Evan, Peter, Lisa

Today Evan recovered the DRMI alignment after yesterday modifications of the input pointing . We could go back to about x25 times single arm build-up, and we observed again as  the other night  that we need more gain than before in the TR CARM input to make the transition from ALS COMM, and that the CARM TF is  pretty bad . A  previous measurement  showed a reasonable ugf @ 200 Hz and 30 degrees of phase margin. The same gains now cause the loop to oscillate at ~350 Hz, because of a bump in the TF which flattened the response from 200 to 300 Hz. Lowering the gain brings us too close to instability on the other side (~100 Hz), and we lose lock when trying to reduce the CARM offset.
So, unless something else is wrong, as far as we can tell we need to go through again the painful CARM loop tuning that happened a couple of weeks ago to move on. 

I have been bothered by the problems we had  the other night  during the locking sequence, while trying to lock with the same parameters that worked very well the day before. Also, there have been several instances in the past in which some tuning parameters where good one day, and totally wrong the day after. I have been saying that it has all to do with being in different "alignment" states, but the changes we were seeing seemed too big to make sense even with this typical excuse. 

However, I think it does make sense, at least for some of the lock acquisition failures. Because we use the transmission DC signals as error signals for CARM, we somehow "force" ourselves to move to some fixed build-up value, and we tune our locking parameters accordingly. However, if for example we are not properly aligned, "forcing" ourselves to some pre-defined CARM offset means that we are essentially moving higher up in the sequence than we should be, similarly to the case in which we had high loss in the arms . Attached is one example of what I mean.  

The first plot shows a successful locking sequence, while the second one shows a failed one. In the first plot, for a power build up of X24 single arm, REFL_DC is still 20 and REFL_9I is ~1800. In the second plot, for a similar power buildup, REFL_DC had already started to decrease (17) and REFL 9I is much higher than before, ~ 2500. That would correspond to a much higher build-up (~100) in our "baseline" working sequence, and consequently a different locking tuning (gains, filters, etc). So, it is not surprise that we break lock.

So, the current plan of improving the initial alignment and closing WFS along the sequence is indeed a good one, and most likely all we need.

no restarts reported. Conlog frequently changing channels report attached.

Lisa, Peter, Sheila

Today we tried to investigate the difficulties with locking last night.  We noticed that the spot position on IM4 had moved durring the low pass filter installation yesterday (alog 16719).  We also noticed that the MC WFS were off all day yesterday (IMC-WFS_GAIN was 0 most off the day).   We moved the PZT offset with the MC locked and the WFS on to bring the position on IM4 Trans back to what it was before the low pass swap.  

We then redid the intial alingment, but haven't gotten a good DRMI alingment before running out of time.  (DRMI has locked with the arms off resonance, but the laingment wasn't good.)

We've had another rotation stage failure.  

We noticed the same behavior on thusday night.  We can type in a requested power, the code calculates an angle, but hiting the go to power button doesn't cuase anything to happen.  On thursday as well as the first time that this happened today (around 23:00:12 UTC on Feb 14th) typing the calculated angle into the requested angle box and going there worked.  Patrick looked at trends from thursday night, and it looked like I had not changed the requested power before hitting go to power, but today I am sure that I did, and we have the same behavoir.  

After that happened, I tried to adjust the power using the requested angle, which then stopped working, so I couldn't move the rotation stage without first going to home.  Now that I've gone to home both buttons seem to be working. 

model restarts logged for Fri 13/Feb/2015
2015_02_13 02:28 h1fw1
2015_02_13 09:32 h1fw1
2015_02_13 21:45 h1fw0

all unexpected restarts.

Evan, Kiwamu, Dan, Peter, Lisa

The goal for tonight was to relock the interferometer on DC readout, properly calibrate the spectrum (see  Kiwamu's entry  with today measurements), and in the process see if the  IO periscope PZT filters  had improved the beam jitter noise that we think is responsible for the bump around 100 Hz in the sensitivity.

The only other (known) change with respect to yesterday was the  new DARM filter  designed to give us better phase margin, with the goal (surprise!) of making the locking sequence more robust.

However, we soon realized that we had to retune the transition from ALS COMM to CARM (common mode board input 1 changed from -16 dB to -7 dB) as the current gains were totally wrong (meaning that corresponding OLTF were extremely marginal). We fixed the gains up to the transition to DARM on AS 45 Q, to discover that there was no chance of moving forward with the CARM TF we measured (Evan will post the offending TF later).  

Even if we believe the new DARM filters had little to do with the problem we were seeing, we decided to go back to the old DARM filters and the Guardian CODE which was running last night, as a sanity check. 

Still, we had to keep the low CARM gain to transition to transmission signals for CARM, then we made a couple of attempts to full lock, but we would lose lock before transitioning to AS 45 Q, this time because the DARM gain was wrong..

So, the message is that the problems we see are really easy to identify (loop oscillation, nonsense TFs) and we could keep fixing the locking sequence, but we don't want to make very radical changes in the gain/filter tuning at this time. Old DARM filters and Guardian code are running (as last night). 

This is a brief and preliminary update of the calibration activity from today. I calibrated the ITMX and ETMX reponses using the usual free-swing Michelson fringe.

If I believe the measurement, I must have underestimated the ESD response by a factor of 5.3 (!?) in the previous calibration which is hard to believe for me. I would like to repeat the measuerment perhaps with different conditions (e.g. opelv on/off, L2P off, linearization off/on, different bias, different frequencies and etc) and on ETMY as well.

(MICH free swing)

The method is the same as what Joe described in LLO alog 14135. To obtain the ASQ_pkpk value, I did not run the fancy matlab code or anything, but I just picked up a highest peak value and lowest one in H1:LSC-MICH_IN1_DQ. The alignment was adjusted beforehand by locking MICH. The pk-pk value was measured to be 27.0 counts. Using the relation, d (ASQ)/d(ITMX) = 2 * pi * ASQ_pkpk / lambda, I get

ASQ optical gain = 1.59 x 10⁸ cnts/m

The input power to IMC was at 9.6 W, measured at the periscope bottom PD. ASAIR_ALF could get to 4550 counts at maximum and ASAIR_B_LF 1500 counts when MICH was freely swinging. The below are some dtails:

• ASAIR_A_RF45 whitening gain was set to 30 dB
• Whitening stage 1 and 2 were engaged (probably this was too much as it seemed already limited by some elecronics noise).
• I attach the time series.

(ITMX L2 stage calibration using MICH)

After locking MICH, I shook ITMX L2 stage at H1:LSC-SUS_ITMX_L2_LOCK_L_EXC with a drive level as high as possible without DAC saturation. I did a swept sine measurement to check how high frequency I would be able to get without loosing good signal-to-noise ratio. It seems that exctiation above 20 Hz is hopeless -- the drive signal dives into sensor noise. From this measurement I picked up one frequency point, 13.05 Hz where the ITM response was measured to be

ITMX L2 response = 8.41 x 10^-18 m/cnts @ 13.05 Hz

• L2 coil driver state was set to 2 to get the maximum range.
• No oplev damping loops were engaged.
• MICH loop correction was applied in the estimation of the L2 stage response as was done in Joe's alog 14135. Abs(1/(1+G)) = 0.817 @ 13.05, measured.
• I did not oberve a DAC saturation during the measurement.
• I attach a dtt file of the open loop measurement and ITMX response measurement.

(ETMX calibration using X arm)

Keeping the 9.6 W incident power, I locked the IR laser to the X arm with a UGF of 100-ish Hz. I did a swept sine measurement on ITMX and ETMX at different times but in the same lock strech. On ITMX, the L2 stage was driven again with the same setting as that of the MICH locking. On ETMX, I had set up the suspension filters such that they are the same as the full locking condition (e.g. drive signal goes not only ESD but also L1 stage and so on). Neverthelss, since my swept sine measurement does not go below 10 Hz, the ETMX response essentially represents the ESD response (with a small effect from the L2 stage which is almost two orders of magnitude smaller than the ESD in terms of displacement).

Taking the ratio between the two actuators, I confirmed that the ratio goes as f^2 as expected in a frequency range from 10 to 60 Hz. The ETMX/ITMX ratio was measured to be

ETMX_L3 / ITMX_L2 = 1.70 x 10² @ 13.05 Hz

ETMX_L3 response = 1.43 x 10^-15 m/cnts @ 13.05 Hz. This is almost 5.3 times stronger than what we have in the CAL-CS calibration.

• IM4 and PR2 were controlled by the REFL WFSs in order to maitain the high build up in the X arm all the time during the measurement.
• ETMX oplev pitch damping loops was on at the L2 stage with a gain of -9500.
• The swept sine excitation was injected at SUS_ETMX_L3_LOCK_L.
• The L2P and L2Y were kept engaged on L3 stage on ETMX.
• The L1 coil driver state was set to 2.
• The bias voltage was set to 9.5 V.
• The ESD liearization was on with a force coefficient of -1.8e5.
• I did not see a DAC saturation on ETMX.
• I attach the dtt files.

I set up the BS bounce mode damping loop, tested it and confirmed that it works very fine. One should be now able to lower the peak height of the mode by a factor of 10 or so in a minute.

During the calibration activities in this evening, I ran into a problem with the MICH locking where I was not able to lock it. It turned out that the bounce mode was too high so that the length control saturated the DACs on the BS suspension. I guess that the mode was excited due to several times of locking attempts as I was trying unusual locking settings for the calibration. The below is a screen shot of the bounce mode damping filter:

The design process was essentially the same as similar to what Jeff described for LLO's violin mode damping loos (LLO alog 15000) but I did not try to drive it as hard as DAC saturation. I am leaving the filter enabled for now. If someone find this filter doing something bad, please feel free to switch off at its output (because turning off at the input seems to create a huge impulse response). I did not explore the gain so much.

Replaced hard coded optic and site strings with macros, and fixed sme errors in the on/off indicator, both of which were introduced when copying the DARM_ERR damping stuff from L1:

/opt/rtcds/userapps/release/sus/common/medm/quad/SUS_CUST_QUAD_M0_DAMP_MODE.adl
/opt/rtcds/userapps/release/sus/common/medm/quad/SUS_CUST_L2_DAMP_MODE_FILTERS.adl
/opt/rtcds/userapps/release/sus/common/medm/quad/SUS_CUST_QUAD_OVERVIEW.adl

Attached is a plot of the residual (in-lock) DARM spectrum, calibrated in meters according to the calibration factor that is used in OAF: 2.86e-7 m/cnt for DARM_IN1. The residual is about 7e-14 m-rms. The residual is dominated by the 0.45 Hz quad pendulum mode. We should suppress this peak further. There is a resonant gain filter in DARM, but it is a little too low (0.43 Hz vs 0.45), and just is not enough. However, there is also a dip in the SUScomp filter at 0.45 Hz, which reduces the gain at that frequency by 10-12 dB. This compensates the damped pendulum response -- but what is the point of including here? It's not like we're going to try and have a ~1 Hz bandwidth for this loop. Seems like the thing to do is to remove this feature (some complex pole and zero pairs around 0.4-0.6 Hz) from the SUScomp filter, which will increase the gain at 0.45 Hz by 10-12 dB.

Next would be further squashing the noise from 1-5 Hz. The L1 DARM boosts have more gain in this band, so we could adopt that design. First we should make sure we are not injecting noise at these frequencies from residual OpLev feedback, or quad local (BOSEM) damping.

08:01 I changed the state bit from Undisturbed to Commissioning. However there is an entry on the card reader for Peter K. entering the LVEA at 6:50 AM. Karen and Cris also entered the LVEA after Peter and before 08:01.

08:04 Corey to the squeezer bay for 3IFO work
08:13 Alastair to the LVEA to check on the TCSY laser
08:49 Thomas and Elli to end X and end Y to reboot PCAL computers
09:34 Thomas and Elli back
09:34 Jim W. working on HAM4 in the control room
09:38 Corey done
09:43 Filiberto and Manny back from end X, were soldering cables
09:45 Filiberto to end X
10:08 Filiberto back
~11:18 Richard, Peter and Kiwamu working on installation of low pass filter for IOO periscope piezo controller
11:53 Richard back
12:42 Betsy running transfer functions on ETMX, optical lever damping turned off
12:49 Travis starting measurement on SRM
13:16 Karen opening outside rollup door in OSB shipping/receiving
13:16 Elli to LVEA to refocus ITM cameras
13:21 Kiwamu to IOT2L table to check if ghost beam is on photodiodes
13:32 Kiwamu back
15:49 Nutsinee to end X to change PCAL camera autofocus from manual to auto

The phase and gain margin in the DARM loop in the final configuration is quite small. In the interest of improving it, I started by comparing what is used on DARM for H1 and L1. The attached plots show the comparison; the differences are:

• The H1 DARM high frequency roll-off starts much lower than on L1 (roughly 80 Hz vs 600 Hz), and this explains the lack of phase margin on H1 compared to L1. At 30 Hz, e.g., there is about 25 deg more phase lag in H1 than L1.
• L1 DARM has a bandstop filter for the violin mode fundamental (500 Hz), but H1 does not. H1 has a notch at 60 Hz, which is not needed in the final configuration.
• The low-frequency boosts are a little different. The L1 boost starts earlier (higher freq), and has significantly more gain than H1 in the 2-5 Hz band. L1 also has more gain below 0.2 Hz. H1 has a resonant gain at 1 Hz that L1 does not have.
• The overall gain for H1 is about 60 dB higher than for L1 (which must be compensated somewhere else than in the filters)

Attached plots:

• L1H1DARMcompare: All filter modules used in the final DARM configuration. The L1 magnitude has been increased by 60 dB for easier comparison.
• L1H1DARMsuscompare: The subset of filters that are used to compensate the suspension response.
• L1H1DARMboostcompare: Comparison of boost and resonant gain fitlers.

I have modified the H1 DARM suspension compensation filter so that it is now just like the L1 version. Previously this filter (FM8) was called 'LLO', but I changed the name to 'SUScomp'. I also changed the 60 Hz notch (in FM7) to be a bandstop for the violin modes. This is not exactly the same filter as in L1 since the H1 violin mode frequencies are a bit lower (though more narrowly grouped) than L1. This filter is now called 'Violin', and is defined to be: ellip("BandStop",4,1,80,480,530)

Of course the new FM8 filter has more gain than the previous one above ~50 Hz. This could be a problem in terms of ESD range while using ALS-DIFF, so there may need to be some tailoring of other low-pass filters that are used in that state of the acquisition sequence.

I haven't made any changes to the boost/resonant gain filters yet. The spectrum of the residual (in-lock) DARM should be looked at to see if these should be modified.

Just to minimize confusion:

- we know that the current calibrated spectra in  H1:CAL-DELTAL_EXTERNAL_DQ  is wrong, so don't look at this channel yet. There is an overall factor that must be wrong by at least a factor of 2, as the high frequency noise is too good given the current circulating power in the arms;

- also, as explained  here , we know that the floor of the high frequency part of the spectrum is shot noise, but there was something wrong about the power scaling and/or scaling with DARM offset in the high power data from last night, and that's why the calibrated spectrum was not scaling as expected.

(More) Accurate calibration measurements are in progress.

The piezo low-pass filter box (D1500001) S/N S1500001 was installed.

Measured voltages at the 15-pin analog interface
pin     before   after           signal
1        0.110    0.016          V-MON-X
2        0.206    0.143          V-MON-Y
3        0.109    0.160          V-MON-1
4        0.206    0.143          V-MON-2
5        0.000    0.000          V-MON-3
6        0        4.664          Servo-1 OFF/ON
7        0        4.656          Servo-2 OFF/ON
8                                GND
9        0.160    -0.476         SGS-MON-X
10       1.431    1.173          SGS-MON-Y
11       0.163    -0.472         SGS-MON-1
12       1.430    1.173          SGS-MON-2
13       -13.15   -13.12         SGS-MON-3
14       4.824    4.821          OFL1
15       4.821    4.819          OFL2

DIP switches 1 & 2 were switched from ON to OFF.

To bring the beam back, most of the adjustment was done with the control slider for pitch, which went went a
couple of hundreds on the slider to ~4000.  The yaw slider was adjusted by a few hundred.




Kiwamu, Richard, Peter

At GPS time 1107854779 [2015-02-13 09:26:03 UTC], H1 appeared to unlock but the ISC_LOCK guardian state didn't change. The attached plot shows the X-arm power and the guardian state to confirm (hopefully).

At the time of writing the guardian is still reporting 'DC_READOUT' as the current state, while there is no circulating power in the arms. The other guardian nodes (e.g. ISC_DRMI) did respond to the lockloss, however, so I presume it's just an oversight in the ISC_LOCK DC_READOUT state definition to check that the other guardians are still nominal, or that the arms are actually locked.

Dan, Jeff, Sheila

Tonight we had some more fun noise hunting.  Turning the power up to 8 Watts was easy, but didn't improve the noise so much since we are mostly not shot noise limited. 

• First, when we got to DC readout we noticed that someone did a nice job damping the ETMY violin mode, reducing the peak in the asd by about a factor of 2.  This allowed Dan to turn on one stage of whitening in the OMC DCPDs, which helped improve the noise floor.  
• We further reduced the gain of the ETMX oplev damping (by a factor of 2 for a gain of -5).  This reduced coherence around 10 Hz and helped the DARM noise floor.
• We have been running all day with the ESD linearization on, this seems to be fine. 
• We had the ISS second loop on for the first long lock of the night, it has been off ever since.  

This gave us a lower noise floor, we sat here from 5:28:10 UTC Feb 13th to 5:41:29 UTC, with the intent bit undisturbed. 

We then made a single destructive attempt to reduce the ESD bias voltage, we'll have to come back to this.  :)

On the next lock we decided to try increasing the input power.  This was suprisingly easy, we have now been sitting at 8 Watts input power since 7:03:43 UTC.  We have been adjusting the BS alignment by hand every 10-15 minutes or at 10 Watts.  We have watched the AS36Q WFS signals while we do this, and can see that they are not always correct (zero does not always give us the best AS90).  BS yaw dramatically improves the intensity noise coupling.  It seems as though the IFO would stay locked like this all night if we sat here and adjusted the BS once in a while.  We are leaving it locked and undisturbed.

In the attached plot, the green trace is earlier today, The blue trace is with 1 stage of DCPD whitening on and the ISS second loop on, which is injecting noise at around 700 Hz.  The red trace is now, with 8 Watts input power, BS YAW oplev damping gain reduced, and the ISS second loop off.  You can see that we are not really shot noise limited, as increasin the power by a factor of 2.8 only gives us a small improvement at high frequencies.  Something is clearly wrong with the calibration.