Reports 1-1 of 1 Clear search Modify search
Detector Characterisation (Glitches)
robinet - 14:09 Monday 18 February 2019 (44910) Print this report
Unknown glitch family at 31 Hz

Looking at the data of last week-end (which are very clean in terms of glitches!), I noticed a persistent family of glitches around 30 Hz. In fact, this family has existed for a very long time. I can see examples in O2 data. The glitch rate is rather small, a few tens of glitches every day, and this is probably why this family has not been investigated (AFAIK). Now the data seems cleaner, this family is in the top of the list.

I had a first look at the glitches (see spectrogram in Fig. 1). This is what LIGO would call a rain drop glitch. I also scan *ALL* virgo channels to search for a coincident glitch: there is none.

I guess this post is a call for volunteers to investigate further this glitch family. I'm posting a list of GPS times of the glitches of last week end. Some of them are so loud they produce an range drop or even a lock loss.

GPS - Frequency [Hz] - SNR
1234311598.3945 31.40 29.14
1234311603.6055 31.40 23.39
1234312342.7695 31.40 52.49
1234312914.2617 31.40 27.40
1234312969.8164 31.40 50.15
1234313468.9648 31.40 73.74
1234314328.2070 31.40 825.42
1234315169.9258 31.40 103.66
1234316898.2930 31.40 13.96
1234322170.1289 31.40 302.52
1234325354.9727 31.40 30.66
1234325356.2070 31.40 11.42
1234325375.5039 31.40 31.51
1234325673.2461 31.40 32.98
1234325935.9492 31.40 15.55
1234326187.6289 31.40 683.17
1234326209.5352 31.40 1232.40
1234326274.2383 31.40 60.53
1234326331.3477 31.40 259.32
1234326367.7148 31.40 11.12
1234326393.8867 31.40 11.43
1234326438.6133 31.40 45.99
1234326451.5586 31.40 56.98
1234326482.2539 31.40 40.20
1234326545.2773 31.40 51.39
1234326554.3242 31.40 32.78
1234326740.0586 31.40 46.47
1234326746.2617 31.40 13.69
1234326819.0430 31.40 381.75
1234326834.5273 31.40 10.92
1234326955.3164 31.40 160.32
1234329917.3945 31.40 28.33
1234331193.4805 31.40 18.52
1234339428.3867 31.40 271.20
1234385870.7852 31.40 13.71
1234391955.5586 31.40 756.79
1234392857.0586 31.40 82.00
1234397370.0117 31.40 10.19
1234400044.5742 31.40 74.66
1234432599.9258 31.40 412.29
1234455907.7930 31.40 14.96
1234457434.4258 31.40 336978.43 (lock loss!)
1234462773.6914 31.40 16.72
1234474319.1055 31.40 473.80
1234477166.6055 31.40 29.01
1234477837.8398 31.40 10.98
1234482283.3789 31.40 24.66

 

Images attached to this report
Comments to this report:
mwas - 12:58 Tuesday 19 February 2019 (44923) Print this report
Figure 1. I have looked at the spectra during one of the loudest glitches, and it has a 1/f^3 shape. This is the frequency shape of a step function (1/f), and then filtered by the mirror suspension (1/f^2).
Figure 2, 3 and 4 are examples of some the loudest glitches. There is clearly a step in the correction sent to the mirrors, at least for figure 2.
In the three examples the glitch correspond to a power drop on B1_PD2 and then an oscillation caused by DARM loop . It would be intersting if the glitch always start with B1_PD2 power going down.
Note that the answer might be different for the very loud glitches (SNR>100), and for the more moderate glitches.
Images attached to this comment
mwas - 10:41 Wednesday 20 February 2019 (44939) Print this report
I have looked at the 12 glitches with SNR>100. 10 out 12 correspond start with the power dropping on B1 (and then a short oscillation presumably the step response of the DARM loop).

The loudest of the glitches seem to be even visible in B1p PD1
Figure 1-4, show the glitches with SNR>500. B1p PD1 is scaled by a factor 185, as it is expected that only 1/185 of the light from the ITF reaches that photodiode.
In the 4 examples, there seem to be a transient also in B1p, with a similar amplitude, but it is slightly faster (30% shorter time scale?).
If a noise transient is added in between the B1p beam pickoff and the B1 photodiode. Then the shape of the transient in B1p, should be the shape of the transient in B1 times the DARM open loop transfer function response.
It would be interesting to convolve the transients seen in B1 with the DARM loop response, to see if it matches the shape of the transient in B1p.

Another interesting probe in B1s2P PD1, this photodiode looks at light rejected by OMC2 that is in P polarization. About 1% of the light that should be going through OMC2, is rejected due to birefrigence mistuning between OMC1 and OMC2.
Figure 5-7 show the same glitches with SNR>500, with B1s2P scaled by a factor 100. The glitch in B1s2P has the same shape as in B1. Which means that the glitch is added up-stream from OMC2.
It would be interesting to automate the shutter opening for B1s1P, this PD has been aligned ~two weeks ago, and should also have ~1% of the B1 beam due to polarization mistuning between the incomming laser beam and OMC1

Images attached to this comment
mwas - 16:30 Wednesday 20 February 2019 (44946) Print this report
Found out why the transient on B1p looks short (faster). I have been looking at the Audio channels, and for B1p_PD1 the analogical high-pass filter of the Audio channel is different. It has a corner frequency at ~15Hz instead of ~2Hz, to help with free Michelson calibration measurements.
If instead I look at the Blended signals for B1 and B1p, and then apply a high pass filter at 10Hz by hand to avoid being dominated by low frequency fluctuation I obtain figure 1 (B1 in blue, B1p in red).
The transient has the same time scale on both photodiodes.
Images attached to this comment
Search Help
×

Warning

×