The objective of the shift was to prepare the swap of the IMC control loop from the RFC error signal to the CARM_FREQ beating signal.
We also wanted to characterize the beating to CARM/DARM loops by injecting noise. The shift was therefore divided in two phases.
We started the shift by locking the ALS loops with reallocation and drift control engaged, and then switched to the CARM and DARM basis using the beating signals.
After engaging reallocation and drift control on the arms, we implemented the updated control filter prepared by Ruggi for the reallocation and CARM DARM loops (switch performed at 16.30 UTC, see figure 6). The filter change involved moving a zerofrom 0.15 Hz to 2 Hz and allowed increasing the gain from the initial +/-50 for NARM/WARM to 240. This seemed to reduce the low frequency (<1Hz) oscillations of these signals and increase their accuracy but figure 5 shows an increase in the high frequency components of the correction signals when the filter was swapped. A further analysis of this swap might be necessary.
To prepare the handoff to the IMC loop, we performed a scan of the setpoints of the control loops for CARM and DARM to find the IR resonance peaks on B7 and B8, thus obtaining a condition analogous to locking the arms on the IR before enganging CARM to SSFS.
At this point, a sudden unlock of the RFC (17:43:30 UTC, Fig.1) provoked the unlock of the green. We tried to lock it again, but the procedure kept failing at the moment of engaging the reallocation.
We concluded that the new filter, while being effective at reducing the noise after the reallocation loop was closed, made the lock more difficoult.
We decided to phase back to the previous filter to engage the reallocation, and let the CARM and DARM loops change the filter, with the new gain, at the moment of their activation before swapping their input signals.
This proved successful and we obtained a stable lock of the green with reallocation which lasted, with few interruptions, for the rest of the shift. Notice that we opted not to engage the drift control.
We then completed the tuning of the CARM/DARM setpoints (around 19:10 UTC, Fig.2) in order to find the IR resonance. We obtained a tramsission for the IR of around 0.4W on B7_DC and B8_DC, later (around 19:24 UTC, Fig.2) increased to 0.6/0.7W with:
- DARM_SET = -118.4685 (corresponds to -36837 Hz on the LSC_DARM_FREQ_ZEROED channel when divided by the calibration factor)
- CARM_SET = -36.0273 (corresponds to -11202 Hz when divided by the calibration factor)
The calibration factor is 6.7e-4*4.8, where 6.7e-4 is the gain found as explained in the logbook entry #50897, and 4.8 is the new gain factor related to the new filter.
At this point Ruggi started injecting noise on the RFC at 19:13 UTC for 200s to obtain a calibration for the signal swap. The swap itself requires to understand the logic of INJ loops in Acl, and for the moment we decided to postpone it and proceed with the characterization of the CARM/DARM loops in the second part of our shift.
We performed the following noise injections. Due to an error in the command sent to ACL to set the noise filters, we used the "flt_NONE" to filter white noise, which has a pole at 10 Hz and a gain of 1e-20. The very low gain of the flt_NONE explains the huge coefficients that we used.
- 20.48.00 UTC: START test real noise injection on CARM.
- 21:12:20 UTC, GPS:1298668358: Amplitude: 2e18 (5 min)
- 21:18:00 UTC, GPS:1298668698: Amplitude: 2.5e18 (1 min)
- 21:19:18 UTC, GPS:1298668776: Amplitude: 3e18 (5+ min)
- 21:27:00 UTC, GPS:1298669238: Noise OFF
In Figure 3 are reported the transfer functions of the two injections of duration 5 minutes (respectively 21:12:20 UTC and 21:19:18 UTC) and in Figure 4 the correspondent FFT of the noise injected.
Due to an error in the DARM noise injections, the NI local controls opened, compromising this operation, that will be planned for the next one.
We left the interferometer with the ALS loops in DOWN and the arms locked on the IR.