In the context of the SRCL2DARM coupling I gave a look at the DARM spectrum in the four GPSs related to the MICH/SRCL subtractions being on/off in the tests reported in entry 65428. While no major difference was found on Hrec (as there is a "offline" subtraction there as well), some difference is present in DARM at very low frequency; in Figure 1:

• purple: MICH2DARM ON and SRCL2DARM ON;
• red: MICH2DARM OFF and SRCL2DARM ON;
• black: MICH2DARM OFF and SRCL2DARM OFF;
• blue: MICH2DARM ON and SRCL2DARM OFF (this last data actually starts at 15:07 UTC as a 25-min glitch happened in the meantime).

In both cases (even with MICH2DARM is off) DARM is better with no SRCL subtraction, so something has changed in the coupling.

Today I profited of both LIGOs not Observing to make a quick test on SRCL2DARM; in Figure 2:

• purple: 18:08:00 UTC + 120 s: SRCL2DARM gain = + 1.3 (default);
• black: 18:12:45 UTC + 120 s: SRCL2DARM gain = 0;
• green: 18:15:40 UTC + 100 s: SRCL2DARM gain = -0.5;
• yellow: 18:17:20 UTC + 120 s: SRCL2DARM gain = -1.0;
• blue: 18:20:20 UTC + 120 s: SRCL2DARM gain = -1.3.

It is confirmed that the current gain makes things worse at low frequency; no subtraction at all is the best scenario below 5 Hz, while some subtraction being engaged (with flipped sign) shows some (little) improvement between 10 and 20 Hz, at the expense of some increase of noise below 5 Hz.

I left the SRCL2DARM subtraction off, and disabled it in the automation for future locks; activity will follow in the future on the topic.