Today (Jul 5) around 18:00 UTC I have set back the ASC_SR_TX_SET to 0.022 with a 2 hours long ramp. As it was reset a few hours before by the unlock, and the BNS range has been lower by a few Mpc since that relock.
Following the analysis done in the past days, which showed correlation between B1p DC (and the B1p beam shape) and the BNS range, we decided to make a scan of the SR tx in order to try to drive this effect.
In the Figure 1 a trend of the test is visible and we can say that for SR tx offset of ~ 0.022
- the B1p DC is minimum
- the OG is lower than the 0 position
- the frequency noise coupling is higher than the 0 position
but the BNS range is maximum; this may lead to a lower mistery noise? to be understood
We leave for this lock a ramp going to the optimal point for SR tx (which will be reset at the next unlock)
Today (Jul 5) around 18:00 UTC I have set back the ASC_SR_TX_SET to 0.022 with a 2 hours long ramp. As it was reset a few hours before by the unlock, and the BNS range has been lower by a few Mpc since that relock.
Figure 1During the step of the ring heater on June 20 the modes order 7 and 11 have moved in opposite direction, while simple simulations of a single arm would expect for all modes to move in the same direction in the arm FSR, with the change in the FSR smaller for very high order modes as they start to be clipped in the arm.
I have reused the Finesse 3 simulation that Augustin has been developping for scattered light (https://git.ligo.org/augustin.demagny/finesse-simulation-04) to look instead at the CARM to B1 transfer function. These simulations include a misalignment of SR by 2urad. I have added appertures of 170mm in radius to add realistic losses to very high order modes, and simulated with the max TEM number of 10.
Figure 2 shows the result around the first FSR for a nominal end mirror radius of curvature of 1683 meters, while Figure 3 shows the same for 1679m radius of curvature for both end mirrors. It shows the same effect that the two HOM spanning around the first FSR of the arm move in opposite directions, which is counterintuitive, but similar to what has been observed experimentally on figure 1. Note that peak in the middle of the figure is slightly above 50kHz, while the analytical calculation predicts an FSR with a spacing of 49'969 Hz (slightly below 50kHz). Does this mean that the peak we see is not the FSR, but something else? What is strange is that it remains in the simulation when the max TEM is reduced to 0, so it cannot be the order 9 mode. The side modes around 44kHz and 56kHz, also remain present for max TEM 1, and disappear for max TEM 0, so this is actually two images of the order 1 mode, and then it makes sense that they move in opposite direction when the RoC is changed.
Figure 4 and 5 shows the frequency range of modes order 1-3, respectively for 1683m on EM and 1679m. These modes behave as expected, moving to lower frequency for a shorter radius of curvature.
Figure 6 and 7 shows the frequency range of modes 4-6, the picture is confused with more modes than one would expect. Some of the modes move to higher frequency for a shorter RoC, while others move to lower frequency. In particular the mode around 33kHz, which could be the order 6 mode moves to higher frequency, so the opposite of the low order modes.
I will add the modified code into git under 'high order mode - CARM.ipynb'.
Figure 1 shows the spectrum for the 56MHz TEM00, both USB and LSB, normalized to be units of RIN, and with the contribution of shot noise and PD electronic noise quadratically subtracted. The spectrum looks the same for both sidebands, and there was no significant change during the 2.5 hours separating the repeated measurement of the USB, which is a sign that the alignment has not drifted in the meantime, the beam jitter peaks are small in all cases. The requirement (VIR-1225B-19) for the RAM servo is 2.5e-8 V/V/rtHz at 100Hz, which should correspond to RIN of the sideband of 5e-8 1/rtHz. So probably RAM is a significant part of that spectrum, but this would need to be checked further, as the RAM servo was better than the requirements from what I remember.
Figure 2 shows the coherence with B1s, it is substantial at ~30%. It means that the sidebands are a major contributor to the dark fringe fluctuations in the 100Hz and 500Hz. The coherence for the red trace is smaller as there is a loud glitch on B1s dominating the spectrum during that measurement.
There is no broadband coherence of the sideband with h(t).
Figure 3 shows the series of measurements with the first measurement being the 56MHz USB, and the last measurement being also the 56MHz USB
Figure 4 shows the spectrum of the higher order modes from 2 to 10, with two measurements for the mode order 9 (vertical and horizontal peaks). The PD electronic noise is 1e-8 mW/rtHz, and for the peaks with low power, the spectrum at 1kHz is very close to that limit. There is about a factor 10 difference in power between the brightest mode (order 3) and the dimmest (order 10). The B1 photodiodes have a lower dynamic (~5mW), which would be sufficient as the brightest mode has 1mW in this scan, and have a lower electronic noise at 6.5e-9, so that could be a simple solution to gain a factor sqrt(2) in SNR for these measurements.
Figure 5 shows the same normalized in units of RIN and with the shot noise and electronic noise quadratically subtracted. In these units the noise level is more similar between the modes, the beam jitter peak height varies by one order of magnitude between different modes. There could be a similar level (within a factor 2) of noise that has a power-law shape around 100Hz, with a slope between 1/f and 1/f^{2/3}. That level of noise is a factor 2 higher than what is measured with the 56MHz sideband, so it is not intrinsinc to the measurement process, but HOM may be more affected than a TEM00 mode by beam jitter.
Figure 6 shows that none of those mode is coherent with B1s down to 0.1% level, which is surprising. It tells us that the EDB OMC is adding new information that is not accessible using only B1s, but it is unclear what that information means. There is coherence in the 20-30Hz band where the SRCL and MICH lines are in LN2 and at ~156Hz where the mechanical mode of SIB1 is.
Figure 7 shows the coherence with h(t), there is no broadband coherence to the 0.1% measurement noise level.
Figure 8 and 9 shows mode 3 and 5 which have the powerlaw slope the most clear, as they have high power and low amplitude for the beam jitter peaks
Figure 10 and 11 shows mode 2 and 4 which have a significant lower noise than modes 3 and 5
/users/mwas/OMC/RINanalysis_20250703/
Today I have locked the EDB OMC again on HOM in LN2. This time without realigning the OMC after the first initial realignment. There is no coherence with any of the HOM that I have tried.
Initially the EDB picomotor process could not be reached. I must have forgotten to switch off the server last week, and it lost connection on Sunday. Piernicola and Bernardo turned on and off the picomotor manualy, which restarted the connection.
aligned EDB OMC on 56MHz USB, first to maximize the power, then ~50
picomotorse in horizontal to reduce the jitter peaks
16:47 UTC (5min) - locked on 56MHz USB
when going through carrier TEM00 there is ~0.1 mW on it. The 6MHz TEM00 are much smaller, around 0.02mW.
Adjusted demodulation phase to put all the signal in the i quadrature, this must have increased the gain by factor 3 or 4 (and the calibration)
17:09 UTC (5min) - locked on carrier TEM00 (very noisy and low power), not sure if the locked actualy worked or just stabilized in temperature around the mode
Reduced the gain by factor 2 (and calibration of error signal, still
probably wrong by factor 2) as the loop was oscillating when locking
on TEM00 56MHz LSB
17:19 UTC (5min) - locked on 56MHz LSB
order 1 mode of 56MHz sideband has about 0.05mW of power, 10 times less than the TEM00
17:30 UTC (5min) locked on carrier order 2 mode, mostly horizontal
17:40 UTC (5min) locked on carrier order 3 mode, 1-vertical, 2-horizontal
17:50 UTC (5min) locked on carrier order 4 mode, 4-vertical, 0-horizontal
17:59 UTC (5min) locked on carrier order 5 mode, 0-vertical, 5-horizontal
18:07 UTC (5min) locked on carrier order 6 mode, 1-vertical, 5-horizontal
18:16 UTC (5min) locked on carrier order 7 mode, 1-vertical, 6-horizontal
18:26 UTC (5min) locked on carrier order 8 mode, 1 vertical, 7 horizontal
18:35 UTC (5min) locked on carrier order 9 mode, 9 vertical, 0 horizontal
18:46 UTC (5min) locked on carrier order 9 mode, 1 vertical, 8 horizontal
18:56 UTC (5min) locked on carrier order 10 mode, 0 vertical, 10 horizontal
19:06 UTC (5min) locked on 56MHz USB TEM00, alignment is still good
Figure 1 shows the spectrum for the 56MHz TEM00, both USB and LSB, normalized to be units of RIN, and with the contribution of shot noise and PD electronic noise quadratically subtracted. The spectrum looks the same for both sidebands, and there was no significant change during the 2.5 hours separating the repeated measurement of the USB, which is a sign that the alignment has not drifted in the meantime, the beam jitter peaks are small in all cases. The requirement (VIR-1225B-19) for the RAM servo is 2.5e-8 V/V/rtHz at 100Hz, which should correspond to RIN of the sideband of 5e-8 1/rtHz. So probably RAM is a significant part of that spectrum, but this would need to be checked further, as the RAM servo was better than the requirements from what I remember.
Figure 2 shows the coherence with B1s, it is substantial at ~30%. It means that the sidebands are a major contributor to the dark fringe fluctuations in the 100Hz and 500Hz. The coherence for the red trace is smaller as there is a loud glitch on B1s dominating the spectrum during that measurement.
There is no broadband coherence of the sideband with h(t).
Figure 3 shows the series of measurements with the first measurement being the 56MHz USB, and the last measurement being also the 56MHz USB
Figure 4 shows the spectrum of the higher order modes from 2 to 10, with two measurements for the mode order 9 (vertical and horizontal peaks). The PD electronic noise is 1e-8 mW/rtHz, and for the peaks with low power, the spectrum at 1kHz is very close to that limit. There is about a factor 10 difference in power between the brightest mode (order 3) and the dimmest (order 10). The B1 photodiodes have a lower dynamic (~5mW), which would be sufficient as the brightest mode has 1mW in this scan, and have a lower electronic noise at 6.5e-9, so that could be a simple solution to gain a factor sqrt(2) in SNR for these measurements.
Figure 5 shows the same normalized in units of RIN and with the shot noise and electronic noise quadratically subtracted. In these units the noise level is more similar between the modes, the beam jitter peak height varies by one order of magnitude between different modes. There could be a similar level (within a factor 2) of noise that has a power-law shape around 100Hz, with a slope between 1/f and 1/f^{2/3}. That level of noise is a factor 2 higher than what is measured with the 56MHz sideband, so it is not intrinsinc to the measurement process, but HOM may be more affected than a TEM00 mode by beam jitter.
Figure 6 shows that none of those mode is coherent with B1s down to 0.1% level, which is surprising. It tells us that the EDB OMC is adding new information that is not accessible using only B1s, but it is unclear what that information means. There is coherence in the 20-30Hz band where the SRCL and MICH lines are in LN2 and at ~156Hz where the mechanical mode of SIB1 is.
Figure 7 shows the coherence with h(t), there is no broadband coherence to the 0.1% measurement noise level.
Figure 8 and 9 shows mode 3 and 5 which have the powerlaw slope the most clear, as they have high power and low amplitude for the beam jitter peaks
Figure 10 and 11 shows mode 2 and 4 which have a significant lower noise than modes 3 and 5
/users/mwas/OMC/RINanalysis_20250703/
The point heating with IPATSiA has changed slightly the position of the HOM in the FSR.
Figure 1 shows the order 2 peaks, these move up in frequency slightly, add additional peaks at higher frequency either appear or become more prominent
Figure 2 shows the order 6 peaks, these move down slightly, and additional peaks at lower frequency either appear or become more promintent .
Figure 3 and 4 shows the same as an ASD, comparing the time with IPATSiA off (purple) and on (blue).
The fact that additional peaks appear can make sense, adding a heatpoint will make the mirror astigmatic, as the heatpoint is not perfectly centered.
The fact that the order 2 and 6 modes move in opposite directions is surprising. The same had happened when the ring heater power was changed https://logbook.virgo-gw.eu/virgo/?r=67078 . I do not have an explanation for this.
Figure 1. In the last few hours there was two glitches where the B1p power increased by a factor 2 for less than 1 second.
Figure 2 shows the trend over the last two days, and there is at least two more glitches like that. For example on June 25 just before midnight.
The glitches are due to events on WE marionette, involving the actuation on TX and TZ. A step is visible on both the correction; it induces a temporary misalignment, well visible on the optical levers. One can easily find a few of them going through the recent data.
There could be one vertical actuator which changes slightly its gain: the actuators are disposed in X configuration, which means that all are equally involved in the control of TX and TZ. The effect could be amplified by the fact that, after the mirror replacement, the correction needed to align TX remained a bit high (4 V). It can easily zeroed by means of the motors. It would be worth to do it and see if the glitches are mitigated.
In case they are not, it would be more difficult to exclude something more worrying, like a mechanical glitch, but I see it quite irrealistic.
It is hard to imagine how a glitch at ~400 can be created out of a calibration noise channel that usually contains only small numbers. One possibility is if the issue is that the channel order become wrong for a few samples. So that the LSC_B8_DC (third channel in the packet) is received instead of LSC_CAL_WE_MIR_Z_NOISE (thirteenth channel in the packet). LSC_B8_DC has a mean value of ~535, so that could create a value around 400, if it is wrong for 3 out of 4 of the 40kHz samples used by the DSP and recorded back in the DAQ at 10kHz.
Figure 1 shows the spectrum of the order 2 mode for three different alignment: initial alignment, slight re-alignment to reduce the beam jitter peaks by factor 10, alignment after doing slight realingments on 9 different modes. That last curve has 5 times less power than the other two, as the alignment has drifted far after many realignments on very high order modes.
Figure 2 shows the same data in terms of RIN, with the expected level of shot noise and PD electronic noise quadratically subtracted. There is an underlying broadband noise that has a slope somewhere between -1 and -0.66, it is present in all three cases, but is higher for the large misalignment (yellow line)
Figure 3 shows the sensing noise subtracted RIN for modes 1, 2, 3, 4 and 5, in that order. Modes order 1 and 3 are more strongly affected by beam jitter peaks. All the spectra hint at a broadband noise with a slope between -1 and -0.66.
Figure 4 shows that low level of noise with this setup can be achieved. It shows a spectrum of a single bounce beam from two months, lock on the 56MHz LSB done the previous day, and the order 1 mode and order 2 mode after realignment. The single bounce beam had noise a factor 4 lower at 100Hz, and had a similar power to the order 1 mode and order 2 mode. The 56MHz LSB had 5 times lower power, so more sensing noise, it has a comparable RIN to the carrier HOM, but a different shape, more flatter.
Figure 5 shows modes 6, 7, 8, 9 horizontal, 9 vertical, 10 and mode 00 after the cumulative misalignment from many changes in alignments for each HOM. The three 20% higher spectra at 140Hz correspond to modes that had a mostly vertical shape (order 6, 7 and 9 vertical mode).
As the realignment to minimize the beam jitter peaks actually degraded the alignment for the TEM00 mode by a factor few, the latter measurements correspond to observing each time a different mix of higher order modes of the beam. As the EDB OMC decomposes the beam on a completely different basis in a misaligned state.
None of the modes was coherent with h(t).
/users/mwas/ISC/RINanalysis_20250624/analyseRIN.m
Continued the study of the HOM noise in LN2. Went through all the modes from 1 to 9.
The attached figure show the mode shape, and the corresponding spectrum, taken during the measurement. Started to take the pictures only starting from mode order 3.
The OMC error signal demodulation phase might be badly tuned, with most of the signal on the q quadrature, so it is better to reconstruct the error signal from the demodulated signals instead of using directly the EDB_OMC1_err channel. I noticed it only when looking at mode order 9.
The data collected below needs to be analyzed.
15:49 UTC (5min) - locked on carrier order 1 mode (mostly horizontal)
15:55 UTC (5min) - increased OMC modulation depth to 0.5V
16:08 UTC (5min) - locked on carrier order 2 mode (mostly vertical)
tried to improvement alignmetn to reduce the beam jitter peaks at a few hundred Hz. It did work, used only the first mirror. The power of the mode also increased slightly
16:19 UTC (5min) - locked on carrier order 2 mode (mostly vertical)
16:25 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 3 and made small vertical realignment to reduce beam jitter peaks
16:37 UTC (7min) - locked on carrier order 3 mode (6 petals), but then around 16:38:40 1 minute jumped to mostly vertical, with higher beam jitter peaks
16:45 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 4 and made small vertical realignment to reduce beam jitter peaks
16:57 UTC (6min) - locked on carrier order 4 mode (mostly horizontal
17:04 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 5 and made small vertical realignment to reduce beam jitter peaks
17:20 UTC (5min) - locked on carrier order 5 mode (5 horizontal lobes, two vertical)
17:26 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 6, beam jitter peaks are already relatively small
17:38 UTC (5min) - locked on carrier order 6 mode (mostly vertical)
at 17:43:05 the registered temperature of the EDB OMC jupmed by 5mK (it is out of loop), also B1s increased by a few percent (PD with turned of Vbias)
17:46 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 7, realigned slightly in horizontal and vertical, beam jitter peaks almost completely disappeared and power of mode increased by factor 2
18:03 UTC (5min) - locked on carrier order 7 mode (mostly vertical)
at 18:07:40 the registered temperature of the EDB OMC jupmed by 5mK (it is out of loop), also B1s increased by a few percent (PD with turned of Vbias)
18:11 UTC (4min) - increased OMC modulation depth to 0.5V
18:15 unlock due to WE DSP glitch, the work around implemented so far does not help
relocked in LN2, locked OMC on moder orde 8, aligned slightly to reduce beam jitter peak and increase power of mode
18:51 UTC (5min) - locked on carrier order 8 mode (a wide blob in both directions), forgot to reduce modulation depth, so it is at 0.5V
another alignment improvement
19:00 UTC (5min) - reduced OMC modulation depth to nominal
locked on mode order 9, realigned slightly in horizontal and vertical, to reduce beam jitter, peaks sometimes disappear, but go up and down in height
19:15 UTC (5min) - locked on carrier order 9 mode (mostly horizontal)
19:22 UTC (5min) - increased OMC modulation depth to 0.5V
relocked onto the vertical mode 9, not trying to realign it, as it is the smaller mode from which it easy to jump into the horizontal one
19:34 UTC (5min) - locked on carrier order 9 mode (mostly vertical)
locked on order 10 mode
19:52 UTC (5min) - mostly horizontal
going to order 2 mode again, when crossing all the sideband and
carrier TEM00 I have noticed the 56MHz TEM00 had low power, the 56MHz
order 1 modes had similar power as the 56MHz TEM00, and there was a
high power TEM00 in between, is it a carrier created from the HOM?
20:13 UTC (5min) - order 2 mode, vertical, has 5 times less power than at the beginning
20:27 UTC (5min) - unexpected TEM00, jitter peaks are relatively high, but I haven't realigned for this mode
Most likely the realignments to reduce the beam jitter peak have
actually completely misaligned the OMC, so it mixes all the modes
together, including all of the HOM into a TEM00.
Will need to repeat the whole experiment, but this time aligning only
once on the 56MHz, and not touching the alignment afterwards, even if
the spectra are spoiled by beam jitter
This is a follow-up of the work started few months ago.
This time, we aimed to test the input power variation after the OMC lock, ideally in the low-noise configuration (see Fig. 1).
However, due to the DSP/RTPC communication issues this morning, we initially conducted the test in the high-power configuration (LN1) to avoid any unlocks caused by the Sc_{N,W}E cards.
In both LN1 and LN2, we applied a ±10% variation in input power (see Figs. 2–5). The interferometer remained locked without major issues.
Some points to note:
-
The PSTAB loop was left closed. As a result, it attempted to compensate for the power variation, causing a small glitch in the power. A possible future test could involve repeating the procedure with the PSTAB loop open to determine whether any residual power glitches remain.
-
No saturation was observed on PD1 or PD2.
-
In LN2, the power variation appears to be associated with a step or movement in the SR angular degrees of freedom. Interestingly, an SR_TX step seems correlated with an increase in power, while an SR_TY step is correlated with a decrease.
One possible explanation could be a difference in the resonant frequencies of the rotator in the two directions, which may interfere with the DCP line in DARM used for the SR AA error signal. -
In LN1, with the SR AA loop open, no spurious SR movement was observed.
Below is the log of the actions performed:
2025-06-24-16h46m19-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-16h46m19-UTC>INFO...-Parsed message*: [4, 1, 200, 50]
2025-06-24-16h46m19-UTC>INFO...-setting channel 4 axis 1 to 200 at speed 50
2025-06-24-16h46m19-UTC>INFO...-writing position -4750
2025-06-24-16h48m06-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-16h48m06-UTC>INFO...-Parsed message*: [1, 1, 500, 50]
2025-06-24-16h48m06-UTC>INFO...-setting channel 1 axis 1 to 500 at speed 50
2025-06-24-16h48m06-UTC>INFO...-writing position 174559
2025-06-24-16h48m06-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-16h48m06-UTC>INFO...-Parsed message*: [4, 1, 50, 50]
2025-06-24-16h48m06-UTC>INFO...-setting channel 4 axis 1 to 50 at speed 50
2025-06-24-16h48m06-UTC>INFO...-writing position -4700
2025-06-24-16h49m51-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-16h49m51-UTC>INFO...-Parsed message*: [1, 1, 500, 50]
2025-06-24-16h49m51-UTC>INFO...-setting channel 1 axis 1 to 500 at speed 50
2025-06-24-16h49m51-UTC>INFO...-writing position 175059
2025-06-24-16h51m17-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-16h51m17-UTC>INFO...-Parsed message*: [1, 1, -2500, 50]
2025-06-24-16h51m17-UTC>INFO...-setting channel 1 axis 1 to -2500 at speed 50
2025-06-24-16h51m17-UTC>INFO...-writing position 172559
2025-06-24-16h51m18-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-16h51m18-UTC>INFO...-Parsed message*: [4, 1, -250, 50]
2025-06-24-16h51m18-UTC>INFO...-setting channel 4 axis 1 to -250 at speed 50
2025-06-24-16h51m18-UTC>INFO...-writing position -4950
2025-06-24-16h55m19-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-16h55m19-UTC>INFO...-Parsed message*: [1, 1, 2500, 50]
2025-06-24-16h55m19-UTC>INFO...-setting channel 1 axis 1 to 2500 at speed 50
2025-06-24-16h55m19-UTC>INFO...-writing position 175059
2025-06-24-16h55m19-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-16h55m19-UTC>INFO...-Parsed message*: [4, 1, 250, 50]
2025-06-24-16h55m19-UTC>INFO...-setting channel 4 axis 1 to 250 at speed 50
2025-06-24-16h55m19-UTC>INFO...-writing position -4700
2025-06-24-17h00m48-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-17h00m48-UTC>INFO...-Parsed message*: [1, 1, -2500, 50]
2025-06-24-17h00m48-UTC>INFO...-setting channel 1 axis 1 to -2500 at speed 50
2025-06-24-17h00m48-UTC>INFO...-writing position 172559
2025-06-24-17h00m48-UTC>INFO...-Received move_rel message from LSC_script_spinicelli_20250624_160808
2025-06-24-17h00m48-UTC>INFO...-Parsed message*: [4, 1, -250, 50]
2025-06-24-17h00m48-UTC>INFO...-setting channel 4 axis 1 to -250 at speed 50
2025-06-24-17h00m48-UTC>INFO...-writing position -4950
Today's shifted was plagued by DSP glitches. Managed to do small amount of measurements of the light rejected by the SDB1 OMC in LN2 using the EDB OMC.
17:13 UTC (5min) - locked on 56MHz LSB (high temperature than carrier TEM00)
17:20 UTC (5min) - increased OMC modulation depth to 0.1V
17:26 UTC (2min) - increased OMC modulation depth to 0.3V
17:33 UTC (2min) - locked on carrier order 1, nominal modulation
unlocked during that measurement which was supposed to 5min
Continued the study of the HOM noise in LN2. Went through all the modes from 1 to 9.
The attached figure show the mode shape, and the corresponding spectrum, taken during the measurement. Started to take the pictures only starting from mode order 3.
The OMC error signal demodulation phase might be badly tuned, with most of the signal on the q quadrature, so it is better to reconstruct the error signal from the demodulated signals instead of using directly the EDB_OMC1_err channel. I noticed it only when looking at mode order 9.
The data collected below needs to be analyzed.
15:49 UTC (5min) - locked on carrier order 1 mode (mostly horizontal)
15:55 UTC (5min) - increased OMC modulation depth to 0.5V
16:08 UTC (5min) - locked on carrier order 2 mode (mostly vertical)
tried to improvement alignmetn to reduce the beam jitter peaks at a few hundred Hz. It did work, used only the first mirror. The power of the mode also increased slightly
16:19 UTC (5min) - locked on carrier order 2 mode (mostly vertical)
16:25 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 3 and made small vertical realignment to reduce beam jitter peaks
16:37 UTC (7min) - locked on carrier order 3 mode (6 petals), but then around 16:38:40 1 minute jumped to mostly vertical, with higher beam jitter peaks
16:45 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 4 and made small vertical realignment to reduce beam jitter peaks
16:57 UTC (6min) - locked on carrier order 4 mode (mostly horizontal
17:04 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 5 and made small vertical realignment to reduce beam jitter peaks
17:20 UTC (5min) - locked on carrier order 5 mode (5 horizontal lobes, two vertical)
17:26 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 6, beam jitter peaks are already relatively small
17:38 UTC (5min) - locked on carrier order 6 mode (mostly vertical)
at 17:43:05 the registered temperature of the EDB OMC jupmed by 5mK (it is out of loop), also B1s increased by a few percent (PD with turned of Vbias)
17:46 UTC (5min) - increased OMC modulation depth to 0.5V
locked on mode order 7, realigned slightly in horizontal and vertical, beam jitter peaks almost completely disappeared and power of mode increased by factor 2
18:03 UTC (5min) - locked on carrier order 7 mode (mostly vertical)
at 18:07:40 the registered temperature of the EDB OMC jupmed by 5mK (it is out of loop), also B1s increased by a few percent (PD with turned of Vbias)
18:11 UTC (4min) - increased OMC modulation depth to 0.5V
18:15 unlock due to WE DSP glitch, the work around implemented so far does not help
relocked in LN2, locked OMC on moder orde 8, aligned slightly to reduce beam jitter peak and increase power of mode
18:51 UTC (5min) - locked on carrier order 8 mode (a wide blob in both directions), forgot to reduce modulation depth, so it is at 0.5V
another alignment improvement
19:00 UTC (5min) - reduced OMC modulation depth to nominal
locked on mode order 9, realigned slightly in horizontal and vertical, to reduce beam jitter, peaks sometimes disappear, but go up and down in height
19:15 UTC (5min) - locked on carrier order 9 mode (mostly horizontal)
19:22 UTC (5min) - increased OMC modulation depth to 0.5V
relocked onto the vertical mode 9, not trying to realign it, as it is the smaller mode from which it easy to jump into the horizontal one
19:34 UTC (5min) - locked on carrier order 9 mode (mostly vertical)
locked on order 10 mode
19:52 UTC (5min) - mostly horizontal
going to order 2 mode again, when crossing all the sideband and
carrier TEM00 I have noticed the 56MHz TEM00 had low power, the 56MHz
order 1 modes had similar power as the 56MHz TEM00, and there was a
high power TEM00 in between, is it a carrier created from the HOM?
20:13 UTC (5min) - order 2 mode, vertical, has 5 times less power than at the beginning
20:27 UTC (5min) - unexpected TEM00, jitter peaks are relatively high, but I haven't realigned for this mode
Most likely the realignments to reduce the beam jitter peak have
actually completely misaligned the OMC, so it mixes all the modes
together, including all of the HOM into a TEM00.
Will need to repeat the whole experiment, but this time aligning only
once on the 56MHz, and not touching the alignment afterwards, even if
the spectra are spoiled by beam jitter
Figure 1 shows the spectrum of the order 2 mode for three different alignment: initial alignment, slight re-alignment to reduce the beam jitter peaks by factor 10, alignment after doing slight realingments on 9 different modes. That last curve has 5 times less power than the other two, as the alignment has drifted far after many realignments on very high order modes.
Figure 2 shows the same data in terms of RIN, with the expected level of shot noise and PD electronic noise quadratically subtracted. There is an underlying broadband noise that has a slope somewhere between -1 and -0.66, it is present in all three cases, but is higher for the large misalignment (yellow line)
Figure 3 shows the sensing noise subtracted RIN for modes 1, 2, 3, 4 and 5, in that order. Modes order 1 and 3 are more strongly affected by beam jitter peaks. All the spectra hint at a broadband noise with a slope between -1 and -0.66.
Figure 4 shows that low level of noise with this setup can be achieved. It shows a spectrum of a single bounce beam from two months, lock on the 56MHz LSB done the previous day, and the order 1 mode and order 2 mode after realignment. The single bounce beam had noise a factor 4 lower at 100Hz, and had a similar power to the order 1 mode and order 2 mode. The 56MHz LSB had 5 times lower power, so more sensing noise, it has a comparable RIN to the carrier HOM, but a different shape, more flatter.
Figure 5 shows modes 6, 7, 8, 9 horizontal, 9 vertical, 10 and mode 00 after the cumulative misalignment from many changes in alignments for each HOM. The three 20% higher spectra at 140Hz correspond to modes that had a mostly vertical shape (order 6, 7 and 9 vertical mode).
As the realignment to minimize the beam jitter peaks actually degraded the alignment for the TEM00 mode by a factor few, the latter measurements correspond to observing each time a different mix of higher order modes of the beam. As the EDB OMC decomposes the beam on a completely different basis in a misaligned state.
None of the modes was coherent with h(t).
/users/mwas/ISC/RINanalysis_20250624/analyseRIN.m
Attempted today to misalign the input beam while keeping the interferometer locked. To do that the input beam loops where first opened and the PR was shifted in X.
15:05 - moving PR in X by 10um with 30s ramp, unlocked just after
15:54 - moving PR in X by 5um with 100s ramp
another step by 5um with 100s ramp
unlock at 15:59:20
16:46 - moving PR in by 30um with 20s ramp, unlocked before the ramp
was finished
The most successful was the quick test earlier today around 11 UTC, purple is with input beam misaligned, blue in normal condition
Figure 1 shows a broadband view where the order 1 mode at 5.3kHz is clearly excited, also the order 3 mode around 16kHz is excited.
Figure 2 shows the order 8 mode at 44.3kHz.
Figure 3 shows the order 10 mode at 55.3kHz.
Figure 4 shows the time series of the quick misalignment around 11:00 UTC that was successful at getting the HOM a little excited, and correspond to a 2% drop in the arm power.
Figure 5 shows that no mode around the order 9 mode location is excited, but the bump already present there is giant. The average of the the order 8 and order 10 frequencies would put the order 9 mode at 49.8kHz, very close to the arm FSR of 49.97kHz.
There are also cases where the interferometer can survive DSP glitches.
Figure 1 shows a case when the alignment of the OMC was kicked by the B1 PD saturation.
Figure 2 shows that before the B1 PD saturation the NE correction signal received by the DSP was slightly different from the one sent by the LSC process.
Figure 3 shows the difference between the sent and recevied NE signal in a more clear way, by shifting by 4 samples the received correction so it is close to be superposed with the sent one, there are two red dots which are clearly not following the blue ones.
This case also happened after the transceiver replacement of this morning.
Piernicola suggested that the step glitches we have could be what in the past was called payload glitches. Following entries from that logbook at first sight these are different, as the one example of spectrum of payload glitch had a 1/f^2.5 spectral shape instead of something between 1/f and 1/f^0.5 that the step glitches have. However, checking if some high frequency mode is excited by the step glitches is a good idea. That idea is also linked to earlier issues of that kind back in 2008.
Looking at (5-10) step glitches from June 21, all of them have an associated glitch at high frequency. Figure 1-4 shows a couple of example of normalized spectrograms of SDB2_B1_PD1_Audio_100KHz. Each time I have selected the frequency range when the largest increase in spectrum compared to the median was present. These are 4 different glitches, each time centered at 64s in the time series. In some
As a control I have checked one random time, and one 25 minute glitch time, and in neither of them there was a glitch above 10kHz.
Figure 5, For some of the step glitches I could also find a glitch at 7814Hz at the same time, but not for all step glitches. 7814Hz is the drum mode frequency of the WE mirror.
It would be interesting to check on a larger sample of step glitches whether all have a glitch at high frequency (between 10kHz and 30kHz) at the same time, and which fraction of them have a glitch at the WE drum mode frequency (7815Hz). It would be also interesting to check the much smaller number of step glitches in late April / early May, if those had also the same property. The drum mode in April/May may be different by a few Hz, as it was a different mirror suspended at the same time.
This makes it probable that the step glitches are due to the WE mirror settling, for exampl with some kind of curing or drying process in the mirror ears. These would be after the suspension fiber, so not necessarily filtered by the pendulum. The more systematic study of which frequencies are excited (in particular how often the drum mode we know is excited) would allow to confirm that. If confirmed, then the only solution is to wait for the glitches to slowly reduce in rate, as they have been doing for the past two weeks.
Back in 2008 the exponential decay in the glitch rate had a e-folding time scale of 2 weeks. That would be the same order of magnitude as the decreased observed over the last two weeks.
/users/mwas/detchar/hoftGlitch_20250620/arches.m
Glitch rate vs. time plot updated until this morning 1000 UTC
Updated plot, using one week more of data (until this morning 2025/06/30 0800 UTC).
Same plot updated with one week more of data.
Figure 1. During the transient the power in the arms increased by 1%. This is most likely the sign that the input beam is not well mode matched with the arms, and that for different end mirror radius of curvature the matching becomes better. As we change only the end mirror radius of curvature, this corresponds to a mode that has the same wavefront curvature at the level of the input mirror (parallel to the mirror surface), but a different beam radius at the level of the input mirror.
Figure 2 shows the effect on the mode order 2 frequency. Purple is before the step, blue and red are roughly when the power in the arms is maximum during the transient (red is RH cooling and blue is RH warming), green is the maximum excursion of the RH change, ie two minutes before the ring heater was put back to the nominal correction. The maximum excursion corresponds to an 800Hz change in the mode frequency, which should correspond to a 14 meter change in radius of curvature. The peak in arm powers corresponds to 500Hz frequency change, or 8 meters in radius of curvature.
Using equation (53) of [H. Kogelnik and T. Li, "Laser Beams and Resonators," Appl. Opt. 5, 1550-1567 (1966) ] an 8 meter increase in EM radius of curvature correspond to 2.5% decrease in the resonant mode radius at the surface of the input mirror. Which means that to match the arm mode in nominal condition the input beam radius at the IM surface needs to be increased by 2.5%.
Figure 3 and 4 shows mode order 6 and 7, they are moving in opposite direction to the order 2 mode, which is not what one would expect from a simple radius of curvature change. And they also move by a smaller amount, about 1kHz instead of 2kHz. This means that the mode order 9 most likely did not move by the expected 3kHz. This effect of modes moving in opposite direction has already been seen for previous steps https://logbook.virgo-gw.eu/virgo/?r=66937. My guess is that it means the ring heater actuation far from the mirror center is not well approximated by a simple radius of curvature change (parabolic change), but that there is a more complicated shape and modes of high order are able to sense that.
Figure 5 in the meantime there was not a significant decrease in glitches during that test at any time, if anything there was an increase for part of that time.
Figure 6 is a simple histogram of the glitches with SNR>10 during that day with bins of 30 minutes, without taking into account that the interferometer is not locked some of the time. There is no significant improvement during the test on that figure either. A year ago such a histogram would have 1 or 2 glitches per bin (the 25 minute glitches).
Figure 1. The temperature of the laser amplifier has been gradualy increasing since Januray, it is now 4 degrees warmer. This isn't a variation due to summer, but a continous increase. A similar trend was present last year and was resolved by several interventions that reduced the temperature by 5 degrees in several steps. Given the gradual increase, the same type of interventions seems needed now.
The temperature increase does not seem to be explained by the degradation of the chiller performance as it used to be in the past.
There is now a dedicated chiller for the neovan head and the value of the flow, even if decreasing, does not seem to be correlated to the increase of temperature.
Instead we can notice a slow drift on the QPD at the output of the neovan as well as a decrease of the output power (plot 3).
It is not easy to understand if this misalignment is a cause or a consequence of the temperature increase.
In order to check if it is a cause, this morning we went in laser lab to try to realign the beam inside the neaovan head to see if it had an effect (plot 1). We tried to bring back the values on the AMP_QPD where they were few months ago.
We used only then mirror in front of the neovan head. We moved the vertical axis by about 40 degrees of the knob clockwise and the horizontal axis by about 60 degrees counterclockwise.
The power at the output of the neovan increased by about 3%.
The power at the output of the PMC increased by about 1%.
We quickly tried to relaign the PMC with the mirror in front of it but could not improve the coupling.
The temperature of the head slightly decreased while the power was increasing but it was relatively low, about 0.2 degrees and it has increased again over the day (plot 2)
We will monitor the situation in the next days to see if it get somehow more stable when it will be thermalized.
Next actions to try to improve the situation might be about trying to increase the flow of the chiller to see if it has an effect (changing filter, flush, reverse the flow... )
To be noted that the pick off for ALS PC and SQZ slightly decreased and we did not realign it.
This will need to be done.
Figure 1. Looking at the CARM correction coherence with DARM and h(t), it is present around 20Hz for DARM but not for h(t). My guess is that the h(t) reconstruction uses precisely measured response of the individual mirrors (NE and WE), and knows that some of the CARM correction is actually applied in the DARM degree of freedom, because the strength of the NE coil-magnet pairs and the WE coil-magnet pairs is not exactly the same.
The WE coil-magnet pair strength has changed due to the mirror replacement, so the CARM to DARM coupling has increased as the driving matrix has not been update. But the results with Hrec show that instead of doing a CARM to DARM subtraction, a more finely adjusted driving matrix should be able to avoid any coupling of CARM to DARM.
There have been many glitches during that test. I have read the glitches with SNR greater than 10 using omicron-print. And plotted them in matlab, with frequency on the y-axis, and log10 of the SNR in color scale.
Figure 1 shows the result, with roughly overlaid on top the three periods of data taking with the CO2 point absorber fixed as a red dotted rectangle, and the two cool down periods with the CO2 beam off as green dotted rectangles. I guess the few minutes in between a red rectangle and a green rectangle are one the CO2 beam was moving, and the small gaps between a green rectangle and a red one are inacurrate positing of the rectangles.
There seems to be more glitches as the CO2 beam is closer to the center. I don't know what that means. It cannot be a simple explanation that the CO2 laser is glitchy and it pushes on the test masses, as in that case the glitch rate would be the same regardless of the beam spot location on the mirror.
omicron-print channel=V1:Hrec_hoft_16384Hz gps-start=`lalapps_tconvert Jun 19 15:15` gps-end=`lalapps_tconvert Jun 19 17:45` snr-min=10 print-q=1 > glitches1.txt
/users/mwas/detchar/hoftGlitch_20250620/plotGlitches.m
Looking back through data I have been able to find when the 3Hz bump in B1/DARM started.
Figure 1 shows the week it has happened. On December 17 there is a change in the DARM spectrum, with noise in the 10Hz-25Hz noise band getting much lower, but a bump with variable height appearing at 3Hz.
Figure 2 shows a couple of spectra before the change (red and cyan) a couple after (blue and green) and the current situation (purple).
Looking in the logbook the origin of the change is clear. There was an ISC shift on subtraction filters https://logbook.virgo-gw.eu/virgo/?r=65832 . The last activity were changes in the CARM2DARM subtraction filter, and in figure 4 of that entry a bump in DARM at 3Hz clearly appears when the subtraction is turned on. This explains why the bump appears in the OMC spectrum, the OMC length has to follow CARM.
This is probably the origin of the more numerous short data segment for Virgo, with B1 saturation chopping data usable by data analysis into short segments, many of them to short to usable by offline pipeline. This have been reported at the Virgo week this week, https://tds.virgo-gw.eu/?r=24945, and apparently the problem has started in December. So most likely it is the same problem.
Looking at the starting point of the Detchar investigation, the problem did indeed start in the middle of December, https://logbook.virgo-gw.eu/virgo/?r=66288
During the lock acquisition around 9:15 UTC I have lower the OMC loop gain by a factor 1.8 to test if that affects the 3Hz bump in the B1 PD audio spectrum that sometimes saturates the audio channel.
Figure 1. This change has reduced the height of the 3Hz bump as seen in the OMC error signal by a factor 3, and reduced the coherence, but did not change the bump in the B1 PD audio spectrum significantly (10% level change). So the OMC loop is not the origin of the 3Hz bump, but it may be making it slightly worse.
The OMC loop gain peaking between 3Hz and 5Hz has reduced, as expected for a lower gain. Keeping this new configuration for now.
The bump at 3 Hz in DARM is never present in CARM NULL. Fig 1 shows a spectrum during a recent long period in CARM NULL, compared to the spectrum when the lock acquisition reaches the final phase. At the beginning, around 3 Hz there is a hole, rather than a bump. The situation changes in the period when DARM passes in DC readout. Fig 2 shows the spectrum in the first 3 min of DC readout, when BS angular control is still in drfit control, compared to next 3 min, after the engagement of BS full bandwidth. Again, we can see a clear change of regime concerning the bump at 3 Hz.
The engagement of BS full bandwidth comes together with other actions. Fig 3 shows DARM band rms close to the center of the bump, superimposed to BS AA flag and to the status of OMC LOCK. The index changes from 30 to 60 a few seconds before the BS full band and at the same time DARM rms shows a first little increase. It is not yet so evident the relationship between the bump and BS angular control, even because BS angular data have nothing special at that frequency. But I think it is worth to take data in different conditions, untill the source of the issue is clearly identified.
This is a nice analysis, it would mean that the quality factor of EM01 has increased compared to the last measurement VIR-0407A-23. Back in 2023 the quality factor was measured to be ~100e3, and now it has increased to ~650e3, which is comparable to the measurement of that same mirror in 2020.
Note that most of the other measurements were done using the ring down of a kick of the mirror. I don't know if the result of the fit of the steady state width of the line can bias the result to be an under or an over estimate.
Figure 1, adding the times of the DAS tuning (many hours) as blue rectanges, and the times of the RH steps (an exponential decay starting at the step) as green dashed lines. There might be a relation between the DAS tuning and the step glitch rate, there doesn't seem to be any for the ring heater steps, which means either that the glitches are independent of the RH steps, or require a much larger step to make a difference.
Today another issue with the rotators on SDB1 may explain what has happened on June 11. The automation drives the OMC shutter on channel 2 axis 1, while at the same time work was done to open the SDB1 Faraday shutter using channel 4 axis 2. There were commands sent for both within a few seconds, if the driver mixed up the commands it might have driven channel 2 axis 2, which is the north wave plate on SDB1 in front of the OMC.
The lesson here is to not drive any rotators on SDB1 by hand when the interferometer goes into DOWN, as at that time the automation closes the slow shutter, and there is only one driver acting on all the Agilis rotators and translation stages on SDB1.
Figure 1. For the unlock related to B1 PD audio saturations. There was a strong 3Hz transient 5 seconds before the unlock which saturated the B1 photodiode audio channel. https://logbook.virgo-gw.eu/virgo/?r=66999
Figure 2. That saturation kicked the SDB1/OMC alignment by 40urad, which then most likely caused the unlock.
