[Obata, Tanaka, Yokozawa, Komori]

Abstract:

We successfully reached the observation state by tuning the DARM gain; however, the lock was maintained for only several tens of minutes.

Details:

Initially, we attempted Q-measurement of the sapphire fiber violin modes in the RF-locked state.
However, we found that the LSC CARM channel, which had been empty and used for the ringdown measurement, no longer had a DQ channel because it had previously been deleted to save DGS memory.
Although the measurement could still be performed using real-time monitoring, we instead focused on recovering the DC readout today.

The IFO was unstable because the ETMX length RMS (ETMX_ISCWD_WD_AC_RMS) frequently exceeded the watchdog threshold of 30,000 counts due to large suspension motion around 2 Hz, particularly at higher input power.
In addition, as reported in klog:35814, the DARM loop tended to oscillate at 30–40 Hz after the handover to the OMC.
This suggests that the DARM gain may have been too low, causing the unity gain frequency to drop to 30–40 Hz, where the phase margin is insufficient.
These issues were the main reasons for the failure to achieve a stable full lock.

First, we changed the target angle of the HWP in the increasing las power state from 183° to 173° in the LSC guardian, which reduced the final input power from 15 W to 13 W.
Second, we increased the gain of ETMY_NBDAMP_L2 by 6 dB while monitoring the open-loop transfer function (green: previous, red: current in Fig. 5).
These changes reduced the frequency of watchdog threshold crossings.

Next, we measured the gain balance between the RF and DC DARM signals at an input power of 13 W.
The transfer function between the two signals was measured around 10 Hz, where it showed high coherence and good reliability (Fig. 1).
We found that the DC gain was smaller than previously, so we increased LSC-OMC_DC_GAIN from 1 to 1.5.

With these changes, we successfully reached the observation state twice, at approximately 10:05 UTC and 11:26 UTC on December 10.
The lock durations were about 20 minutes and 40 minutes, respectively.
The open-loop transfer functions after these modifications are similar to those measured previously (DARM: Fig. 2, PRCL: Fig. 3, MICH: Fig. 4).
Further investigation will be required to achieve longer lock durations.