On Friday (alog 55108) I measured the LSC feedforward filters that we'll need when we re-try increasing our DARM offset.  I've now got some candidate SRCL FF filters, although as usual they are tricky to fit. 

This SRCL FF filter would inject excess SRCL control into DARM above ~150 Hz, since the fit is not good above there.  It's hard to convince the fit of the feedforward to maintain high fidelity in the mid frequencies (where we really care) and also roll off the high frequencies.  So, an alternate solution might be to more sharply cutoff the SRCL LSC loop.  According to our most recent SRCL OLG measurement (albeit from May 2019) we have more than 50 degrees of phase margin in SRCL, so should have some room for additional rolloff. 

I have created a candidate new SRCL LSC cutoff, and calculated how the change in gain peaking we will expect to see.  It seems not so bad, so I propose that before we start our DARM offset test tomorrow morning we measure the SRCL OLG (to confirm that that phase margin is correct), and put in the new cutoff filter.  Then we will increase the DARM offset, turn on the new SRCL FF filter, do a quick scan to ensure that we're at the optimal squeezing angle, and then compare this 14 pm DARM offset configuration with our current nominal 10 pm DARM offset configuration.

In the first attachment I show the current SRCL LSC cutoff in red versus the candidate cutoff in blue [ ellip("LowPass",4,0.3,20,150)ellip("LowPass",2,1,20,650)gain(1.16145) ].  The candidate filter takes away an additional 13 degrees of phase margin than the current filter.

In the second attachment I show the SRCL OLG measurement from May 2019 in red/orange, and then what that measurement would look like if the cutoff were replaced with the candidate cutoff in blue.

In the third attachment I show G / (1 - G) for both of the cases from the second attachment (current situation in red/orange, and candidate situation in blue). 

I think that it should be fine to try the new SRCL cutoff. 

All that said, the fourth attachment is my candidate fit for SRCL FF at the higher DARM offset. This is an iterative filter, so if the currently in-use filter were perfect, this iterative filter should be unity.  The blue points are the measured TF to fit, and the green is the filter output (IIRrational plus some hand adjustments to make the low frequency and high frequency less egregiously bad).  The fit above 150 Hz isn't good, which is why I'm interested in making the SRCL loop cutoff more sharp.  The fit also isn't so great below 10 Hz, but at least that part is out of the GW band, and shouldn't hurt us too much when testing the higher DARM offset. The bottom half of the 4th attachment is a representation of how good or poor the fit is to the data.

Spent 30 minutes updating amplitudes for DHARD & CHARD signal injections in 'userapps/asc/h1/scripts/sensingMatrix/run_sensmat.py' - previous values were ~1000x too high.

(Kyle R, Gerardo M)

Cooling lines were attached and cooling system was tested for leaks, flow was verified.  Oil was added per instruction manual to the gear and bearing chambers.  Power was applied to the hepta controller.

The hepta pump was started for a few seconds and stopped manually, the ON time was enough to determine that the direction of the motor is correct, and the pump ran smooth.  However we did note a red warning light ON, phase sequence, troubleshooting will be done later about what this error light means.

We did a second start, but this time we allowed for the trouble/error to shut the pump OFF, it almost took the same time as when done manually, once again the pump ran smooth but once again showing the error mentioned above.

Cooling lines were shut off and power was turned OFF for the controller.

TITLE: 02/20 Eve Shift: 00:00-08:00 UTC (16:00-00:00 PST), all times posted in UTC
STATE of H1: Observing at 119Mpc
OUTGOING OPERATOR: Jeff
CURRENT ENVIRONMENT:
    SEI_CONF state: WINDY
    Wind: 9mph Gusts, 8mph 5min avg
    Primary useism: 0.02 μm/s
    Secondary useism: 0.19 μm/s
QUICK SUMMARY: Just relocked, calm environment.

Started and stopped a few times around 1325 hrs local

Evan G., Jeff K.

We analyzed the data that Jeff collected at the start of a lock stretch (see LHO aLOG 53951) to understand how the sensing function evolves and impacts the response function as the IFO thermalizes. There were indications that the low frequency part of the sensing function was evolving quite a bit more than we expected. To investigate, Jeff had injected a comb of lines using the PCAL and at the DARM excitation point so that we had a measure of the loop suppression to extract the sensing function.

This data was analyzed offline using gwpy to compute transfer functions at 2 minute intervals in order to plot the changes as a function of time. Typically the sensing function transfer function swept sine measurements are made when the IFO has thermalized. These injections were at fixed frequencies so that we could track the evolution.

Attached are 8 figures:
1) For the 4 frequencies where the sensing function is measured, we plot the time evolution of the sensing function compared to the pyDARM model predicted using the modelparams_H1_20190909.py file and GDS computed f_cc and kappa_c
2) For the 18 frequencies where PCAL is injected, we have measured the response function evolution as a function of time compared to the pyDARM model predicted using the modelparams_H1_20190909.py file and GDS computed f_cc and kappa_c
3) For the 18 frequencies where PCAL is injected, we plot the spread of the data points compared to the pyDARM model predicted using the modelparams_H1_20190909.py file and GDS computed f_cc and kappa_c
4) For the 18 frequencies where PCAL is injected, we plot the spread of the data points compared to the GDS predicted values (this serves as a check that comparing to pyDARM and the GDS computed values is fine). Note that GDS has a high-pass filter of ~10 Hz
5) Sensing function at intervals of 10 minutes to show the evolution as a function of time (BLUE is the start of the lock, YELLOW is the thermalized IFO state)
6) Response function at intervals of 10 minutes to show the evolution as a function of time (BLUE is the start of the lock, YELLOW is the thermalized IFO state)
7) Sensing function at intervals of 10 minutes to show the evolution as a function of time if one also applies the f_s value as computed by GDS, assuming a Q value of 20 (BLUE is the start of the lock, YELLOW is the thermalized IFO state)
8) Response function at intervals of 10 minutes to show the evolution as a function of time if one also applies the f_s value as computed by GDS, assuming a Q value of 20 (BLUE is the start of the lock, YELLOW is the thermalized IFO state)

The plotting script is in the CAL svn: ^/trunk/Runs/O3/H1/Scripts/FullIFOSensingTFs/process_sensing_20191217_darm_comb.py

The bottom line is in figure 6: at 20 Hz near the beginning of a lock, h(t) is would have a systematic error of ~8% in magnitude and negligible phase. This evolves over the course of ~2 hours, reducing the systematic error close to zero by the end of the 2 hours.

We need to compare this with our current uncertainty estimates and evaluate how to proceed if any interesting triggers are occurring in this 2 hour window at the start of a lock stretch.

   Good Observing shift so far. Environmental conditions remain favorable. Range is averaging around 119Mpc. No issues of concern at this time.  

   The FMC crew has finished tumbleweed clearing along the X-Arm for the day.   

• Topped off Crystal Chiller with 175 mL. ​​​​​​
• Diode Chiller OK.
• Filters look fine (slight yellow tint of Crystal filter [known item]).

WP8536 Reduce Guardian logs lookback

Jamie, TJ

The Guardian log lookbacks were reduced to 2 months. TJ verified that this improved log retrieval times.

h1iscex IO Chassis new DC power supply

Richard, Keita, Fil, Dave:

The h1iscex front end and IO Chassis were powered down and the IO Chassis was powered from a different +24V power source. Later in the day it was noted that the 21st channel of the second ADC (adc0-ch20) was misreading. Instead of reading -500 counts it was reading +2000. When the cable connecting the ADC to the AA chassis was disconnected, all channels in this ADC went to zero except for chan20 which increased to 10,000. Fil boosted the power and after a second IO Chassis cycle the problem was resolved.

EX Beckhoff hardware restored

Richard, Fil, Dave:

After a week with some photodiodes disconnected from the EX Beckhoff slow controls system, these were restored. The Beckhoff status on the CDS overview went GREEN, I added EX back into the full system summary.

TITLE: 02/19 Owl Shift: 08:00-16:00 UTC (00:00-08:00 PST), all times posted in UTC
STATE of H1: Observing at 118Mpc
INCOMING OPERATOR: Jeff
SHIFT SUMMARY: Other than squeezer range issue
LOG:
12:01 Out of Observing to relock squeezer, see previous alog.

 Just noticed that the range has been degrading for the last 2 hours or so. Range had drifted down to about 108 mpc, no apparent environmental causes, wind is low, microseism hasn't changed. Following Sheila's instructions in alog 54967. Seems like the range has recovered, but it took several minutes after relocking the squeezer to see that. 

Out of Observe at 12:01 utc, back 12:04.

TITLE: 02/19 Eve Shift: 00:00-08:00 UTC (16:00-00:00 PST), all times posted in UTC
STATE of H1: Observing at 121Mpc
INCOMING OPERATOR: Jim
SHIFT SUMMARY: Once the EX ISC IO chassis was power cycled and the ALS interlock was delt with, it has been smooth sailing.
LOG:

• 0013 iscex model restart
• 0040 Keita, Fil back from EX
• 0045 Keita, Fil back to EX
• 0125 Keita, Fil back

After Keita and Fil were able to get the EX ISC IO chassis working again (alog55171 & alog55173) and then get around the interlock issue (same alogs), it was easy sailing up to Low Noise.

I accepted one SDF that I made adjusting the TCSY CO2 power in calibration.

Camilla pointed out earlier today that the TR offsets that we have hard-coded in the guardian looked like they weren't so good any more.  I thought they'd be fine, but then TJ's relocking attempt got stuck because the offset was so bad.  So, I updated the values, and now the normalized transmitted arm powers actually look like zero when the arms are not locked in IR.

WP 8537
alog 55041

Ongoing noise hunting efforts at EX continue. Last week we moved auxiliary beckhoff electronics away from the ESD electronics (alog 55041) and powered down the Baffle Photodiode Amplifier chassis.

Today the ISC IO chassis 24V power was moved to same power supply that feeds the SEI IO chassis. Previously the 24V feeding the ISC chassis was also powering the slow controls Beckhoff electronics.

The Baffle Photodiode Amplifier chassis that was turned off last week was powered back on. Leaving unit off showed no change in our lines.

F. Clara, R. McCarthy

Sheila, Jenne, Keita

Summary: We have some evidence that our sensitivity can be better with a higher DARM offset.

Details:

Last Thursday I took some data with the squeezer off at different DARM offsets, 55086 to compare with the data that Jenne took 54969. 

• The first attachment is just a GDS calib strain spectrum, and show data from the two DARM offsets that we used on Thursday and some data from a no squeezing reference time at our nominal 10pm DARM offset (from 54822).  The calibration lines at 17Hz, 410Hz, and 1.1kHz all line up to within 1% of each other in these three times in GDS. 
• The second plot shows the coherence between strain and the LSC control signals. The SRCL feed-forward is only well tuned for our nominal DARM offset, as expected.  The PRCL coherence is slightly increased by changing the DARM offset and MICH FF seems not to depend much on the DARM offset.  Jenne remeasured the feed-forward on Friday with the DARM offset at 14pm 55108.
• The third attachment  shows the same GDS spectra as the first attachment, but with an estimate (best possible) or the sensitivity we could get with perfect SRCL subtraction for all three cases (strain*sqrt(1-coh)).  There are several frequencies where the 14pm sensitivity is just a bit better than 10pm.

The high frequency noise is consistent with the measured change in optical gain and the dark noise:

• The 4th attachment shows kappa c and the DCPD current plotted against the x0 offset (the darm offset in picometers used for the DCPD normalization.)  Kappa C (the normalization for the sensing function used to compensate for time dependence in optical gain) is calculated by demodulating DARM err, which is measured after the normalization by x0.  This means that ideally, if the DCPD normalization by x0 was perfect, kappa C would not change with the DARM offset.  Kappa C does decrease for the 5pm DARM offset, which could be because of junk light that isn't sensitive to DARM.  In the lower panel of the 4th attachment the curve which is drawn there is a parabola defined by the 14pm measurement, just as a visual aid.
• The 5th attachment is shows how the sensitivity is improved at high frequency. Only the 5pm (gray colors) and 14pm data (reddish colors) are shown here.  The sum of dark noise and shot noise explains the degradation of the high frequency sensitivity for the low DARM offset, mostly because of the loss of optical gain.  This is from DCPD sum data calibrated into strain using pyDARM and correcting for changes in the fringe offset, kappa C, and the cavity pole frequency.  With an FFT length of 10 seconds, the calibration lines agree within 3%, the plots attached are with a 3 second FFT so that we can get more averages for the cross correlation. 

The last two plots show the low frequency sensitivity and the impact of SRCL subtraction, comparing our nominal DARM offset of 10pm to the candidate new offset of 14pm. 

• The 6th attachment shows the DCPD sum data while the 7th attachment shows the same thing for the DCPD cross correlation.  You can flip back and forth between them to see the difference if you like. 
• The SRCL noise is estimated based on coherence with the DCPD sum and is the same in both plots. 
• For the gray (10pm) case, the dashed line showing the measured sensitivity is directly on top of the sensitivity with the estimated SRCL noise subtraction, for the 14pm case there is a significant improvement from the estimated subtraction.  Where the solid maroon traces are below the gray traces, there is some improvement in the sensitivity with the reduced DARM offset. 

No obvious cause.

Vlad, S.Karki, D. Barker

We pushed the Pcal Simulink model and MEDM screen that Shivaraj generated at LLO (LLO alog 49839). The updated suspension filter files (LHO alog 53155) have been loaded into the corrponding Pcal filter banks as well. 

The new calibration coefficients have been uploaded to the appropriate EPICS variable using the txt files linked below:

aligocalibration/trunk/Runs/O3/H1/Results/CALCS_FE/lho_yend_pcal_epics_created-20191111.txt

aligocalibration/trunk/Runs/O3/H1/Results/CALCS_FE/lho_xend_pcal_epics_created-20191111.txt

While quacking the susmodel we encounted a number of issues:

1) The filterfile had to be copied to a local directory and qucked there, as the quacking script would give errors concerining the new SOS coefficients (reason unknown)

2) The quack script only updates gains and SOS coeffcients of the respective filter modules. It does NOT update the desgng strings (the commented out lines that describe what the filter is). Attemting to run "foton -c" (something that needs to be done to parse the foton filters correctly) on the filter file would result in foton detecting a difference between the design string and the real filter, and would take the design string and overwrite your brand new filters.

The workaround to the second problem (thanks to Shivaraj) was to delete the design string before running "foton -c", forcing foton to regenerate it.