I'm not sure if the ESD changes today (alog 25468) are at fault, or if something else has changed today, but I cannot lock ALS Diff using the nominal loop.
In order to lock ALS Diff, I have included a gain of 0.001 in the DARM2 filter bank (usually just has gain of 1, no filters yet). Also, I have commented out the line in the Diff guardian that turns on DARM1 filters 2 and 3 (an integrator and resG respectively). I can turn the DARM2 gain up to 0.002 or 0.004 most of the time, but at 0.006 the DARM loop starts to ring up over 2-3 seconds until we saturate the ETMX L3 coils. Even at the very lowest teensy-tiny gain, I can't turn on either the integrator or the resG.
There is so very little gain here that I can't measure the DARM loop in the 10-100 Hz band.
Unsurprisingly, this means that Diff is too noisy to hold the Yarm anywhere in particular, so I can't move on with the locking sequence (with the thought that it would all be okay if I could get far enough to transfer to RF DARM).
I am reverting my changes (uncommenting line 228 in ALS diff guardian and gain to 1 for DARM2 bank).
I'm calling it a night since I'm running low on ideas.
Also, Kiwamu straightened me out about the DARM filter module screen situation - the auto-generated screens live in the OMC folder. He'll fix the overview screen tomorrow.
One should check for high frequency saturations. The ALS DIFF signal is fairly noisy and with the new wide bandwidth ESD we might have too much noise at ~kHz..
The digital compensation logic for the high-range ESD state is not correct.
In the high range state, the low-noise path is disconnected by a switch, so the analog PI RC network does not enter the overall TF of the driver. Therefore, we do not need to compensate its p/z pair (now 3.2 kHz / 67 kHz). However, the digital logic will still engage this p/z compensation unless you manually disable it (i.e., turn off the antiAcq filter in the ESD SFMs).
Turning off this filter allows DIFF to lock.
During the run, we ran with the binary I/O logic disabled (state request = −1), and the antiAcq filter off. This was changed during maintenance day on 19 Jan, after which point we were incorrectly compensating for this nonexistent p/z pair. This probably also explains the continual EX saturations that appeared during DRMI locking a few weeks ago.
It should have been digital sign flip.
Old filter was also incorrect in EX and was much more aggressive than the new one, so the HV driver saturation is not the reason that the IFO didn't lock.
The real issue is, when I changed the number of poles (from 1 to none because it went too high due to the hardware change), the sign of the digital TF unintentionally flipped. It's hard to read the sign flip by just looking at the gain-normalized zpk expression in foton (i.e. "n (Hz/norm)"), you need to look at something else e.g. "s" or "f" expression or the bode diagram:
Old one: zpk([152], [3250], 1, "n") in Hz/norm = zpk([-955], [-20420], 21.38) in s (red in the attached)
New one: zpk([3225, [], 1, "n") in Hz/norm = zpk([-20263.27], [-32768], -1.617) in s (blue in the attached)
This of course applies to both EX and EY, so even if with the filter disabled for EX, handing off to EY would break the lock.
New filter with correct sign was made, saved and loaded to the frontend: zpk([3225, [], -1, "n") in Hz/norm = zpk([-20263.27], [-32768], 1.617) in s.