Tried the new feedforward path, adding the DBB Q1Y signal to the DARM IN1 signal. It's working.

First, I measured the transfer function from Q1Y to DARM (no noise injection), and fitted it. The result is shown in the first figure (I removed from the fit the 2kHz sharp feature which is clearly a misfit artifact). The filter is implemented in the LSC-JITTERFF bank.

The second plot shows the improvement in the DARM signal and the reduction of coherence whe the feedforward is on (blue reference with FF off, red with FF on). All the peaks remaining in the DARM signal are not due to the HPO jitter.

We shall see how stable this is over time.