Two weeks ago an excavator working near the central building provided a persistent and maybe useful noise injection. The effect on the sensitivity was quite big (fig 1). One could consider this normal and forget about it, but the shape of the noise (fig 2) is so similar to the curve in quiet condition, that maybe it is worth to ask ourself what is the feedthrough of this noise towards the ITF.
The shape of the ground noise sensed by the seismometer is shown in fig 3. We can see the typical effect of the anthropogenic noise, namely the increase of the 3 Hz region, but also a relevant component around 10 Hz, that rarely are interested by the variations of environmental conditions. Looking at the rms of the disturbance in time domain, one can recognize a series of short seismic events, producing very often a similar variation in the rms of ITF noise (fig 4). In fig 5, the coherence between the two rms signals, in order to certify that the seismic events in CEB were actually the source of noise (in that period, there was also agricultural activity near WEB).
In fig 6 the 'polinomial decomposition' of the noise is shown, following the method shown in the previous entry. It is not obvious to me why I can use exactly the same components of the quiet noise, just rescaling the 1/f^4 and the 1/f^1.5. And, really amazing, the scaling factor looks the same for the two components. In fig 7, the quiet data, taken half a hour later, are fitted having divided by 2 the low frequency components.
One possible explanation: the events do not add noise, but have an impact on the working point, so that the same noise have a doubled coupling factor. I say it, but it seems to me definitely not realistic, because the working point cannot change in the timescale of a glitch. Moreover, there is nothing in the typical parameters sensible to the working point, like B1p_DC (fig 8).
I would propose to 'excavate' more around this topic.
