Looking at data from last night to evaluate any improvements from adding the z tilt decoupling, I found that the changes made things worse above 100mhz. I'm speculating that this maybe due to the actual tilt coupling having some frequency dependence (like we've seen with the suspensions) but the matrix that we use for these corrections only allows me to use a dc gain. Attached plot shows the coherence for ITMX, first plot shows red Z/RY (which got decoupled) coherence is "worse" than blue, second plot shows the Z/RX (which is didn't get decoupled) coherence is unchanged. Because we used the Z T240 as a witness for the decoupling measurements, I think that pushing the blends lower should reduce the coherence.

We have run in this configuration before briefly (45mhz blends in Z, broad band sensor correction in Z on HEPI), but I switched away from it when I saw this made RZ move more during an earthquake (via the BSC Z to RZ coupling, even when using our Z drive subtraction), but I think the grounds displacements at the time were several microns (well above the 1-1.5microns rms that seems to be our upper limit). Since then I have done a couple measurements with HEPI putting in similar displacements to what we have seen during the biggest earthquakes we've survived and it doesn't look like 45mhz Z blends make RZ any worse than the 90mhz blends when I "simulate" an earthquake with HEPI.