We worked on the external SQZ bench to tune the MM telescope by observing SQZ level and B1_7MHz magnitude, and to check the effect of alignment on noise scattered light noise in Hrec.

We started at 9:50 UTC by increasing the bandwidth of SQZ AA loop with dither lines, in order to quickly recover the alignment while tuning the MM telescope.

Then we slightly mistuned the SDB1 FI waveplate in order to have more optical power (about a factor 10) reflected out from B1 to the external SQZ bench, and to amplify the noise from scattered light. In the first attachment Hrec spectra before and after waveplate mistuning.

Then we tuned the CC phase to anti-squeezing, Fiodor went to DET lab and tuned the translation stage of the output mirror of the MM telscope on external SQZ bench. After each step of the translation stage we let the AA with dither lines stabilize the alignment of SQZ on OMC, and we checked the phase to maximize the anti-squeezing level.  We moved the MM telescope mirror back and forth by approximately +/-3 mm; in the end we tuned the CC phase to squeezing and did one last step of translation stge to optimize SQZ level. The final position was about 1 mm away from the original one. Figs. 2 and 3 show the evolution of anti-squeezing and squeezing level during the tuning, together with B1_7PD1_MHz magnitude and one channel of the slow AA sensing. The tuning apparently resulted in a slight improvement of the squeezing level by about 0.2 dB, see Fig. 4.

After the last setting of the MM telescope axial position, we slightly tuned the horizontal angular tilt of the MM mirror to reduce the noise from scattered light in Hrec.

To understand the effect of scattered light we took a reference Hrec spectrum with the slow shutter closed. We finished the tuning at 13:21 UTC and we tuned back the SDB1 FI waveplate, but we later realized that the tuning was not optimal - we tuned again after the calibration at 19:20.

In Fig. 5 the Hrec spectrum at low frequency is compared in different conditions:

red: SQZ injected, in the morning before the start of the tuning

green: slow shutter closed

blue: SQZ injected, after the tuning.

No significant difference above 20 Hz.

We found two issues with the SQZ_PLL_AA server:

1) quite often some commands are note executed, both if sent through VPN or via python interface. In particular, it's not easy to act on the dither lines frequencies.

2) it seems that the demodulation phases at lines 53, 61, 67 and 73 of AA Matrix are not applied