What vacuum activity is Hugh doing at X-end?

Replacing a Parker valve.

Hepi - not vacuum.

Also, here is a video of the ASAIR camera when MICH is in lock: https://alog.ligo-wa.caltech.edu/aLOG/uploads/13665_20140829034359_ASAIR_20140829.avi

The motion of the entire beam is due to SR2 and SR3 whose suspension and isolators are not well tuned yet.

Attached are two plots:

 - screenshot of QPD servo settings with sensing and actuation matrices

 - screenshot of OMC QPD signals showing the loop suppression.  Dashed references are with the loops open, current traces are loops closed.  We get about a factor of ten below a few Hz, we can do better!

Note: if the gain slider is set too high (more than ~0.3, with the current loop settings) then the servo actuation begins to saturation the DAC output to the OM3 coil driver.

[Dan, Nic, Koji]

The detailed description of the calculation had been missing. And we found a mistake in calculating the output matrix.
Here is the updated version of the matrices. This new setup should be tested when the IFO time is available.
In addition, we are going to update the calibration so that the servo inputs show the beam displacement and angle
in um and urad.

1. Sensing matrix

The ISC alignment input of OM1/2/3 was excited at 3.9Hz and 2.9Hz for Pitch and Yaw, respectively.
The spot motion was read out by QPDA and QPDB.
Measured sensing matrix was

| H1:OMC-ASC_QPD_A_PIT_OUT |   | T_OM1P_QAP T_OM2P_QAP T_OM3P_QAP || H1:SUS-OM1_M1_LOCK_P_IN1 |
|                          | = |                                  || H1:SUS-OM2_M1_LOCK_P_IN1 |
| H1:OMC-ASC_QPD_B_PIT_OUT |   | T_OM1P_QBP T_OM2P_QBP T_OM3P_QBP || H1:SUS-OM3_M1_LOCK_P_IN1 |
 
                               | -1.05e-3 -1.64e-3 -6.24e-4 || H1:SUS-OM1_M1_LOCK_P_IN1 |
                             = |                            || H1:SUS-OM2_M1_LOCK_P_IN1 |
                               | -6.38e-4 -1.86e-3 -9.55e-4 || H1:SUS-OM3_M1_LOCK_P_IN1 |
 
| H1:OMC-ASC_QPD_A_YAW_OUT |   | T_OM1Y_QAY T_OM2Y_QAY T_OM3Y_QAY || H1:SUS-OM1_M1_LOCK_Y_IN1 |
|                          | = |                                  || H1:SUS-OM2_M1_LOCK_Y_IN1 |
| H1:OMC-ASC_QPD_B_YAW_OUT |   | T_OM1Y_QBY T_OM2Y_QBY T_OM3Y_QBY || H1:SUS-OM3_M1_LOCK_Y_IN1 |

                               | -2.21e-3  3.05e-3 -1.35e-3 || H1:SUS-OM1_M1_LOCK_Y_IN1 |
                             = |                            || H1:SUS-OM2_M1_LOCK_Y_IN1 |
                               |  1.11e-3 -3.13e-3  1.84e-3 || H1:SUS-OM3_M1_LOCK_Y_IN1 |

Here we define the combined matrix T:

T=
| T_OM1P_QAP T_OM2P_QAP T_OM3P_QAP 0          0          0          |
| T_OM1P_QBP T_OM2P_QBP T_OM3P_QBP 0          0          0          |
| 0          0          0          T_OM1Y_QAY T_OM2Y_QAY T_OM3Y_QAY |
| 0          0          0          T_OM1Y_QBY T_OM2Y_QBY T_OM3Y_QBY |


2. Spot size ratio

Since OM3 is a flat mirror it is straight forward to use it as a scanner to infer the beam size on the QPDs.

When the QPD signals are normalized by the sum, the pitch and yaw output signals becomes proportional to
the spot displacement normalized by the spot size. i.e. Intensity distribution

I(x) = sqrt(2/pi)/w Exp[-2 (x-dx)^2/w^2]

gives us the signal

s(dx) = Int_(-Infinity)^0 I(x) dx + Int_0^(Infinity) I(x) dx
      = Erf(sqrt(2) dx / w)

ds/dx|dx=0 = sqrt(8/pi)/w

Therefore

(H1:OMC-ASC_QPD_A_PIT_OUT) =   Sqrt(8/Pi)/wA * [(2 L_QPDA) * theta_OM3_PIT(f) * (H1:SUS-OM3_M1_LOCK_P_IN1)]
(H1:OMC-ASC_QPD_B_PIT_OUT) =   Sqrt(8/Pi)/wB * [(2 L_QPDB) * theta_OM3_PIT(f) * (H1:SUS-OM3_M1_LOCK_P_IN1)]

(H1:OMC-ASC_QPD_A_YAW_OUT) =   Sqrt(8/Pi)/wA * [(2 L_QPDA) * theta_OM3_YAW(f) * (H1:SUS-OM3_M1_LOCK_Y_IN1)]
(H1:OMC-ASC_QPD_B_YAW_OUT) = - Sqrt(8/Pi)/wB * [(2 L_QPDB) * theta_OM3_YAW(f) * (H1:SUS-OM3_M1_LOCK_Y_IN1)]

Here wA and wB are the spot size at QPDA and QPDB, L_QPDA and L_QPDB are the lever length from OM3 to each QPD,
and theta_OM3_PIT(f) and theta_OM3_YAW(f) are actuator response of OM3 from the actuator count to the physical angles,
respectively.

Note that the negative sign for the fourth formula comes due to odd number of reflecting optics in the
OMC QPDB path.

Unfortunately, the absolute actuator calibration of theta_OM3_PIT(f) and theta_OM3_YAW(f) were not known.
Also, because of the mode mismatch, we don’t know actual wA and wB. Therefore we decided to compensate
the difference of the spot size between QPDA and QPDB.

Using the optical path length diagram, we obtained L_QPDA = 0.520 [m] and L_QPDB = 0.962 [m]

(wB/wA)_PIT =   (T_OM3P_QAP / L_QPDA) / (T_OM3P_QBP / L_QPDB) = -6.24e-4 / -9.55e-4 * 0.962/0.520 = 1.21
(wB/wA)_YAW = - (T_OM3Y_QAY / L_QPDA) / (T_OM3Y_QBY / L_QPDB) = -(-1.35e-3) /  1.84e-3 * 0.962/0.520 = 1.35

We took the average of these two values and used 1.3 as w_B/w_A.

3. Input matrix

We want to convert the basis of the signal from QPD basis to the beam angle/position basis with regard to the waist:

| (H1:OMC-ASC_QPD_A_PIT_OUT) |        | wB/wA   0  | | 1  L_QPDA_WAIST | | V_POS |
|                            | propto |            | |                 | |       |
| (H1:OMC-ASC_QPD_B_PIT_OUT) |        |   0     1  | | 1  L_QPDB_WAIST | | V_ANG |

| (H1:OMC-ASC_QPD_A_YAW_OUT) |        | wB/wA   0  | | 1  L_QPDA_WAIST | | H_POS |
|                            | propto |            | |                 | |       |
| (H1:OMC-ASC_QPD_B_YAW_OUT) |        |   0    -1  | | 1  L_QPDB_WAIST | | H_ANG |

Again the additional negative sign for the QPDB YAW comes from the odd number of mirrors in the OMC QPDB path.

Looking at the diagram attached to the original entry the distances from the QPDs to the waist position
are L_QPDA_WAIST = 0.0434 and L_QPDA_WAIST = 0.484. Taking the inverse matrices of the right hand side,
we obtain

| V_POS |        |  0.845 -0.0985 | | (H1:OMC-ASC_QPD_A_PIT_OUT) |
|       | propto |                | |                            |
| V_ANG |        | -1.75   2.27   | | (H1:OMC-ASC_QPD_B_PIT_OUT) |

| H_POS |         |  0.845  0.0985 | | (H1:OMC-ASC_QPD_A_YAW_OUT) |
|       | propto |                | |                            |
| H_ANG |         | -1.75  -2.27   | | (H1:OMC-ASC_QPD_B_YAW_OUT) |

i.e.
| H_POS |        |  0      0       0.845  0.0985 | | (H1:OMC-ASC_QPD_A_PIT_OUT) |
| V_POS |        |  0.845 -0.0985  0      0      | | (H1:OMC-ASC_QPD_B_PIT_OUT) |
|       | propto |                               | |                            |
| H_ANG |        |  0      0      -1.75  -2.27   | | (H1:OMC-ASC_QPD_A_YAW_OUT) |
| V_ANG |        | -1.75   2.27    0      0      | | (H1:OMC-ASC_QPD_B_YAW_OUT) |

This matrix is defined as IN.

4. Actuator selection

Which mirror combination we should use? Nic and Dan checked Gouy phase at the location of the mirrors.
That suggested that OM2 and OM3 has -70deg and +70deg with regard to the Gouy phase at OM1. Therefore
we decided to use OM1 and OM3. That means we convert 4 inputs to 6 outputs using the following ACT matrix.

      | 1 0 0 0 |
      | 0 0 0 0 |
      | 0 1 0 0 |
ACT = |         |
      | 0 0 1 0 |
      | 0 0 0 0 |
      | 0 0 0 1 |


5. Output matrix

Now we want to set the output matrix to make the roundtrip open loop matrix diagonal.

That means we want to make

IN.T.ACT.OUT = I
i.e.
OUT = Inverse(IN.T.ACT)

We want to make the actuation orthogonal in terms of the POS/ANG basis. In order to obtain the
actuation matrix, we need to take the inverse of the product of the matrix in 3. and 1.

OUT = Inverse[IN.T] =
|    0  -1019     0   410 |
|    0  - 369     0  -781 |
| -405      0   214     0 |
| -301      0  -392     0 |

The actual output matrix of the MEDM screen is a combination of ACT and OUT, that is
          |    0  -1019     0   410 |
          |    0      0     0     0 |
          |    0  - 369     0  -781 |
ACT.OUT = |                         |
          | -405      0   214     0 |
          |    0      0     0     0 |
          | -301      0  -392     0 |

6. Servo control

The servo filter did not have the slow integrator which suppresses the DC component.
An integrator below 0.1Hz was added to FM6.
 

Title: gains moved around in OMC servo

The OMC NORM output was not ~1.0, this was because the input to the normalization was less than 0.1, and the denominator has a lower saturation at that point.

I put a gain of 10 into 'H1:OMC-DCPD_NORM_FILT_GAIN' and 'H1:OMC-DCPD_NORM_GAIN'. Thus bringing the denominator above 0.1 and allowing the normalization to work. There was a gain of 1000 in 'H1:OMC-DCPD_NORM_GAIN' which I moved into 'FM8' of 'H1:OMC-LSC_SERVO' (called 60dB).

Finally, the gain change due to the normalization fix had to be corrected by putting a gain of 1 into 'H1:OMC-LSC_SERVO_GAIN'.

Old pictures.

Here are images of a mode-scan of the OMC, and spectra that show the control signal, the normalized DCPD Sum (called DCPD Norm, in units of RIN), and coherence between some interesting channels.  The noise on the DCPDs is limited by the OMC, not the intensity noise from the IFO; only a little bit of the noise on IMC_TRANS is making it to the DCPDs.  Note that the ISS is currently disabled.  The two DCPDs are coherent so we're not shot-noise limited.

I took 60-second averages of the sum of OMCR_A with the OMC locked and unlocked.  Unlocked the sum was 9316.68, locked was 1834.33.  The visibility/mode-matching into the OMC is about 80%.  (A small but nonzero fraction of this is due to the power in the sidebands, the modulation depth is 0.3.)

A text file for the mode scan can be found here.  The columns are [time, PZT_VMON, DCPD_SUM].

Note, all of this data was taken with a single bounce off ITMX., with one stage of whitening on the DCPDs.

Also I've attached a figure of the OMC open-loop gain measurement.  UGF is 90Hz.

Nice!

A few things in reply to Dan's comment:

1) I wonder why the mode scan looks so messy. Ramping the PZT over the full range should deform the cavity slightly, so we usually see a couple-percent difference in transmission from mode to mode, but the variation seems much wider here. Was the alignment not stable? Also, what's going on with those PZT readback saturations?

2) Was this RIN plot from before the NORM calibration was fixed? If not, it seems crazy high. It looks like your input beam is pretty noisy, since you see some coherence with IMC TRANS, but I guess this is somewhat expected at lower frequencies with ISS off. However, there is no way the OMC should be adding noise at that level.

3) Now would be a good time to balance the DCPDs. I believe Keita made a precise measurement of the electronic TFs which can be used for frequency-dependent correction, and then Koji should have the responsivity numbers for the diodes. Those should take care of most of the difference, and then the rest can be done with the balance slider (we needed 0.6% gain bias at LLO). The easiest way to do this is to put an intensity modulation line in and cancel it in the NULL signal.

I believe this was done with a single bounce of ITMX.
ITMY had an oplev issue at the time as you can seen in https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=13654

See the blue trace in the attached time series. The trace represents the sum of the optical lever of BS. It shows a funny drop every 70 sec.

As Jim B. pointed out, the end X Beckhoff PLC 2 has gone haywire. The new medm readback of its system time is being rapidly overwritten by random numbers. From a trend of H1:SYS-ETHERCAT_X1PLC2_CURRENTTIME_WSECOND, it appears to have started around 3:18 PDT this morning.

Dataviewer shows the X1_PLC2 time display going bad at 03:19 PDT Aug. 28.

Further information:
H1:ALS-X_LOCK_STATEREQUEST got changed to 256 at ~ 3:18.

Restarting the end X Beckhoff computer has not helped.

Upon Daniel's suggestion I did a restart of TwinCAT and then a burtrestore of PLC2 to 08:10 yesterday morning. PLC2 then crashed and I got the error in the attached screenshot. I acknowledged the error and then clicked 'Activate and run' for PLC2. This appears to have fixed it.

It appears that an automatic Windows update restarted the computer at 3:15:29 AM PDT.

The relevant events from the Windows system log are attached.

Dan, Kiwamu

We tried PRX, PRY, MICH and PRMI. PRMI did not lock, but the other configuration locked. Also I am worried that either ITMs or BS may be moving a lot.

(MICH)

• ASAIR_A_RF45_Q
  • demod phase = -91 deg
• whitening gain = 18 dB, stage 1 and 2 ON.
• servo gain -200
• we did not change the digital filters at all (i.e. the same old filters were used)
• UGF --> forgot to measure, must be higher than 10 Hz.
• The BS pitch was wobbly with an amplitude of about 2 urad in peak-to-peak, observed by oplev. Engaging the oplev damping servo successfully got rid of the pitch wobbliness.

(PRX and PRY)

• REFL_A_RF45_I
  • demod phase = -2.2 and 12 deg for PRX and PRY sideband resonance respectively
• Also checked the demod phase for REFL_A_RF9_I (out-of-loop)
  • demod phase = 77.0 and 79.7 deg for PRX and PRY sideband lock respectively
• whitening = 0 dB, no whitening stages engaged.
• servo gain = -200
• we did not change the digital filters at all (i.e. the same old filters were used)
• UGF --> 60-ish Hz

(PRMI)

• tried a smaller gain in PRCL and MICH by a factor of 500 and 20 respectively according to LLO's locking instruction (DCC-T1300991).
  • I did not see a momentary lock at all
• fiddled with the gains. No momentary lock was observed

(ITMX mysterious oscillation in pitch)

At the beginning, we saw a quite steady oscillation in pitch of ITMX at 0.5 Hz with an amplitude of 2 urad according to the uncalibrated oplev. We suspected a hidden excitation somewhere and therefore cleared all the excitation using the diag command. This did not solve the oscillation. Then I increased the pitch damping gain by a factor of 3 to suppress the oscillation and this worked well. Setting the gain back to the nominal value (i.e. -1), I saw no ring up in pitch and the oscillation still stayed small. So I left the damping gain as nominal. Strange. Since ITMY oplev is off from the QPD, we did not check the angular motion in ITMY.

(I feel that the mirror motion is bigger than it used to be. Maybe ?)

Since the last PRMI commissioning in this winter, we have kept a couple of limiters in LSC and SUS for preventing the servos from unnecessarily kicking the optics during lock acquisition. The limiter values and their associated filters had been carefully adjusted so that they don't limit the in-lock DC values but they limit some impulsive signals. However, the limiters were limiting the in-lock feedback signal at DC when PRX(Y) or MICH was locked. I had to increase some of the limiter values by more than a factor of 5 or take them out last night.

I have not done a quantitative spectral analysis yet, but I am feeling that the motion of the suspensions are greater than before.

As for the limiter issue, we need only two limiters to be enabled, which is BS_ISCINF and PR2_M3_LOCK (see this old entry 10559). In addition, I had a wrong limiter on at LSC_PRCL last night.

Anyway, the one limiting the DC value yesterday was BS_ISCINF. This used to be 5 x10⁵ and it was 2 x 10⁵ in eariler yesterday for some reason. I then increased this limiter to 2 x 10⁶ in order to avid the signal being cut at this limiter.

Jeff, Krishna, Robert

here is the calibrated (against seismometer) amplitude spectral density.

That doesnt look quite right. Lets debug via email. -Robert

That should read

ITMx (not y)

and

ASSY-D1001838-V6 #002 (upper ring heater)

not #005. typos.

Note, when the 40kg optics are installed and removed from the QUAD lower structures that these ring heaters are mounted to, there is some torquing of the structure (and therefore the RH).  Possibly this adds to these failure modes.

One more fix. The lower ring heater should be:

ASSY-D100195-108

Made a mistake with a V5 and V6 part.