Philippe Nguyen, Corey Austin, Sharan Banagiri, Kara Merfeld, Anamaria Effler, Robert Schofield
O3 initial PEM injections took place mainly the weeks of March 25 at LLO and March 18 and April 15 at LHO. Hundreds of injections were made at tens of locations including acoustic, magnetic, shaking, impulse and RF injections (DTT files: https://lhocds.ligo-wa.caltech.edu/exports/pem/19aMarPEMinjections/LHO). A preliminary report was prepared, https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=47881, we reference it below when it contains more detail then presented here.
A. Vibration coupling
1) Worst site: LLO EX transmission monitor
The large drops visible in LLO’s range as anthropogenic vibration levels increase, are associated with the EX transmission monitor. The noise has a higher SNR in the pitch and yaw signals from the quad diodes than it does in DARM. Thus, the diodes witness the noise before it is combined with other noise in DARM. Candidates include scattering and servo noise. https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=46147
2) Next worst sites: at LLO are EY and HAM5/6 and, at LHO, HAM5/6 and the PSL followed by EY
Figure 1 indicates the sites at LHO and LLO where the ambient vibration level is estimated to make the greatest roughly linear contribution to DARM for each frequency. A rough summary of the LHO plot is that, the greatest ambient vibration contribution to DARM is in the HAM5/6 area below 100 Hz, and in the PSL area above 100 Hz. The rough summary for LLO is that the worst sites are HAM5/6 and EY, and both are a significant contribution to DARM below 100 Hz.
The plots are made by first calculating a coupling function (meters of DARM per meter of motion at sensor) for each sensor for each of multiple injections, and then producing a single coupling function for each sensor, using an algorithm to select the best of the multiple injections, usually the loudest relative to other sensors. The coupling function is multiplied by the ambient vibration level to produce an estimate of the ambient contribution to DARM. At each frequency in the plot, the sensor or sensor region with the highest estimated contribution to DARM is indicated by color, and the estimated contribution plotted.
These estimates are for banded linear coupling. When we notice upconversion, we study it separately, injecting in narrow bands to find the motion frequencies responsible for the upconversion. The estimates in Figure 1, however, may be somewhat distorted by upconversion: if the upconverted noise is within the injection band, that noise is assumed to be produced by the injection at the frequency that the upconverted noise appears at in DARM. This tends to be more of a problem for scattering, since it is so non-linear, than for beam jitter. We did not find cases of upconversion, other than at the ETMX transmission monitor, that would alter these estimates of the worst coupling.
These plots, other summary plots, and coupling functions for every relevant PEM sensor are located at PEM.LIGO.ORG, press the “coupling functions” button.
3) At both sites, HAM5/6 coupling is likely at the septum
The coupling at HAM5/6 was narrowed down to the septum at both LHO and LLO, using impulse injections. https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=48886
4) EY coupling at LLO is likely in the manifold, at LHO, it may in the ETMY chamber. More evidence is needed to be confident that baffling the rest of the periscope and the Pcal beam nozzles will fix the LLO EY coupling problem.
Impulse injections suggest that the vibration coupling site at LLO EY is in the manifold (near the Pcal periscope) and, with less certainty, that the coupling site at LHO EY is in BSC 10. Resonance structure of the signals suggest that we don’t have accelerometers on the coupling sites. There are some glints visible from the un-baffled part of the LLO EY periscope, but other sites have similar glints. There is the possibility of scattering from the Pcal beam nozzle, but shaking was inconclusive. More evidence is needed in order to be confident that baffling the rest of the periscope and the Pcal beam nozzles will fix the EY coupling problem. https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=46208
5) HAM5/6 coupling at the septum window?
The coupling at the septum may be at the window: a bright scattered light spot is visible on the septum window at LHO, at a viewpoint 3 degrees from the beam. https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=48965
However, no spot was observed at LLO at 28 degrees (the 3 degree view was not available), https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=45980 , and updated scattering estimates from Peter F. suggest that the window could not account for the scattering unless there is some imperfection https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=46272 .
One mitigation possibility would be to remove or angle the window and hedge our bets by also baffling the light-accessible regions of the septum.
Other beam spots on LLO HAM6: https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=45378, https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=44985
OMC REFL beam not the cause of the main HAM5/6 scattering: https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=46010 .
6) IO jitter coupling at both sites
Figure 2 shows jitter coupling for both sites.
7) Comparison of LLO EY and EX in-lock photos
Relevant to the source of coupling at both the TMSX and EY, are photos taken at the LLO end stations: https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=45469
The photos show a beam spot on the mirror that sends the red beam to the ALS table, M13.
8) Shaker injections at EY produce noise similar to some anthropogenic noise https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=46089
9) Other vibration coupling
Beamtube shaking https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=45710
Coupling at LHO BS chamber, LLO BSC1, and at the reduction flanges by the ITM optical levers at both sites is shown in Figure 3.
48 Hz peak at LHO The 48Hz peak at LHO appears to be driven acoustically (though modulated by alignment or other source). See the consistently lower amplitude when the HVAC is off https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=48919 . Acoustic and impulse injections suggest that the source is in the vertex area, possibly in BSC2, but more work is needed.
Reduction of 58 Hz chiller peak at LHO: https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=47881
10) Vibration coupling not seen (at least 10 below DARM)
|
Site
|
Location
|
Injection Type
|
Relevant to
|
|
LHO, LLO
|
mids
|
Acoustic/shaker 10-100 Hz
|
Beamtube scatter
|
|
LHO, LLO
|
IOT2L (IMC table)
|
Acoustic/shaker 10-100 Hz
|
Local coupling
|
|
LHO, LLO
|
IMC tube (HAM2-3)
|
Shaker 5-100 Hz
|
Local coupling
|
|
LHO
|
HEPI EX, EY, BS, IX, IY
|
HEPI Inj 1-100 Hz, beam dof
|
ACB, BS baffles, scatter
|
|
LLO
|
ISI all BSC and HAM
|
ISI Inj 1-5 Hz, beam and r dofs
|
Daytime scatter
|
|
LLO
|
Arm beamtube
|
Shaker ~58 Hz
|
Beamtube scatter
|
|
LHO, LLO
|
SRC tube (HAM4-5)
|
Shaker 5-100 Hz
|
Local coupling
|
11) Vibration coupling estimates from PEM injections correctly predict environmental coupling
We usually expect the coupling functions to be good to within a factor of about 2. For the known vibration features in DARM, the 48 Hz and 90 Hz peaks, as well as several of the jitter peaks at LHO, the estimate in Figure 1 is good to within a factor of 2. In addition to this comparison to stationary features, the coupling predictions have been tested against several transients, mentioned below.
LLO DARM glitch near S190510g is correctly predicted from PEM injection coupling functions. It was produced by thunder-driven vibration at EY https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=46025
Range reduction from LHO HVAC is roughly predicted from estimates for HAM5/6. https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=48912
Range reduction from rain at LHO is roughly consistent with PEM coupling functions https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=49495
Range reduction from wind at LHO is roughly consistent with PEM coupling functions.
B. Magnetic Coupling
Figure 4 shows that magnetic coupling is at least a factor of 10 below DARM at both sites. The coupling function for the stochastic group estimates of inter-site correlation from the LEMIs is here:
We only have one large coil working so these estimates of magnetic coupling are made using comb injections instead of broad- band injections (because of limitations in the field amplitude we can produce with the current system). Resonances observed in magnetic field coupling from the large-coil, broad-band injections (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=47881 ), justify the ongoing installation of large coils so that we don’t miss similar resonances between combs. We are now dominated by electronic and cable coupling, so the coupling functions are more complex than when the coupling was dominated by permanent magnets on optics.
Weekly injections. Because cable and electronics coupling can vary with electronics work, magnetic coupling is expected to vary during the run. In O2 the coupling varied at LHO by a factor of several between the beginning and end of the run. To better understand and identify changes in coupling, we have set up weekly automatic injections for each Tuesday before maintenance. The results are here: https://lhocds.ligo-wa.caltech.edu/exports/pem/WeeklyMagneticInjection/output/
Weekly injections have also begun at LLO: https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=45731
C. RF Coupling
Figure 5 shows that we can detect 9 and 45 MHz injections at both sites with at least two orders of magnitude greater SNR on our radio receivers than with DARM.
D. Site activities
Figure 6 shows several activities in the control room that show in DARM, an animated crowd walking, rolling a chair with a person in it across the control room, and dropping a large super ball from 4 ft. I suggest that we be a little more careful with site activities in this high-detection rate era. There is a chance that new or rare noise sources, that are not well accounted for in background estimates, may result in wasted telescope time before we get a chance to retract alerts based on PEM.
