Yesterday we performed some tests on the Guralp sensor at the NE building.
These tests were triggered by Paolo. He observed that the seismic noise detected by the NEB guralp sensor worsened and changed since December 2019 till the end of the O3 run and never recovered to the "good level" measured in April 2019. The seismic noise worsening/changes occurs above ~30Hz, and specifically they happened on Tuesdays, the Dec 10, Jan 14, Feb 11, and March 17. This analysis is illustrated in the first attached document.
We first suspected the LN2 truck going to NEB every two weeks for refilling the external tank. But, the arrival and departure times of the truck, that are clearly identifiable in the suspension signals, demonstrate that the truck already left NEB when the noise in the Guralp arose.
Note the guralp sensor is placed in a corner of the tower pit floor, covered by an insulating box. See the attached picture.
First, we did an Huddle test putting one meggit accelerometer at 1m distance from the Guralp sensor (see the picture). We trust the meggit accelerometer to have a flat acceleration response from a few Hz to some hundred Hz. We measured a transfer function between the two sensors along the three directions, orienting the meggit in turn along the vertical, then N, then W, using one viton cube support. These TF are reported in figures 2, 3 and 4 respectively for the vertical, North and West axis.
The vertical TF looks fairly smooth up to 100Hz. The North and West axis TFs instead start losing coherence above 50Hz. This is associated to excess noise in the Guralp signal, evidenced by the red curve in Figures 3 and 4.
secondly, we checked what happens to the Guralp signals when gently kicking the guralp box, like it might have happened during cleaning operations of the pit.
We observe that, once the signals has recovered from the transient, the quiet guralp signals above ~50Hz look quite different: some stationary spectral noise structures changed amplitude. Figures 5 and 6 (linear scale)
This might resemble what Paolo observed.
Concluding:
- We make the hypothesis that the sensor has been hit occasionally during cleaning operations during O3, each time faking a noise increase or noise change in the signals.
- ... indeed, as just found out, all the identified dates correspond to cleaning operations at NE (reported in the cleaning operation logbook https://wiki.virgo-gw.eu/Commissioning/Operations/CleanFirm)
- This might be a symptom of the sensor masses being not centered (centering the masses is not straightforward and we are working to set up the operation)
- The observed noise (ref. to Paolo attached document) are to be considered not due to a real seismic source, but just sensor noise.
- The sensor cannot be trusted above 30Hz. Indeed, the Guralp sensors specs. guarantee its operation only in the range up to 50Hz.
- We need to improve the sensor installation, in such a way to protect it from accidental kicks.