Now we have a faster DHARD yaw loop during CARM offset reduction, analogous to the fast DHARD pitch loop that Elli, Ryan and I commissioned last week (LHO#16544, LHO#16520). Injecting a step response at the error point shows a loop time constant of a few seconds.
With the loop open, I injected ASC-DHARD_Y_EXC with bandlimited noise from 0 to 3 Hz and measured the transfer funcion ASC-DHARD_Y_EXC → ASC-DHARD_Y_IN1. Since we have no yaw oplev damping on the ETMs, the optomechanical plant has two moderately high-Q resonances, which I estimate to be at 0.6 Hz with a Q of 18, and at 1.36 Hz with a Q of 45.
The attachment shows the plant measurement, along with the analytical system used for the plant inversion. Here I've intentionally lowed the second Q from 45 to 20, to avoid excessive ringing in the filter step response.
Just to be clear, these are not truly DHARD loops, since we're not actuating on the ITMs. Really it's just the differential angular motion of the ETMs.
Both DHARD loops are meant to be engaged with FM3, FM4 (giving a 1/f shape overall), FM8 (plant inversion plus high-frequency rolloff), and F10 (30 Hz elliptic low-pass). For pitch, gain is 16 ct/ct, and for yaw, gain is 200 ct/ct. In contrast to what I said in LHO#16520, nothing should be engaged in the ETM L2 lock filter modules.
Additionally, these loops currently require us to disable the M0 and L1 pitch/yaw lock filter banks on the ETM suspensions. This prevents the DHARD control signals from being offloaded to the upper stages.