. Note that we've rolled of these inversion zeros at low frequency of two real poles 0.1 Hz. 
- The 328.7 Hz is the new f_c, and we retain the same high-frequency roll off of 7000 Hz.  
- The gain of gain(9574.81) which is the correct normalization gain to get the over-all gain to be unity at 100 Hz, which is the frequency at which I matched the no-detuning gain of the (unused) 329:7000 filter in FM3.
The attached PDF shows that the ER9 gains agree between detuning and no detuning, and as expected, the overall optical gain is ~20% lower that O1 because we've not yet digitally compensated for the ~20% lower optical gain (because of 20% lower PRC gain; see LHO aLOG 28133) in the DARM loop.


Further Discussion on why I've only updated the Sensing Function:
All actuation strengths are within ~5% of there O1 values (see LHO aLOG 28130), and we've compensated all electronics better (see LHO aLOGs 27180, 27150, and 28087), so we need not update anything in the actuation chain. Rana and Evan H. have changed the local, top-mass damping loop filters for the QUADs, so nominally the QUAD dynamics have been changed, but that should be a small effect in the GW band. We'll update for O2, but no need for ER9. There have been several changes to effects at high-frequency, but all of those are covered in the GDS FIR filters which absorb the CAL-CS output and acausaly correct for these super-Nyquist frequency effects.

J. Kissel, E. Goetz

Having a few new breakthrough ideas on the UIM actuation system (see LHO aLOG 24914), we explored whether we are modeling the [m/N] L1/UIM force to L3/TST displacement transfer function incorrectly. This was done by driving the UIM out to 600 [Hz] and measuring the response in DARM. Not only did we find the expected-but-not-yet-modeled wire violin modes at ~330 [Hz], 420 [Hz], and some at ~500 [Hz], but we found several bending-mode resonances at 111 [Hz] and 166 [Hz]. Indeed, upon first glance, we think the 111 [Hz] resonance is the remaining missing frequency dependence that explains the turn-up seen at 100 [Hz] in all previous measurements of the UIM to TST transfer function.

We'll process in more detail some time in the future, but check out the attached screen shots and be amazed at how not 1/f^6 the L1 to L3 transfer function is.

--------
We'd started by exciting the L1 stage via awggui in a broad-band fashion such that we could catch all of the wire violin mode frequencies watching DARM. As Evan mentions, we had suspected that these wire resonances -- documented in T1300876 -- were the source of the deviation from 1/f^6, and they'd just not been included in the SUS dynamical model -- [[EDIT -- Brett has now included them in the model, and they are a non-negligible effect; see LHO aLOG 24915]]. This broadband TF is shown in the first attachment. Black is the with excitation ON, and red is ambient (to distinguish the ~505 [Hz] fiber violin modes from the ~495 [Hz] TOP to UIM wire violin modes, and the Beam Splitter violin modes & 331.7 [Hz] calibration line from the Sus. Point to TOP wire violin modes). Only the peaks of the wire violin mode resonances are visible above the DARM noise; driving them any higher breaks the IFO's lock.

Just in case, we drove down to the same ~80 [Hz] region, and BINGO! We also found new, unexpected resonances at 166 [Hz] and even as low as 111 [Hz]! (see second and third attachments) Our best guess for the source of these resonances are imperfect actuators. Perhaps the bending of the L-bracket that mounts the OSEM coil to the reaction chain's UIM (see D060375, see page 1 for the total assembly, lock at pg 14 for the L-bracket). Though, if it were these L brackets, I'd suspect there would be 4 independent resonances... also it doesn't look like enough moving mass to have resonance frequencies as low as ~100 [Hz]... dunno, will think more.

Finding something at 111 [Hz], we then took a careful swept sign measurement covering it, and indeed, it looks strikingly like another piece of the UIM puzzle. (see 4th attachment). We also grabbed a PCAL to DARM transfer function over this frequency vector, so we can turn the TF into an absolute calibration later.

For future reference, the templates for the swpet sine TFs are here:
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/PostO1/H1/Measurements/FullIFOActuatorTFs/
2016-01-12_L1toDARM_FullLock.xml
2016-01-12_PCALYtoDARM_FullLock.xml

and I attach screenshots of the awggui sessions used to excite L1 and PCAL in a broad-band fashion (DARM_IN1 ASDs during the broad-band excitations were taken using the standard wall FOM for DARM).