Ivey, Edgard
Follow up to Oli's estimator test [LHO: 85615].
Ivey took the data for the three tests that Oli ran and got some spectra to show that the estimator is working as expected. We can see three traces in each one of the figures attached:
- The Blue traces are using light damping, that is, using the OSEMs as we have them right now, but dividing the M1_DAMP_Y gains by a factor of 5. [Light damping]
- The Red traces are the current settings. [ OSEM only damping ]
- The Yellow traces are the estimator damping. This tries to get the best of both worlds by using the HAM5 GS13 and the models we posted in [LHO: 85446] [ Estimator damping ]
The first figure shows the M1 Yaw signals as seen by the M1 OSEMs. As you can see, the spectrum is completely dominated by sensor noise, except at the suspension resonances. Under light damping, the OSEM actually sees the suspension modes ring, but under 'OSEM only' and Estimator damping, the resonances don't show up anymore, as expected. We note that there is a bit of the 3 Hz resonance that seems to not be fully damped by the estimator.
The second figure is one of the money plots. It shows the M3 Yaw as seen by the SR3 OPLEV. In it, we can see that the OSEM only, and the Estimator damping are both successful at damping down the resonances of the suspension. This means that the Estimator is working as intended. We note that in the estimator case, there is a bit of excess noise at 2 Hz.
The last figure shows the total drive request for the M1 OSEMs in drive cts. As expected, the Estimator is requesting less drive above 5 Hz when compared to the OSEM only damping (which is the current setting). Since this is completely aligned with our expectations, we believe this implies that the estimator is working as intended and we are likely reducing the SR3 M1 Y OSEM noise injection into M3 Y by a factor of 5 at 10 Hz and above.
We also note that the drive request at 2 Hz has a bump likely due to model mismatch. This effect is likely the cause for the excess 2Hz noise seen in the OPLEV signal. We might need to model it to see why it would lead to a coherent addition of motion that can be captured by the M3 OPLEV.
Since this test was so successful, our expectation is that we will be trying to affix the OSEM calibrations, a Yaw estimator, and then a Pitch estimator for SR3 in the next few weeks. Stay Tuned!
