Same plots with the addition of more recent data from February 16÷22 shown in light yellow, which confirm earlier data with 15 cm aperture (February 10÷15).
The last three plots show noise vs dark fringe power, and are are limited to data from February 16÷22 with DCP frequency above 380 Hz. Strain noise and net RIN are plotted vs total and carrier dark fringe power, where the only difference is the ~65 mW constant power from RF sidebands.
From a check of the etalon control filter actually running in the past days, we understood that the we were overestimating a bit the gain in the model. The correct open loop is shown in the attached plot, and the conclusion are about the same.
This morning I've changed the controller for the Etalon.
the new controller is:
# Etalon north boost 03
ACL_FILTER_SET Etalon_north_boost_03 1 21210 1 10
ACL_FILTER_POLES Etalon_north_boost_03 2e-05 0
ACL_FILTER_ZEROS Etalon_north_boost_03 2e-06 0.5
ACL_FILTER_POLES Etalon_north_boost_03 0 0
# Etalon west boost 03
ACL_FILTER_SET Etalon_west_boost_03 1 21210 1 10
ACL_FILTER_POLES Etalon_west_boost_03 2e-05 0
ACL_FILTER_ZEROS Etalon_west_boost_03 2e-06 0.5
ACL_FILTER_POLES Etalon_west_boost_03 0 0
ITF found in Commissioning mode and LOCKED_ARMS_BEAT_DRMI_3F state; activity on NI RH test ongoing.
The activity went on up to 19:36 UTC; ITF left in LOCKED_ARMS_BEAT_DRMI_3F as requested by the commissioning team.
SUSP
At 15:45 UTC, after an unlock, the WI LC were opened by the guardian; properly closed.
The goal of today’s activity was to verify whether the ITF can survive the switch-on of the NI ring heater.
The test procedure is described in the Git issue created by Michal:
https://git.ligo.org/virgo/commissioning/commissioning-tasks/-/issues/179
At 05:52 UTC, the ITF was manually unlocked.
A relock in CARM NULL 1F was requested and reached at 06:04 UTC.
At 06:40 UTC, the NI outer ring heater power was decreased by 10% (STEP 1).
The NI etalon loop was switched off at 08:43 UTC.
After that, the NI RH was switched on at 08:47 UTC (V = 6.1 V, P = 1 W).
ITF unlocked at 10.28 UTC.
A relock was achieved on the first attempt at 10:39 UTC, but ITF immediately unlocked due to a loop oscillation.
In agreement with ISC, we decided to park the ITF in CITF and wait for the transient to subside.
At 13:18 UTC, the ITF was successfully relocked in CARM NULL 1F, with no changes applied on the ISC side.
The ITF remained locked for 1 hour before unlocking at 13:58 UTC.
The ITF then unlocked multiple times, all exhibiting the same behavior.
It was therefore decided to decrease the DAS outer ring power by an additional 10% (STEP 2) in order to recover the 12 MHz sideband, with the aim of improving the stability of the ITF.
This step was performed while the ITF was down at 14:49 UTC. An increase in the 12 MHz sideband was observed, however, this did not have any impact on the CARM NULL 1F behavior, where the same type of unlock was observed.
In order to further increase the 12 MHz sideband, an additional 10% reduction was applied at 16:15 UTC (STEP 3).
None of these steps allowed the ITF to remain locked in CARM NULL 1F for more than two minutes. All unlocks showed the same behavior.
The behavior of the main signals is shown in the attached figure.
After that, the RH was switched off at 18:23 UTC, and the original DAS settings were restored at 19:35 UTC.
The activity concluded at 19:36 UTC.
ITF Found in CARM_NULL_1F State and COMMISSIONING Mode with AUTORELOCK_FAILSAFE Engaged.
All times are UTC.
05:51:51 Manual ITF Unlock to relock ITF in CARM_NULL_1F to do planned Task 179 (Nardecchia).
The work on TCS NI RH is ongoing (Nardecchia, Casanueva, Derossi).
08:43:34 Etalon NI Loop turned STOP via VPM > ISC > LSC_Etalon_Acl (Operator).
08:47:35 NI RH ON (Nardecchia).
10:27:52 ITF Unlock from CARM_NULL_1F during TCS transient.
10:38 ITF in CARM_NULL_1F.
10:41:17 ITF Unlock from CARM_NULL_1F.
10:48 ITF in LOCKED_ARMS_BEAT_DRMI_3F as indicated in Task 179 if the ITF was unable to relock in CARM_NULL_1F.
12:30 ITF Unlock from LOCKED_ARMS_BEAT_DRMI_3F.
13:03 Manual ITF Unlock (Operator, Casanueva, De Rossi).
13:05 Requested CARM_NULL_1F (Operator, Casanueva, De Rossi).
13:16 ITF in CARM_NULL_1F.
13:58:54 ITF Unlock from CARM_NULL_1F.
14:09 ITF in CARM_NULL_1F.
14:11:45 ITF Unlock from CARM_NULL_1F.
14:23 ITF in CARM_NULL_1F.
14:25:44 ITF Unlock from CARM_NULL_1F.
14:36 ITF in CARM_NULL_1F.
14:38:29 ITF Unlock from CARM_NULL_1F.
14:45 ITF in LOCKED_ARMS_BEAT_DRMI_3F.
ITF left in COMMISSIONING Mode and LOCKED_ARMS_BEAT_DRMI_3F State. Activity on NI RH (Task 179) is still ongoing.
In the latest days the etalon control has been working in a stable condition on both the arms, allowing a new attempt to fit the data with a model. I'm taking into account the period FEB17 14:00 - FEB23 9:00.
The reconstruction of the temperature measured by the in-loop temperature sensor, shown in fig 1 and fig 2, is based on two inputs and two model of response. One input is the correction applied by the loop: this is the square of the channel LSC_Etalon_NI/WI_HB_cmd. The signal is filtered using a model of the response of TM temperature to the power applied by the heating belt: two simple poles at 2.4e-6 Hz, gain 0.005 @ 0. The second input is the power in the arms, normalized in order to be 1 when the arm is locked in CARM NULL. The signals is filtered using model of the response of TM temperature to the power applied by the YAG: simple pole at 2.2e-6 Hz, gain 0.23 @ 0, for NI; simple pole at 4e-6Hz, gain 0.23 @ 0, for WI. The third component of the temperature trend is genereted by the environment, but this is unknown. Here we assume a constant variation and we fit a slope which optimizes the reconstruction. The parameters of the models used for this fit are slightly different from the ones used previoulsy, but the model of the loop does not change irs behaviour in a relevant way.
The plots show separately the effect of the YAG: it is clear the frequent long down time of the ITF was the main source of inaccuracy of the control. There is also the evidence that the loop is ineffective to reduce the temperature oscillation due to lock/unlock period lastig several hours, which was the typical condition we had last week. This behaviour of the control is foreseen by the loop model (fig 3): at the UGF, 25uHz (period 12 hours), the fase margin is low and the effect of the loop is an amplification of the noise by more than a factor of 2.
The phase margin is limited by the presence in the controller of a pole at 10 uHz. When the controlled was changed from as simple integrator to a more complex design, a safe choice in terms of roll-off was done, having no idea about the possible impact of an increased high frequency component in the correction signal. I think we should try to move upwards the pole. In fig 3, we see the effect in the phase and the suppression if the pole is set at 20 uHz: the overshoot is reduced and moved to higher frequency, where the amplitude of temperature oscillation due to lock/unlock would be lower (shorter period means less time to accumulate temperature variation). The new filter has a gain 4000 @ 1Hz, instead of 2200.
From a check of the etalon control filter actually running in the past days, we understood that the we were overestimating a bit the gain in the model. The correct open loop is shown in the attached plot, and the conclusion are about the same.
Attempting SRCL detuning directly in LN3, as there was not much
automatic adjustment during the last test in LN2, just a 10% gain
reduction for the MICH loop
18:45 UTC turned off SRCL SET servo
18:46 UTC SRCL SET -10 with a 5min ramp
19:09 UTC SRCL SET -15 with a 5min ramp
19:32 UTC SRCL SET -20 with a 5min ramp, DCP at 350Hz
19:56 UTC SRCL SET -25 with a 5min ramp
20:16 UTC SRCL SET -30 with a 5min ramp, DCP at 300Hz
This time the SRCL steps correspond to larger reductions in the DCP
frequency as measured by Hrec, and larger phase changes in the 491Hz
line on B1.
In this configuration where the 1/f^{2/3} noise is reduced by the SRCL
detuning squeezing could have a major contribution to the BNS range,
by choosing the phase to squeeze the noise between 50Hz and 300Hz, and
give up on higher frequencies, where anti-squeezing will be added.
Leaving the inteferometer for the night in the last step, with SRCL detuned.
Figure 1. Shows the trend of many channels during the steps,
Figure 2. Calibration seems to be understimating the noise by up to 4% in the bucket between 100Hz and 150Hz. So the real BNS range may be lower by 2Mpc, at 50Mpc instead of the 52Mpc recorded in the data.
Figure 3. The optical spring seems to be around 9Hz. And the B1p power decreases from 190mW to 150mW.
Actually the used signal is DARM_CORR_raw.
Figure 1. For an additional small improvement in the SNR of these dither signals one could use the LSC_DARM_CORR_raw channel instead of LSC_DARM_CORR, this is the DARM correction before the addition of the noise of the MICH feedforward, so before with add a motion of the arms to compensate for the motion of BS induced by the MICH sensing noise.
The signals used in loop at low frequency for setting the alignment working point of the arms are supposed to put the optical axes in the center of the test masses, minimizing the coupling from rotation to length. They obtained demodulating the arm locking signals at the frequency the lines injected on all the angular dofs, in the frequency region between 7 and 9.5 Hz. The strategy works since the beginning of AdvVirgo, but the quality of the signals worsened in O4 commissioning: having also CARM SLOW locked, the arm locking signals started to contain some additional noise in the region of the angular lines. In order to have enough SNR on the dither signals, we used to keep the amplitude of the lines higher than in O3. This is one of the reasons of the big increase of DARM level below 10 Hz, but also a relevant worsening of Hrec between 10 and 20 Hz, because the second harmonic of the lines appeared above the other noises.
The problem has been fixed today, changing the signals to demodulate: instead of the locking signals of each arm, now we use only DARM correction. This can be done only at CARM NULL: at the beginning, when each arm is locked by its own signal, the old strategy needs to be used. A flag in the DSP has been implemented in order to recognize the status and switch the strategy as soon as it is possible.
The noise of the demodulation has been reduced as expected, giving the possibility to reduce the amplitude line (0.005 -> 0.002). In fig 1 and fig 2 we see the effect on Hrec and DARM. The unlock reverted this setting: the amplitude needs to be updated in the automation.
Yesterday a preliminary test has been performed: almost all the arm angular lines have been turned off. Only the dither used for COMMp have been left, otherwise there was a drift not compatible with the low noise operation. In fig 3 and fig 4 we can see that the level of Hrec and DARM at low frequency was a bit lower than the one obtained today: it means that there is still a margin of improvement. In order to reduce more the amplitude of the lines, we should try to remove from DARM some low frequency noise coming from the coupling of MICH and SRCL noise.
We also can notice that the level of noise above 20 Hz, dominated by a mystery 1/f^4, did not change at all reducing the noise coming from the lines.
Figure 1. For an additional small improvement in the SNR of these dither signals one could use the LSC_DARM_CORR_raw channel instead of LSC_DARM_CORR, this is the DARM correction before the addition of the noise of the MICH feedforward, so before with add a motion of the arms to compensate for the motion of BS induced by the MICH sensing noise.
Actually the used signal is DARM_CORR_raw.
Two more figures to illustrate the subtraction of quantum and thermal noise on the data set from February with 15 cm aperture. The plots show the noise vs DCP frequency and optical gain respectively.
- black points: BRMS Hrec noise in 110÷133 Hz
- blue points: after quadratic subtraction of quantum and thermal noise
- purple points: BRMS quantum noise in 110÷133 Hz
- brown points: BRMS thermal noise in 110÷133 Hz
For better clarity I attach the same plots with points grouped in similar colors corresponding to the size of SRM diaphragm:
- green points: 34 cm aperture (light green October 2 2025; dark green December 8 2025)
- bue points: 19cmx15cm aperture (dark blue January 14 2026; light blue January 23 2026)
- yellow points: 15 cm aperture (February 10 2026).
I removed the points from January 8, as Hrec was not working properly at that time.
I included two plots with the net Hrec noise after subtraction of quantum and thermal noise.
Data with 15 cm aperture offer a good chance to check the effect of contrast defect on broadband noise, as dark fringe carrier power varied over one order of magnitude. The seventh plot shows dark fringe carrier power versus DCP frequency. Dark fringe carrier power is computed as the product of calibrated B1p DC power and the fraction of carrier from B1p phase camera. The last two plots show the noise dependence with dark fringe carrier power for the data from February with 15 cm aperture. Broadband noise in power units changes by ~40% for a factor 10 change in dark fringe carrier power.
One thing to keep in mind is that during O4 the angular control sensing noise was a significant contribution to the low frequency noise, especially for DIFFp TX, probably because the SR misalignment is increasing angular noise coupling. It was also often coherent with h(t).
An example of the angular noise budget from O4 https://logbook.virgo-gw.eu/virgo/?r=65286 can be compared to the case when SR is aligned in January this year https://logbook.virgo-gw.eu/virgo/?r=68516
ITF found in LN3_Aligned.
At 7:06 UTC I put the ITF in Maintenance mode; at 8:11 UTC I manual unlocked the ITF to perform the planned tasks.
Below the list of the activities communicated in control room:
- DAQ Maintenance activities;
- cleaning of the experimental areas by external firm;
- closing of the SQZ shutter on SDB1 (Zendri);
- civil works measurements at terminal building (Paoli);
- Replacement of current sensors for DET HVAC humidity generator ;
- work at NE/WE suspension (Ruggi);
- from 11:00 test on WE UPS;
- tasks performed by the operator:
- lock and scan of the OMC in single bounce;
- TCS - chillers checks and refill;
- TCS - thermal camera images reference;
- TCS - power checks:
| CH [w] | Outer Ring [w] | Inner Ring [w] | |
| WI | 0.334 | 0.26 | 0.045 |
| NI | 0.58 | 0.565 | 0.08 |
After the maintenance the ITF was relocked in LN3_Aligned and Paolo started to work; the civil works measurements at terminal building are still in progress.
LSC
Under request of Maddalena I changed the Etalon setpoint:
2026-02-24 09h44m31 UTC berni 'NI:NI Set point [NI_RH_SET=18.695,rampTime=10800]' sent to LSC_Etalon_Acl
2026-02-24 09h44m54 UTC berni 'WI:WI Set point [WI_RH_SET=18.802,rampTime=10800]' sent to LSC_Etalon_Acl
Small SRCL detuning that changes the bandwidth of the detector from 430Hz to 380Hz seems to reduced the 1/f^{2/3} noise and improve the sensitivity. A question is what a detuning like that will do to squeezing. I have used GWINC to try to answer that question. The detuning correspond to about 7 degree of SRCL in units used by GWINC.
Figure 1 shows the result of GWINC for the current configuration, SR aligned and no squeezing, as a solid black line. In addition I have added the results of several other configurations. In green with SRCL detuned by 7 degrees and frequency independent squeezing with the phase adjusted by hand to make an improvement in shot noise between 50Hz and 300Hz. In blue the ideal case of SRCL tuned and FIS. And last in red a made up model of the SR misaligned configuration of O4 by adding losses and changing the SR reflectivity.
With SRCL detuned we cannot have squeezing at ~100Hz and at high frequency at the same time, trying to do frequency dependent squeezing with a detuned filter cavity doesn't help either, as the current FC is too narrow band for that case. But by adjusting the phase of FIS one can get in the 50Hz-300Hz roughly the same performance as for the FIS in the SRCL tuned configuration, and loose sensitivity at high frequency compared to O4 only above 600Hz.
Without squeezing there is not difference in shot noise between the SRCL tuned and detuned configuration below 600Hz, and above the shot noise is ~10% higher for the SRCL detuned configuration due to the lower bandwidth (380Hz vs 430Hz)
This morning I replaced two current transducer model LEM AP 200 B420L with the model LEM AP 50 B420L (50A full scale value).
Those sensors measure the RMS value of the current used by DET HVAC humidity generators.
As requested by ENV team, names have been changed from INF_CEB_HUM1_CURR to INF_DET_HU_SPARE_CURR and
from INF_CEB_HUM2_CURR to INF_DET_HU_MAIN_CURR.
The Virgo sensitivity in the 20÷30 Hz band is limited by an unknown noise source with 1/f^4 spectrum. This was already the case during O3. The attached plots show how the noise level changed during O3 and O4. The first plot is the Hrec spectral density at 30 Hz during the two science runs, with O3 displayed in red points and O4 in black points. Data in the few months before the start of O4 are also included. Each points represents an average over 1000 s.
The second plot shows the Hrec spectrum at four different times, representative of high and low noise (@30 Hz) in O3 and O4 as highlighted by coloured arrows on the first plot. A few comments: the noise in the 20÷30 Hz band
1) was generally lower in O3 than in O4;
2) changed by ~20% during both science runs, and in the best conditions during O4 was similar to the worst conditions during O3;
3) was lowest at the end of 2023 in the O4 configuration.
One thing to keep in mind is that during O4 the angular control sensing noise was a significant contribution to the low frequency noise, especially for DIFFp TX, probably because the SR misalignment is increasing angular noise coupling. It was also often coherent with h(t).
An example of the angular noise budget from O4 https://logbook.virgo-gw.eu/virgo/?r=65286 can be compared to the case when SR is aligned in January this year https://logbook.virgo-gw.eu/virgo/?r=68516
The deployment of the last Fd version (v8r65p1) and Fr version(v8r49p0) has been done for
- all the Fm servers using Fm v4r10p4
- all the FFLMoni servers using Fd v8r65p1
- all the FdIODy servers using Fd v8r65p1
- on the storage-stol01 part for the RAW stream
- on the storage-stol03 part for the Trend100, the Trend, the RDS and the RAW_FULL streams
The operations has been performed between 2026-02-24-05h26m28-UTC and 2026-02-24-06h04m32-UTC
During the maintenance period, in agreement with the commissionig coordinators, we completed the Fd v8r65p1 deployment
- on the FdIO base servers, Fbs, FdIOServer, FdIOMerger, MoniServer, running on olserver52 and olserver53
- on the TolmFrameBuilder servers running on the rtpcs
- Operations performed between 2026-02-24-09h12m05-UTC and 2026-02-24-09h34m44-UTC
At the beginning of the shift I found the ITF in LN3_ALIGNED, planned activity of WI reallocation filter, CARM->DARM coupling, other lowFreq investigations ongoing (Pinto, Ruggi)
15:05 UTC Etalon setpoints changed: time ramp 10800 s, NI set to 18.68, WI set to 18.8
17:39 UTC unlock
18:52 UTC ITF in LN3_ALIGNED
19:30 UTC end of commissioning activity
22:00 UTC end of shift, ITF left in LN3_ALIGNED
ITF found in LN3_Aligned with Alain already working from remote at the planned activity Acl update with v1r39p17; activity concluded at around 8:00 UTC.
After that started the planned activity of WI reallocation filter, CARM->DARM coupling, other lowFreq investigations; still in progress at the end of the shift.
Acl update with v1r39p17
As described here , the first block has been done this morning in 2 parts
- First part
- Second part
- SNEB_rtpc Acl servers update
- operations performed between 2026-02-23-07h32m01-UTC and 2026-02-23-07h42m37-UTC
- After these operations
- the SNEB LC loops, the SNEB SBE loops and the NEB NCal loops were successfully closed
- The NEB_NNC_Tilt server running parameters (set poin and gain) were restored but the loop remains open
- The ITF has been successfully locked upto LOW_NOISE_3_ALIGNED
Update with Fd v8r65p1
- the storage-stol02 's Fd servers have restarted with latest v8r65p1 Fd version
- operations performed between 2026-02-23-08h03m58-UTC and 2026-02-23-08h09m26-UTC