My goal was to see if we can build a consistent model of the scattered light noise coming from OMC ASC motion, based on Georgia's injections (47446). Hopefully the A2L improvement described by Jenne (47488) will mitigate the problem.

For future reference, below the times of 0.5 Hz line injection:

'ANG_X': [1236416153, 1236416985, 1236416435, 1236416612],
'ANG_Y': [1236398014, 1236398260],
'POS_X': [1236398867, 1236398529],
'POS_Y': [1236399517, 1236399737],
'quiet': [1236422949]

The model is the same I used in 47423, but this time I used a linear combination of all four OMC ASC signals ('H1:OMC-ASC_ANG_X_INMON', 'H1:OMC-ASC_ANG_Y_INMON', 'H1:OMC-ASC_POS_X_INMON', 'H1:OMC-ASC_POS_Y_INMON') as the witness for the scatterer motion. So we have four k coefficients and two f coefficients:

where x_i are the OMC signals, "band-passed" ad 0.5 Hz as described in 47423. 

I considered two possible approaches to the fit of the parameter model:

1. treat every injection separately, and fit the optimal parameters independently for each of them
2. fit the same parameters to all injections simultaneously

In both cases I did not include the quiet time in the fit.

Fit each injection separately

The plot below shows the best fit of the scattered light model above for each injection separately. You can see that the model fits all injections quite well, but the optimal parameter in each case are quite different. In all cases it seems the largest contribution to building the witness channel comes from H1:OMC-ASC_POS_X_INMON. But the value of the k coefficient that converts the witness to scatterer motion varies a lot, as well as the coupling coefficients f_1 and f_2. The blue curve is DARM, the orange curve is the modeled scattered light noise. The parameters for each fit are listed in the box in each plot.

Applying each of these fit parameters to the quiet period gives us a busy plot, that shows how most of the fits overestimate the noise in the quiet case.

Fitting all injections all together

Assuming that the same scattering physics applies to all the injections (even if they're along different directions and separated by hours), we can force the fit to use the same parameters for all injections. The results are shown below. Now the very last plot shows the projection for the quiet time using the best parameters. The POS injections are fit well, while the ANG injections aren't that good. At least with these parameters (shown in the plot) the quiet time projection isn't overestimated anymore. 

I also tried other approaches, like allowing the per-filtering of the OMC signals to be a parameter (both as a band-pass Butterworth or a generic second order stage), without any more success than the global fit above.

In any case, it seems safe to assume that in normal operation, scattered light from OMC ASC motion is close to the measured sensitivity.

Some thoughts on why the same model parameters don't work for all injections

1. the up-conversion physics might be different from simple beam scattering: for example there can be a modulation that comes from clipping
2. maybe the OMC ASC sensors are good only at some frequencies, or we need to account for frequency dependent pre-shaping, for example due to the ASC loops
3. we haven't quite found the right witness sensor: maybe when we shake the OMC ASC we are inducing a motion of the real thing, but we don't have a good measurement of what that is

Screenshots attached. The wind sensor terminals are not in OP. I believe this also happened a week or so ago and was fixed by restarting the Beckhoff computer. I have not tried doing this yet. It seems that the issue may be more fundamental than I first thought.

There are also a lot of CRC errors that I believe have been there for some time. The red circles in the first screenshot indicate these.

End Y looks fine.

When attempting to relock this morning it was noticed that the SRC was pretty off in alignment. The AS air spot was noticeably high in the camera image. Heating of SR3 caused a rather large change in the pitch of SR3 which also seemingly effected some yaw as well. Below are before and after trends of the M3 witness sensors, OpLev, and alignment sliders.

During last long lock, there's high coherence right in the bucket with PRCL and REFL_A_RF45 both I and Q. Not much coherence with MICH, some with SRCL.

https://ldas-jobs.ligo.caltech.edu/~gabriele.vajente/bruco_lho_1236519918/ 

This coherence might not see high, but at 50 Hz it projects into DARM at 1e-20 m/rHz, whereas DARM is 4e-20 m/rHz. That's about a 5% contribution in amplitude to DARM.

Topped off Crystal Chiller with 100mL.  Diode Chiller OK.  Filters look fine (mentioned slight yellow tint of Crystal filter to Jeff B & he said it is OK).

There was some problem with the high frequency photon calibration lines not lining up with DARM, so I remeasured PCAL to DARM and the DARM OLG and measured and fit the DARM plant.
The DARM plant fit was:
Optical Gain = 3.26e+06 cts/m
DARM pole    = 423.61 Hz
Delay        = 4.45e-05 s
Spring Freq  = 3.67 Hz
Spring Q     = 14.45


I updated the front end calibration filters according to this fit.  The high frequency lines seem better now.  

Then Georgia and I tested the calibration with a BB pcal injection.  There was a large phase discrepancy at low frequency, so we flipped the L1 actuation gain in the front end calibration, this seemed to help us.

We then tuned the actuation stages to get a decent front-end calibration.  We are good in the bucket to 6% with wiggles due to slightly improperly compensated actuation phase.  The reported range fell from 100 Mpc to 95 Mpc due to the changes.

We noticed that the 10-30 Hz noise is worse recently compared to our Feb 28 reference. I ran Gabriele's A2L script on ITMX (the only test mass without a dither line). It barely changed the A2L gains (Y2L went from 0.366 to 0.259, P2L from -3.926 to -3.9819), and there wasn't any change.

I also drove the OFI (transverse) at 1Hz to see if it affected our 48 Hz scatter or any other lines, it did not. Attachment shows DARM spectra and OFI error signals with and without the excitation, excitation times are shown in red, quiet times in grey.

Georgia, Dan, Sheila

We changed the ESD bias voltages on ITMX, ITMY, and ETMY to see how it affected DARM - it didn't.

[Jenne, Anamaria, Sheila]

Sheila reminded us that back in 2015 (alog 19691) they adjusted the P2L coefficient for the OMC ASC, which helped improve scatter from the OMC.  We tried to do that again today, and think we were sucessful with P2L.  We then thought we should give Y2L a try, and that has us pretty confused. 

We drove the OMC SUS LOCK P filter bank at 0.2 Hz with 4500 counts.  I then checked the value in the P2L filter bank to ensure that the fringe wrapping was minimized.  The old P2L value from 2015 was 6.0.  I tried both sides of that (5ish and 7ish), and found that I could increase the scattering shelves by some amount at 5.0, and by the same amount with 7.7.  I'm leaving the value at 6.35, which is halfway between those values.  This is indistinguishable from the 6.0 value, but is halfway between the 2 matching values I found.  Since we're not actually sending our OMC ASC signals through the LOCK filter banks and the drivealign matrix, in order to propagate this value of P2L decoupling, we put values in the OMC ASC output matrix.  So, the pos and ang Y (Y is y-axis here, not yaw...) outputs get fed to the SUS_P as you'd expect for P2P, and then we also feed those outputs to SUS_L with matrix elements 6.35 times the P2P matrix elements.  The PosY -> SUS_L element used to be -670, and it is now -711.  The AngY -> SUS_L element used to be 310 and is now 327. 

We then switched to driving the OMC SUS LOCK Y filter bank at 0.2 Hz with 1000 counts.  With a Y2L value of 0 (we have never used any Y2L decoupling for the OMC suspension), I saw giant scattering shelves.  I then changed the Y2L gain to minimize the scattering shelves, and saw a huge improvement, which you can see in the attached figure.  The Y2L value I need to minimize those shelves is 35, which seems awfully high, but I was able to convince myself that maybe it needed to be large since the input beam to the OMC breadboard is not near the geometric center of the breadboard.  We then started stepping up the OMC ASC output matrix values for PosX and AngX to OMC L, to move them toward being 35 times the PosX and AngX matrix values.  However, we started to see (not nearly as big as the injection scattering) scattering.  This was repeatable, in that I took away the 'decoupling' matrix elements and the scatter went away, put them back, the scatter comes back.

Anamaria and I checked that there aren't any DC gains that are different in the OMC ASC path versus the LOCK filter banks, so it's not clear to me why the Y2L decoupling value that was good for the Lock Yaw drive isn't good for the OMC ASC Yaw drive. For now, we are leaving the Y2L elements of the OMC ASC output matrix at their nominal values of 0 until we understand more what is going on.  One supposition is that it matters what frequency we are moving the OMC at, and that we should be doing this at some other frequency.

This morning all the laser got tripped and the CO2s went down. I used this opportunity to switch on the mask from a cold (ring heater only) state to get the actual OPD of the mask. Beforehand the CO2 cooling down was interfering with OPD measurement of the mask.

We now see an OPD more in line with the prediction (top right plot in main figure). Using 1.1W of CO2 on the mask gives similar level of OPD change compared to Aidan's model, which originally required 450mw. The test was cut short because we needed to get the thermal state ready for locking.

Opening up the AA chassis (S/N S1102788) we confirmed that the AA boards D070081 (chn 16-23 S/N S1202203; chn 24-31 S/N S1202202) have AD8622 stuffed. On channels  20-27 all of them (input and output) were replaced by ADA4075-2ARZ. This incudes the channels for the LSC_REFL_SERVO readbacks, the IMC-REFL_SERVO readbacks and the SQZ-CLF_REFL_RF6 readbacks. Attached are some measured transfer functions for a channel with the ADA4075-2 (reference trace with notch) and for a channel with the AD8622 (no notch). Varying the drive voltage from 1mV to 5V made no difference for the ADA4075-2, For the AD8622, one can clearly see that it cannot drive fast enough and at frequencies around the notch it looks anything but a notch.

Plots: AD8622 transfer function for 5V, 1V, 100mV, 10mV and 1mV drive compared against the ADA4075-2.

Given that H1 is still in commissioning mode, I hesitate to report on lines that may be due to ongoing work, but I'm seeing a very strong 1-Hz comb in DARM as of last Wednesday, which I suspect is unintended. Below are spectra of 20-120 Hz and its 20-Hz sub-bands from yesterday's data (typical of daily spectra since Wednesday). 

Note that the lines are on exact 1-Hz multiples (to ~0.5 mHz resolution), but they have flat shoulders of O(0.01 Hz) full width. An arbitrary but typical example at 34 Hz is shown.

Also, the amplitudes of the lines do not fall entirely monotonically with frequency. There are clusters of lines with a spacing of roughly 7-8 Hz.

Any ideas on what changed, perhaps during the last Tuesday maintenance? 

J. Kissel (with help from L. Sun, E. Goetz, L. McCuller, E. Bonillla, R. Kumar)

I've been processing the actuation function data from 2019-03-01 (see LHO aLOG 47206), and have some updates. Not were I want it, but since the heat is on I'll give a status update.

Recall from LHO aLOG 46806,
Steps to success:
    (i) Find out why ETMX L1 and L2 doesn't work, and fix it, such that we can go forward with full ETMX DARM actuation. 
        DONE (see fixes to UIM and PUM crossovers in LHO aLOGs  47164 and 46861)
    (ii) Measure the PUM and UIM coil drivers, and update the compensation. 
        DONE (see LHO aLOG 47167)
    (iii) Truly identify ETMX all 8 violin mode fundamentals and their harmonics to update the PUM dynamical model (needs IFO time), finish fitting the UIM transfer function data to update the UIM dynamical model (needs Kissel time)
    (iv) Remeasure all stages of ETMX with the full IFO 
        DONE (see LHO aLOG 47206)
    (v) Update the front-end calibration (including reference model parameters at calibration line frequencies such that time-dependent correction factors are accurate), and 
    (vi) confirm success with the full IFO (needs IFO time).

So, I'm currently working on (iii) [with the help of Edgard, Rahul, Borja (#TeamSUS) along with Lee and Lilli #TheFittingCrew] and this aLOG is reporting the progress on fitting the results from the 2019-03-01 measurements, i.e. (iv) [working with Evan #TeamPyDARM].

Check out the attached .pdfs from each stage below, which are the output of the following script, 
    /ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Scripts/FullIFOActuationTFs/process_actuationmeas_20190301.py
after creating a new pyDARM model parameter file,
    ^/trunk/Runs/O3/H1/params/modelparams_H1_20190301.py

TST: MCMC Results 
    Force per count of longitudinal DAC request (just down stream of DRIVEALIGN bank):
    Gain = 4.734e-12 (N/ct)
    Force per actuator signal (current or voltage):
    Gain = 4.433e-11 (N/V**2)

    Actuator gain, H_c (N/ct)              | 4.734e-12 (+2.09e-15,-2.102e-15) or (+0.04415%,-0.0444%)
    Residual time delay, tau_A (usec)      | 10.79 (+0.7573,-0.76) or (+7.02%,-7.045%)


Comments: Except for the sign flaw in the "intermediate results" plot, I'm very happy with the state of the systematic error, I believe the MCMC fit, and I think we're ready push to front-end.

PUM: MCMC Results 
    Force per count of longitudinal DAC request (just down stream of DRIVEALIGN bank):
    Gain = 6.223e-10 (N/ct)
    Force per actuator signal (current or voltage):
    Gain = 0.03038 (N/A)

    Actuator gain, H_c (N/ct)              | 6.223e-10 (+3.339e-13,-3.33e-13) or (+0.05367%,-0.05352%)
    Residual time delay, tau_A (usec)      | 4.096 (+1.568,-1.571) or (+38.28%,-38.34%)


Comments: There's still something very fishy going on below 20Hz. I don't understand it, and this is under investigation by Evan and I.

UIM: MCMC Results 
    Force per count of longitudinal DAC request (just down stream of DRIVEALIGN bank):
    Gain = 7.498e-08 (N/ct)
    Force per actuator signal (current or voltage):
    Gain = 1.597 (N/A)

    Actuator gain, H_c (N/ct)                   | 7.498e-08 (+2.842e-11,-2.869e-11) or (+0.03791%,-0.03826%)
    Residual time delay, tau_A (usec)      | 57.81 (+1.218,-1.215) or (+2.107%,-2.101%)

Comments: the low frequency data (below 10Hz) is pretty un-informative, and this is where the UIM matters -- BUT I'm not sure we've ever got any better. I'm dividing out the high frequency dynamics of the OSEM bracketry and UIM to PUM violin modes based on old H1 ETMY data for now CSWG aLOG 11212, and that has significantly flattened out the systematic error, so we can now use all the data from 10 to 100 Hz for the fit of the actuation coefficient instead of what we did in O2 which was only use 3-5 points between 10 and 20 Hz. I'm working with Lee to get an updated fit on data that I took on 2019-01-25.

Finally, I show one of the plots from the latest output of 
    https://svn.ligo.caltech.edu/svn/aligocalibration/trunk/Common/pyDARM/darm_loop_critique.py
which, unlike the scripts from above is a function, which I called using the following command line,
    python3.6 darm_loop_critique.py --run=O3 --IFO=H1 --DARMmodelfile=modelparams_H1_20190301 --modelFunction=modelPars --outputfile=2019-03-01_critique

This shows the relative contribution of each stage. to the overall actuator to give you a feel for what matters where (now that things have been updated to reflect Sheila's latest work with the PUM crossover).

Comments: 
    - Only the phase is shown, so take the plot with a grain of salt. 
    - In the calibration band (a bit larger than the detection band, between 5 and 5000 Hz), excitingly, the UIM is now *very* unimportant to get perfect. However, I'm interested to see how this plot shapes up once we've fixed the dynamical model (which is currently all this plot has to go on)
    - It is far more important to get both the PUM and the TST stage right.
    - We'll need to resolve this reported frequency-dependent systematic error in the PUM, since it's right in the phase/magnitude mixing region ("the crossover" implies the hand-off between stages only happens precisely at a single frequency. No bueno.)

So -- I continue my work on the PUM as priority.

We also need to retake a sensing function sweep suite, so I can compare the results against and open loop gain model.