Keita, Georgia, Jenne, Sheila
We spent a bit of time today investigating the lockloss which occurred when we tried to re-AC couple the ISS second loop. During lock acquisition the second loop is engaged with its AC coupling servo on. Then in LASER_NOISE_SUPPRESSION we DC couple it, by holding the output of the AC coupling servo. In order to increase the input power to the IFO once we've already reached nominal low noise, we need to be able to re-engage the AC coupling servo, which it turns out is very fiddley.
In December last year Stefan cleaned up a guardian state to do this, it worked then when he tried it with the IMC. What this state actually does is:
- servo the offset of the reference_servo such that the output is zero
- servo the offset of the AC_coupling_int (integrator) such that the output is the same as the previously-held output (i.e. the value that we held the output to be when we DC coupled the ISS)
- re-engaging the AC_coupling_drive, but with a 1Hz high-pass filter in the AC_coupling servo (FM6), which essentially AC-couples the AC-couping servo, i.e. leave the ISS DC coupled
- Turning off FM6 (note it has a ramp time of 15 seconds), so the ISS is now fully AC coupled
Checking out the guardian log (first attachment), today we survived the first 3 steps, but lost lock straight after the 4th step of turning off FM6 to actually AC couple the loop. Keita did some sleuthing and found out why. There appears to be an offset on the input to the AC_COUPLING_SERVO (brown trace in second attachment). When we turn off FM6 (blue trace shows the switch stat), the output of this servo (orange) is changes to make the input zero. Unfortunately this drive to the AC couple servo causes a large change in the output of the AC coupling drive (red), and coupled to the ISS first loop, driving the IFO input power down.
Tomorrow we want to step through the guardian state by hand and manually add offsets (eg to the SECONDLOOP_EXCITATION_OFFSET channel) to bring the AC_COUPLING_SERVO input closer to zero before turning off FM6. If that works by hand we can add another adding-offsets-to-drive-something-to-zero type servo to the guardian.
I seem to remember there is also an output offset which depends on the gain slider.
It looks like we do increase the second loop's output gain slider when we DC couple / do the LASER_NOISE_SUPPRESSION state, so we should bring that back down by 5dB to be the same settings as when we originally did the DC coupling.
[Georgia, Jenne]
We tried to run the AC coupling state as Georgia writes above, doing each step by hand, including reverting the output gain slider as Daniel suggests. This time, turning off the FM6 worked fine, but we lost lock when we turned the reference servo back on. We realized that when the ISS gets DC coupled, the reference servo's gain is significantly increased, by changing the value of the input matrix from 0.5 to 125. So, when we were re-engaging this servo but the ISS was AC coupled, this reference servo had 200x too much gain. We now have the gain set back to the lower value in the AC coupling state, and have been able to successfully go back and forth between DC and AC coupling in an IMC-only state at 35W.
We also noticed that there was a 20 second sleep between turning off FM6 (which has a 15 sec ramp time) and engaging the reference servo, and during this time the diffracted power was going concerningly close to zero. It seems that this wait should be non-zero, but not that long. We tried once with a 5 second wait, and that seemed better, so we've shortened it down to 1 second.
We believe that we should be able to re-AC couple the ISS once at full power now. (Although, we've thought that before....)