I got 2 horizontal L4Cs under HAM1 this week and was able to get measurements to try designing some filters for X  feedforward this week. It's possible to get nice improvements in HAM1 motion above 1 hz to beyond 10 hz, with some improvements in ASC over those frequencies, but it's difficult to get filters that don't make the low frequency motion worse. When I tried a filter like Huyen's , I got broad excess noise below .1hz. When I high pass the filter, I get different injection at the high pass, I think I can put this in a place that doesn't affect the IFO, but it will take some tuning.

First plot compares the LLO filter (yellow, scaled to my HAM1 data), the filter I have running in HAM1 right now (red) and the ratio of tfs for the ff filter design (blue).

Second plot compares some asds with the current filter running. For the top subplot, red and brown are with X ff off, green and light blue are on. For the other 2 subplots, red is X ff off, green is X ff on I'm able to get good improvements above 1hz in the HEPI channels, and CHARD pitch even sees some improvement in places, maybe a factor of 2ish around 6hz. Some of the other ASC seem to see small improvements as well. But the feedforward also causes some excess noise below 1hz, which all of the asc channels seem to pick up.

After talking with Arnaud a bit, I think I may have just found a cause for the noise below 1 hz. The hepis all have a "twist" path which is a feedforward path that is needed to compensate for the HEPI actuator drives causing the crossbeams to bend, tilting the horizontal L4Cs. This path is only used on HAM1, because HAM1 is the only chamber that uses the L4Cs in-loop. The error point for this should be the cartesian drive and it is currently read at the output for the isolation loops, shown in the third image. Because we've never used the 3dl4c path before no one had thought about the fact that this path should readout after the add block which sums the output of the iso loops and the output of the 3dl4c feedforward path. This means the total drive going signal to the twist path doesn't account for the 3dl4c drive signal, so it's probably not being fully subtracted from the HEPI blended l4cs. Fourth attachment shows how this path should be wired up. HAM1 has an active ECR, so I'll fix HAM1 for on Tuesday (which has it's own library part), but this should just be fixed in the general HEPI master part.