Dan Brown, Danny V., TVo

Note for the morning crew, after the frontend restarts from the dolphin crashing, we cleared all the affected watchdogs for SEI and SUS.  However, when we tried to relock, the green WFS wouldn't engage on both arms until we had to do some realignment and then WFS_PUM worked.  It seems like H1:LSC-TR_Y_NORM_MON has an elevated setting, maybe a dark offset somewhere that hasn't been reset?  Also, it seems like the IMC WFS are not engaging as well at first, but once I lowered the threshold on the ASC IMC trigger to 30 instead of 40, they resumed. 

I'm not sure what re-start procedure needs to happen in order to recover from such a dramatic front end bootfest, for now we're leaving the ISC LOCK guardian in down.

Dave, Thomas, Danny, Jamie, Dan Brown

While attempting to re-lock the IFO, the dolphin network crashed and all the codes needed a restart.  we called Dave and he gave us instructions on how to log in and restart:

ssh controls@{computer}

then we run /etc/startWorld.sh

Everything is green on the cds overview screen now.

Looks like h1seih16 has had another kernel-panic event. This time it has caused the DAQ data for related front ends to become marked as bad (CDS overview attached).

Similarly to earlier crashes like this, nearly but not all of the front end computers sharing the same 10GE port on the DAQ data concentrator has bad DAQ data. In this case h1seih16 shares its h1dc0 port with psl, all corner station sus and seismic, and corner station sus-aux (except susauxb123). So it is not known why sush2b and susauxh56 are OK.

At this point there is no reason to think that the real-time models are not running correctly, we have just lost EPICS controls of h1seih16 and the DAQ data from those affected systems.

Would the first person to arrive on site please call me and we can recover the systems.

Frequency noise injections
Into CARM Servo Board EXC A via LSC-EXTRA_AO_2_EXC (Pic 1)

Inj Amp     Start Time     End Time
-------------------------------------
1e-2        1225337909     1225338605
3e-2        1225338652     1225340457
3e-3        1225340607     1225341956
1e-3        1225341989     1225344477


Intensity noise injections
Into PSL-ISS_SECONDLOOP_EXCITATION_EXC with the low pass disengaged

Inj Amp     Start Time     End Time
-------------------------------------
1e3         1225346150     1225347492
3e3         1225347687     1225348986
1e4         1225349022     1225350095
3e2         1225350140     1225351784
1e2         1225351821     1225353267


MICH Feedforward
I changed the actuators for the MICH FF to DARM measurements from the ITMs to the ETMs per Sheila's suggestion.
I did not have permissions to write to /opt/rtcds/userapps/release/lsc/h1/templates/, so I created two new templates in /opt/rtcds/userapps/release/lsc/h1/scripts/feedforward.
Have not had a chance to fit the MICH FF because I unlocked the IFO during the test.  

That is the only reason we lost lock this evening, had 13 straight hours of IFO time.  

I also altered ALS guardian so that X arm green WFS don't come on.  Mediocre alignment makes the WFS cause locklosses during LOCKING_ALS.

Attached are some plots from an updated noise budget, with several measurements retaken in the last couple of days. 

Last time we posted a noise budget 44840 we were missing an intensity noise measurement, here you can see that intensity noise is making a significant contribution about a few kHz. The second attached screenshot shows the intensity noise injection, which only goes up to 3kHz.  This could be contributing to OMC length noise observed here: 44874

PRC2Y gain reduction

The third attachment shows the ASC noise contributions to DARM, only the PRC2 P+Y measurements have been updated since the last noise budget here. Since PRC2Y was previously dominating the ASC noises, I reduced its gain by a factor of 2 and engaged an 8Hz elliptic low pass, now PRC2Y is not showing up in the plot because it is more than a factor fo 10 below DARM.  

Other notes on the noise budget:  

I also reran the HAM3 injections as well as MICH and SRCL today, those curves are updated in the plot.  

I am still using a guess and check method for the optical gain and cavity pole since we don't have the kappas yet.  This time we are using kappa c = 92% and a cavity pole of 440 Hz. 

Jamie, Sheila, Craig

We were updating our noise budget and noticed that the spectrum of GDS calib strain we calculate using pwelch with a hanning window and 5 second FFTs had spectral leakage.  For comparison we took some spectra in DTT and found that with a bandwidth of 0.5Hz we can reproduce the problem in DTT.  The attached screenshot shows spectra taken with different bandwidths and windows. 

It seems like we need more attenuation of the microseism.

We have had another DAC issue on h1seih23/45 which will require a restart of all the models on these front end computers.

Would the first person into the control room today please call me on my cell phone and we can get these systems going again.

I have performed all the diagnostics data gathering; as before the first DACs on both systems are reporting empty FIFOs. 

On Tuesday I can install new IOP software on these systems to run through empty fifo errors.

Georgia, Craig Sheila

• We moved the compensation plates tonight and Georgia posted photos of the change in the spots on the balast mass from the PR2 camera here:  44992
• We had trouble engaging ASC when we first started locking, so we reset the green QPD and camera offsets.  
• We ran a2L, which helped the low frequency noise a lot (it must not have been run in a while).
• Based on 44991 we reduced the SRCL gain by a factor of 2, and saw an improvement in the DARM noise from 60-100 Hz, as we would expect (see first screenshot).  This gave us a couple of Mpc.  
• Our DARM noise is now similar to what it was before the Montana EQ above 60 Hz (2nd attachment).
• SRCL OLG with reduced gain is attached. 
• We added in the PRC2 cut offs, but we think this caused a lockloss. 
• The next lock we didn't use the PRC2 cut offs, and our noise was worse although we don't know why it is worse.  We turned off the DCPD whitening because the violin modes were high, and this caused our noise to be worse. 

In our DARM noise budget, both SRCL and MICH noises are higher than they were during O2, and are limiting our noise right now.  Here I am just summarizing some information about the SRCL noise.  

In the first attachment, you can see that the coupling of SRCL control signal without feedforward on is very similar now to what we had in June 2017 (blue and green traces).  The brown and pink traces show the coupling with feedforward on, you can see that the feedforward now is actually doing a slightly better job than we were doing in O2 from 40-100 Hz.  

In the second attachment, you can see that the SRCL control signal is now larger than it was in O2.  The lower panel shows the signal out of the sensor.  We changed the whitening gain from 30dB to 21dB, (we now compensate for the whitening gain in the digital gain but we didn't in O2, the gain change is corrected for in this plot). 

Above 70 Hz, the control signal is about a factor of 2 above what it was in O2.  This makes sense, because our UGF is about a factor of 2 higher than it was in O2.  (See 3rd and 4th attachments, and the OLG posted by Evan Hall here: 24933) 

Below 70Hz or so, there is excess noise.  Much of this noise is due to HAM3 motion (see Georgia's measurement in 44393 and the comment from Hang).  We were able to reduce this noise in SRCL by tuning the PRCL subtraction in the input matrix (44893), however, the tuning seems to be changing from lock to lock so that we are not able to use a consistent input matrix to get rid of this noise.  

We did some FINESSE modeling to see the power/noise/optical response at the REFL/POP port in the NLN state.

=====

In the first plot we show REFL port power as a function of power-recycling gain (PRG). In addition to the total power, we also show the contributions from the carrier, RF9, and RF45 individually. The simulation result is very similar to the analytical estimation did in LHO:44410 (the RF45 contribution in this entry is lower than previously estimated as we assumed the NLN here and thus here we have extra 6 dB reduction in RF45 mod. depth). However,  the measured REFL power (LHO:44947) is lower than the model by about a factor of 2...

Also note that as the carrier is no more critically coupled, the shot noise level (in [W/rtHz]) will go up by almost factor of ~ sqrt(4) = 2 due to the increased amount of REFL power. However, the signal strength for, e.g., CARM will only go up by ~ sqrt(46/33) ~ 1.2. Therefore we might expect to lose a factor of 2/1.2 ~ 1.7 in CARM sensitivity...

=====

For the LSC_POP DC power the model and measurement seem to agree well. See the second plot. 

=====

In the third plot we show the expected optical response for POP9 to PRCL (top), and POP45 to SRCL (bottom). Those values are calibrated to [W/m]. For PRCL we see the response decreases as the PRG increases. This is expected because as the PRG goes up, the difference in the PRC finesse as seen by the carrier field and by the RF9 filed gets reduced. For POP45 to SRCL, the signal strength increases as we improve the PRG.

=====

In the forth plot is the shot noise level of the POP45 signal in [W/rtHz]. 

We also consider how well we can invert the response matrix of POP9/POP45 to PRCL/SRCL as the PRG changes. The condition number of this response matrix is shown in the fifth plot. Here to compute the condition number, we normalized the response such that the max. resp. to a dof is normalized to 1 (such normalization is equivalent as we having a freedom to choose an overall scalar at the input matrix). For example, the normalized resp matrix at PRG of 46 is given below,

which leads to a condition number of 1.2. The larger the condition number, the harder the inversion of matrix (and thus the higher the sensing-noise-limited sensitivity).

=====

In the last two plots we showed the SRCL noise. In the second last one we simply take the POP45 noise in [W/rtHz] divided by its response to SRCL in [W/m]. In this case we see that the shot-noise-limited SRCL sensitivity improves slightly. However, this assumes that we do not need to do any matrix inversion, which is unrealistic as we do not have a significant gain hierarchy between the corner dofs. 

In the last plot we show SRCL noise, taking into account that the condition number gets worse as the PRG goes up (by multiplying the single dof shot noise with the sqrt of the matrix condition number as an approximation). However, the SRCL noise is worsen by only a few %. Effectively it should not change from now to O2. 

=====

Parameters used in the simulation:

T_prm = 0.031, T_itm = 0.014, T_srm=0.325. The PRG variation is achieved by varying the common losses of the arms. 

For the modulation depth, we assumed Gamma_9 = 0.19 and Gamma_45 = 0.25 * (-6 dB) = 0.125, as in the NLN we reduced the RF45 mod. depth by 6 dB.

We assume that 1.25% of the power refl'ed from the ifo is delivered to the refl PD. For POP, we assumed T_pr2=230 ppm and 9% of the pop beam goes to POP QPD. 

Perfect mode matching is assumed for this study. 

[Sheila Cheryl Keita Georgia]

After discovering the coupling between HAM3 and DARM we were curious about the beams that we see on the PR2 camera and whether they can help us identify sources of clipping and backscatter. We see a few beams in this camera image that we would like to identify.

We haven't zoomed or moved the PR2 camera since we last tried to identify these spots in July. I used Gabriele's script to superimpose the camera image we see now, at 20W, with the visible image. See attached screenshot.

There are 4 spots visible on the PR2 scraper baffle: it looks like 3 at first glance but I think the central spot has a smaller spot slightly above it, the intensity of this smaller spot fluctuates during the lock. To check that these spots were not coming from the MC2 transmission we locked the mode cleaner at 25 W, and did not see these spots with anywhere near comparable intensity.

In the visible image, looking through the hole in the scraper baffle, with think we can see the outline of a round mount which could be the 3" mirror directly behind PR2. Visible in the 3" mirror we think we can see the reflection of a mount that could be the first to the first mirror on the POP QPD path. I'm not sure if any other beams we see on the camera are actually coming from these mirrors though.

There is also some light visible on the balance mass, which Keita and Sheila hypothesize could be a reflection off the AR surface of an ITM, which is wedged such that the thin part is on top, which is reflecting off PR3 and is not getting caught by the scraper baffle. When ALS is locking/locked we also see spots on the balance mass, from both the X and Y arm green. This is shown in the second attached screen shot. The other spots visible in the second screenshot are also green.

On the left we also see some spots which might be coming from the beamsplitter elliptical baffle.