After cleaning up the code for offline matlab data cleaing to make it more systematic and use different data for measuring filters and applying filters (otherwise the subtraction is not comparable to what can be done online), there are two more improvements that could be done online.

Figure 1 shows h(t) raw in blue, h(t) raw cleaned offline in red, and h(t) cleaned online in yellow. The spectra of noises used for the offline cleaning are the other lines. In addition to the difference at 90Hz-150Hz, there is a bump at ~19.2Hz present in the online cleaned data and not in the offline cleaned data. This is because in the online cleaning there is a notch at 19.2Hz for the SWEB_B8_DC subtraction, so there is no cleaning at that frequency. It would be good to check that if that notch is really needed (or if it is needed only for SNEB_B7_DC, or not needed at all).

Figure 2 shows the ratio of the offline cleaned data with the online cleaned data in red. In between 1kHz and 2kHz there is a sligthly better spectrum for the offline cleaned data, this because the subtraction of the 56MHz RIN using B1s1_PD1_Blended stops at 1kHz in the online cleaning, it might be interesting to extend it till 2kHz.

Figure 3 shows an omegascan of the h(t) online, and Figure 4 of h(t) cleaned offline. The glitch at ~65 seconds that Florent mentioned disappears with the offline cleaning, I expect it comes from the notch in the subtraction at 19.2Hz, so when the scattered light arch passes through 19.2Hz it is not being subtracted in the online cleaning.

Code used for plots is in /users/mwas/calib/noiseSubtraction_20190730/

Following these entries, I have restarted HrecClean at 14h50 UTC with a different noise subtraction configuration than Hrec: frequency separation between SPRB_B4_56MHz_Q and 1:SIB2_B2_8MHz_I moved from 90 Hz to  150 Hz ; maximum frequency of V1:SDB2_B1s1_PD1_Blended increased from 1 kHz to 2 kHz ; notches around 19.1 Hz removed from the subtraction of V1:SNEB_B7_DC and V1:SWEB_B8_DC. At first glance, the range is improved by ~0.5 Mpc.

HRecClean, HrecPD1 and HrecPD2  were restarted with an updated version v3r09p1 instead of V3r09) allowing the process to reach the Golden state (and the process to be green in VPM). HRec process will be restarted with this new version when the ITF will be unlocked during the on-going calibration period.

The configuration of the main HRec process will be updated in the coming days after tuning of the noise subtraction parameters.

The process HRec has been restarted this morning with the updated version (v3r09p1) to reach to the Golden state.

Its configuration has also been updated to better subtract some noise following the previous entries by Michal. 

Last night, HRec process was running with its usual configuration, while HRecClean process was modified at the level of noise subtraction configuration:

• the BNS range is improved by 0.5 Mpc with the new configuration (see figure 1), mainly related to better noise subtraction in the range 90 to 150 Hz (see figure 3 in particular).
• between 1 and 2 kHz, the benefit is not clear, and there is at least one period when using B1s1  in this range adds noise instead of reducing it (see figures 4 and 5). Note that in general the bumps seen in figure 5 are not present and no difference was seen between Hrec and HrecClean, and that few minutes earlier the data shown here, the bumps seen in figure 5 were present in both Hrec and HrecClean channels, and in this case were slightly reduced in the HrecClean channel.

As a consequence, HRec configuration has been updated and the process restarted this morning around 9h UTC (while ITF was unlocked during commissioning period) with the following modifications:

• edge frequency between B4 and B2 noise subtraction moved from 90 to 150 Hz
• remove the notch at 19.1 Hz in the B7 and B8 noise subtraction