Reports 1-1 of 1 Clear search Modify search
Detector Characterisation (Broadband noise)
mwas - 12:22 Friday 12 April 2024 (63944) Print this report
Electric noise coupling to h(t)

Looking at Bruco results from yesterday I have noticed broadband correlations that I don't remember seeing last time I looked a few weeks ago.

Figure 1. NEB IPS CURR is coherent between 20Hz and 60Hz. Did the NEB IPS CURR noise level increase in the past few weeks?

Figure 2 WI FF50Hz P ERR is coherent between 15Hz and 100Hz. I am surprised to see coherence with the 50Hz feed-forward very far from the 50Hz frequency.

It would be important to look at this in more detail. A more complete list of which channels become coherent with h(t), when did it become coherent, and if the corresponding environmental channels increase in noise level at the same time.

Images attached to this report
Comments to this report:
direnzo - 16:17 Friday 12 April 2024 (63947) Print this report

It seems that the coherence estimation by BruCo got fooled by one of the 25-minute glitches, happening at the end of the analyzed interval, UTC 14:20:25 2024/4/11 + 900 s. The glitch is recorded by omicron at 14:33:39.68. So, the spectral estimation is biased: don't trust this result.

The coherence mis-estimation is documented in this git issue.

Figure 1: coherence spectrogram of 1 hour of data around the interval of yesterday BruCo run. The effect of the glitch is visible at 14:33 UTC as an excess of coherence. The left-hand side panel shows the estimated coherence in the 1-hour interval using the median method, which is more robust to glitches. Except for the 50 Hz line and its harmonics, no suspect coherence value is visible. Since v3r2, BruCo has available the --medianpsdestimation option to estimate the spectrum using the median instead of the average (Welch method).

Figure 2: Q-scan and whiten time series of Hrec and V1:ENV_NEB_IPS_CURR_R_2000Hz in an interval of 1 second around the time of the glitch. The current channel shows absolutely no excess noise.

But then why the coherence between the two?

Figure 3: there is in fact a glitch in V1:ENV_NEB_IPS_CURR_R_2000Hz 2 seconds after that in Hrec, and of a vaguely similar shape, as visible from the spectrograms. The similarity in the spectrum has triggered a larger coherence value in the time bin of figure 1 embracing both glitches.

To confirm that this occurred just by chance, I plotted the spectrograms for other 25-minute glitches for both Hrec and V1:ENV_NEB_IPS_CURR_R_2000Hz. For all, no glitches in the latter channel have been observed.

Figure 4: Q-scans of Hrec and V1:ENV_NEB_IPS_CURR_R_2000Hz for another glitch belonging to the 25-minute family.

Images attached to this comment
fiori, tringali, paoletti - 16:59 Friday 12 April 2024 (63950) Print this report

The increased coherence with ENV_NEB_IPS_CURR seems associated to the accidental presence of a glitch in hrec (one of the 25-min glitches) and a glitch in the ENV_NEB_IPS_CURR sensors.

The two glitches are NOT coincident indeed but a few sec away (the IPS glitch occurs after), but they both fall in the 10s window of the Bruco computation.

Figure 1: coherences are computed 120s before the glitch (purple) and in a 120s window containing the glitches.

Figure 2: spectrograms show that the two glitches are a few seconds away. The COHETIME plot implemented in dataDisplay measures a sort of "anti-coherence" (the coherence goes to a lower value in the time window that contains both glitches).

It is not clear what is peculiar to the NEB_IPS_CURR sensor glitch to produce coherence with Hrec?

Images attached to this comment
Search Help