Jeff, Evan
The current CW hardware injections into DARM control are too loud to allow transitioning from ETMX to ETMY. So we turned them off around 21:00:00 by turning off the input to CAL-INJ_HARDWARE.
On ETMX, the loudest line (at 1395 Hz) has an rms of about 2 counts. For ETMY (with the low-pass filter on the low-noise driver), one must actuate 50×(50/2.2)2 ≈ 26000 times harder at high frequencies. That means the ESD will have to push 50000+ counts rms. The attached plot shows the DARM drive at a few points along the actuation chain (only ETMX was used as an actuator). The hardware injection does not show up in DARM_OUT, or in the version of DARM_OUT that it sent to the CAL model. It does show up in the ETM drives. The attached screenshot helps us remind ourselves of the relevant model topology.
Also note that this is happening at frequencies above the Nyquist of the quad drive DQ channels, which is currently 1 kHz. That means it would have been much harder to catch this if we had been trying to do a lockloss analysis after the fact. The same goes for instances where the drive is being saturated by higher-order violin harmonics, as has happened in the past.
The line amplitudes were bumped up by a factor of 10 for last night's test. Can we remove that factor of 10 to see if you still see problems? We can go down still further for the higher-frequency lines.
Yes, factor of 10 lower is probably fine. I turned injections back on with a gain of 1 instead of 10.