Alexa, Elli, Sheila, Kiwamu, Rana, Evan
With CARM controlled by a combination of sqrt(TRX+TRY) and TR_REFL9 (i.e., REFL9I normalized by TRY), we've achieved a power buildup in each arm that is 800× the single-arm buildup. We believe this corresponds to a CARM detuning of 5 pm.
At this point, REFL_A_LF is at about 2.3 mW, compared to 40 mW when the arms are anti-resonant. If this is truly an indication of the power reflected back from the PRM, this means that the visibility of the interferometer is in excess of 90 %. We do not have TCS on today.
One major roadblock to further progress seems to be angular instability that is seen at the AS port. We have ASC feedback from AS_B_45_Q to ETMX and ETMY, but the bandwidth is too low to suppress the 1 Hz motions that we see. We may need to spend some time increasing the WFS bandwidth, or resort to more aggressive oplev damping.
A laundry list of things we've done today to troubleshoot the transition:
Based on the measured losses in the arms (an average of 110 ppm), we expect an interferometer recycling gain of 32 W/W and a coupled cavity pole of 0.6 Hz (i.e., 9 pm) [using eqs. 6 and 8 from Fritschel et al. 2001]. Therefore, we believe that the arm buildup should be about 1000× the single-arm power with CARM at 0 pm. This is consistent with earlier estimates made by others (LHO#15390, T1000294).
A list of transition steps not yet implemented in the guardian:
Our last three attempts are 11:24:48 UTC, 11:00:27 and 10:37:35 Jan 31.
On oplev damping:
* I turned on the usual f^1 velocity damping, but since it measures the mirror angle and feeds back to the PUM, this is not a very good loop. Turning up the gain excites the 2nd pitch mode.
* In some of these suspensions, there are bad, high-Q filters called 'invP2P' or 'invY2Y'. These kind of invert the plant, but cause us much trouble since they ring for so long. These are not appropriate for the kind of system we have where the plant Q changes (due to damping loop changes) and frequency shifts (due to radiation pressure). With a little sloppy plant inversion we could have a higher BW and squash all of this excess pitch motion during the lock acquisition. Probably much easier than getting high BW WFS at this point.
On the third attachment above, you can see the unfortunate wire heating issue: the SR3 drifts by ~0.5 urad in 4 minutes after the lock loss. It also exponentially drifts up once we ramp up the power in the IFO.
The dark port is not very dark here, but still this seems strange. Some related info on this from LLO by searching for the phrase 'PR3 drift'. Hopefully we can combat this with some 'cage servo' or WFS rather than make an SR3 wire baffle.