This is a comparison of data in LN3, for the analysis of the new DIFFpTX control.
Fig 1: this morning, in the latest lock before the change of DIFFpTX control, the microseism was a bit higer in the critical region around 0.75 Hz. The following comparisons should take into account this difference.
Fig 2: The locking correction of West Arm sees this difference: the amplitude of the payload resonances are lower by a bit less than a factor of 2.
Fig 3: Pitch on COMMp angular dof is lower too; here the difference is a bit larger.
Fig 4: On DIFFpTX the difference is more evident: up to a factor of 3. This should be due to the improvement of the control filter.
Fig 5: One should expect an improvement of dark fringe stability, but not so much.
Fig 6: In the morning data, the 2D graph DIFFpTX Vs B1p seems to say that B1p fluctuation is entirely due to DIFFp TX. The same analysis applied to the afternoon data does not agree with the previous one: it seems that the coupling has changed a lot.
Fig 7: There is a simple explanation: ASC_DIFFp_TX_CORR is the signal after the loop gain, but the gain has changed by a factor of 2. I had forgot about the automatic adjustment of the gain. In principle, this adjustment should guarantee a sort of constant calibration for ASC_DIFFp_TX_CORR: for this reason I prefer to use the signal after the gain to make quantitative comparisons. But in this case the assumption is clearly wrong.
Fig 8: The 2D anaysis done with DIFFp_TX before the gain. Here the calibration is more realistically the same. The large improvement of DIFFp_TX and B1p_DC is not only due to the change of control filter, but also to the increase of the gain.
We have learned that the gain was lower than the optimal one for a good control accuracy. It is not clear why with different controls the adjustment of the gain works in a different way. By the way, looking at older data, it seems that in the recent data the gain was lower than usual.
The last point: the new filter has a roll-off not well refined and the noise at 20-40 Hz could be too high. A new version has been downloaded. No difference is espected in the control band, but we will check in the next lock acquisition that everything works fine.