Here is some of the data from Saturday's jitter injections (29780), plotted differently.  The main point is that we should be able to reduce the yaw jitter coupling by at least a factor of a few, and this should help the DARM spectrum, but not enough to get the noise a factor of 10 belpw the quantum noise with the current level of jitter.  

• the first attachment is a transfer function from the IMC WFS signals to DCPD sum, which is directly comparable to the transfer functions Evan posted in 25653.  For PIT, the coupling in this measurement was similar to what Evan measured in Febuary, but this coupling was reduced by about 8dB on saturday which should have brought it almost back to the September level. For Yaw, the coupling now a factor of 3-4 higher than it was in the September and Febuary.  
• Since Saturday I have tried moving PR2 PITCH along with the POP offset to reduce the pitch coupling (no sucsess), and moving the POP YAW offset (also nothing).  Soft offsets have not been tried. 
• The last plot has projections from Saturday based on the IMC WFS DC and the ISS PDs, as well as a DARM reference and the GWINC curve from our DARM FOM for reference.  These are from before the PIT coupling was improved, so the situation should be a little better now. 

Related to 29870, 29804,

I have taken open loop transfer functions of the 1st loop in this morning to see where the UGF currently is.

Based on the measurements, I set the gain slider in the medm to 9 dB which gives us a UGF of 50-60 kHz.

Also, the raw data is attached as a zipped file.

Apart from these measurements, I confirmed two features as follows.

• The gain slider in the medm is miscalibrated. An X dB change in the gain slider introduces a 2X dB gain change in the open loop transfer function.
• The overall gain does not linearly scale with the gain slider any more if the UGF goes above roughly 70 kHz.
  • Also, a significant drop starts showing up at around 80 kHz due to presumably some saturation/nonlinearily somewhere in the circuit.

During the measurements, I kept the suspiciously high offset value of 20 for the 1st loop. The second loop was off.

I added 125 mL to the crystal chiller.  I noted that the fill port water is very turbulent, swirling and full of air bubbles.  I recall that it is usually calm.

FAMIS 6489.

Dave, Darkhan, Jeff:

A code bug in the file cds/common/src/BUFFER_AND_AVERAGE.c was fixed and h1calcs was restarted. No DAQ restart was required. This is the only model which uses this file.

Dan, Jim W, Sebastien, Dave:

While Sebastien was visiting LHO during the past week, we performed a test install of the MIT (Sebastien and Michael C.) earthquake arrival predictor code (seismon). Here is a summary of the install work so far:

We are using the spare CW injection computer, Dan M. upgraded this system from SL6.7 to SL7.2 last week.

The software install is in two parts: the USGS client system which periodically downloads seismic data from USGS, and the python code which analyses this data and issues local alters.

The USGS client code (based on java) was installed last Friday and have been running continuously since then. The connection between CDS and USGS is initiated internally for security reasons.

Today we got the seismon_traveltimes python code running, which for earthquakes with magnitude > 5.0 provides predicted arrival times (for each of; p-wave, s-wave, 3XRayleigh waves), predicted maximum velocity and surface distance from eq-source to LHO.

I'm working on scripts to run this continuously with a time resolution of about a minute, and to export these data as EPICS channels.

After a period of testing, we will decide how this system will be installed permanently on site (e.g. dedicated computer, integrated EPICS IOC)

John, Chandra

Overfilled CP4 today while monitoring exhaust pressure and flow (from control room).

-Manually set LLCV to 66% open (double nominal value)
-Took 17 min. to fill to 100% (started at 90%)
-Took another 60 min. to see a drastic change in flow and pressure beyond 100% full (i.e. cooling exhaust line) *constant 66% open
-When exhaust pressure spiked to ~6 psi (flow meter spiked to ~480 slpm) we terminated the experiment and lowered LLCV to 20%, then 10%
-Kyle reported the new plastic flex line froze and all exhaust pipe was cold to the touch. Flow meter read -5decC.
-John and I measured uninsulated exhaust line temp afterward:  -8degF where pipe exits building (lots of frost build up) and ~26degF several feet downstream

We will repeat experiment tomorrow morning after exhaust line warms up but this time run until LN2 flows out end of exhaust line. We will install one or a few thermocouples along exhaust line.

SudarshanK, TravisS, RickS

Yesterday, we went to Xend to assess and optimize the Pcal beam locations on the ETM surface.

Since O1, we have known that the beams were clipping somewhere on the receiving side.

Unsing an updated method of assessing the beam locations that is based on the work that Miranda Chen (REU student from UTRGV) did this summer (basically using the ESD electrode pattern rather than the edge of the ETM) we found that the Pcal beams were quite far from their optimum locaitons (see first image attached below). The offests from the optimal locations (111.6 mm above and below the center of the face of the optic) were U [ 1.71, 12.27], L[10.02,  -7.27] where U and L denote the upper and lower Pcal beams and the coordinates are [x offset, y offset] in millimeters with the usual conventions, positive x to the right, positive y up.  Because we use 2" diameter optics in the Pcal periscope structure and because the receiveing-side mirrors are twice the distantance from the ETM, an offset of 12 mm at the ETM would correspond to an offset of ~24 mm at the periscope, likely the cause of the clipping we have observed. The reflected power measured by the power sensor in the receiver module increased by about 50% from about 4 V to 6.2 V.

We adjusted the beam locations using the steering mirrors in the Pcal transmitter module and ended up with positions that are within a mm of the optimal locations (see second image attached below): U [ 0.15, 0.43], L[-0.18,  0.05].

The third image below shows the Pcal spots on the ETM surface (inside green circles), visible when the chamber illuminator is off.

We are in the process of modeling the expected calibration errors resulting from bulk elastic deformation induced by the Pcal forces being so far from the optimal locations, but we expect errors as large as 20-40% at frequencies above 4 kHz for offsets as large as what we had before correcting the positions yesterday.

11am local

28 sec. to overfill CP3 with 1/2 turn open on bypass LLCV (bypass exhaust valve remains open)

Next fill due Fri.

It looks like the code restarts yesterday made BRSX upset. Right now, the DRIFTMON channel is swinging about 10k counts, p2p. Krishna suggests "several thousand" is too high for the damper to be useable. It sounds like I'll need to re-center the table, after the BRS has settled down some. We'll keep monitoring, but for now the SEI_CONF manager needs  to be left in a state that doesn't use BRSX, currently in WINDY_NOBRSX, though VERY_WINDY and some of the earthquake states will work as well. TJ has confirmed that with the BRSX_STAT node in error, SEI_CONF will not complete a transition to a state that uses the BRSX, but a code restart or INITing SEI_CONF could get around that.

Attached trend shows that trouble started around 18:30 UTC yesterday, which I think is about the time we did several restarts of the BRSX Beckhoff.

Attached is the OLTF in two configurations with 7dB VGA gain.

50W (actually 48W), AC coupling OFF, and boost ON (solid line)

50W (actually 48W), AC coupling ON, boost OFF (dashed).

UGF is about 9kHz with phase margin of 40 deg or so (boost ON) or 50 (boost OFF).

Boost buys us about 9dB more gain at 1kHz.

I asked Kiwamu to do an analog measurement up to higher frequency on the floor.

J. Kissel, K. Izumi

Kiwamu and I sat down again this morning to review the suspected sign error in the L3 / TST / ESD path in the calibrated output (see LHO aLOG 29860). He's convinced me -- just like what had happened in July (LHO aLOG 28396) -- after the most recent flip of the ETMY L3 ESD bias sign (LHO aLOG 29397), although I had accounted for all of the digital changes in the L3 path, I did not change the replica of the actual to reflect the real actuation strength sign flip. 

Several decisions have been made so as to prevent this problem in the future:
(1) We've decided against changing anything in the CAL-CS front-end calibration for future ESD bias sign flips.
In the real ETMY L3 path, when the ESD bias sign is flipped, it changes the sign of the real-world ESD actuation. As such we compensate that change by changing the sign of the gain in the L3_DRIVEALIGN_L filter bank. In doing so, that means there's no change in the overall LSC super-actuator transfer function. Because there is no change in the overall transfer function, we therefore do not need to change the replica of the path in CAL-CS.

In the past, we've changed the CAL-CS replica to save our sanity aesthetically -- such that, for example, we can look at the DRIVEALIGN matrix gain in both the actual ETMY SUS and in CAL-CS and they match. But if one does so, one MUST change the sign of the real ESD (H1:CAL-CS_DARM_ANALOG_ETMY_L3_GAIN). This is what was not done in July, nor was it done in late August, because there's no digital equivalent in the control system to remind me. 

In light of us (well, me) failing to keep this replica path correct, forgetting to code up a setting change here or there every time, we find the cons of keeping things aesthetically the same out-weighing the pros. 

We will no longer touch the CAL-CS calibration settings when we flip the bias sign in the future.

(2) We've removed all CAL-CS changes from the ISC_LOCK guardian.
While assessment and control system compensation for the bias flip still live in the ISC_LOCK and SUS_PI guardians, we've removed all modification to CAL-CS settings from the ISC_LOCK guardian. We've remove execution of these steps by commenting out the relevant lines in the ISC_LOCK guardian:

        # Check for BIAS flip on ETMY
        # Note that SUS_PI guardian also is affected by ETMX Bias Flipping. 
        #            There are checks and appropriate actions in there too. -Kissel, Hardwick 08/31/2016
        if ezca['SUS-ETMY_L3_LOCK_INBIAS'] < 0:
        #    ezca['SUS-ETMY_L3_ESDOUTF_UL_GAIN'] = +1   # Commented out after finding that compensating for the 
        #    ezca['SUS-ETMY_L3_ESDOUTF_LL_GAIN'] = +1   # Bias flip down stream of the linearization just does
        #    ezca['SUS-ETMY_L3_ESDOUTF_UR_GAIN'] = +1   # not work.
        #    ezca['SUS-ETMY_L3_ESDOUTF_LR_GAIN'] = +1
            ezca['SUS-ETMY_L3_DRIVEALIGN_L2L_GAIN'] = -30.0
            ezca['SUS-ETMY_L3_ESDOUTF_LIN_FORCE_COEFF'] = -124518.4
        #    ezca['CAL-CS_DARM_FE_ETMY_L3_DRIVEALIGN_L2L_GAIN'] = -30.0          # Commented out after second occurrence
        #    ezca['CAL-CS_DARM_FE_ETMY_L3_ESDOUTF_UL_GAIN'] = +1                 # of overall sign error in L3 path of 
        #    ezca['CAL-CS_DARM_FE_ETMY_L3_ESDOUTF_LIN_FORCE_COEFF'] = -124518.4  # CAL-CS. LHO aLOG 29868.
        #    ezca['CAL-CS_DARM_ANALOG_ETMY_L3_GAIN']= -1
        else: # flip everything.
        #    ezca['SUS-ETMY_L3_ESDOUTF_UL_GAIN'] = +1   # Commented out after finding that compensating for the 
        #    ezca['SUS-ETMY_L3_ESDOUTF_LL_GAIN'] = +1   # Bias flip down stream of the linearization just does
        #    ezca['SUS-ETMY_L3_ESDOUTF_UR_GAIN'] = +1   # not work. LHO aLOG 29397.
        #    ezca['SUS-ETMY_L3_ESDOUTF_LR_GAIN'] = +1
            ezca['SUS-ETMY_L3_DRIVEALIGN_L2L_GAIN'] = +30.0
            ezca['SUS-ETMY_L3_ESDOUTF_LIN_FORCE_COEFF'] = +124518.4   
        #    ezca['CAL-CS_DARM_FE_ETMY_L3_DRIVEALIGN_L2L_GAIN'] = +30.0          # Commented out after second occurrence
        #    ezca['CAL-CS_DARM_FE_ETMY_L3_ESDOUTF_UL_GAIN'] = +1                 # of overall sign error in L3 path of 
        #    ezca['CAL-CS_DARM_FE_ETMY_L3_ESDOUTF_LIN_FORCE_COEFF'] = +124518.4  # CAL-CS. LHO aLOG 29868.
        #    ezca['CAL-CS_DARM_ANALOG_ETMY_L3_GAIN']= +1
        ezca['SUS-ETMY_L3_LOCK_BIAS_GAIN'] = 1


Attached is a screen shot of the relevant MEDM screens to aide discussion.

• Guardian issues will be addressed by Jaime upon his next visit. (next weekish)
• SAFETY MEETING TODAY
• ER10 will NOT begin at the end of next week. More about this after tomorrows JRPC meeting.
• Thursday WILL be used for maintenance. Duration not specified. (PCal work; Kyle at EX; opportunistic CDS electronics work)

SEI - BRSX is severly rung up since last night. Team is working on it.

SUS - Nothing to report

CDS ELT - Everything seems ok

VAC - only 1 PT140 now. CP4 exhaust flow measurements being done today. New vacuum gui

PSL - Nothing to report

FAC - FRS maintenance

COMMISH - intensity stabilization is working as intended

I've stepped all four ring heaters consecutively tonight by 4 W each to identify which test mass belongs to the new 17783 Hz/47.7 kHz PI mode (alogs 29820 and 29838). The shift in frequencies of the test masses is dominated by a change in the Young's modulus which we can alter via the ring heater power; by stepping each ring heater and finding which causes a larger relative frequency shift of the mode in question, we can identify which test mass the mode is from.  See T1600080 and LLO alog 18002 for more info. All power changes below were applied to both top and bottom ring heater. All times UTC.

ETMX: 0.75 --> 2.75 @ 9:29 --> 0.75 @ 9:48

ETMY: 0 --> 2 @ 9:49 --> 0 @ 10:08

ITMX: 0.25 --> 2.25 @ 10:08 --> 0.25 @ 10:29

ITMY: 1.25 --> 3.25 @ 10:31 --> 1.25 @ 10:41

Note we just lost lock around 10:42, probably due to these ITM steps. 

Driving the ITM ESDs to saturation in a single quadrant raises peaks in DARM.

We've only briefly (and at first accidentally) investigated. I drove ITMY and ITMX ESD UL to saturation at 17783 Hz and we see a broad peak in DARM over an order of magniture above noise floor at 81.9 Hz for each. Also drove ITMX same quadrant at 15006 Hz and saw several peaks in DARM, most notably at 55 Hz and 110 Hz. We'll do proper injections next. 

Kiwamu, Terra, Lisa

We had a hard time dealing with bounce and roll modes tonight and that slowed us down.

After keeping them under control, we followed directions from Daniel and we engaged the  ISS boost  (gain increase from 5dB to 7dB, boost 1 on -- we believe this is supposed to give us a factor of 2 suppression at 2 kHz, with higher gain at 100 Hz). 

The higher suppression of intensity noise around 100 Hz exposed a suspicious shape in the noise curve, so we decided to triple-check the calibration by making complete calibration measurements (see Kiwamu's entry).

With the new calibration filters, the suspicious shape at 100 Hz disappeared, and we got back a more familiar curve...and some Mpc, SensMon is around 60 Mpc.

The "jitter/intensity/(frequency?)" noise coupling to DARM stabilizes in a bad place after a while at 50W (yellow is O1 reference, all the other colors are curves taken at different times during the lock in low noise), making DARM worse between 500 Hz and 1 kHz, and even more clearly above 3 kHz. So there is more work to do to figure out how to keep the coupling low.


We leave the IFO to Terra for PI tests.


Some details:

- bounce and roll modes are becoming a problem, they required constant attention during lock acquisition;

- we had an error message once complaining that the ISS third loop was not engaged, and we couldn't move on in the locking sequence; Kiwamu bypassed the problem by forcing power increase (we also had some problem closing the ISS second loop from the medm screen (?)) 

- there have been (at least) a couple of ISS instabilities that prevented us from staying locked in DRMI; we "fixed" them by increasing the diffraction power, that happened at least twice today

- we reversed the change to the DARM gain, now it is back to 1400 

I took a quick look at six hours of 30-minute SFTs automatically
created by FScans this morning (thanks for setting the intent bit!),
to see if recent line mitigation attempts have reduced the amplitude
of the 1-Hz and other combs at low frequencies. Unfortunately, 
the low-frequency noise floor is too elevated to say much yet
about that mitigation success, but in the band 28-32 Hz, where
last night's noise is lower, there do seem to be fewer and lower-level
lines, including reduced amplitudes for the 1-Hz comb with a ~0.25-Hz offset.

Please note that the 56.84-Hz comb reported in the ER9 data
(present but weaker in O1 data)  remains quite strong.
There are other strong lines present, too, but I gather noise-hunting
is just getting started. The 56.84-Hz comb suggests a DAQ problem
somewhere, as noted before.

The attached plots show different bands of O1 data (narrow black curves)
superposed on last night's data (broad red swath). In each case, an 
inverse-noise weighting is used, to suppress periods of elevated noise.

Figure 1 - 10-2000 Hz
Figure 2 - 10-28 Hz
Figure 3 - 28-32 Hz
Figure 4 - 32-50 Hz
Figure 5 - 50-100 Hz
Figure 6 - 100-200 Hz
Figure 7 - 200-1000 Hz
Figure 8 - 1000-2000 Hz