No significant coherence were detected in those measurement.
No significant coherence were detected in those measurement.
DARM's 270 Hz peak is moving toward a few Hz larger over a day.
* Hveto
I noticed that there are several PEM sensors have coincidences with line glitches appearing between 50 Hz and 60 Hz in DARM.
This line glitch is recorded in the microphones located in the BS_BOOTH, POP, POS, and IMC_REFL tables. It is also recorded on seismometers located at MCF, BS. The glitch appeared on all these channels for a duration of approximately 6 seconds.
This time, I added LSC_MICH for testing.
I also changed the analysis start frequency of the strain channel from 0 Hz to 10 Hz because the low frequency side is buried in many line noises.
> Is it similar behaviour reported in klog30314?
Yes to the fact that PLL lock by guardian is prevented by the jump of IMC temperature feedback.
No to the reason why IMC temperature feedback jumped.
In the previous case, IMC guardian executed "clear history" because the feedback output became lager than software threshold. As a result, IMC temperature feedback jumped to 0ct. In this time temperature feedback jumped from ~-5000ct to ~-6000ct. So it's not due to "clear history".
In any case, PLL guardian must become to find a proper temperature bias on its own.
Abstract:
Quick estimation of DAC noise is not consistent with the DAC noise estimation by turning off dewhitening filter.
It is necessary to measure all TFs from each coil to DARM for all Type-A suspensions for further precise projection.
Detail:
I measured the TF from K1:VIS-ITMY_TM_COILOUTF_H1_OUT to DARM when PRFPMI was locked with OBSERVATION condition (1-stage dewhitening filter was turned on).
The result is about 3.2e-17 * (f/100Hz)^-2, so if there is no dewhitening filter, the TF should be 3.2e-16 * (f/100Hz)^-2.
According to the DAC noise measurmenet results, typical DAC noise above 10 Hz is 4e-4 cnt/rtHz at ADC.
To convert the value at ADC to the value at DAC, we need to multiply 2, so the DAC noise level at DAC is 8e-4 cnt/rtHz above 10 Hz.
So, DAC noise from each coil can be estimated as 3.2e-16 * (f/100Hz)^-2 * 8e-4 = 2.5e-19 * (f/100Hz)^-2 above 10Hz.
Each test mass has four coil, so DAC noise contribution from each test mass is 2* 2.5e-19 * (f/100Hz)^-2 = 5e-19 * (f/100Hz)^-2, if we asuume DAC noise is independent with each other.
However, according to the DARM spectrum measurement results when turning off dewhitening filters were about 2e-18 * (f/100Hz)^-2 (klog31710 etc).
So, the my estimation results seems inconsistent with the mesured DAC noise level.
Since I have no time to measure all TF from each coil to DARM, I just use ITMY TF as a representative.
So, it is necessary to measure all transfer functions for further precise estimation of DAC(coil driver) noise.
Is it similar behaviour reported in klog30314?
If so, it is well known issue to be solved for long-term stable operation.
Temperature feedback and guardian state transition of IMC and PLLs around this large jump are shown in Fig.1.
After this jump, PLL repeatedly failed with GRUB_RESONANCE.
So I re-tuned PLL temperature bias around 17:00 in order to recover PLL lock.
The reason why temperature feedback of IMC has so large jump hasn't investigated yet.
Is it similar behaviour reported in klog30314?
If so, it is well known issue to be solved for long-term stable operation.
> Is it similar behaviour reported in klog30314?
Yes to the fact that PLL lock by guardian is prevented by the jump of IMC temperature feedback.
No to the reason why IMC temperature feedback jumped.
In the previous case, IMC guardian executed "clear history" because the feedback output became lager than software threshold. As a result, IMC temperature feedback jumped to 0ct. In this time temperature feedback jumped from ~-5000ct to ~-6000ct. So it's not due to "clear history".
In any case, PLL guardian must become to find a proper temperature bias on its own.
It looks fine, and Miyoki-san's prediction that the alert bell was the cause of this issue seems to be correct.
Figure 1 shows some signals from 9:00 JST to 21:00 JST yesterday.
Large RMS cannot be seen after finishing maintenance work (around t=-3h)
Because sampling rate of DQ channels for PMC feedback and microphone is lower than 4kHz, they cannot detect this issue (left panels in Fig.1). Only error signal (MIXER_MON_OUT_DQ) detected because its sampling rate was 16kHz. On the other hand EPICS channels without Anti-Aliasing detected as a aliasing noise (right panels). This fact means there is unfortunately no witness to see time series waveform of this disturbance.
Figure 2 shows the zoom up around 10:30 - 12:20. According to the EPICS channels of microphone signals, external disturbance disappeared around 11:44. Even if there is no enough sampling rate, EPICS channels without anti-aliasing is enough for checking just timing coincidence. PMC control was disengaged around 11:41 for the maintenance work. This is reason why the large RMS disappears on PMC control signals faster than on microphones.
The beam dampers in the PSL are maybe this type (https://www.ophiropt.com/ja/f/l250w-bb-50-water-cooled-thermopile-sensor).
Tips on circulating water. (https://www.youtube.com/watch?v=eGc7seR_QK4&t=4s)
I am now considering using this type with an additional SiC plate or so (https://www.kawaso-texcel.co.jp/product/brazing_heatsink_standard.html). The SiC or other plate should be polished?
~ 100W heater power was added in EYV at ~ 15:30 on the 22nd. Then, the F0 position increased to zero area.
The EY BF position increase stopped because of Takahashi-san's heater temperature reduction.
By the way, we have never(?) stopped several FFUs at TMSXY to check their influence on sensitivity. If trans-beam data is used in the present alignment control, is it better to try to turn off and on FFUs?
Yes. Technical staff would turn the bell off after the end of the fire drill. So the off of bell could be after 11:00 or so. How is the present situation of the spectrum ??
Abstract
Though past DAC noise measurement was done with zero output, a possibility that noise level is different between situations with zero output and with non-zero output was pointed out.
So we measured DAC noise again with non-zero output.
Because we cannot see any difference between zero output spectrum and non-zero output one at the output of Anti-Imaging chassis, past measurements seems to be reasonable as DAC nosise.
On the other hand, because engaging de-whitening filter didn't make a 10 times reduction of DAC(?) noise in the past measurement, there may be a noise larger than DAC noise coming from downstream of de-whitening filters.
Details
Quick check of DARM contamination by DAC noise was done in klog#31706 (PRs), klog#31707 (BS+SRs), klog#31708 (IX), klog#31710 (IY), klog#31712 (EY), and klog#31716 (EX). In these measurements, we can know that de-whitening filter for some of actuators makes sensitivity improvement. In usual, we assume that a DAC noise with de-whitening filters in the unit of DARM displacement can be suppressed by factors of 1e-1, 1e-2, or 1e-3 above 10Hz for the case of engaging 1, 2, or 3 stages of de-whitening filters, respectively. On the other hand, past DAC noise measurement (see also link below) shows more small improvement of DAC noise by engaging the de-whitening filters.
- klog#24725 (IY), klog#24726 (IX), klog#24919 (EY), klog#25011 (EX),
- klog#24916 (BS), klog#28069 (SR2), klog#26129 (SR3), klog#26230 (SRM),
- klog#25009 (PR2), klog#25886 (PR3), klog#26205 (PRM).
If de-whitening filter doesn't make a improvement by a proper factor for the number of engaged filter stages, current noise limit may come from the downstream of the de-whitening filter stages instead of DAC noise itself. As another scenario, a possibility that noise level was different between situations with zero output and with non-zero output was pointed out. In former case, we may need to replace high power coil driver to low power one. In latter case, noise projection should be done with the spectra measured in the case of non-zero output. To know which case we are in now, we made a DAC noise measurement with the non-zero output.
Measurement was done for TM actuators of ITMX with the same manner as past measurement. At first, we measured the noise level with connecting only whitening filters and Anti-aliasing filter to ADC as a background noise of our measurement. These result are shown as blue curves in Fig.1. Upper left, lower left, upper right, and lower right panels show the spectrum of DAC channels for H1, H2, H3 and H4 coils, respectively. Next, we connected DAC and Anti-imaging filters to the input of whitening filters. We already know DAC noise is smaller than ADC noise, so we applied whitening filter gain as 36dB to see DAC noise on ADC channels. DAC noise with zero output and 1000ct output are represented as green curve and brown curve, respectively. There is no significant difference each other. So noise level in the case of zero-output did't seem to be a underestimate.
We considered a possibility that though DC output doesn't make larger noise than a case of zero output, AC output may make it. So we made a 0-3Hz Gaussian noise for the DAC output shown as magenta curve. Seeing this noise on ADC channels, we can obtained red curve. Because DAC output (magenta curve) is larger than the DAC noise (green or brown curves) below 100Hz, we can see the injected output on ADC (red curve) below 100Hz. On the other hand, red curve shows same level as green and brown curves above 100Hz. This fact means that AC output also doesn't make lager noise than a case of zero output.
According to these results, past measurements seems to be fine for the estimation of DAC noise and noise coming from the downstream of de-whitening filter is probably larger than DAC noise. Several questions still remains, but it can be investigated by using Anti-Imaging output without coil drivers. So the rest of works can be done on Mozumi standalone system.
Request from Ushiba-san and Aso-san
[K1OMC0]
model: k1omc
Continued work on updating k1omc from the other day.
#1 Changed AS_POW to separate the input and output a triangle wave with offset.
#2 Changed from K1:OMC-PZT-HV1_OFFSET to K1:OMC-PZT_HV1_OUT16.
#3 The value in #2 above may cross 0 when the triangular wave is added.
Then, the positive and negative offsets are repeatedly reversed.
To prevent this, the offset and amplitude of the triangle wave was changed to take into account.
#4 Same reason as #3 above. The offset addition position was moved to after K1:OMC-PZT_HV1.
#5 The same thing as #3 occurs when overshoot occurs. Therefore, the threshold was changed to have an upper and lower limit.
k1omc_New.png
=> #3 and #4 above are the same process, so I modified them for next time. No updates have been made to this model.
We confirmed Xend Wi-Fi routers. #11 is installed in Xend parking area. #05 is installed EXA area. Today we moved #05 router to the management network.
After the cabling work of heater power supplies (klog 31657), we connected these cables to the network switch at EX0 rack. The information is summarized to the excel sheet on JGWdoc.
I tried to lock IMC to the NEOLASE laser.
Firstly, I put a half-wave plate for the new laser to adjust the polarization to S for PMC and EOMs downstream.
PMC can be locked with the NEOLASE with 9A current like the last time. This time I injected almost 20W to the PMC which is similar power to the current FB laser to make it easy for gain adjustment or something.
Then I switched some cables and modified the guardian code as klog:25843. After some adjustment of wide-band EOM gain, IMC could be locked for at least several minutes, but only SLOWBOOST and 2 of 3 COMMONBOOST could be engaged. I tried higher or lower gain for wide-band EOM, but it did not work.
I need more careful measurements for open-loop TF, cross-over frequency, UGF, etc with somebody's help next time.
[Kimura, Yasui and Tanaka]
We switched from TMP to ion pumping for vacuum pumping at SRM.
The following is a record of the pressure change in the vacuum chamber during the pump switchover for reference.
11:38 SRM=2.6 x10^-6 Pa, OMMT=7.3 x 10^-5 Pa
11:40 Close GV in front of TMP
12:18 SRM=5.5 x10^-6 Pa, OMMT=9.6 x 10^-5 Pa
13:14 SRM=6.7 x10^-6 Pa, OMMT=9.8 x 10^-5 Pa
13:26 SRM=6.7 x10^-6 Pa, OMMT=9.9 x 10^-5 Pa
TMP stopped
13:30 TMP turned off and flexible tube between the TMP and the dry pump was disconnected.
SRM=5.6 x10^-6 Pa, OMMT=9.9 x 10^-5 Pa
Is it related to the sound of a bell inside the PSL room because of today's fire drill? I guess that 10:33 is just the time to push the warning button.
So I checked time series and spectrum for checking when this problem started and which frequency has large RMS, respectively.
According to the time series plot, it's occurred when IFO lock was recovered around 10:33. Precisely, it occurs ~5 second after LSC_LOCK guardian entered OVSERVING_WITHOUT_LINES.
By checking the spectrum of PZT feedback, RMS comes from the 2400Hz (and sideband?) series.
Because of the today's maintenance work, I need to postpone the investigation of this issue.
If someone can takeover, please check it more detail.
TTFSS gain seems to be fine (~9dB for both common and fast gains).
So it's not a same problem as klog#31662, at least.
Is it related to the sound of a bell inside the PSL room because of today's fire drill? I guess that 10:33 is just the time to push the warning button.
Yes. Technical staff would turn the bell off after the end of the fire drill. So the off of bell could be after 11:00 or so. How is the present situation of the spectrum ??
It looks fine, and Miyoki-san's prediction that the alert bell was the cause of this issue seems to be correct.
Figure 1 shows some signals from 9:00 JST to 21:00 JST yesterday.
Large RMS cannot be seen after finishing maintenance work (around t=-3h)
Because sampling rate of DQ channels for PMC feedback and microphone is lower than 4kHz, they cannot detect this issue (left panels in Fig.1). Only error signal (MIXER_MON_OUT_DQ) detected because its sampling rate was 16kHz. On the other hand EPICS channels without Anti-Aliasing detected as a aliasing noise (right panels). This fact means there is unfortunately no witness to see time series waveform of this disturbance.
Figure 2 shows the zoom up around 10:30 - 12:20. According to the EPICS channels of microphone signals, external disturbance disappeared around 11:44. Even if there is no enough sampling rate, EPICS channels without anti-aliasing is enough for checking just timing coincidence. PMC control was disengaged around 11:41 for the maintenance work. This is reason why the large RMS disappears on PMC control signals faster than on microphones.
[Komori, Michimura]
To summarize, the discrepancies observed can be explained as follows.
For the PD plot (DCPD power-based shot + dark vs. measurement with ~10% discrepancy):
- Presence of additional noise just below the current measurement (frequency noise is likely; Currently, out-of-loop CARM sensor PD3 receives 0.6 mW and is limited by dark noise. Typically RFPD has dark noise equivalent to 10 mW shot noise, so actual frequency noise contribution could be 1/4 of the green sprectrum in this plot in klog #31719, and this can explain 10% discrepancy)
- Mis-calibration of the DC PD to mW (only square root dependence)
- PD quantum efficiency eta=0.93 is wrong (only square root dependence)
For the displacement plot (IFO parameter-based shot + dark vs. measurement with ~20% discrepancy):
- IFO to the PD loss is more than 15% (likely due to OMC alignment etc.)
- Incorrect calibration of PROC_DARM_DISPLACEMENT to meters
- Not all of the Pin=1.4 W couples to the IFO
To evaluate the total DAC noise in current sensitivity, we calculated the quadratic sum using several results.
As described in klog31706, klog31707, klog31708, klog31710, klog31712, klog31716, TM stage and MNV stage of the Tyep-A suspension had several excess.
Abe-san calculated the quadratic some for each results (1/10 for TM stage and 1/100 for MNV stage) and compared with current sensitivity.
Several coherence can be detected in both pit and yaw.
