Today we closed the SQB1 bench control loop using the signals provided by the PSDs.
At the beginning we had were not able to have the filter cavity locked due to activities on the ITF.
The strategy we implemented is to add the error signal due to the displacement of the beam on the PSD to the LVDT error signal.
We did the test only on TX and TY, using the reconstruction matrix B.
ACL_SUM_CH LC_TX_err_pre "" 1 1 LC_TX -1 LC_TX_set 0 NOISE_flt 0 spiral_TX
ACL_SUM_CH LC_TY_err_pre "" 1 1 LC_TY -1 LC_TY_set 0 NOISE_flt 0 spiral_TY
ACL_SUM_CH LC_TX_err "" 1 1 LC_TX_err_pre -1 LC_TX_PSD_err
ACL_SUM_CH LC_TY_err "" 1 1 LC_TY_err_pre -1 LC_TY_PSD_err
At the beginning we centered the PSDs with the picomotors and we adjusted the set point by hand to the TX and TY recontructed with the PSDs, in order to have TX TY signal reconstructed with the PSDs close to zero.
We added the buttons on the vpm to engage the loop on SQB1_LC process:
- Open PSD bench control
- Close PSD bench control
We were able to close the loop with the Green pointing control OFF (5 mins of data):
2022-05-30 13h11m05 UTC bonnand 'Close PSD bench control [-1:Disable,1:Enable=1]' sent to SQB1_LC
We tried to close the loop also with the Green pointing ON:
2022-05-30 13h16m41 UTC polini 'GREEN AA Pointing:ON [-1:Disable,1:Enable=1]' sent to AFC_Ctrl
But it is not working since the two loops are working one against the other and the bench is drifting when M7 moves.
We tried to characterize the PSD loop by measuring the transfer function. We added the possibility to add the noise
ACL_FILTER_SET "flt_NOISE_PSD" 1 1 0.01 10
ACL_FILTER_POLES "flt_NOISE_PSD" 0.1 1
ACL_NOISE_CH NOISE_corr_PSD "um" 1 LC_LOOP_FREQ 5.0 0
ACL_FILTER_CH NOISE_flt_PSD "V" 1 LC_LOOP_FREQ NOISE_corr_PSD 1 "flt_NOISE_PSD"
ACL_SUM_CH LC_TX_PSD_err "" 1 1 LC_PSD_TX_err 0 LC_TX_Blended 1 NOISE_flt_PSD
ACL_SUM_CH LC_TY_PSD_err "" 1 1 LC_PSD_TY_err 0 LC_TY_Blended 0 NOISE_flt_PSD
We injected noise in TX PSD at 14h55m43 UTC with amplitude 5.
We found an UGF < 10^-3.
We stopped the noise injection on the PSD error signal at 15h56m UTC (1 hour of data).
We left the bench controlled only with the LVDT.
Next steps:
- to understand how to make the pointing and the bench control with PSD working together.
----------------------------------------------------------------------------------------------------------------------------------
We also implemented the code on SQB2, once we optimize the loop on SQB1, we will go on also with SQB2.