This morning I locked the IFO up to ENGAGE_SRC_ASC (so the soft degrees of freedom were still uncontrolled).

I then measured the response of the transmission QPDs to CSOFT, DSOFT and TMS motions. Results are below

Pitch response

Yaw response

The inverse matrixes, which should provide us the signals for the soft loops that are decoupled from the TMS motions are below:

Pitch inverse matrix

Yaw inverse matrix

New sensing matrix

With the measurement above, and rescaling to maintain similar values, the sensing matrix should be

This new matrix has been uploaded to the guardian code (in PREP_ASC_FOR_FULL_IFO)

X arm ALS lock glitchy

The X arm green lock looks much more glitchy than usual. This is an intermittent problem, it seems to fade away after a while the cavity is locked.

ALS DIFF find IR failing

This morning the ALS DIFF find IR state failed a few time, without being able to find the Y arm resonance. I had to tune the DIFF offset manually.

Soft and SRC angular loops

Yesterday we noticed that when the soft loop were closed, the SRC1 and SRC2 yaw loops were kicked out of zero, and took a long time going back. Therefore I locked the IFO up to ENGAGE_SRC_ASC, and did a couple of tests:

• I slowly moved the soft degrees of freedom to zero the error signals. The IFO stayed locked and I saw no oscillation in any of the ASC loops.
• Then I characterized the SRC1 and SRC2 yaw loops, by adding offsets to the input and looking at their impulse response. They were extremely slow. I could increase the gains from 1 to 4 (SRC1_Y) and from 20 to 100 (SRC2_YAW), getting a reasonable response and no sign of instability
• I did the same test with the SRC1 and SRC2 pitch loops, although we did not see any large excursion with them when closing the soft loops. I could increase their gains to 4 and 100 as in yaw, so I put those values in the guardian as well for symmetry

The new gain settings for SRC1 and SRC2 yaw loops are in the guardian.

I also measured the response of the transmission QPD signals to the soft degrees of freedom and the TMS, and computed a new sensing matrix, which is now in the guardian. See 43749 for details.

Full IFO alignment is inconsistent with initial alignment

We have now seen at least twice that if we adjust the soft degrees of freedom when locked in full IFO, and optimize for the best recycling gain, we end up with an alignment which is not good when we lost lock. Every time we lost lock after moving the soft degrees of freedom the IFO ends up in a alignment configuration that does not work for DRMI lock.

Attached graph shows pressure "blip" resulting from having vented the turbo-side of the closed 1 1/2" AMV pump port valve.

I forgot to post at the time, but I updated the HAM4 loops on the 14th. Main change was backing off the boost gains some to get more phase margin at the new 54-ish hz feature, especially in the rx/ry dofs. I'll just claim I was busy with the HAM1 loops. I also reverted the guardian change I made in alog 43205, so the high bandwidth loops are now engaged automatically by the chamber guardian again. This should satisfy FRS ticket 11177.

BS ST1 H1 and H2 spectra appear slightly elevated but probably not to the level of concern.  All others appear normal.

HAM3 H2 CPS spectra appears elevated in the high frequency region.  All others appear normal.

The attached plot shows the pre-modecleaner transmission and the out-of-loop photodiode (PDB) of the first loop ISS.
That flat periods are when the second loop output switch is off.  The first wiggle period is with the second loop
output switch on and the second loop 10 Hz pole engaged.  The second wiggle period is with the same but with the
10 Hz pole off.  These were taken with the second loop gain slider at 20 dB.

    At gains less than 20 dB, the first loop ISS pretty much rails straight away from which it seems to require
human intervention to bring it back.

    The symptoms of when the first loop rails, is that in the diffracted power plot on the MEDM screen the minimum
diffracted power is close to 0%.  Or there is a large separation between the minimum and maximum plots.  Typically the
diffracted power displayed in the numeric field is ~3.3%, and the AOM voltage displayed is 0.233 V.  This would suggest
that the control signal to the AOM driver is either oscillating faster than EPICS can monitor or that the AOM driver
input is saturated.  To remedy this, slide the REFSIGNAL slider all the way left, which then results in the AOM voltage
coming away from the 0.233 V.  The diffracted power will probably read something >= 11%.  Then lower the REFSIGNAL to
something like -1.90.  The diffracted power displayed should start coming down to something between 4 and 5%.  Then
slowly decrease REFSIGNAL to something like -2.  Ideally the REFSIGNAL should be adjusted so that the out-of-loop DC
voltage displays 10 V.

    At this point, it is not obvious to me if the saturation problem lies with the servo card or the AOM driver.

Sheila, Craig, Georgia

CHARD Y

We have increased the gain of CHARD yaw by 30dB.  At this point in O2 we were increasing the gain by 47dB, but that would be giving us too high a gain currently, so I reduced the filter gain.  Attached is a measurement taken with the CHARD Y gain at 10 (without the 30dB filter on), and no boost. From the measurement it seems as though we should be able to increase the gain of CHARD Y by about 20 dB, and engage the boost. This doesn't work. 

MICH Y

As Jenne's measurement from earlier today showed (43729) the MICH Y loop phase crosses instability at around 1.5 Hz.  When we tried to engage the boost for CHARD Y we lost lock due to a 1.5 Hz instability, which could have been from CHARD Y, but actually looking at the lockloss the instability shows up much more clearly in the MICH control signal.  We manually turned off the MICH Y low pass and things seemed a bit more forgiving (even though we still can't engage the CHARD boost and keep things nice and stable, we didn't loose lock by trying once the MICH low pass was off). 

CSOFT P

The stability of this loop around 0.5 Hz depends on both the optical lever damping and the HARD loop shapes.  Since we turned off the optical lever damping a few weeks ago it isn't surprising that this loop goes unstable and causes locklosses now.  The only need gain at around 0.5 Hz so that we can suppress the radiation pressure instability due to the suspension cross coupling, I cahnged the filters around and edited the guardian so that this loop should come on with a configuration more similar to DSOFTP, the only difference being 20dB lower gain.  We haven't tried this yet but in the morning in order to give this loop a gain equal to DSOFT pitch just increase the gain by a factor of 10 and use the filters that are in the guardian (They match DSFOT). 

Earlier today Gabriele looked at the TMS QPD combination that we are using, which for TMS Y is different for DSOFT and DSOFT.  We should fix this, but haven't been able to lock long enough to engage soft loops since then. 

POP_QPD before SRC ASC

We changed the order of these states since it seems to make the ASC engaging go more smoothly. 

CHARD P 

earlier today we were able to increase the CHARD P gain from 0.3 to 1, but not 1.5

Other locking things today

We have had a few intermittent problems with ALS today, with some glitches that have come and gone.  We have been seeing for a while that ALS DIFF noise is much higher when the COMM PLL is locked, so we edited the guardian to unlock the COMM PLL, lock DIFF, then lock the PLL before locking COMM.  We should probably try to fix the cross talk. We also had a time tonight and earlier today when the X arm PDH would not lock.  Craig and I went to EX to try to investigate, but the problem was gone by the time we got there, and there is no sign of the laser mode hopping. 

Earlier in the day we got rid of the CARM boost, after experiencing many  locklosses a minute or so after transitioning to ANALOG_CARM.  Without the boost things have improved.  It might be that with the extra acoustic noise around the input arm we can't use the boost. 

We've had many locklosses at the TR_CARM and RESONANCE states today.  

Started ring heater test for TCS

Ran the script that Georgia requested at 5:41:30 UTC. 

python /opt/rtcds/userapps/release/tcs/common/scripts/ring_heater_schedule.py ETMY -s now -d 3.0 -p 0.5

Since we optimized the crystal position (alog43647) thought it'd be a good idea to double check the threshold power.

Sadly, we are still at 20 mW. The input pump power has mode mismatched taken into account (70% of off-resonance green refl goes into the OPO).

Haocun, Nutsinee

-1.3 dB squeezing observed for the first time today. Not everything was optimized. Tiny but a good start!

Details

We went in today hoping to just test run CLF with lower power and new LO power we decided on making sure 3MHz doens't saturate. So our fringe visibility wasn't at its optimal (85% good enough for what we planned to do). With 10 uW of CLF we can push LO to 1mW without saturating the homodyne. This already cleared the dark noise by 20 dB. 

(shotnoise with just 1mW LO compared to dark noise, homodyne balanced)

The signal looked decent enough. LO locked (almost) without issues. Currently we are limited but ~34kHz PZT resonance. Maybe we should replace the 6kHz notch slow option on the common mode board with a 34kHz notch filter? LLO had one (not on the common mode board I was told) LLOalog39503 

(Here's the transfer function. Plenty of phase margin at UGF.)

Now that we closed the LO loop again (seemed stable this time). We thought why not try to see squeezing? 

Green input power to the fiber coupler was ~17mW. This corresponds to roughly 5mW into the OPO. We didn't optimize the temperature for the best non linear gain. We were also using one out of two phase delay to adjust the relative phase between LO and CLF at the time. 

The best we observed was 1.3 dB squeezing and ~3-4 dB of anti-squeezing. Sorry the screen shot doesn't quite reflect that. 

(Note that the roll-off starting at ~20kHz might be due to the fact that we ran the cable from SQZT6 all the way to the ISC rack where SR 785 was parking. I don't think that's real, compared to the dark noise clearance plot from above, which I brought SR785 right next to the HD)

So there're lots of room for improvement here without getting into phase noise hunting just yet. We will try to make a measurement again tomorrow more carefully/thoroughly. 

We also measured the homodyne shot noise with LO at its nominal operated power (as of now). Will post that in a separate alog. 

Note that TTFSS wasn't locked during all these measurements (pump laser not locked to PSL). With the TTFSS locked the noise hump around 2kHz became visible with no obvious improvement at high frequency. Given what we see today I don't think it's beneficial to keep going with the current TTFSS we have. I think we are ready for the modified board.

The report is at T1800386. Attached at 2' intervals over 4 feet (three attachments), it took a bit over 400lbs to rip thru the fabric at the attachments.  I suspect the fabric is also now permanently distorted.  Attached is a photo that pretty well shows the test.

TITLE: 08/29 Day Shift: 15:00-23:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Preventive Maintenance
INCOMING OPERATOR: None
SHIFT SUMMARY:  Commissioning all day long.
LOG: See attached .txt. file.

UGF = 37.5 kHz
Phase Margin = 46 degrees

This morning we optimized the soft loop alignment, with all other ASC loops closed, and updated the transmission QPD offsets

We have re-run the script that measures spot positions in the IMC.  (The input beam moved during PSL work yesterday according to Cheryl's camera's 43710)

The spot positions' we got are: 

DOF   Mean Spot Position [mm] 
MC1 P     0.21  
MC2 P     1.595   
MC3 P     0.2713  
MC1 Y     0.1672 
MC2 Y    -0.805   
MC3 Y     1.0628  

These are not very different from what Craig posted here

[Jenne, Craig, PatrickG]

We have a new plotting / calculating routine that will plot a measured DTT transfer function, that has been measured via noise injection.  As EvanH points out in alog 27518, data points with only modest coherence will be biased toward unity if you are just looking at the traditional DTT IN1/IN2 transfer function.  So, now we have a command line tool that will extract the measurement data from DTT, and calculate the open loop gain from the cross spectral densities, and save a plot.  Then, if you want to know what the loop shape would look like if you added some other filters, you can use the second tool to extract filter module info from python, apply those filters to the measured OLG, and plot the result. 

As is often the case with low gain loops (like our ASC right now), we can't push hard enough to excite with noise all of the frequencies that we are interested in.  So, I have measured the CHARD loops with several bandpassed noise injections.  The DTT extractor cannot get info about references, so each measurement must be saved in a separate .xml file.  For example, I have 3 separate .xml files for the CHARD_P measurements that I took yesterday. 

The output from the first plotter shows that CHARD_P, as it was yesterday, had very very low gain - see first attachment. Each color is the data from one of the frequency bands, for data points where the coherence between (IN1 and EXC) and (IN2 and EXC) are both greater than 0.5. 

The output from the second - see second attachment - is the measured TF from above, with the indicated filters applied (here, FM1 and FM3, which is what guardian wants to have turned on for the power up).  You can see that if we were to include both of these filters, we'd be unstable around 9 Hz. 

Things that would be nice to add: a user-chosen coherence threshold, and the possibility of a user-defined gain value, and plotting of the loop suppression, so we can see if there is any expected gain peaking in the candidate OLG.  Farther-future would even be to add error bars to the measurements, given the coherence.  Also, a more generic location for the scripts.

For use right now (in a folder that contains your DTT .xml files):

>> /ligo/home/jenne.driggers/git/dtt_tools/bin/stitchTFs.py *.xml       (In the case that you want all .xml files from this folder.  Otherwise list them explicitly, space separated)  This extracts the measurement data, saves the plot and the calculated OLG. 

>> /ligo/home/jenne.driggers/git/dtt_tools_bin/plotFilters.py filename.h5 FM# FM#    (The previous command saves the .h5 file, so pass that file name to this script.  Then list the FMs that you want to add, space separated)  (ex. filename.h5 FM1 FM3)