[Jenne, Gabriele]

On Tuesday (belated alog....) we installed the postprocessing version of NonSENS on the oaf computer, so that we can tune up our feedforward, without having to continue doing measurements and fits (which are challenging).  We added IPC senders to h1asc (for ASC postprocessing feedforward) and to h1cal (so that we can do this postprocessing feedforward on the calibrated version of DARM).  h1lsc already had the SRCL, MICH, and PRCL senders to OAF that we needed. 

Today, we trained the LSC feedforward model using some data from last night, and turned it on, and we're getting a factor of up to ~2x subtraction around 10 Hz!

The new channel that we can start to look at is H1:OAF-CLEAN_DELTAL_EXTERNAL_DQ (and, of course, the non-DQ version).

Since Elenna is doing some noise budget injections, we've also got a spectra showing the subtraction working during a time of SRCL injection, which helps us see how effective the subtraction really is.

The improvement we're seeing is above 10 Hz, since that's what we've asked the subtraction to focus on.  Since this is postprocessing subtraction, we don't really care how we do or don't contaminate the data at frequencies lower than our search pipelines look.  The subtraction is much more effective if we don't force the filters to do a lot of work at lower frequencies, and so this is resulting in what looks like noise contamination below 10 Hz.  But again, the biggest consequence may be that we want to add a little extra lowpassing to the data we save, for dynamic range reasons.