The plots attached to this elog show the trend of the LHO detector noise over the O1 time span so far. Each plot shows the band-limited RMS of the CAL-DELTAL_EXTERNAL signal, in a selected frequency band. I didn't apply the dewhitening filter. The BLRMS is computed over segments of 60 seconds of data, computing the PSD with 5s long FFTs (Hann window) and averaging. The orange trace shows a smoothed version, averaged over one hour segments. Only time in ANALYSIS_READY have been considered.
The scripts used to compute the BLRMS (python, look at trend_o1_fft_lho.py) are attached and uploaded to the NonNA git repository. The data is then plotted using MATLAB (script attached too).
I think there's a lot of interesting information in those plots. My plan would be to try to correlate the noise variations with IFO and environmental channels. Any suggestion from on-site commissioners is welcome!
Here are my first comments on the trends:
- 10-20 Hz and 20-30 Hz noise can vary by a factor about 2 on average, with a somewhat fast time scale (hours)
- There is a sudden drop of the noie in the 30-40 Hz region on 9/23 UTC. After that the noise stayed lower. I don't recall what happened at that time...
- The "unknown noise" in the 70-100 Hz region is rather non-stationary, varying on a daily basis of some 50-70%. Similar trends in the 140-160 Hz region.
- The noise in the 200-220 Hz region increased on 9/23. It seems to be the same time of the drop in the noise in the 30-40 Hz region... Same thing for 250-260 Hz, but there is clearly more variability here
- The jitter peak bumps (340-360 Hz) are varying quite a lot. As Evan pointed out, they are almost always small at the beginning of the lock, and then they never go back to the same level. Their variation seems quite slow, with a time constant of the order of a day. This could be something thermal in the IFO.
- Noise is the 400-450 Hz and 700-900 Hz regions are reasonably stationary, being shot noise. The 400-450 Hz noise showed an increase, again on 9/23, small but consistent.
So what happenend on September 23rd around the end of the day, UTC time?
S. Dwyer, J. Kissel, G. Vajente Gabriele notes a drop in BLRMS that is most (only) visible in the 30-40 [Hz] band, and questions what happened around Sept 23rd. Sheila (re)identified that I had made a change to the calibration that only affects DELTAL_EXTERNAL around the day, see LHO aLOG 21788. This change, increasing the delay between the calibrated actuation and sensing chains just before they're added, would indeed affect only the region around the DARM UGF (~40 [Hz]), where the calibrated actuation and sensing functions are both valid, and roughly equal in contribution to the DELTAL_EXTERNAL signal. Indeed, the (6.52 - 6)/6 = 0.086 = 8% drop is consistent with the 8% amplitude change expected that Peter describes in his motivation for me to make the change of the relative delay, see LHO aLOG 21746. Gabriele has independently confirmed that he's using DELTAL_EXTERNAL (and not GDS-CALIB_STRAIN which would *not* have seen this change) and the change happened between Sept 22 2015, between 15:00 and 20:00 UTC (*not* Sept 23rd UTC that he mentioned above). I've confirmed that the EPICs record was changed right smack in between there on Sept 22 17:10 UTC. Good catch, Sheila!