It has been confirmed the 29.2Hz line is gone from Hrec (figure 1). The 92 Hz bump was previously confirmed to have disappeared (38738) we just turned on the fan yesterday and observed the change in enviromental sensors to verify the source then it was switched back off. So, there is not a recent lock with the fan on for comparison, but none the less the 92Hz region is flat. Yesterday the operator recieved a flag for the ENV_WE_CT_ACC_Z being out of it's thrshold. The top right of figure 2 shows the significant reduction in the overall noise when the scroll pump was switched off, which resulted in the flag.

Both the 29.2Hz and 92Hz structures seen in elog 38738 and 38803 have been eliminated from the environmental sensors (thanks to the VAC team). Firstly the 29.2Hz line has been associated with the west end  scroll pump (on the floor near the cryotrap), it has been determined that the fan responsible for the 29.2Hz is not neccessary. Secondly the 92Hz source is a result of a small fan on the signal recycling tower (again can be kept off). It is worth noting that there is a similar fan on the power recycling, and a switch off of that fan will follow. Once we have an interferometer we can check these structures are gone from LSC_DARM as well.

Looking at a spectrogram with very high frequency resolution (0.001Hz) over the course of about an hour and a half you can see a slight drift of the 29.2Hz line in ENV_WE_CT_ACC_Z. There is an identical drift seen in DARM as well. This confirms two things the line seen in WE_ACCs&SEISs and DARM are from the same source, and the line is not a calibration line or caused by some resonant (because of the slight drift).

Originally the 29.2Hz line seen in LSC_DARM was believed to be the A/C in the WE, however during the switch off of the A/C the line is still present (figure 1). Looking at the spectrogram of LSC_DARM we can also see the line is stable in amplitude and frequency. Bruco shows good coherence with WE accelerometers and seismometers. Looking at a FFT of the WE ENV sensors we can see (figure 2) the line is strongest in the Cryotrap Accelerometer, but the line can be seen in all WE ACC and SEIS. It is worth noting that the line is not present in the microphone. Just as a double check coherence was between DARM and all the sensors with the lines present (figure 3), as expected we see very good coherence.

Is there anything else we can switch off in the WE? 

A further analysis of the mysteriously disappearing 92Hz bump in DARM has been done. Attached are Figures 1,2, & 3 which show spectrograms of DARM, MICH & PERCL respectively. There is clearly a decrease in noise around 92Hz in all of them. Figure 4 looks at the accelerometeters around the PR tower before and after the disappearing bump. There is a line at 92Hz in the PR accelerometers that has disappeared. Lastly figure 5 is a spectrogram of the SPRB_LINK_ACC which was used to pinpoint the time the line disappeared to 14:19:49UTC ±1s (1184595607). Correlated in time to some VAC switch offs (38736).

It is also worth noting there appears to be 6 drifting lines in PERCL (3 pairs of lines all drifting in the same way), the clearest lines are Figure 3 around 115Hz. By eye these do not seem correlated to any F7 temperature sensors and are drifting significantly over a day. These lines can be very faintly seen in LSC_DARM. Is this an issue with aliasing even though we don't see mirror images? There is also some change around 80Hz in Figure 1 (LSC_DARM) that is not yet well understood.

Summary: today we performed a switch off of several infrastructure machinery and electronic devices. At the of the day seems we improved DARM at the 45Hz and the 90Hz bumps, and we also understood the origin of some lines. BUT we worsened the DARM noise at and around 50Hz.

The switch off was done in sections, with this sequence:

1. between 7:35 and 9:01 UTC: air conditioning of WEB, NEB, MCB and CEB was switched off (including Wi-Fi in experimental halls, the photon calibrator and SWEB, SNEB fans - SPRB was already off): see detailed timing. Then all HVAC stuff, fans and the photon calibrator was switched back on between 9:12 UTC and 9:40 UTC (this last time is rough). Wi-Fi was switched back on at 14:25 UTC.
2. 9:45 UTC (end time): VAC stopped the cooling fans of turbo pumps at PR, BS and PR-BS link
3. 10:45 UTC (end time):  VAC stopped the HVAC cooling fan of the 'remote pump' room in CEB (restarted at 13:30 UTC)
4. 11:30 UTC (end time): VAC stopped all the remaining active cooling fans of CEB turbos: IB; DT;SR;NI;WI (both types, industrial ones and 'home' type ones) (see also here for VAC times). Then VAC restarted all the fans (finished at 14:30 UTC) in particular implementing some seismic isolation of the cooling fans of the PR turbo and of the PR-BS link turbo.
5. switched off some un-necessary electronics in CEB, NEB, WEB (time list to be added)
6. between 13:05 UTC and 13:24 UTC: some TCS devices in TCS room and also on NI and WI TCS benches (see timing here)

The switch off #1 (HVAC) produced the disappearing of lines at 6.5Hz, 24.6Hz, 36.3Hz (photon calibrator) and 207Hz, as reported in origin of some lines

The switch off #2 (PR, BS and PR-BS turbo p. fans) produced the reduction of a bump around 45Hz in both DARM, MICH and PRCL, associated to the removal of a pair of seismic lines around 45Hz in the accelerometers at PR, PR-link, BS, PR B4-GHOST and SPRB-link: see Figure 1.

The switch off #3 (HVAC fan of remote scroll room) did not produce macroscopic effects in DARM, yet possible effects in ENV sensors have to be investigated.

The mitigation of the turbo pump fans did produce some effect (seismic peaks are still visible in ENV_B4_GHOST_ACC put about 10 times smaller). see VIM spectrogram in Figure 2. The Bump in MICH is still present but reduced: Figure 3.

The switch off #4 (other VAC cooling fans) did not produce a macroscopic effect in DARM (ENV sensors yet to be analyzed)

At 13:03 UTC the ITF was relocked, after the switch off #5 and #6 (electronics and TCS), and we noticed an increase of the 50Hz line and the bump around it. Figure 3.

Finally, a few minutes after 14:00 UTC, a seismic line at 92Hz disappeared from CEB sensors (see VIM spectrogram in Figure 4), correspondently the bump at  90Hz disappeared from DARM, MICH and PRCL: see Figure 5.  The switch off of the 92Hz source is not part of the controlled switch off. To be investigated.

Also to be investigated is the origin of the 50Hz extra noise. The actual plan is to perform a selective switch off of electronics devices to localize it. Possibly tomorrow.

Concerning the disappearence/reduction of 45Hz and 90Hz bump, this seems consistent with the hypothesis of scattered light from B4-GHOST and indications of the projections.

Looking through the spectrogram of LSC_DARM we see a few lines disappear during the A/C switch off. Most of these 'lines' are farily broad small bumps in DARM. Correlating the time of the disappearance of lines with the switch OFF times. We can make the following correlations

• @10:43LT 6.5Hz disappears (figure 1) -> 6.5Hz is a result of the UTA Hall A/C
• @10:38LT 24.6Hz disappears (figure 2) -> 24.6Hz is a result of the DAQ A/C
• @10:05LT 36.3Hz disappears (figure 3) -> 36.3Hz is the Photon Calibration Line 
• @11:01LT 207Hz disappears (figure 4) -> 207Hz is a result of the second splitter (inj ele-room) //could also be the UTA A/C (laser lab), seems to be very high in frequency though

Asside from these lines there seems to be no other obvious lines that disappear. It is worth noting that in these are all relatively small bumps, with the exception of the Photon Calibration LIne. 

I. Fiori first noticed a bump around 45 Hz (figure 1), which she recognized to be around the same frequency of the turbo pumps of the towers (elog 25104). This bump is consistent with a noise as a result of scattered light. It was then suggested to look at the B4_GHOST_ACC (supposed location of scattered light) around 45Hz to see if there is any apparent coupling. In fact we do see a line very similar to the lines associated with the turbo pumps from elog 25104. Then in an attempt to find out which tower is the source of the 45Hz line in B4_GHOST a coherence was run with all the central bulding tower accelerometers. Figure 2 shows the B4_GHOST_ACC is most coherent with the DT, suggesting the source of the line may be the DT turbo pump.

It is hard to say conclusively that the 45Hz bump seen in DARM is really correlated to the turbo pump, because of the non-linearity of scattered light. Tomorrow during the switch off the tower turbo pumps will be switched off to see if there is any effect in the sensitivity. It is worth noting that there is a similar bump seen in LSC_DARM at 92Hz, but there seemed to be no obvious correlation with any lines in B4_GHOST_ACC

As a further check of the removal of the drifting lines, I took the plot of LSC_DARM used before with the families marked and generated the plot from last night. Every line is gone...

Below is a link to the working list of lines with potential sources (currently up to 500Hz). This list can be edited and updated by anyone with the link. The idea is that once we find a potential source of a line it is posted not only in the logbook, but also updated in the list with the respective logbook entry number and the name of the editor.

https://docs.google.com/spreadsheets/d/1ZXZlyU1A7G2xxtGiqJa3LrVFr_vrB9OXbsAhH96_1Yk/edit?usp=sharing

For future reference the link to the list is posted to the EnvMon wiki page as well.

Note this list is derived from ER11a->ER11b, so any drifting line may have since moved to a different frequency.

We performed a few tests:

• Figure 1: with the ITF in LOW noise 1 we applied a DC current to all four WE mirror coils. The temperature probe at the coil measured an increase of about 0.25°C, while no measurable effect is sensed by the Temperature probe at the TCS RH. No drifting of the lines was observed during this period.
• Figures 2 and 3: we applied a current to WE ring heater (30V, as instructed by Viviana, which corresponded to 0.5A read on the driver display). The current was applied between 17:31UTC and 17:37UTC. The Temperature of the RH  rised promptly and kept rising also after we turned off the current. Overall the temperature of RH rised by about 0.5°C (we notice that also the temperature of the COIL probe rised but the trend is different: the COIL temperature having about a twice steeper rise while the RH was on then decreasing right after the switch off of the current. The RH temperature instead kept increasing for half and hour afterwards followed by a slow decrease). We observed a prompt (just a minute or so delay) drifting of the "WE-family" lines which crealry follows the temperature evolution of the RH probe. The NE and WI-NI family lines did not move. The frequency response we induced in the lines was of about 3 Hz / °C, which is less than what we measure in the interferometer (about 10 Hz / °C, see 38327).
• We switched off the SWEB air flow fans between 17:48 and 17:55 UTC. This did not produce an appreciable effect in DARM.
• At 19:50 UTC we unplugged the cable between the RH driver and the tower from the driver back panel (the driver is still on).

We also moved the magnetic probe named ENV_WE_RACK_MAG from the DSP rack to the DAQ box rack on its left: many peaks are visible, not coherent with DARM.

Finally we recovered the  ENV_WEB_MAG_V magnetometer (which was off since last Tues) by reconnecting the ADC cable which was found loose.

**EDITED 16:15LT 7/7/17 The original post had the spectrogram (figure 3) from the wrong day it has since been replaced**

Attached are figures 1 and 2 that show LSC_DARM, ENV_WE_RACK_MAG, and their coherence before and after the fan speed increase. There is clearly a change in noise in the MAG when the fans are at max, however there is little to no change in the coherence aside from the 50Hz harmonics. We'd expect to see a significant 'jump' in the frequency of the drifting lines, but as you can see in figure 3 (a known family1 group of drifting lines) these lines seem stable throughout the fan test. (There is a slight drift in the 125Hz line, but that occurs over the course of 24 hours and is unaffected by the test)

Looking through the entirety of the LSC_DARM fft there are no significant changes. Figure 3 is zoomed into the region where there was believed to be a change in the coherence (figure 2 in  38470), however upon further investigaiton the figure from 38470 there seems to be no change in coherence.

All that being said I wouldn't say this is 100% conclusive. The drift in lines we see (figure 4) is over the course of 10+ days, while this test was only an hour. Granted the fan speed was likely moving slowly over the course of 10 days and the speed was moved quickly within the hour, but there may be some coupling we are unaware of that retards the drift of lines in time. It may be worth trying this test again over the course of a few days?

Upon further investigation by Soumen Koley we've found another famliy of drifting lines. Running a brute correlation shows this family to be associated with the temperature in NE (38466).

Previously we had two families one correlated with the temperature in the WE and another with WI/NI (aka central building). Now we have a family for each building WE, NE, and central building (WI/NI). Attached are some plots in an attempt to visually classify the lines:

• Figure 1: All lines classified into families (including drifting and non-drifting)
• Figure 2: All drifting lines classified into respcetive famliy
• Figure 3: Family 1 drifting lines (WI/NI family) divided into families of lines that are correlated and anti-correlated with temperature
• Figure 4: Family 2 drifting lines (WE family) divided into families of lines that are correlated and anti-correlated with temperature
• Figure 5: Family 3 drifting lines (NE family) divided into families of lines that are correlated and anti-correlated with temperature
• Figure 6: Lines that drift in frequency and seem to disappear at times
• Figure 7: Stationary Lines

Also attached is the exhastive list (including some description of each line) written by Soumen from a spectrogram generated from ER11a->ER11b. It is worth noting that all the lines belonging to the families are very stable in amplitude through time. There are a few lines denoted  in the .xlsx (not belonging to the families) seem to either change in amplitude or entirely disappear from ER11a->ER11b, it may be interesting to further investigate these lines as well.

There have been tests to attempt to find the source of the drifting lines:

Test 1: Lengthening the F7 suspension (testing if it's an issue of the lengthening of the suspensions with as temperature increases)

              Prelimenary analysis shows no changes in frequency

Test 2: Changing the speeds of the fans in all the buildings

             Analysis to follow....

We have temporarely placed a magnetic probe

https://scientists.virgo-gw.eu/EnvMon/List/FluxGateMagnetometer/fluxGateMagn_manual.pdf

in between West End electronic suspension racks.

We see a forest of lines, probably all due to the many fans of the crates (eg. 36.5Hz, 73Hz ...); none of these seems having coherence with DARM (see first plot).

Nevertheless there are frequency regions, all around Mains harmonics, that have high coherence (see second and third plot, as example).

It would be better to do this test again when the interferometer will have a lower noise.

Attached are plots highlighting the lines previous noted to be moving in frequency. These figures are divided into 100Hz bands from 0 to 500Hz. The lines with the box outline correspond to the family potentially associated with WI, and the filled box is the family potentially associated with the WE. It's worth noting the initial list of lines generated was only from 0 to 500Hz, so there may be more lines in the families above 500Hz.

Attached are two files corresponding the families of drifting lines seen throughout ER11/a and /b. These families have been associated qualitatively with WE and WI temperature sensors respectively. However, a brute correlation is being run with all temperature channels to provide a quantitative confirmation.

A visual inspection of DARM drifting lines over a 10 day period seems to confirm the correlation with in-tower temperature sensors that was found by the brute force correlation analysis done over ER11/a.

Attached are some spectrogram (**) shots of the forest of drifting peaks: Figure 1, Figure 2

We tentatively found two families of peaks:

• peaks whose frequency drift correlates (or anti-correlates) with the temperature of CEB towers. Among which the one that qualitatively matches best are sensors at the level of F7 or RH inside WI tower  (Figure 3, Figure 4)
• peaks correlating (or anti) with the  WE tower  temperature sensors (Figure 5). A deeper insight among WE tower temperature sensors shows a visually good correlation with the TE sensor at the WE ring heater (Figure 6)

These qualitative observations have to be verified with correlation analyses.

--------------------------------------------------------------------------------------------

(**) this is magic Didier's command to get this spectrogram:

$ROOTROOT/$UNAME/bin/root.exe -l /opt/w3/vim/spectro/spectro_auto.C\(\"LSC_DARM\",1181180000,864000,0.1,0,500,0,0,0\)

plus you need to have the file .rootrc in your home dir:

cat ~fiori/.rootrc
Unix.*.Root.DynamicPath:    .:$(ROOTSYS)/$(UNAME):$(ROOTSYS)/lib:$(FRROOT)/$(UNAME)
Unix.*.Root.MacroPath:      .:$(ROOTSYS)/macros:$(FRROOT)/root
Rint.Logon:              $(FRVROOT)/root/FrvLogon.C
Rint.History:            $(HOME)/.root_hist

An analysis of data from the probes placed by F. Paoletti has been done. The first figure shows the coherence between LSC_DARM and ENV_SDB2_ACC_Y. As you can see there is very little coherence aside from the 50Hz harmonics. Figure 2 zooms in around ~24 Hz where we do see good coherence and a similar structure in both DARM and the ACC. However, figure 3 shows a similar structure in all ACC in the central building, meaning this is not localized to SDB2, but rather the central building. It is worth noting that the line seen at ~24Hz in LSC_DARM is very small.

The last figure looks at ENV_SDB2_MAG. When compared with LSC_DARM there seems to be very little coherence other than the 50Hz harmonics.

We further analyzed the effects of the turbo pumps throughout the central building.

• Brown- Both turbo pumps on
• Pink- SR Link turbo pump off detection tower pump on
• Blue - Both turbo pumps off

Figure 1: Everything fully unzoomed... There is a significant decrease in seismic noise when the SR_Link turbo pump is switched off. We believe the seismic noise is a result of the SR_Link pump sitting on the floor. There is no noticible change in the seismometers when the DT pump is switched off. Another intersting area is the DT_ACC is around 2400Hz, this if further investigated with figure 4.

Figure 2: Zoomed on the 456Hz comb which is only present when the SR_Link TP is switched on as shown in previous posts to have been removed from LSC_DARM.

Figure 3: Zoomed around 401Hz the believed frequency of the DT turbo pump, but we see no difference when the pump is on or off.

Figure 4: Zoomed around 2400Hz there is a significant decrease in noise only in the DT_ACC when the DT turbo pump is switched off. It is worth noting that this a localized effect and is not present in the other ACC further from the turbo pump. Figure 5 further analyzes the effects in LSC_DARM around 2400Hz and there seems to be no changes from when the pump is on or off.

Figure 6: A plot of LSC_DARM before and after the turbo pump switch off shows no broadband changes and the only observable line to disappear is the 456Hz comb.

--Conclusion--

Other than the 456 comb, which is caused by the SR_Link turbo, there seems to be very little effect on LSC_DARM by either turbo pump.

We also saw good coherence between LSC_DARM and ENV_BS_ACC_Z at 456Hz (see figure). While this is the only line that disappeared in LSC_DARM further analysis of any boradband changes us underway.

At about 11:30 LT we added these temporary probes at SDB2:

• ENV_SDB2_MAG, magnetometer positioned underneath SDB2 vacuum tank, picture 1
• ENV_SDB2_ACC_Y, accelerometer attached with double-sided tape at the link towards SDB1, see picture 2