I've attached the oplev trends for PIT ITMX and ETMX over the last 12 hours, (7urad, pp)  these can be compared to the trends in T1300563

(page 11 and 16).  You can see that we had about 1/5 the fluctuations in pitch durring HIFOY that we do now.   After Hugh's work on friday the longitudnal motion is reduced (now we see about 2-3 fringes per second) to levels where wwe should have enough range to lock stably.  As was documented in the alogs 9384 and 9381, the unexpected low finesse of the ETM means that higher order modes ( caused by misalingment or mode mismatching) cause offsets in our locking signals.  As a result we are more sensitive to pitch fluctuations that we should have been, and our lock is not really stable yet.  So if SUS/SEI can work on reducing the pitch fluctuations, we should give them whatever time they need with the optics.  Also, we should simulate if our locking signals would be OK with the level of pitch fluctuations seen durring HIFO Y. 

Last, I am going into the LVEA to look for parts in the east bay and work on ISCT1. 

Kiwamu, Stefan

With the BS L2P and L2Y decoupling in place we were successfully able to hand off a PRY lock to the BS, turning off any PRM and PR2 feed-back.

We needed an output matrix element of -0.05.

The attached open loop gain plot shows a lock on PRM & PR2 in black, the original BS lock in orange, and the final BS lock after adding an additional z30:p100 lead filter.

I then noticed that the BS is still ringing up during lock-acquisition because of an asymmetric saturation in the M2 coils. Adding a low limiter (10000cts) before the f^2 filter fixed the problem, but that was not enough low frequency range once in lock.
Thus I added a z1:p20 filter in the ISCINF filter, with a 2e5cts limiter, followed by a z20:p1 filter in the M3_LOCK filter. This way I can limit high-frequencies earlier, but still have the low frequency range.

With that the PRY was acquiring using only the BS actuation (StripTool trace attached) - success!

Next on the to-do list: power-recycled Michelson.

Koji, Kiwamu, Stefan

We finished setting the BS drivealign matrix. Attached are the transfer functions of the two filters.
To test it we put in a 0.5Hz drive, and took a snapshot of a StripTool, starting without the drivealign elements, and ending with the L2P and L2Y drivealign elements engaged.

Note that we don't understand why the L2P coupling seems to fall faster than 1/f^4 above 3Hz. Since the measurement has very low coherence, this might actually be a measurement error. However, since the drive is so weak up there, this might actually not matter in practice.


The raw data is in
/ligo/home/controls/sballmer/20140115/BSTFdata
BSTF_v5measDone.xml
BSTFpitch_v5.xml
BSTFyaw_v5.xml

For reference, here are the actual filters used:

BS_M2_DRIVEALIGN_L2P:
zpk([0.0988999+i*0.419892;0.0988999-i*0.419892;0.0016662+i*0.747769;0.0016662-i*0.747769;
    0.77626+i*1.58279;0.77626-i*1.58279],[0.0545374+i*0.369777;0.0545374-i*0.369777;
    0.330458+i*2.11871;0.330458-i*2.11871;0.0573483+i*1.115;0.0573483-i*1.115;0.033991+i*0.481048;
    0.033991-i*0.481048],-0.010800065,"n")

BS_M2_DRIVEALIGN_L2Y:
zpk([0.0375136+i*1.25343;0.0375136-i*1.25343;0.185383+i*1.14346;0.185383-i*1.14346],
    [0.0211465+i*1.07707;0.0211465-i*1.07707;0.255547+i*1.35495;0.255547-i*1.35495],-0.00289650892,
    "n")

For now it relieves the IAL drive to top mass lock filter banks.

asc/h1/scripts/xarm_IALrelieve
SVN revision 6911

I had left the IAL dither loops running for the last 12h. There is still frequent mode-hopping to the 10 mode. But when It jumps on the 00 mode, the x-arm build-up in transmission seems to consistently be up to 650 cts.

The BS Y2Y coupling measurement finished and was saved in controls/sballmer/20140115/BSTFdata/BSTFyaw.xml

The PRM P2P coupling measurement finished and was saved in controls/kizumi/binary_stars/20140116_PRMdiag/PRM_P2P.xml

The PRM Y2Y coupling measurement was started. It is now at 0.7Hz and running. It was temporarily saved in controls/kizumi/binary_stars/20140116_PRMdiag/PRM_Y2Y.xml

I tuned the dither alignment for the x-arm tonight.

YAW:
 - Dithering PZT1 at 410Hz using oscillator 2. Driving ETM using control bank 2.
 - Dithering PZT2 at 470Hz using oscillator 3. Driving TMS using control bank 3.
 - Oscillator 1 and control bank 1 are currently not used.

PIT:
 - Dithering PZT1 at 380Hz using oscillator 2. Currently not driving anything.
 - Dithering PZT2 at 440Hz using oscillator 3. Driving TMS using control bank 3.
 - Dithering ETMX at 1.7Hz using oscillator 1. Driving ETM using control bank 1.


I didn't use the PZT1 pitch dither because there is a DC offset in the demodulated error signal. Note that PZT1 is input pointing position (PZT2 is angle).
I was able to zero the dither offset byputting offsets into the green QPDs:
H1:ALS-X_QPD_A_PIT_OFFSET = -0.464
H1:ALS-X_QPD_B_PIT_OFFSET = -0.084
Those offsets were chosen such that they only lead to a different PZT1 drive, leaving PZT2 untouched.

Attached is a snapshot of all relevant screens.

Still to do:
 - Relieve script or servo (or move integrators to the suspensions).
 - Automatic turn-off of the servo if the arm power drops below a threshold.
 - Actuation node dither for ITM centring. (I would prefer less mode-hopping for this.)

Sheila, Stefan

The low-finesse in the green cavity poses a problem for stabilizing the x-arm length:
 - The PDH error signals of 00-mode and 01-mode overlap.
 - Alignment fluctuations lead (to the first non-trivial order) to a fluctuation in 01-mode power. I.e. the relative contribution of the 10-mode-PDH-signal to the main PDH-signal varies.
 - If the 10-mode-PDH-signal at the main locking point is non-zero, this will couple into frequency fluctuations (relative between arm length and green light). These can be a significant fraction of the line width (about 1.5kHz ?).
 - To avoid this we need the 10-mode-PDH-signal to have a co-located zero-crossing, leading to only gain modulation, but no frequency fluctuations.

 - I can think of two strategies to achieve that:
   1) Set the modulation frequency to exactly 1 transverse mode spacing off the resonance. This lines up the zero-crossings of the PDH signals.
      Setting the demodulation phase exactly will guarantee that the large q-signal does not couple.
      Disadvantages: - The slope of the 10-mode-PDH-signal is large, leading to large gain fluctuations, and requiring an exact knowledge of the transverse mode spacing. (Problem with thermal heating later on?)
                     - The q-signal is maximal - setting the demod phase accurately is crucial.
   2) Pick the modulation frequency at some different value, but set the demod phase such that on the locking point the contribution of the 10-mode-PDH-signal is zero. This will need some more modeling.

For now we chose solution 1): We set the modulation frequency to 24.424281MHz (5099Hz below the co-resonant point of 24.429380MHz). 5099Hz is our current best guess of the transverse mode spaceing - we will need to measure that accurately. Using our phasing table from alog 9386, we set the delay phase shifter to 18.8125nsec (switches up: 1/16, 1/4, 1/2, 2, 16; down: 1/8, 1, 2, 4). As expected, when locked on the 00-mode, we now get a big q-phase offset.

Sheila, Stefan

The low-finesse cavity made it non-trivial to set the demod phase correctly. Today we used the trick that any Q-signal disappears when the side-band is exactly anti-resonant. We then made an interpolation table to set the phase accurately within about 0.5nsec (~5deg) at any frequency.

1) First we identified the frequencies at which side-band and carrier are co-resonant (no demod signal):
   The 621st co-resonance we found experimentally at 23'303'644Hz, the 681st co-resonance at 25'555'116Hz. Fitting this to line up we get an FSR of about 37525.93Hz +-0.1Hz, or an arm length of 3994.47m +-1cm. Compare this to Daniel's alog: 3994.485m. (And I swear I did the error analysis before seeing Daniel's number.)

2) Next we hit the frequencies where the side-band is exactly anti-resonant, and phased the signal there. The delay numbers correspond to the phase shifter setting, assuming the delay is in with the switch up (i.e. 16.5nsec = 16sec and 1/2nsec switches up, rest down).
   23.922799MHz    31.5000nsec
   24.147946MHz    26.5000nsec
   24.260520MHz    23.3125nsec
   24.373093MHz    20.0000nsec
   24.448142MHz    18.1875nsec
   24.560716MHz    15.5000nsec
   24.785863MHz    10.5000nsec

Doc D0901920 lists the cavity length as 3994.485 m (as built, anyone?). The attached Mathematica snipped calculates the axial location of the sidebands.

Written by Yuta, posted by Koji, while he is waiting for renewal of his ligo.org account:

In the entry alog #9381, Sheila explained how the PDH signal is distorted by the broadened resonance
of the higher-order modes due to low finesse of the cavity.

Here in this entry I explain how the shape of the PDH signal can be modified by changing the sideband frequency.

[Motivation]
ETMX transission for green was larger than designed(designed:5% -> measured:36%) and cavity length lock does not stay long.
The PDH signal looks strange(see alog #9381). To explain the situation and see how we can improve the PDH signal,
we calculated PDH signal including HOMs.

[Method]
1. Calculate PDH signal including the effect of carrier HOMs and sideband HOMs.
2. Change sideband frequency to see how PDH signal changes.

The parameters I used are the same as the ones listed in alog #9381.

Calculated TMS is 5.076kHz and the sideband frequency before we have changed last night was 24.407079MHz.
HOM content and vertical scale are arbitary in the following plots.

[Result]
1. HOMPDH_sb24_4MHz.png and HOMPDHIQ_sb24_4MHz.png show the calculated transmission, PDH signals,
and XY plot of I-phase and Q-phase PDH signals, when the sideband frequency was set to the original value. This
IQ plot is very similar to what we have seen(see video in alog #9381) and agrees well with Sheila's calculation.

2. HOMPDH_sbonres+0_5TMS.png and HOMPDHIQ_sbonres+0_5TMS.png are the plots when one of the sideband
is at the middle of the TEM00 and TEM01/10 resonances. PDH signal gives zero crossing at TEM00 resonances,
but it also gives zero crossing at other HOMs. So, mode hopping rate is expected to be high.

3. HOMPDH_sbonres+TMS.png and HOMPDHIQ_sbonres+TMS.png are the plots when one of the sideband is
at the TEM10/01 resonance. PDH signal does not give zero crossing at TEM00 resonances, but if the correct offset
is given, mode hopping rate should be low. The IQ plot will be somewhat simple "8" shaped plot in this case.

See alog #9379 for what we have done using these results.

Stefan, Koji, Kiwamu,

We newly installed a L2P correction filter on BS which reduces the coupling to the bottom stage pitching from the penultimate length actuator drive.

Stefan completed the L to P and P to P measurements on BS yesterday. Then Koji fitted the data using LISO and I punched in the fitted poles and zeros of the ratio L2P/P2P in foton and loaded the filter in M2_DRIVEALIGN_L2P. We did a quick test in which we had a big excitation in the M2 length drive and monitored the pitch motion by the oplev with and without the L2P correction. It looked successful although we did this test at only two frequencies, namely 1.1 and 0.8 Hz, because a Y to Y measurement was running in background which we didn't want to screw up at that moment. The coupling looked successfully reduced at least by a factor of 10 at both frequencies as far as we can observe it on a StripTool screen. Note that in order to get the cancelation right, the gain of the L2P  correction chain needs to be -1. If one wants to further test the performance, he or she should run a swept since measurement extensively.

* * *  the L2P correction filter * * *

- - - - freq [Hz]   Q - - - - - -
pole 0.3737768342347  3.4267960418
pole 0.4822474954282  7.0937407121
pole 1.1164786260     9.7341833818
pole 2.1443263848     3.2444753770

zero 0.4313822619174  2.1809027959
zero 0.7477704984560 224.3944177073
zero 1.7628959088     1.1355062310

gain factor -0.0108000650535

With the difficulty we had been having locking the green arm stably, there was some speculation about what the difference in the ETM transmission from the design of 5% would mean for our locking.  Here are a few plots of pdh signals and resonance profiles for the cavity we have. 

First, the IQ scatter plot we see doesn't look anything like a PDH signal I am used to.  There was some speculation that this is caused by Doppler fringing.  Using the equation 20 from P000017 the critical velocity for doppler effects with our etm is 200um/second, (about 800 fringes per second!) so we were never in this regime.  (if the transmission of the ETM was 5%, this velocity would have been 3um second, which would have been reasonable with the motion that we had on thursday and before)  Since Hugh's changes to the ISIs yesterday we are seeing about 3 fringes/ second when it is fast, (more often 1 or 2 fringes) so our cavity length is changing at about 0.8um/second at most.

Here  is an IQ plot of the PDH signals (units not meaningful) for an ETM transmission of 1% (normal looking), 5% (lobes from sidebands are becoming large) and 38% (our situation).  

Even though that looks strange, it still seems like an OK signal for locking to if the demod phase is tuned correctly: Plot  (In this plot I have used a different demod phase for the low finesse cavity than the one showing the design)

We also have been wondering how we are going to measure the modulation depth, since we do not have an OSA that works for green light.  At some point we tried to lock on a sideband to have a look at the transmitted power relative to the 00 mode.  The power we saw would have suggested a ridiculously high modulation depth.  I believe that we might have been fooled by the camera image and were really locked onto a first order mode.

Keita suggested that I check if we really lock to the side band resonance when by flipping the sign of our servo.  We would if the demod phase is tuned correctly and we didn't have higher order modes. Plot  (This plot shows the shape of the resonance profiles for the carrier and each sideband against the PDH signal, the vertical scale is not meaningfull.  The lower panel shows what happens with the wrong demod phase).   I made the same plot for a higher finesse (Tetm=1%) where the sidebands are resonant when the cavity is locked with the sign flipped, even for the wrong demod phase.

We also wondered if our low finesse would mean that higher order modes would be resonant when the cavity is locked.  Here is a plot of resonance profiles for the first 7 higher order modes.  The top panel shows Tetm=5%, the bottom panel Tetm=38%.  The only of these modes that is resonant with the carrier is the lower sideband of the 5th order modes, which we should be able to keep small once we have a good alingment.  

The last few plots are what higher order modes do to our locking signals. Here is an IQ scatter plot and the locking signal with resonance profiles of the carrier and sidebands when 10% of the power is in a second order mode.   With a mode mismatch like this we still get reasonable locking signal.  

I had a look at some of our cavity transmission peaks , and made a rough approximation of the higher order mode content.  Here are plots of the PDH signals (with my approximation of the transmission profile in the bottom panel) and an IQ plot.   These are both with the demod phase tuned for the 00 mode with our finesse (by using the plot without higher order modes).  

Daneil and I took a video when we were at the X end last night of our actual IQ scatter plot, which looks about as crazy as the prediction above. 

The next two plots are what the IQ plot and PDH signals would look like with my approximation of our higher order mode content and the intended ETM reflectitvy of 5%.  The higher order mode content doesn't totally confuse the locking signal the way that it does with the lower finesse.  

I think that Yuta has some similar plots, and maybe has tried to actually match some of the PDH signals that we see.  

In all of these plots I used: 

Since the fringe velocity is so much less than it has been ( 1-3 per second) after Hugh's changes yesterday, I decided to see if I could adjust the demod phase by looking at the open loop gain.  I went to the end station from about 1:45 to 2:11.  This wasn't verry sucsessfull since the cavity wasn't well alinged and I couldn't see any cameras to know what mode it was locked to.  I'm back at the corner now.

I tried the PZT dither today:
- The green arm transmitted signal is now available in the ASC system
  (ADC card 3, channel 24-27 [25-28], are connected to output 1-4 form the the ALS PD concentrator 3) 
- I used the demodulated green arm transmitted signal for all the loops I tried today.
- Dithering PZT2 provided a nice error signal for TMS pointing.
- Dithering PZT1 gave an error signal that I couldn't zero out - not sure why yet. (TMS position offset?)
- Dithering ITM at 1.7 Hz provided a good loop for aligning ITMY.
- The loop did a good job aligning the cavity, but mode-hopping was still hampering the effort.


Still to do:
- Try higher frequency ITM dither (lower stage drive?)
- Investigate PZT1 dither error signal offset - Try Green QPD pointing offset?
- Do yaw loops.

Alexa, Stefan, Sheila and Kiwamu

Yesterday, we set up an optical spectrum analyzer on the PSL table to measure the characteristic of the EOM. The results look good except for the 24 MHz resonance . The resonance for the 24 MHz sideband is apart from the modulation frequency by 250 kHz or so. However the IMC doesn't need a large modulation depth and the IMC locking has been OK, we are good. Also, I took some impedance data and will put those data together into a DCC document for a record purpose.

The plots below are the measured modulation depth when driven hard as a function of the modulation frequencies. The vertical dashed-lines represent our actual modulation frequencies i.e. f1 = 9099471 Hz , f_mc = 24078360 Hz and f5 = 5 x f1.