In a follow-up to what TJ did yesterday(73685), I alternated the CARM gains (H1:LSC-REFL_SERVO_IN{1, 2}GAIN) between 6 (what it was nominally at) and 12 every two minutes.
Times when we were at each gain setting (in UTC):
12: 20:08:17 - 20:10:17
6: 20:10:29 - 20:12:29
12: 20:12:44 - 20:14:44
6: 20:14:58 - 20:16:58
Attached is an ndscope over this timeframe looking at the CARM gain and SQZ BLRMS channels (same as the channels that TJ and Camilla had been looking at yesterday(73682)).
Attached is our very high frequency monitor of the DCPDs. (Since this channel is not saved, I also upload the DTT file).
The red / pink traces are at our nominal CARM gains of 6dB on each slider (H1:LSC-REFL_SERVO_IN1GAIN and IN2GAIN), and the blue / teal traces are the higher values of 12dB on each slider. All data is 10 avgs at 1 Hz BW, 25th Oct 2023. Times of the data are in the legend for each trace.
It seems that increasing the CARM gain is beneficial to our noise between 6 kHz to about 25 kHz. From about 25 kHz to a little above 100 kHz it increase our noise a bit. Above 100 kHz we're limited by some other noise, and so not seeing the effect of frequency noise.
I think it seems fine to leave CARM with the higher gain.
I've just modified LASER_NOISE_SUPRESSION to increase the sliders to 12dB (it used to increase them up to the previous nominal of 6dB at the end of the state). I've just increased the sliders before we go back to Observe, so we'll have lower frequency noise for all our upcoming locks.