I tried to feedback a part of the PZT correction (sent to the main laser to lock it to the FC) to the end mirror. The goal is to reduce the cavity motion at the logitudial resonance frequency (around 1Hz), acting on the mirror in order to reduce the correction sent to the laser. In fact in this region, the correction increases the frequency noise of the main laser, inducing higher PLL noise.
The PZT correction is taken from the channel "pzt mon" of the rampeauto (which is 100 times smaller that the real correction), it is sent to the ADC of the DGS where it is digially filtered and then the correction is summed to the one already sent to the end mirror coils (from anguar damping).
We observed that the signal seen by DGS has a very strong 50 Hz, which is not present when locking at it with the oscilloscope.
In order to select the part of the spectrum aroun the 1Hz, I used a filter like this:
| Freq (Hz) | Q | |
| zero | 0 | simple |
| pole | 1 | simple |
| pole | 20 | simple |
| pole | 20 | simple |
The simulated openloop TF is shown in pic1. Where as a mechanical TF I used the one measured in entry #1506
Pic.2 show the spectrum of the PZT correction with (bue curve) and without (red curve) test mass feedback engaged. It can be seen that the correction to PZT is reduced by a factor ~10 in the 1Hz region when the feedback is engaged (blu curve).
To be done:
- Amplify the signal before the ADC with a stanford to better exploit the ADC dynamics.
- Check the RMS improvement and optimize the correction filter (an overshoot is now visible at 2-3 Hz).
- Add a notch filter to avoid feeding back the 50 Hz and to be able to check the signal improvemt in time (currently it is fully dominated by the 50 Hz).
- Check wheather the PLL noise is reduce when the loop is engaged.
