J. Kissel, E. Hall, J. Driggers

Back in July 2016, Evan Goetz and Evan Hall redesigned the ETMY L2/L3 (PUM/UIM) actuation hierarchy filters to improve the rolloff of the PUM stage (see LHO aLOG 28746), as a result of the calibration group's critique (G1501372). As a refresher, I attach two select .pdfs from Evan's entry that shows the difference that the new crossover makes.

These filters were installed and turned on the same day as the design aLOG (July 29 2016 21:57 UTC), and ran until late August (Aug 27 2016 10:57 UTC -- Friday Night / Saturday Morning at 3:57a local time), when they were somehow reverted back. Sadly, the configuration with the better filters seems to then come and go for the next few months, 
  - Initial Install Jul 29 2016 21:57 UTC (Friday  , Jul 29th, at 02:57pm PDT)
  - lost on         Aug 27 2016 10:57 UTC (Saturday, Aug 27th  at 03:57am PDT)
  - fixed again on  Sep 10 2016 19:32 UTC (Saturday, Sep 10th, at 12:32pm PDT) 
  - lost again on   Sep 16 2016 03:09 UTC (Thursday, Sep 15th, at 08:09pm PDT)
  - fixed again on  Sep 30 2016 15:12 UTC (Friday,   Sep 30th, at 08:12am PDT)
  - lost again on   Oct 01 2016 17:50 UTC (Saturday, Oct  1st, at 10:50am PDT)
and it has stayed this way since. Notice that none of these correspond to a maintenance Tuesday; they correspond to afternoon/evening/late-night commissioning times.
 
It looks like, currently, the old configuration (see attached for MEDM ScreenShot aide showing this bad configuration) is currently hard-coded into the run portion of the ISC_LOCK guardian's LOWNOISE_ESD_ETMY state, 
    3540        yl2da = ezca.get_LIGOFilter('SUS-ETMY_L2_DRIVEALIGN_L2L') 
    3541        yl3da = ezca.get_LIGOFilter('SUS-ETMY_L3_DRIVEALIGN_L2L')
     ...
    3567        yl2da.only_on('INPUT', 'FM2', 'OUTPUT', 'DECIMATION')
    3568        yl3da.only_on('INPUT', 'FM3', 'FM4', 'OUTPUT', 'DECIMATION') #KI ellptic FM7 added Aug-17-2015
     ...
    3587        self.yl2da = ezca.get_LIGOFilter('SUS-ETMY_L2_DRIVEALIGN_L2L') 
     ...
    3592        self.yl2da.switch('FM7', 'FM8', 'ON')
The latter two lines are switched on after a counter is set. 

I attach a 60-day trend of the bitword that represents the status of the filter banks,
          Good                    Bad
L2 L2L    37984 (FM6, FM7)       38082 (FM2, FM7 ,FM8)
L3 L2L    37912 (FM4, FM5)       37900 (FM3, FM4)
Unclear why L2, in its bad state, indicates a filter change -- as far as the current guardian code goes, the (bad) filter configuration does not change anywhere in the acquisition sequence. But, it helps identify when the changes were made. The trend is also annotated

Naturally, this is too far back in time for a viable aLOG search (I tried keyword searching for a few relevant buzz words and looking for mention of such things in the aLOG for a few days surrounding Aug 27), so I asked Jenne, who didn't remember.  As of yet, it's unclear if this was a conscious reverting and hard-coding because the IFO was non-functional in the new configuration, or if this was accidental, but it is clearly related to commissioning activity and guardian code. During commissioning phases, guardian code is not committed and quick / "small" changes are made without aLOG all the time, so I don't expect we'll ever be able to find out what happened unless someone remembers.

But, I suggest we switch back to the good configuration right after the holiday break, if and when we have an engineering run to recoup after the holidays.

Note -- the lack of upgrade has been correctly accounted for in any/all O2 calibration models, so there's no problems there. It's just sad.