We have an earthquake mode that allows us to move into higher bandwidth interferometer controls during an Earthquake. Those controls also cause extra noise, as noted in this alog, so we are trying to subtract the injected linnear noise during this mode. Here is a trend of the seismic environment state and the sensmon range channel detector sensitivity. The plot shows that once we transitioned into earthquake mode, we were in earthquake mode for about 15 minutes before a lock loss. It is also visible that the gains are changing on the channels.

Elenna and I decided to try subtracting the noise from the interferometer controls during those 15 minutes starting at GPS 1443544132 and ending at 1443544932, see this alog for reference. I am comparing this GPS time for this earthquake to one hour before so we can know what the sensitivity should look like if we subtract all of the noise. We decided to subtract the channels whose gains changed. The channels subtracted were

H1:ASC-DHARD_P_OUT_DQ,

H1:ASC-CHARD_Y_OUT_DQ,

H1:ASC-DHARD_Y_OUT_DQ,

H1:ASC-CHARD_P_OUT_DQ,

H1:ASC-DSOFT_Y_OUT_DQ,

H1:ASC-DSOFT_P_OUT_DQ,

H1:LSC-MICH_OUT_DQ,

H1:LSC-PRCL_OUT_DQ,

H1:LSC-SRCL_OUT_DQ.

The channel with the greatest noise contribution was H1:LSC-MICH_OUT_DQ, which I determined through taking the coherence of all the LSC and ASC channels and the witness channel. The final subtraction and BNS change are visble in this GWSubtract file. Therefore, if we run a noise subtraction during earthquake mode we know we can get back at least 11 Mpc of sensitivity. This would be useful if we made a detection during earthquake mode, however we could still do better by around 70 Mpc. I plan to make this into a more usable tool and look into what is causing the rest of the noise next. I also plan to look into more channels to see if others might be causing noise as well. Since this is an earthquake mode time, it is possible the earthquake is modulating some of these channels, so a nonlinear noise subtraction may make more sense.