Earlier today the TCS team was working to center the WI central heating, and we noticed a tremendous excess of noise after the WI had thermalized with ~1W of CO2 power. The sideband power degraded and the SSFS UGF went down, and the range dropped due to broadband noise across the entire spectrum. See Figure 1 for a timeline. After the central heating was turned off, the range recovered, but dropped again - not sure what was going on here.
Based on this hint (that some of our noise can be changed by adjusting TCS), we examined the TCS differential tuning. With a simple Michelson locked, we turned on the WI central heating, and we were able to minimize the contrast defect with 0.28W of power, judging from the pickoff. This is in agreement with the prediction from previous measurements and from modeling, to correct for the cold lens in the substrate. Eventually we achieved a contrast of around 0.1%. Then, we tried to cross-calibrated the NI central heating power, by spoiling the contrast and then correcting with more WI central heating. We found that ~0.74W of NI central heating was balanced with about 0.9W of WI heating. However, the CO2 beams appear to be misaligned on both mirrors: the WI central heating introduces a horizontal asymmetry in the B1p camera image, and the NI central heating generated a vertical asymmetry. It is probably easier to align the central heating using this figure of merit (contrast defect pattern), rather than the hartmann sensor. Paolo enabled the BS drift control so we maintained a good alignment throughout the shift. Figure 2 tells the story of the differential tuning and the cross-calibration.
* Note, during the MICH locking we rang up the BS mode at 1.8kHz, this generated some confusion as we turned on the TCS because the downsampled B1p_DC power was monotonically increasing along with the mode height. We doubled up the notch in the MICH actuation at this frequency and the mode decayed quickly.
With the TCS still engaged we tried relocking the ITF, but the gains for the MICH offset reduction sequence had changed and we could not make it to an offset of 0.1. This was a struggle, and while we believe that it would be interesting to explore central heating to see the effect on the noise, it will clearly be no fun to re-tune the locking sequence. After some sleep we might approach this in a more systematic way, e.g. measure loop shapes at each step and understand how the control signals are changing.
We have turned off the TCS (dialed down the waveplates and flipped the mirrors), and reset the gains in the locking sequence. The normal behavior was immediately restored, and we reached DC readout. --> TCS is powerful, but reversible.