TITLE: 10/14 Owl Shift: 07:00-15:00 UTC (00:00-08:00 PST), all times posted in UTC
STATE of H1: Commissioning
INCOMING OPERATOR: Travis
SHIFT SUMMARY:  Very windy (gusts to over 60mph) since near the beginning of my shift with high microseism (~1 um/s).  Not great conditions for locking regardless of other issues we ran into.  Several dust alarms in the PSL and end stations.
LOG:

9:50 DIAG_MAIN notification that the Noise Eater was out of range.  Went to LVEA and reset Noise Eater.  Seemed to fix the issue

10:05 Another notification that Noise Eater is out of range.  Also, both ALS nodes went into fault with the same 'Beat note strengh' error (typo is in Guardian, not my log) that we saw earlier when Kiwamu found the FSS locked on the wrong fringe, aLog 30524.  I again went to the LVEA and reset the Noise Eater, then attempted to toggle the Autolock off/on per Kiwamu's aLog.  This did not fix the FSS issue.  I don't know what else to do here. 

10:23 Put Guardian into DOWN.  WIll require further investigation by PSL crew/commissioners before locking will happen.

 

Kiwamu, Terra

One thought for tracking PI mode frequency drift over a lock stretch (as seen here) is to utilize the drumhead modes: these are reliable in amplitude/presence and by monitoring their freq drift we could predictively change PI filters, etc. to handle drifting of higher modes.

Eigenfrequencies of the test masses have nominally been defined as proportional to square root of Young's modulus: f ~ E^1/2, where E := Young's modulus. The Young's modulus for fused silica has some temperature dependence ( measured to be (dE/dT)/E = 1.52e-4 / K ); as temperature increases, Young's modulus increases, so the mode frequency increases over a lock stretch. In this simple case we would expect all modes in a given test mass to have the same relative drift. Instead we see that del_f has mode dependence as well, i.e. the ~ 15007 Hz differential drumhead mode always drifts more than the ~ 8160 Hz drumhead mode of the same optic (relative to their frequency). I suspect this is due to different modes having more or less torsional versus longitudinal motion and so frequency dependence has both E (Young's) and G (shear) dependence. Peter also suggested nonuniformity of temperature in the test mass volume (as different modal motions interact with certain parts of that volume differently). To this end we define, for an arbitary mode m:

f_m = g(T) , where g is some mode-specific temperature dependent function involving E,G

     ==>  f_m + del_f_m ~= g₀ + g₁del_T + . . .  

     ==> del_f_m / f_m = g₁del_T / g₀ = A * del_T

Similarly, for another mode n: f_n + del_f_m ~= h₀ + h₁del_T  ==>  del_f_n / f_n = B * del_T where h is some function with some different temp dependence

Then the ratio of the change of frequency over frequency of those specific modes is constant: 

(del_f_m/f_m) / (del_f_n/f_n) = A / B = constant

I've quickly looked at three longer recent locks, specifically at change of drumhead modes compared to change of 15007 Hz modes of the ETMS. ETMY seems to agree on a constant while ETMX a bit wandery . I'll look more systematically tomorrow to see if I can narrow it up.

ETMY (del_Drumhead / Drumhead) / (del_15007 / 15007) = 1.38, 1.31, 1.31   |    ETMX (del_Drumhead / Drumhead) / (del_15007 / 15007) = 1.29, 1.45, 2.45

We were basically ready to start TCS tuning tests, but we can't relock the IFO :(  Since it took me a while to get these settings, this will basically be a note-to-self of what I had set up, so that I can do it again more quickly tomorrow. 

AS port OSA

• Sheila helped plug in the OSA that is at the AS port to the roaming wireless router (that belongs with the gpib setup - only having one of these is a problem!).
• We can now use Jamie's script (documented in alog 29563) to pull the data.
  • tektronix download -c 'ch2' ~/LHO_work/2016_10_13_OSA_AS/ 0 60
  • In the above, 0 tells it to keep downloading data until we crtl-c to quit.  60 tells it to get data once every 60 seconds.  The help file is very detailed.
• We forgot to set the vertical axis ranges for channels 2 and 3 differently, so right now we can only see the 45MHz peaks clearly.  We should zoom one of the channels in so that we can see the 9MHz peaks.
• Gabriele has a script that will pull all the data from a folder, find the peak height and plot the peaks as a function of time.
  • Currently it lives in /ligo/home/gabriele.vajente/2016_10_13/sideband_amplitudes.m
  • Give the script the peak location for both upper and lower sidebands in the first data file, as well as about how much they move between data sets, and his script will find the max value in a range based on that. 
  • We ran it for about 45 min while I was still setting up, so without any TCS changes, and saw that the 45MHz sidebands started out pretty imbalanced, but were moving toward being more balanced as the IFO thermalized.
• The goal will be to tune TCS such that the 45MHz peaks are well balanced.

Frequency line

• Similar to what was done in September (alogs 29553, 29509), I used awggui to inject a line at 900Hz into the Refl Servo, 100 counts amplitude.
  • H1:LSC-EXTRA_AO_2_EXC is plugged into Exc A of the Refl Servo's common mode board, and H1:LSC-REFL_SERVO_COMEXCEN is enabled.
• I used the LSC OSC1 to demodulate signals at this frequency.
  • The frequency was set to the same 900 Hz, but the LSC output matrix didn't send the oscillator signal to any optics. 
  • Demod 1 looked at OMC DCPDs, Demod 3 looked at REFL9I, and Demod4 looked at REFL45I (this is the same convention as in alog 29553).
  • cdsutils servos were used to change the phase of the demodulators to zero the Q-phase signal.
    • cdsutils servo -r LSC-LOCKIN_1_DEMOD_1_Q_OUT16 -g -100 LSC-LOCKIN_1_DEMOD_1_PHASE
    • cdsutils servo -r LSC-LOCKIN_1_DEMOD_3_Q_OUT16 -g -100 LSC-LOCKIN_1_DEMOD_3_PHASE
    • cdsutils servo -r LSC-LOCKIN_1_DEMOD_4_Q_OUT16 -g -100 LSC-LOCKIN_1_DEMOD_4_PHASE
• The goal will be to tune TCS such that the frequency line is minimized in DARM.

Intensity line

• I used awggui to inject a line at 860Hz, 10,000counts into the ISS second loop.
  • H1:PSL-ISS_SECONDLOOP_EXCITATION_EXC
  • H1:PSL-ISS_SECONDLOOP_EXC_SWITCH_MON enabled, but first H1:PSL-ISS_SECONDLOOP_REFERENCE_VALUE set to zero so that I wouldn't give the ISS a giant kick. 
• I used the LSC OSC2 to demodulate, but only Demod 1 for OMC DCPDs.  Same situation as the freq line - set the frequency of the LSC oscillator, but don't send the signal anywhere in the output matrix.
  • cdsutils servo -r LSC-LOCKIN_2_DEMOD_1_Q_OUT16 -g -1 LSC-LOCKIN_2_DEMOD_1_PHASE to zero the Q-phase signal.
• The goal will be to tune TCS such that the intensity line is minimized in DARM.

Jitter line (pitch)

• I used awggui to inject a line at 820 Hz, 1000 counts into the PSL PZT in pitch.
  • H1:IMC-PZT_PIT_EXC
• I used the LSC OSC3 to demodulate, but only demod 1 for the OMC DCPDs.
  • Since trying a larger amplitude for this line broke the lock, I didn't set the cdsutils servo.
• The goal will be to tune TCS such that the jitter peak is minimized in DARM.

Obviously, the goal is to find a perfect TCS place where all the couplings are minimized at the same time, and the giant peaks in DARM are minimized.  We'll see if that is possible.

Attached are screenshots of my 3 desktops, to remind myself of any other settings that I've not written down.  Note that the IFO was not locked when the screenshots were taken.

Travis, Jenne, Terra, Kiwamu,

A brief report on a rare fault mode associated with FSS and ALS.

At one point tonight, we noticed that both end station green lasers did not lock to the arms. It turned out that both the end stations lost the beatnote against the PSL light. Looking at various channels, we found that it was actually the reference cavity which locked to a next fringe, resulting in a large shift in the NPRO frequency. This corresponded to a temperate crystal feedback of -0.25 K (compare this to nominal of 0 K). A strange thing is that the temperature sweep for the FSS auto-locker had been set to [-0.1 K 0.1 K] which means that theoretically the chance of hopping to the next fringe is very small. Anyhow, relocking the FSS back to the nominal fringe by disabling/enabling the auto-locker fixed the issue. Both green lasers re-acquired the PLL without an issue.

[Sheila, Jenne, Kiwamu, Patrick]

We lost lock about 8 hours ago, and haven't been able to lock since.  Sheila suggested we look at the Yarm ALS PDH control signal, since that has caused trouble in the past.  Indeed, we saw glitches.

We were looking at H1:ALS-Y_REFL_CTRL_OUT_DQ, and saw very obvious glitches at about a 1Hz rep rate.  At the same time, we see dips in the ALS transmitted power (H1:ALS-Y_TR_A_LF_OUT) down to 0.5 or lower from a nominal locked value of 1.0.  We also see much smaller glitches in the Xarm, often at the same time as the Yarm glitches, but not always. The Xarm power dips are also much smaller than those for the Yarm. The first attachment shows the REFL CTRL signals.  We don't save the transmitted powers, so I don't have a good plot of those. 

We have certainly seen these glitches before, and there are rumors that it seems to happen more often when it's raining.  Not totally clear how that's related - is the extra humidity doing something to us??  Anyhow, in the past the solution has always been to wait it out, and it'll go away.  Colloquially the problem is usually gone by the time anyone drives down to the end station. 

We went down to the end station and tried several things while looking at the PLL beatnote, but none of them changed the glitches.  At a few times, Sheila noted that there were peaks around the beatnote, 8MHz above and below.  We're not totally sure what those are. 

• I changed the temperature of the crystal, and Sheila could see the beatnote moving around, but saw no evidence of mode hopping in the laser.
  • To do this:  On the ALS_Overview screen in the Yarm state machine, select "unlocked". 
  • On the Yarm PLL screen, top left corner, select "off" for the servo.
  • Change the value in H1:ALS-Y_LASER_HEAD_CRYSTALFREQUENCY - this will move the temperature.
  • To go back to regular locking: On the ALS_Overview screen in the Yarm state machine, select "PLL locked".
• We toggled the ALS laser's noise eater, then relocked the green arm.  No change in the glitch behavior.
  • To do this:  From the Yarm PLL screen, top left corner go to ALS Laser sub-screen.  Toggle noise eater button off, then on again.
• We turned off the ALS laser, then turned it back on and relocked the green arm.  Again no change in the glitch behavior.

Suspecting that the PSL was somehow involved since we often see glitches simultaneously in the X and Y arms, we closed the light pipe shutter for the PSL ALS beam that goes to ISCT1.  We were hoping that if there were some kind of feedback from the Yarm to the PSL that was then getting sent to the Xarm through the PLL, this would prevent that.  However, no change.  The light pipe's shutter was re-opened.

We then went back to the Yend to lock the green arm, but bypassing the PLL, thus eliminating the PSL light from the system.  We didn't see any tell-tale dips in the transmitted power while we did this, but when we put the system back to normal, the problem seemed to have already resolved itself.  So, it's not a 100% conclusive test. 

• To do this, we unplugged the PDH common mode board's input (lower of the 2 common mode boards) and plugged it into the PLL common mode board's input (upper of the 2 common mode boards). 
• On the PLL servo board, we changed the Input1 polarity to positive (it's usually negative), and increased the gain to above 10dB.  (16dB worked for a while, but then it seemed like it was too much and causing locklosses).  The arm locked in green very easily.  If it unlocked, sometimes we toggled the polarity, and sometimes it seemed to acquire better if we ramped the gain down then back up again.
• We were locked this way from about 06:42 UTC to 06:50 UTC 14Oct2016. 
• Since the refl control signal wasn't in use, we just monitored the transmitted power and said that we didn't see any dips in the power.
• Assuming that the glitches would have still been there in the normal configuration, we think that this absolves the ALS laser and PDH sensor and PLL servo board from blame. 
• However, when we put things back to normal, the glitches were basically gone.  So, maybe we didn't learn much of anything :/

We're going back to trying to lock for now.

I have changed the whitening filters for the calibrated DRMI spectra. They are changed from zpk([1;1], [100;100], 1) to zpk([0.3; 0.3; 0.3; 0.3; 0.3], [10; 10; 10; 10; 10], 1). The window, which had been flat-top (23090), is now put back to Hann. The dtt template is also update accordingly.

See attached screenshots of OpLev trends. 

Trouble keeping ISS locked through beginning of shift. Had to keep relocking by hand. IMC kept losing lock. Found glitching on H1:ALS-Y_REFL_CTRL_OUT_DQ and H1:ALS-X_REFL_CTRL_OUT_DQ. Jenne, Sheila and Kiwamu tried to isolate the cause. Cause not found, but glitching eventually gone. Attempting to lock. Handing off to Travis.

23:18 UTC restarted video0
23:31 UTC lock loss (Jeff, Jenne)
23:35 UTC Chandra to LVEA, RGA between HAM5 and HAM6
01:16 UTC Initial alignment
01:42 UTC Initial alignment done
02:32 UTC ISI config changed to WINDY_USEISM, never moved past not all nodes arrived
02:46 UTC ISI config changed to USEISM_NOBRSXY
04:31 UTC Jenne and Sheila to end Y to check on ALS laser
05:23 UTC Jenne and Sheila back. Jenne to LVEA to block ALS light pipe.
06:14 UTC Jenne and Sheila at end Y to replace PLL signal with electronics.
07:10 UTC Jenne and Sheila back

FAMIS request 6492

Added 50mL H2O to PSL crystal chiller. It was hard to tell if it needed any as the level was fluctuating quite a bit. No alarms on either the crystal or diode chiller. The water filters appear clean.

I've looked at the OMC length noise. A factor of 30 increase in RMS is enough to create broadband noise with a slope of 1/f^2 below 200 Hz.  Based on predictions for the noise without any excitation on, it seems like OMC length noise should not be limiting DARM at low frequencies.

Relevant alogs and documents: 19212 18034 G1100903 Section 5.4 of Dan Hoak's thesis

Estimating OMC length noise

To get an out of loop measurement of the OMC length fluctuations at high frequencies, I used the method described in the above documents, with a few changes. 

With the interferometer locked at 2Watts, before transitioning to DC readout, I locked the OMC on both the carrier and sidebands.  The first attached figure shows DCPD power spectra with the OMC locked both on resonance and somewhat detuned.  With the OMC on resonance, the coupling of OMC length to transmitted power should be quadratic, so the orange on resonance traces show us the level of intensity noise that is the noise floor of these measurements.  With the OMC slightly detuned, a linear coupling from OMC length is introduced, so at frequencies where the blue detuned traces are above the orange, we have made a decent measurement of OMC length here.  A few quick points:

• The OMC angular controls should be switched to the QPDs for this measurement, and the DARM boost on.  
• It is best to make the measurement for the carrier with the full interferometer locked to take advantage of the IFO's intensity noise filtering.  
• The measurement made locked on the carrier is also sensitive to DARM, so the shot noise imposed on DARM by the RF DARM loop dominates the noise up to 2 kHz.  If necessary we could try to improve this measurement by reducing the bandwidth of RF DARM, but it seems adequate as is.  
• For the sideband measurement, we have to turn the ISS 2nd loop gain up to match what we would have at 50 Watts (there is no gain scaling with input power for the current generation of ISS, so to emulate the loop shape at 50 Watts when locked at 2 Watts: turn of the AC coupling, turn the gain up by 30dB, and engage the 2 boosts.  Reverse order to undo before powereing up)  The noise is then limited by the shot noise of the ISS second loop, so this measurement could possibly be improved by doing it at 50 Watts single bounce.  
• I made the detuned measurements using the dither lock for OMC length, and put offsets in the servo, not by using the side of fringe lock used before.  This means that the side of fringe measurement is out of loop, and we don't have to worry about correcting for the loop gain.  It also means that you cannot lock at a half fringe.  I didn't want to try to write math in the alog, so the attached PDF shows how to find the fringe offset and calibrate the side of fringe measurements for any offset.  

At low frequecies, I used the in loop error signal to estimate the OMC length noise.  The equation for the optical gain of the dither loop is in the attached PDF, I used a dither amplitude of 0.8pm (there is a fudge factor here).  The dither loop is currently not shot noise limited, because we have large frequency noise in DARM around the dither frequency of 4100Hz which changes over hours times scales.  To estimate the length noise imposed by the dither loop, we need a model of the loop gain, shown in the second attached png (thanks Koji for the PZT model).  For frequencies where the error signal is dominated by sensor noise, the estimate of the actual length noise is measured error signal*OLG/bandpass filter upstream of demodulation, where it is dominated by real length residual we can just use the measured error signal (ignoring the bandpass).   

The third attached PNG shows the four different ways of estimating OMC length noise (carrier side of fringe, sideband side of fringe, sensor noise from dither error signal, residual length from dither error signal), so you can see how the calibrations agree.  I've shown the projection for sensor noise when we are locked on the sideband, which is higher than in low noise, so that you can see that it agrees with the out of loop sideband measurement fairly well.  Most of the measurements here were taken with a 1 count exctiation to PZT2 at 12 Hz.  At 12 Hz, the low noise residual error signal should agree with the out of loop sideband measurement,  but there is about a factor of 2 discrepancy.  The closed loop supression for the 2 cases are similar to within 1dB, so changes in the loop gain don't seem to explain this discrepancy. 

To combine these measurements into one estimate of OMC length noise, I used some time domain filters to choose the frequencies where each measurement is valid.  Making time domain filters to combine these signals is a little tricky, I have started with complimentary comlimentary filters to blend the sideband and carrier measurements around 1 kHz, and the sensor noise and residual estimates around 20 Hz, but if you look closely at the combined estimate there are places where it is off by 40% or more from what seems reasonable.  The resulting estimates, are shown in the 4th attached png. When there is no excitation, the RMS of 1pm is dominated by the dither line, with a 0.3 count excitation the RMS is 9 pm, and with a 1 count exctation the RMS is 29 pm.  

Projections to DARM

With an estimate of the OMC length in the time domain, using the first equation in the attached PDF it can easily be used to predict OMC RIN.  The 5th png attached shows DCPD RIN when I did injections with the IFO at low noise.  The largest injection (which increased the RMS by a factor of almost 30) caused a braodband increase in noise below about 200 Hz, with peaks at the 2nd, 4th, and 6th harmonics of the injection frquency.  The model predicts most of the features of DARM within about a factor of 2, although not the upconversion around the 4th harmonic, and not the upconversion around the 4th harmonic.  

The last png attachment is the prediction for the noise in DARM caused by OMC length without an excitation.  It is 2 orders of magnitude below DARM at 100 Hz, the only place OMC length might explain the level of noise in DARM is around the dither frequency.  

No unexpected restarts over these 7 days. h1fw0 gives occassional retransmission requests.

model restarts logged for Wed 12/Oct/2016 No restarts reported

model restarts logged for Tue 11/Oct/2016
2016_10_11 09:14 h1hpibs
2016_10_11 09:24 h1hpiitmx
2016_10_11 09:31 h1hpiitmy
2016_10_11 09:39 h1hpietmx
2016_10_11 09:41 h1hpietmy
2016_10_11 09:46 h1hpiham1
2016_10_11 09:52 h1hpiham2
2016_10_11 09:53 h1isiham2
2016_10_11 09:59 h1hpiham3
2016_10_11 09:59 h1isiham3
2016_10_11 10:04 h1hpiham4
2016_10_11 10:04 h1isiham4
2016_10_11 10:06 h1hpiham5
2016_10_11 10:08 h1isiham5
2016_10_11 10:11 h1hpiham6
2016_10_11 10:11 h1isiham6
2016_10_11 10:34 h1calcs
2016_10_11 10:37 h1susitmx
2016_10_11 10:39 h1fw2
2016_10_11 10:41 h1broadcast0
2016_10_11 10:41 h1dc0
2016_10_11 10:41 h1fw0
2016_10_11 10:41 h1fw1
2016_10_11 10:41 h1nds0
2016_10_11 10:41 h1nds1
2016_10_11 10:41 h1tw0
2016_10_11 10:41 h1tw1
2016_10_11 10:45 h1fw2
2016_10_11 10:55 h1fw0
2016_10_11 11:02 h1fw1
2016_10_11 11:36 h1tcscs
2016_10_11 11:41 h1psliss
2016_10_11 11:43 h1psldbb
2016_10_11 11:52 h1lsc
2016_10_11 12:01 h1lsc
2016_10_11 12:09 h1broadcast0
2016_10_11 12:09 h1dc0
2016_10_11 12:09 h1fw0
2016_10_11 12:09 h1fw1
2016_10_11 12:09 h1fw2
2016_10_11 12:09 h1nds0
2016_10_11 12:09 h1nds1
2016_10_11 12:09 h1tw0
2016_10_11 12:09 h1tw1
2016_10_11 12:24 h1lsc
2016_10_11 13:00 h1susetmx
2016_10_11 13:03 h1broadcast0
2016_10_11 13:03 h1dc0
2016_10_11 13:03 h1fw0
2016_10_11 13:03 h1fw1
2016_10_11 13:03 h1fw2
2016_10_11 13:03 h1nds0
2016_10_11 13:03 h1nds1
2016_10_11 13:03 h1tw0
2016_10_11 13:03 h1tw1
2016_10_11 16:15 h1sysecatc1plc1sdf
2016_10_11 16:17 h1sysecatc1plc3sdf

Maintenance day. New SEI models, new PSL ISS (bug fix) and DBB LSC (jitter feed-forward) models, new SUS ITMX,ETMX HWWD models. Updated beckhoff SDF, various DAQ restarts.

model restarts logged for Mon 10/Oct/2016 No restarts reported

model restarts logged for Sun 09/Oct/2016 No restarts reported

model restarts logged for Sat 08/Oct/2016 No restarts reported

model restarts logged for Fri 07/Oct/2016
2016_10_07 10:22 h1fw2

Testing latest daqd on fw2 (md5 sum filename change)

model restarts logged for Thu 06/Oct/2016
2016_10_06 14:52 h1psldbb
2016_10_06 14:53 h1broadcast0
2016_10_06 14:53 h1dc0
2016_10_06 14:53 h1fw0
2016_10_06 14:54 h1fw1
2016_10_06 14:54 h1fw2
2016_10_06 14:54 h1nds0
2016_10_06 14:54 h1nds1
2016_10_06 14:54 h1tw0
2016_10_06 14:54 h1tw1
2016_10_06 18:15 h1fw2

PSL-DBB model change with DAQ restart. Testing new daqd on fw2.

model restarts logged for Wed 05/Oct/2016
2016_10_05 15:58 h1asc
2016_10_05 16:01 h1dc0
2016_10_05 16:01 h1nds0
2016_10_05 16:02 h1fw1
2016_10_05 16:03 h1broadcast0
2016_10_05 16:03 h1fw0
2016_10_05 16:03 h1fw2
2016_10_05 16:03 h1tw1
2016_10_05 16:04 h1nds1
2016_10_05 16:05 h1nds1

new asc code with daq restart.

State of H1: locked in Nominal Low Noise

Activities:

• Kyle to MY and back
• Christina to EY (outside) and back
• Gabriele worked on alignment during the mid-day lock 

H1 locking:  all changes / issues resolved or being worked on

• REFL WFS have remained on
• after lock loss, some alignment was needed for the arms, BS, PRM, and SRM
• POP Offsets remain good
• Jenne - working on automating the CSOFT loop offset
• Tara - working on PI mode 26 that again did the hopping thing

In brief, no success. In details, here are some alignment tweaks trief over the last couple of days:

• moving SRM in pitch reduces the jitter peaks at 200-400 Hz, at the price of worse sideband powers
• moving IMC DOF_1_P (offset of +400) reduces a lot the peaks, but a the price of decreasing the IMC transmission. SRM had to be moved to keep reasonable sideband buildup
• moivng MICH both in pitch and yaw had not clear effect
• going to the best CSOFT offsets that Jenne found helps a bit.

The decrease of range in the last lock was due to a misalignment of the SRM.

issues getting here:

• REFL WFS centering not coming on in Inital alignment for PRC and SRC, so in both cases those didn't complete until engaged by hand (guardian?)
• alignment
• POPA OFFSETS - Evan looks at these - not sure they're OK but in the end they seem to work
• ASC - Evan toggled PRC1_P on and off (3 seconds each) to save the lock and also allow PRC1 P to converge 
• POPA sum got noisy in ENGAGE SOFT LOOPS until I engaged the CHARD offset of 0.1 by hand in Engage Soft Loops, then the signal improved
• I set the ramp to 120 seconds, and at 120 seconds there's a kick that can be seen on CSOFT_P
• PI mode 2 rang up - easily damped
• PI mode 26 rang up and did a bouncing thing until I lowered the gain by half from -5000 to -2500, and then tuned the phase

Using the crystal chiller that is currently sitting in the mechanical room, the on/off flow rate
was recorded as a function of time.  The data is that reported via the RS-232 interface.  When
switched off, the flow rate drops to zero in 3.245 seconds.  This seems a lot slower than the
drop reported by EPICS, which seems to suggest the flow rate drops to zero in less than a second.

Summary:  I have moved the POP_A and SOFT offsets to find better recycling gain (31.2ish rather than the 29.5 we had been getting).  I was hoping that this would help eliminate the peaks in the 200-900 Hz region, and I think it did a bit, although not as much as I'd hoped.  It did help the high frequency noise "tail" though.

History:  Some time ago, Sheila and I moved the POP_A offsets to improve the recycling gain from 25ish, and that worked very well and has been very consistent.  At the time, we stopped where we did because the POP_X centering PZT was hitting its rail, not because we thought we had found the best possible location.  Then, we elected to move on to trying to get to low noise rather than continuing to chase alignment offsets.  Now that we're at low noise though, I wanted to see if continuing to move the QPD offsets would help get rid of some of this jitter / frequency noise coupling. 

What I did:

• Moved POP_A offsets to maximize recycling gain.  Watched to ensure POP_X PZT wasn't railing.  Moved SRM alignment sliders to keep POP90 low.
• Upon relock, let the DRMI ASC move the IFO (including SRM) closer to the place that was good for the new POP_A offsets (i.e., just leave the offsets in permanently).  This was good, and meant that SRM didn't need to be moved by hand later.  I think it does make the CHARD engaging do a little more work, although that can be ameliorated by resetting all of our initial alignment setpoints.
• Upon relock, also noticed that when the SOFT loops came on, we went through a max for the power recycling gain then came down a bit, all before powering up.  This indicated that I needed to move some SOFT offsets.
• Moved the SOFT offsets, while moving SRM to keep POP90 low.
• Went back to re-try POP_A offsets, ended up adjusting pitch a little more.
• Ran A2L for the test masses.

In the end, the power recycling gain is about 31.2, whereas it used to be about 29.5 in NomLowNoise before this work.  Also, as Jeff points out in alog 30481, it looks like this did good things for the DARM cavity pole and the optical gain. 

At about 06:49:15 UTC, I had just about the lowest value for the broad peak at about 440 Hz.  Going back to the offsets I had at the time did not reproduce that low of a peak though and the recycling gain wasn't at its maximum, so I ended up sticking with the offsets that maximized the power recycling gain.  I may go back and look at the alignment of all the optics at that time to see if there was anything drastically different. 

I also turned back on Gabriele's Jitter feedforward with the same gain of -1, and it still seems to be doing quite well.  I haven't looked at the coherence, so I don't know if this is quite as good as when he and I were tuning it earlier today, but it still made a significant improvement in the 80 Hz - 250+ Hz region.  I have this feedforward turning off in the Down state of the ISC_LOCK guardian, but it must still be turned on by hand.  Once we're satisfied that it does good consistently, we can add it to NoiseTunings.

For now, I'm leaving the POP_A offsets in place, but the SOFT offsets will be turned off upon lockloss.  I want to make sure that including them during the acquisition sequence isn't harmful to lock acquisition before accepting them permanently.  I'm hoping though that they're fine, so that the ASC will handle all the SRM moving and we don't have to do anything by hand.  To be checked tomorrow.

Offsets that I like in the end are:

• POP_A_P: 0.52
• POP_A_Y: 0.44
• DSOFT_P: 0.00
• DSOFT_Y: 0.04
• CSOFT_P: 0.06
• CSOFT_Y: -0.04

Ideas: 

• A2L for recycling cavity optics
• Re-look for SRM ASC error signal - maybe there's a bit more hope if the overall IFO alignment is better?

This is a low level sanity check and a part of the recent delay study (e.g. 29259). I have measured a transfer function between DARM2_OUT and SUS-ETMY_L3_ISCINF_IN1 using dtt while the interferometer was locked last night. The measurement agrees with 1 cycle user model delay (=61 usec). See the attached.