Jennie, Jim
We tried Gabriele's newest version of the HAM1 ASC feedforward. New filters were installed last week, but we ran out of time during the commissioning window. It seems like we can still only get good subtraction from the pitch degrees of freedom, we got a lot of 10-ish hz injection when we engaged the yaw degrees of freedom.
To start we turned off the HAM1 ff, by setting H1:HPI-HAM1_TTL4C_FF_INF_{RX,RY,Z,X)}_GAIN to zero, then shutting off the H1:HPI-HAM1_TTL4C_FF_OUTSW. While collecting that data we switched to the new filters, and zerod the gains for all of the individual H1:HPI-HAM1_TTL4C_FF_CART2ASC{P,Y}_{DOF}_{DOF}_GAIN filter banks. We then tried turning on all of the pitch feedforward first, by ramping the gains in a couple of steps from 0 to 1. It seemed like it worked well, but I don't think we actually got to full gain of 1. We then tried turning on the yaw FF, but that pretty quickly started injecting noise around 10hz, first attached spectra, red traces are with the new ff, blue are with ff off.
Second image are trends where Jennie tries to reconstruct the timeline, which is how we found the first pitch test wasn't complete. Sheila ran an A2L measurement, then we tried the pitch ff again, spectra in the third plot. All of the red traces are new P ff (1397415884), blue is the old (1397407980), green is with the ff off (1397405754). This worked well, we got some slight improvements around 15hz, new chard p gets rid of some noise injection around 2hz. We didn't see the pitch ff affect the yaw dofs.
We left the new pitch filters running, and accepted in the Observe SDF. The yaw filters were left off.
I looked at the attempt of engaging the HAM1 yaw FF:
1) the filter that were loaded were correct, meaning they were what I expected
2) retraining with a time when the HAM1 pitch FF were on yields yaw filters that are similar to what was tried, so it doesn't look like it's a pitch / yaw interaction (Jennie also pointed to some evidence of this in the alog)
I suspect there might be cross coupling between the various yaw dofs. I would suggest that we upload the newly trained filters (attached) and try to engage the yaw FF one by one, starting from CHARD that is the one we care most
