[Yuzurihara, Yamamura]

We are checking the Gaussianity of the strain data.

We used the data of the silent run on 04/06, reported by Yokozawa-san.

Fig 1 ~ 5: the results by using our tool

• two red bands can be seen, which means the Gaussianity was bad at that time.
  • we don't know the reason

We also checked “Rayleigh statistic” and "spectrogram" on the summary page.

• confirmed that similar behavior appeared in "Rayleigh statistic" (Fig 6,7)
• however, in "spectrogram", we could not find anything.(Fig 8,9)

Tomorrow, we will check the details of the timeseries data.

[Yuzurihara, Yamamura]
We investigated the two red bands further.

From Yokozawa-san's advice, we checked the ISS state.

• The band at around 04/05 21:40 (UTC)
  • No coincidence in the ISS state and raw timeseries data. (Fig.1)
  • Any lockloss did not occurr at that time in any other guardian state. ( SummaryPage )
  • The whitend timeseries was the only coincidence we have found. (Fig.2)
• The band at around 04/06 03:15 (UTC)
  • There were ISS lockloss at that time. (Fig.3)
  • Also, the whitend timeseris had a peak. (Fig.4)

We don't understand especially the first one, so we hope someone can give us some advice.

Thank you for your check.

For first glitch, you can find the sudden gap in time series of h(t), that would be a reason.
See JGWDoc11751, and P11 of the JGWDoc11990

[Yokozawa, Yuzurihara, Yamamura]

We investigated further in order to understand how GauCh behaves when such phenomenon appear.

• First glitch (Fig.1)
  • This may be caused by the sudden gap in h(t), as Yokozawa-san already reported.
  • We checked the ASDs because GauCh's result is calculated by the distribution of them. 
  • We comfirmed that if the gap is in h(t), the ASD will get high value and straight structure. (Fig.2 : NO gap data, Fig.3 : gap data)
• Second glitch (Fig.4)
  • This coincides with ISS instability. (Fig.5)
  • We cheked the other channels which is relevant to ISS. (Fig.6 ~ 8)
  • It may be the similar phenomenon with #klog24753.

By the way, other than such apparent shocks in the spectra, we know that the arm transmission power or the contrast at the GW detection port periodically gets degraded during the lock time. The period of these events is once per about 10-15 sec. In principle, the light power in the arms is directly related to the optical gain (i.e. the conversion ratio from GW to light), so it should directly affect the sensitivity, and that would looks like non-Gaussian behavior. I am intereseted in whether such deviation from the quiet state (or in the most Gaussian situation) would be seen with your non-Gaussainity study.