OMC1_err: 3.3e-7 +/- 0.1e-7 m
OMC2_err: 7.0e-7 +/- 0.1e-7 m
Figure 1, in purple is a time without an offset added to the error signal and in blue is with the offset of 5e-5 in OMC1_err. This offset of 5e-5*3.3e-7 = 1.65e-11 m is consistent with the previously estimated l_RMS of 2e-11m.
Clearly B1_f1_i (from which OMC1_err is derived), goes from a mean of zero to some offset value, as expected.
But with the offset the height of the 119Hz calibration line in B1 PD1 goes down by a factor 10, which would suggest that B1_f1_i doesn't have its zero in the right place.
However at the same time the height of the 119Hz line *increases* in B1s1 (and the offset in B1s1_f1_i increases), so the two error signal (transmitted and reflected), don't agree on what is the right offset to reduce the coupling between OMC length and transmitted power fluctuations.
The value of that offset has not changed when decreasing the two OMCs modulation depth. Something we haven't studied is if the needed offset to minimize the coupling depends on frequency, we should try next time to inject broadband noise and see if offset changing has the same impact on the length to power coupling in the whole 10Hz-1kHz band.
For the last calibration done in the 2 minutes starting from 1207502570:
OMC1
l_RMS = 2.1e-12
l_cal_mod = 2.9e-13
l_err_mod = 8.4e-13
OMC2
l_RMS = 2.5e-12
l_cal_mod = 6.8e-13
l_err_mod = 5.1e-13
So there seems to be some residual offset for both OMC1 and OMC2 that dominate the estimated l_RMS, but the modulation of the calibration line and the error signal are only a factor 2-3 smaller than that.
Figure 2 and 4 shows the calibrated spectrum, where clearly the OMC2 lock is much poorer with a precision of 1.1e-12 m instead of 2e-13 m for OMC1.
Looking at the time series 3 and 5 there is some slow oscillation (period ~ 1minute) in the OMC2 lock.
Figure 6, looking at half an hour of data, this oscillation is present in both OMCs, but happens to be in a larger swing for OMC2 during this calibration measurement (from 17:22:30 to 17:24:30). The two oscillations are quite correlated, but not fully, so it is probably not an issue with the frequency of the input light, but maybe gain peaking at the cross-over between Peltier and PZT actuation. To be further investigated.
