Tonight's shift saw some progress, here is what we did:

• first thing, we optimized the demodulation phase of the B2_8MHz signal (see Figure 1)
• initially we had a few issues with the SSFS, as we noticed after an unlock that the SSFS_B4_Corr_DSP signal didn't go to 0 after the ITF unlocked; we found that the cause was most probably the fact that the Acl filter channels used to compute the RTPC correction were never reset, so the integrators could grow indefinitely and lead to big spikes at following attempts to close the SSFS loop; in order to heal this, we modified the SSFS_Ctrl process and added the SSFS_TRIG relay channel, which is triggered asimmetrically on the value of the main switch SSFS_ON; this relay is used to drive a reset condition to the filters, which then we saw behave correctly;
• we locked the PRITF at half fringe again, and we gave a look at the B2_18MHz_I_PNORM signal, which is B2_18_MHz_I / (B1p_DC+B4_DC); this seemed good enough, even if noisier by nature, to hand-off PRCL to it; this was successfull, using a weight of -10 in the PRCL sensing matrix; afterwards, we modified the PRCL loop filter's rolloff, which is now done at 300Hz instead of 500Hz, due to the noisier signal (you can see in Figure 2 that now the loop is working much better); this worked well, so we decided to try and close the PRCL loop with this 3f signal from the start, and this has worked very nicely all night long. So now PRCL uses the B2_18MHz_I_PNORM signal (3f) by default; this has been already automated, although still in a "WIP" way: we are still using both the old PRCL gains and the new sensing weight, when things will be more fixed we will remove this duplication and rescale the gain; also, the Metatron state for this lock is the same used before, we hijacked it for the time being.
• while the PR mirror is realigning, now we engage the boost filter for the MICH loop; this has helped a bit at low frequency, and the phase loss at higher frequency is not worrying;
• as we moved towards dark fringe, we tested Michal's functions to close/reopen the B7 and B8 photodiode shutters; this has worked very nicely from the start, and now is default. In particular, the check whether the low power PDs (PD1) are open is done is the NARM_LOCK and WARM_LOCK nodes, while the automatic closing of the same photodiodes is done in the main sequence of the ITF_LOCK node, just after the switch between PD1 and PD2 in the B7B8_PD_SWITCH state;
• at this point, we managed to get to darker fringe, eventually reaching again 90% dark fringe as it was done on monday (new MICH offset = 0.1) (se Figure 3); the lock did not last much in this condition, but we saw repeatedly that the whole lock acquisition up to MICH_SET = 0.3 is rather smooth now;

As of now, as we're still building it, the procedure we found working best is the the following:

1. using Metatron, you can go up to the REDUCE_MICH_OFFSET state, which will bring the ITF to half fringe with PRCL on B2_3f, MICH on the new rescaled DC signal, CARM on SSFS and DARM still on normalized on B7-B8;
2. at this point you can reduce the MICH offset to 0.3 by hand, while tweaking the gains:
   • cm_send('LSC_Acl','AcConstChSet','PRCL_GAIN',-600.0 ,20.0,'load')
     cm_send('LSC_Acl','AcConstChSet','DARM_GAIN',0.8 ,20.0,'load')
     cm_send('LSC_Acl','AcConstChSet','MICH_GAIN',175 ,20.0,'load')
     cm_send('LSC_Acl','AcConstChSet','MICH_SET',0.3 ,20.0,'apply')
3. at this point the PD switch is needed, and this can be done with Metatron with the B7B8_PD_SWITCH;
4. now you can reduce the MICH offset to 0.2, by hand again:
   • cm_send('LSC_Acl','AcConstChSet','PRCL_GAIN',-650.0 ,15.0,'load')
     cm_send('LSC_Acl','AcConstChSet','DARM_GAIN',1.0 ,15.0,'load')
     cm_send('LSC_Acl','AcConstChSet','MICH_SET',0.2 ,15.0,'apply')
5. finally, you can reduce the MICH offset to 0.1:
   • cm_send('LSC_Acl','AcConstChSet','PRCL_GAIN',-700.0 ,15.0,'load')
     cm_send('LSC_Acl','AcConstChSet','DARM_GAIN',1.25 ,15.0,'load')
     cm_send('LSC_Acl','AcConstChSet','MICH_GAIN',200 ,15.0,'load')
     cm_send('LSC_Acl','AcConstChSet','MICH_SET',0.1 ,15.0,'apply')

The procedure will be automated as soon as possible, once the sequence is fine tuned and robust enough.

Three side notes:

• we still see the MC mirror's Z position still move a lot over time: in Figure 4 you can see the trend during the last two locking shifts;
• we still see quite a number of fast unlocks, probably due to the SSFS, as the IMC is the first thing to unlock; we didn't feel the need to adjust its gain as of now, as we didn't notice any particular instabilities; you can find an example at  3:10:29 UTC;
• we didn't use the UGF monitors so far, as we were almost constantly not in a steady state, but the monitors are very useful for live tweaking of gains, and the servos will be definitely useful once we get to a steady state.