As a sanity check, I put DC offsets in IP L, T, Y test filters and checked the LVDT L, T, and Y displacements.
In all cases, positive actuation offsets yielded negative displacements as shown in the IP LVDTs.
I also checked the displacement at the BF and they seemed to be in phase with the IP displacements.
I then put DC offsets in individual H1, H2, and H3 coils and checked the individual LVDT H1, H2, and H3 readouts. They are negated as well.
This is very possibly the coils signs are inverted. It would be nice if we have extra sensors like the optical lever to confirm this.
But, for now, I will negate the gain at IP_COILOUTF.
So,
IP_COILOUTF_H1_GAIN: -1 -> 1
IP_COILOUTF_H2_GAIN: 1 -> -1
IP_COILOUTF_H3_GAIN: 1 -> -1