Since LHO is getting walloped by the remanants of a Pacific storm, the winds are high and the microseism is high, preventing locking. A while ago RichM had suggested that we try lowering the St1 RX blends when the wind was high, and it seems like this might be a good idea, under the right conditions. I started by switching ETMY, first attachment is ETMY ground vs ETMY RX during high winds, refs are 250 mhz (nominal, high)  blends, live measurement is with the 90 mhz, lower blends . There is a large improvement from .1 to 1 hz. We don't normally run this way because ground tilt is usually below T240 noise, but not today. To check that this wasn't making things worse at ETMY, I also checked the Y motion and it was similarly improved, second attachment, again, refs are 250mhz (nominal, high)  blends, live measurement is with the 90 mhz, lower blends. Again, there is some improvement in the .1-1hz band, low frequency doesn't seem to be any worse. If we look at the CPS as a low frequency witness (below the blend frequency) going to a lower blend doesn't seem to do anything bad, under these very bad, no good conditions, third attachment. Yet again, refs are 250mhz (nominal, high)  blends, live measurement is with the 90 mhz, lower blends. The brown trace shows the Y cps is moving somewhat less than the blue Y cps, so there is at least enough real low frequency signal that we are not injecting T240 RX noise into the Y loop.

Sheila and Evan were doing modecleaner measurements, so I didn't try to get any arm cavity signals. It would be nice if commissioners would give this configuration a shot while the environment is terrible.

I have left the ITMY and ETMY RX loops in these lower blends because it sounds like commissioners are probably packing it in. While winds and microseism are this high (20-50mph(?) wind, 95th percentile(?) microseism) I think we should try this configuration. When winds settle down the ITMY and ETMY ST1 RX blends should be switched back to the Quite_250 blends. 

State of H1:  breifly locked PRMI but unstable, Commissioners doing other work, currently Sheila has IMC at 50W

Details:

• IMC locking - lower stage running out of range before upper stage kicks in
• BRS EY rung up
• winds back to 15-40mph
• useism is now at 1.0 in the Corner Station
• dust in the PSL alarmed all day, so with JeffB's approval, I set the alarm level to 800 on 0.3um particles
• FSS and ISS have unlocked many times, sometimes turning the auto-locker off breifly will help
• NPRO Noise Eater has been reset at least 3 times today, not a surprise, but might need to be tonight
• locking X arm in green finicky and needed a lot of alignment help to get over 1.0
• handing off the JeffB
• Sheila has the IMC

You know it's windy when...

IMC_F.png shows the 4 occasions from last night when the noise eater oscillated.  Anecdotally this tends to happen whenever
the IMC is trying to lock.  My conjecture is that rapid transients in the FAST actuator cause larger than normal changes in
the stress-induced refractive index change in the NPRO crystal, which in turn steers the beam out of the crystal a little
differently.  Since the photodiode used for the noise eater is relatively small, the beam could steer off the photodiode.
The reduction in photodiode signal then causes the noise eater to oscillate.  This does not happen all the time however and
I cannot explain why.  Zoomed_IMC_F.png shows that the IMC was not trying to lock at the time.

    Looking at the first time the noise eater misbehaved.  FAST_v_NoiseEater.png indicates that the noise eater oscillation
coincides with the third rapid increase in the FAST actuator output.  The long stretch where the FAST actuator is pegged at
~11.7V indicates that the FSS was not locked at the time.  SLOW_v_RCTPD.png shows that the FSS did not acquire lock at this
time but acquired lock during the second burst group of activity.  Of note is Kiwamu's observation that the FSS acquired lock
at a point outside the range where the SLOW actuator is limited to in software +/- 0.1V.  The FSS acquired with a SLOW
voltage of ~-0.25V.  The delay in settling down is the time taken for the so-called "temp loop" to be activated, presumably
by the autolocker.  Why it acquired outside this range might be due to the noise eater oscillating.

    Perhaps since the autolocker did not acquire on two flashes through the reference cavity, Travis disabled the autolocker
and began the hunt for the fringe manually.  Not realising that ALS likes having the SLOW voltage between -0.1V and +0.1V,
the crystal temperature kept being adjusted until a fringe was found.  The next fringe was found outside this range and the
FSS acquired at -0.25V, at which point the autolocker was re-engaged.

State of H1: in Initial Alignment, struggling to lock arms in green, have only had breif locks of X arm in IR

Details:

• winds between 10mph and 45mph across the site
• BRS at EX and EY turned off
• X arm in green locked but dipping below 0.5
• IMC on relock is pushing on MC2 so much that it's wise to pause in Ready to let it settle
• NPRO Noise Eater had to be reset twice, this seems like a apttern, like possibly the first reset doesn't really restore the system to a nominal state, and another reset is necessary
• useism is high, attampt to change BRS to WINDY_USEISM but didn't work as expected, set X by hand, Y is rung up so not being used
  •  small guardian code issue resolved by TJ, guardian behavior resolved

Site Activities:

• 16:15-16:45UTC - Chandra to both EX and EY VEAs to colect serial numbers
• 16:40UTC - Bubba to EY to add glycol to equipment in the mechanical room
• 16:45UTC - EvanG injected into Pcal
• 17:15UTC - I reset the NPRO Noise Eater
• 17:29UTC - I reset the NPRO Noise Eater
• 17:52UTC - Bubba back from EY

Summary:
The hardware injection inverse actuation filter has the correct amplitude and sign. This is tested using a known sinusoidal waveform and comparing the Pcal RXPD readback with the input to the inverse actuation filter. See attached figure.

Details:
Similar to my investigation of the sign of the inverse actuation filter (LHO aLOG 29072), I injected a 100 Hz signal, 1e-23 in strain amplitude, 0 phase using awgstream into H1:CAL-PINJX_TRANSIENT_EXC.

To verify the injection has the right amplitude and sign, I read out H1:CAL-PINJX_TRANSIENT_IN2 and H1:CAL-PCALX_RX_PD_OUT_DQ. The time-series data for both channels is bandpassed with a filter centered around 100 Hz. In this measurement, I did not turn on the [:1,1] filter (FM7) for the PCAL readback channel. Instead, I scaled the readback signal by -1e-4 (=-100^2). Also, I had to make an offset since there is a DC component to the Pcal signal. The TRANSIENT_IN2 has its time-series scaled by 4000 to convert between strain and meters.

The attached figure shows that the amplitude is nearly spot on, the sign is correct, and the known phase offset (~240 usec) is understood from the inverse actuation filter for the TRANSIENT injection path.

Thus, the inverse actuation filter for hardware injections is correct in amplitude and sign.

In case you need to retune the filter used for subtracting the HPO jitter (DBB QPD) from DARM, here are the steps

1. take some data with the feed forward off (set H1:LSC-JITTERFF_GAIN to 0, the jitter feed forward MEDM screen can be accessed from the LSC menu in the sitemap)
2. use the script attached below to fit the transfer function. Some notes
   1. set the GPS time and duration at the beginning. Typically 10 minutes of data are enough.
   2. the script will shows the measured coherence between DBB Q1Y and DARM, and the measured transfer function. Typically points with coherence above 0.05 are good for the fit
   3. you might have to tweak the frequency range and the coherence threshold for the fit, in order to get a good result
   4. sometimes the fit algorithm add a pole/zero pair with very high Q close to the edges of the fit range. Normally they are fake and you can remove them
3. the script will output the list of poles and zeros, upload them to the LSC-JITTERFF filter bank. Note that there is a gain of 8e14 in the DBB_FF_QPD1Y filter bank. This is added such that the typical gain H1:LSC-JITTERFF_GAIN is close to 1.

Notes:

• the transfer function should be compensated for the DARM closed loop response. For the moment being I haven't done it, since the subtraction matters only above ~100 Hz and there the DARM loop does not have a significant effect, at least not at the level of accuracy needed for the subtraction as done so far.
• the filter which is loaded now (named 16_10_13) was measured after the IFO was locked for few hours. It might be enough to retune a bit the gain to achieve reasonable performances

Last week KrishnaV and I found that there was a high pass  filter in the tilt subtracted sensor correction path that was limiting the endstation ISI performance at the microseism. ConorM-L suggest we try just turning these filters off, as the sensor correction filter is already rolled off pretty well at 8mhz. Cheryl had the endstation sensor correction turned off for a while this morning so I took the opportunity to try this out. So far it seems to be running okay. Attached plot is the EX ISI and ITMY for comparison. Green and pink are the EX ground supersensor and STS, red is the ST1 T240. Black is the ITMY STS, purple is the ITMY ST1 T240. Before the ETMs were getting very little suppression at .1-.2 hz, and now both ETMs are generally doing as well as the corner ISIs. We'll keep an eye on this for the time being, but I think we should run like this.

J. Oberling, B. Weaver

I checked on the TCS chiller this morning and added 350mL of water to bring the level from 5.0 to 8.9.  I then noticed that the mesh filter seemed to by pushing up out of the reservoir by a good bit.  I reseated the filter and noticed the level had dropped to 6.3.  There was still 50mL of water in the cup, so I added that to the chiller and observed the mesh filter.  Sure enough, the filter puffed up.  There is a large air gap between the top of the water in the reservoir and the spot that the filter seats into.  What I think is happening is when we fill the chiller, that air between the water and the filter has no where to escape quickly (probably due to the amount of water moving through the filter).  This creates an air bubble between the water and the filter that then influences the fill reading on the front of the chiller (hence why the reading went down when I simply reseated the filter).  I have a suspicion that the chiller has not been losing water, we just haven't added enough since the system flush to completely fill it, and when we do top it off we're creating an air bubble that influences the fill level reading.  We then think we've topped the chiller off when in actuality we haven't; as that air bubble slowly works its way out the level reading "drops," thereby making us think we're losing water.

I ran this by Betsy and found she was starting to suspect something similar.  We went out and removed the mesh filter and then topped the chiller off, then replaced the filter.  It took 600mL of water to move the indicator from 6.3 to 8.3, which is much less than we've seen in the past; if you look at the log on top of the chiller it can be seen that there are instances where we fill w/ 250mL of water and move the indicator from 5.0 to ~9.0, which is much less than the 600mL is took to go from 6.3 to 8.3.  This suggests that what I wrote above is correct, we haven't been fully topping off the chiller but have been fooled into thinking we have by an air bubble of our own creation influencing the reading of the chiller fill level.  One of us will check the chiller at the end of the day to check the chiller water level and see if we still need to add water (water has been added every morning and evening for every day this week).

Total water added this morning was 1000mL, and this moved the indicator from 5.0 to 8.3.

I copied Gabriele's JITTERFF MEDM screen, replaced H1 with $(IFO), and saved it as $(userapps)/lsc/common/medm/LSC_CUST_JITTERFF.adl.

I added a link to that screen in the LSC overview.

9:30am local

13 sec. to overfill with 1/2 turn open on bypass LLCV. Note LLCV is still set to 16% open. Exhaust piping is completely defrosted. Left bypass exhaust valve open. Next fill due Monday.

14 day plot, upper plot is corner station winds, lower plot is from the PSL, and at about 17mph the peaks in wind match peaks in the PSL dust monitor.

See attached screenshots of OpLev trends. 

Sheila, Jenne, Kiwamu

Attached is a spectra of IMC-F in different configurations.  (MC locked at different powers, DC readout, low noise)  From 100 Hz to about 1 kHZ, the spectrum of IMC F doesn't change much at all in all of these different configurations. So the IMC control signal is not dominated by REFL9 sensing noise in full lock, and probably represents the real frequency noise at the input to the IMC.  

We can do a better job later, but if we assume this is really frequency noise we can roughly calibrate this into Watts on REFL 9I:

At 1kHz:  0.1Hz/rt Hz Frquency noise arriving at IMC (which is roughly consistent with measurements in P1100192, Fig 8) Suppresion of IMC loop: 1/200 (alog 22188) Supression of CARM loop (alog 22188, our ugf is now more like 8kHz) roughly a factor of 1/30.  We can scale the DC optical gain of 0.017W/Hz used in 22188 by sqrt(2) to account for the factor of 2 increase in input power and the 6dB modulation index decrease since then.  Taking into account the coupled cavity pole at 0.5 Hz give another factor of 1/2000:

0.1Hz/rtHz(1/200 Hz/Hz IMC supression )(1/30 Hz/Hz CARM suppression) (0.024*0.5/1000)W/Hz  = 2e-10 Watts/rt Hz signal on REFL 9I or 1.7e-5 Hz/rt Hz of residual frequency noise expected.  

We can repeat this at 400 Hz:

0.03Hz/rtHz(1/600 Hz/Hz IMC supression )(1/300 Hz/Hz CARM suppression) (0.024*0.5/400)W/Hz  = 5e-12 Watts/rt Hz signal on REFL 9I or 1.7e-12 Hz/rt Hz of residual frequency noise expected.  

Comparing this to Evan's in loop measurement of the CARM noise using REFL control, (here) it is close at 1 kHz but not at 400 Hz.  You can also compare it to the transfer functions from REFL 9I to DARM posted here, and see that at 1 kHz the expected frequency noise is of the order of 5e-20 m/rt Hz at 1 kHz.  

The main message: It is probably worth making a projection for frequency noise in DARM using IMC-F to estimate the frequency noise after the ref cav, because a very rough estimate says it could be within a factor of 2 of DARM at 1kHz.