Note: After rerunning the MCMC again, the weird patches in the sensing corner plots disappear (Oct 12 measurement). This might be caused by the poor measurement. Keep in mind and check in the future.

The high uncertainty at low frequency definitely begs for the need of better precision measurements of the actuation stages at low frequency. We had that during O2, but right now we only have the single, quick-n-dirty measurement of the actuation stages.

Hey Ansel,

The Hartmann camera sync frequency is set to 1Hz for both cameras at the vertex. 

Hi Daniel,

Right, some of us were wondering in an email thread if that frequency could be set to something different for a period of time, ideally with DARM in low noise, but that's not essential for things to be seen in magnetometers.

thanks,
Keith

Keith and Ansel, 

I adjusted the ITMY and ITMX HWS sync frequency to 55 Hz from gps 1224538388 to 1224539253. Please don't hesitate to let me know if you need anything else related. 

Yes, this couples very strongly into DARM. A line at 54.8ish Hz appears with the change (spectrum attached).

Hi Daniel - yes, that definitely had an effect. I checked both DARM, and also the magnetometer channel where the comb was showing up most strongly. The comb disappears in both channels during the time segment you indicated, and reappears afterward.

Indeed this is a known problem with the HWS cameras (FRS 4559). And the problem was solved by placing the cameras on alternate power supplies in 2016. I suspect that following the post-O2 disconnecting and reconnecting of the HWS table and cables, the cameras got plugged into the chassis power supply again. We should do the following for all 8 HWS:

• Confirm that the HWS cameras are plugged into the chassis. If so,
• Get the plugged into a separate power supply, and
• Lock out the chassis power supply to the cameras to prevent plugging them back in there in future, and
• After consultation with Rana, look at getting them plugged more permanently into a dirty power supply (like the one for the vacuum system) and check they're properly isolated from the clean power
• Also - enable EPICS control & read-back of the camera frame rate

This should be done for the ETMs as well (where the 57Hz is an issue). The original isolation fix does not appear to be robust enough.

• Mitigation of 1Hz at LLO ITM HWS: https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=31339
• ITM HWS ground isolation at LLO: https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=30350
• LHO appearance of 57Hz ETM HWS lines: https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=44495
• Reappearance of lines from LLO ETM HWS: https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=41248

Block diagrams (DCC is down so I can't get hyperlinks right now)

• Corner station: E1100892
• End stations: E1100891

"normal" means RF9 not reduced.

(We've looked at REFL45 though it's not used for LSC as when the RF opamp for 45 is going crazy it could affect the input side.)

We will repeat this with the reduced RF9.

Nominal modulation index for 9.1MHz: Γ~0.191
Nominal modulation index for 45.5MHz: Γ~0.251

For the above measurement the modulation index for 45.5MHz was reduced by 3dB to Γ~0.177.

Side-by-side numbers for H1 and L1 are posted in LLO log entry 41547

RF slider settings during this measurement.

thanks; REFL45 is not being used so I turned the whitening gain from +12dB to 0dB and the saturations have stopped. It would be good to put this into Guardian.

Josh, Andy

There are also overflows in the ASC model, FEC 19. The most overflowing channels there (overflowing 101 and 122 times) are 1_11 and 1_30, ASC REFL A RF 45 Q2 and ASC REFL B RF45 I4. Shortly after that are other I quadrant channels. REFL 45I is in loop for pitch. This can couple the overflow to DARM, but it might also be that glitches are showing up in DARM and overflowing the ASC.

It's probably worth checking the levels on the ASC REFL PDs at 45 MHz, and see if they can be stopped from overflowing in low noise. These do seem to be short overflows, not ones caused by the signal going slowly out of range. We haven't checked other signals yet; these were first on the list because they overflow most commonly.

Attached are the overflow times from yesterday for REFL_B_45 I4, and the simulink for reference. Then we have plots of the REFL 45I inmons for two glitches, and the associated spectrograms in DARM.

in the above entry, ITM is a typo, it should be ETM Hartmann.

After putting in the notches for the heartmann noise, we were able to plot the BLRMS of DARM using the OAF-CAL LSC BLRMS (attached).

The frequency bands are 20-30 Hz, 38-60 Hz, 60-100 Hz, 100-450 Hz, and 450-950 Hz. The color coding follows the rainbow: red is the lowest frequency and purple the highest.

Since Georgia added the notches, we no longer see the 25 minute period that was characteristic of the TCS/Harmann noise, but rather the underlying breathing of the noise. Most troubling is the Orange (38-60 Hz) band. This one and the green one show all the breathing that is in our mystery noise band.

I think it would be illuminating if someone could find correlations with the orange and green bands and some other channels; it may shed light on what is causing the coupling. Note that since this is a RMS, one should find a correlation between the absolute value of (whatever) and these BLRMS channels.

I'm also attaching a plot showing DRMI error signals and coherence with DARM. The PRC signal is ~50x larger in the 20-30 Hz band than it used to be in O1. This also shows up in MICH/SRC and so there's a large coherence with DARM. I wish we had a noise budget for PRC/MICH/SRC, not just DARM.

I tried to run 'cdsutils audio' from the command line to listen to some band limited darm (to see if it sounds like rubbing or scattering) but the 'pygst' and 'gnuradio' modules are not available from this terminal. It would be useful to have that functionality added if possible.

I had a look at the DARM BLRMS during this morning's long lock, it looks like the noise in the 38-100 Hz band is much quieter, see attached plot showing 10000 seconds of data. The extra noise in RLP3 band at 1224538440 is an injection (I think).

Two differences between last night and this morning is the initial alignment has been redone, and the BNS range which decreases over the duration of a lock. Second attachment shows H1 BNS range. The BLRMS plot from last night was taken at T-16, the BLRMS plot from this morning starts around T-4.

Out of interest I also looked at the suspension witness monitors (3rd attachment) and test mass op levs (4th attachment) over the duration of this morning's the lock. In these plots pitch is on the left and yaw is on the right. PR2 PIT stands out as being extra noisy, and SRM and PRM drift the furtherest over the duration of the lock.

Thanks for pointing to these glitch times, Jenne! Just to let you know, a few of us from DetChar have started looking into them, and we'll hopefully have more to report tomorrow.

I took a look for glitches in a 2 minute window around the times Jenne points out. Since most of the times Jenne posted pointed to obvious loud glitches, that's what I started looking for. I couldn't find anything similar to this for around 15:08, 15:13 and 15:15 though. For the rest of the times I found rather loud glitches, however they don't all look the same. Most are clearly overflows (if you just look at the spectrograms). Below I list the times of the glitches I found, and if an overflow which model it points to (sorry I'm rubbish at parsing simulink). I got the overflow information from the summary pages:

1224427963.30 - overflow in 98 2_0,3 (UR and LR ETMY L2 coil) and 88 3_1-4 (all quadrants of ETMX ESDs)
1224428469.55 - overflow in 98 2_0,3 (UR and LR ETMY L2 coil) 88 3_1-4 (all quadrants of ETMX ESDs), 10 0_12,13 (LSC) and 8 0_12,13 (OMC length dithers)
1224429544.87
1224430949.50 - overflow in 88 3_1-4 (all quadrants of ETMX ESDs)
1224431938.35 - overflow in 88 3_1-4 (all quadrants of ETMX ESDs)
1224432254.1 - overflow in 8 2_18 (OMC)
1224432439.6 
1224433667.31 - overflow in 98 2_0-3 (all quadrants of ETMY L2 coils) and 88 3_1-4 (all quadrants of ETMX ESDs)
1224439031.89 - overflow in 98 2_0-3 (all quadrants of ETMY L2 coils) 88 3_1-4 (all quadrants of ETMX ESDs), 10 0_12,13 (LSC) and 8 0_12,13 (OMC length dithers)
1224440021.65 - overflow pretty much everywhere. This is ~63 seconds before a lockloss (98 2_0-3 (SUSETMY), 88 2_0-3, 88 3_1-4 (SUSETMX), 30 2_0-4 (SUSITMY), 29 2_4-9 (SUSITMX), 19 1_8-15, 19 24-31 (ASC), 10 0_7 (LSC), 8 0_7 (OMC) 

I have now put the 9MHz reduction into the main acquisition sequence path.

The attached plot shows that we are really seeing more power circulating in the arm cavities (and also less power at the AS port) when the 9MHz modulation depth is reduced.  So, there must be some offset somewhere that we're reducing. 

Also, I tried reducing the modulation depth by 8dB rather than just 6dB, and the IFO gets noticeably more glitchy when I do the extra 2dB.  So, it seems like 6dB of reduction is a reasonable place, and we can work on finding what is causing this circulating power change.

Since we're reducing the 9MHz modulation depth from 0.2ish to 0.1, we're changing the 9MHz power from 4% to 1%, so should have ~3% more carrier power.  That is consistent with the increase in circulating power that we see.  However, the apparent shot noise reduction implies a much larger increase in power, so something is still not quite hanging together.

It would be worth checking the RF levels on the other LSC RFPDs used for LSC control (if you haven't already), as was done yesterday for REFL9.

Premature to say we're gaining something, as I don't see the same reduction in uncalibrated DARM nor in OMC DCPD.

In the first attachment , red, blue and brown are from the single lock stretch corresponding to Jenne's red, blue and brown. No improvement at 1kHz at all, and the frequency noise part (f>2k or so) is worse when 9MHz was reduced. In Jenne's plot the improvement was pretty much 15 % or so over the large frequency region.

The second attachment is later in the morning. Blue is small RF9, green is large RF9.

In the latter there seem to be a difference at 100Hz but I don't know if this was due to high/low RF9 mod index.

Optical gain difference between high/low RF9 was no larger than a few % in both of the lock stretches.

Update (Jenne, Keita): Things makes more sense now.

In the attached, you should compare red (reduced RF9) with brown (not reduced) from the same lock stretch, or blue (not reduced) with pink (reduced) from another lock stretch. Legends are in UTC. In both of the cases, smaller modulation index increases the frequency noise in high kHz but seems to somewhat reduce noise at 100Hz.

It didn't make sense at first because there was an error in the legend of Jenne's plot.

Details:

Turns out that the legend for the brown trace (15:59:07, 9MHz back to normal) in Jenne's plot was incorrect, it was neither UTC nor local time, it was actually from 09:53:02 UTC, i.e yesterday. This means that all of her "9MHz reduced" traces are from today and all of "9MHz normal" traces are from yesterday.

But she intended to look at 15:59:07 UTC for brown trace, which was from today when 9MHz was reduced (but the CM gain setting was not changed to compensate). In the attached, red and brown are the same as Jenne's red and what Jenne intended to show in brown, these are from the same lock stretch.

Blue and pink are from another lock stretch later in the morning. (In this case, CM gain setting was changed to compensate for the optical gain.)

Low RF makes frequency noise worse at high-kHz due to lower S/N (pink VS blue). In the case of red VS brown, overall CM gain was lower, making the difference larger than pink VS blue.

To confuse the matter further, somehow at some point in yesterday the shot noise level seemed to have improved according to  Sheila, and that is clearly seen in green trace from yesterday.

There is at least a bit of a change in the range during the on/off test that I meant to plot the times of from this morning.  See attached.

I've modified EvanH's old stepping modulation depth script so that it will change the modulation depth, wait 10 min, then change it back, repeating 5 times.  If the IFO is locked when the last person leaves for the night, please launch this (attached, and in /ligo/home/jenne.driggers/LHO_work/2018_10_24_9MHz_reduction/step_9MHz_many.py)

I modified Jenne's modification of step9.py so that the user can CTRL+C the skip at any point and the PD gains will all be returned to their original values when the script starting running.  Useful for when we lose lock during the test.  

Pressing Ctrl+C while the gains are changed and the interferometer is locked is not recommended: the script will instantaneously return all gains to original values.

Code lives in:
/ligo/home/craig.cahillane/utils/step9mod.py

Started a run of this code at Oct 25 2018 09:52:44 UTC (1224496382).

Yet another update:

For the moment I take back my statement about lower modulation index VS high kHz frequency noise, the coupling itself is slowly changing with time and I might have been tricked.

The 56.84 Hz comb is a blast from the past.  Here is an entry from March 2013 concerning H2 one-arm data suggesting that the comb could be purely DAQ-related:

https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=5806

This one is close enough to the 57Hz ETM HWS that we've seen in the past that I thought I should mention it here. However, we've recently overhauled that system and I've not confirmed what the new frame-rate is yet.

Here's the old analysis: https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=14778

I Looked at an empty EX ADC channel (H1:PEM-EX_ADC_0_08_OUT_DQ) and found that the 58.64 Hz comb is there too, indicating that the comb comes from some source independent of the interferometer (i.e. in the DAQ or RF pick-up). I have attached ASDs of the empty ADC in 0.1mHz bins both for the entire spectrum (0-1024Hz) and near to the 14th 56.84 harmonic (~795.76Hz). I was slightly surprised that a second nearby line (that would explain the amplitude modulation) was not seen even with the 0.1mHz resolution. 

Yes - the ETM HWS are running at 57Hz. This doesn't mean that this is the issue since it appears to be DAQ related, buuuut .....

aidan.brooks@zotws11:~$ caget H1:TCS-ETMX_HWS_SYNC_FREQUENCY

H1:TCS-ETMX_HWS_SYNC_FREQUENCY 57 Hz

aidan.brooks@zotws11:~$ caget H1:TCS-ETMY_HWS_SYNC_FREQUENCY

H1:TCS-ETMY_HWS_SYNC_FREQUENCY 57 Hz

I've searched EX channels and found some channels where amplitude modulation is seen:

H1:SUS-ETMX_L3_DRIVEALIGN_L_OUT_DQ
H1:SUS-ETMX_L3_ISCINF_L_IN1_DQ
H1:SUS-ETMX_L3_LVESDAMON_LL_OUT_DQ
H1:SUS-ETMX_L3_LVESDAMON_LR_OUT_DQ
H1:SUS-ETMX_L3_LVESDAMON_UL_OUT_DQ
H1:SUS-ETMX_L3_LVESDAMON_UR_OUT_DQ
H1:SUS-ETMX_L3_MASTER_OUT_LL_DQ
H1:SUS-ETMX_L3_MASTER_OUT_LR_DQ
H1:SUS-ETMX_L3_MASTER_OUT_UL_DQ
H1:SUS-ETMX_L3_MASTER_OUT_UR_DQ

Attached plots are BLRMS and zoomed spectrum of H1:SUS-ETMX_L3_DRIVEALIGN_L_OUT_DQ as an example.
The vertical line (orange, dotted) in the zoomed spectrum is at 568.404 Hz (10th harmonics).

NO

we moved the Hartmann sync frequency and the lines moved accordingly - its Hartmann, not DAQ.

Why does the Hartmann sensor so strongly couple to DARM?

With regards to the coupling, the ETM HWS are served by the same power supply as the ring heaters. It's a long shot but it might be worthwhile to try disconnecting the ring heaters from the driver while the to see if the coupling of the 57Hz to DARM is changed.

https://dcc.ligo.org/LIGO-E1100891