It's been locked for more than 30 minutes, and the beam spot is much brighter than yesterday! Another victory.

I'll attach my shoddy cell phone video later so you can see the beam motion on the ETMY. It's still moving, but it's not as bad as yesterday thanks to SUS people's effort (and of course that was essential to this long lock).

Detalis:

SUS people supplied us an updated analysis with new boost filters for each of big resonances. I played around with the gain of PIT and LONG gain a bit. When I increased the LONG gain by another factor of 10 (on top of a factor of 10 that I put in earlier), it immediately started oscillating and tripped the watchdog, but a factor of 2 (i.e. total of 20) was doing good, so I ended up with this:

H2:SUS-ETMY_M0_DAMP_L_GAIN: -20 (originally -1)

H2:SUS-ETMY_M0_DAMP_P_GAIN: -0.033 (originally -0.1)

I didn't touch Y gain even though it is with a new boost filter.

After this, I was able to see that the beam motion was reduced by maybe a factor of 1.5 or 2 by a highly scientific method of eyeballing the beam on the cage using my eyes. The DC level of the reflected beam was still fluctuating by about 20%, though (as opposed to 50 to 60% from yesterday) due to the beam motion.

Jax, Elli and I looked at the table and confirmed that the DC level of the Hartman path, which is just picked off of the main beam path upstream of the Faraday, does not fluctuate. 

After touching alignment, I felt as if the demod signal amplitude is much smaller than yesterday even though I was confident that the alignment was already reasonable. Yesterday we were getting something like 3-4Vpp, while today I got only 300-400mVpp.

I ignored that (modulation depth might have changed for some stupid reason), optimized the demod phase by using toggle switches on the delay line front panel, and plugged the CM board output to the VCO tune input. It didn't lock at first as the overall gain was too small (because the error signal is smaller), so I increaded the gain in Beckhoff world and voila!

Right now I can lock to 00 mode and 01 mode, and 01 mode is also sort of bright. Apparently we can refine alignment further, and it would be easier tomorrow as we can lock to 00 mode for however long we want.