After you applied whitening filters and so on for the IMMT1T, I checked the relations of the signals. Fig 1 shows the spectra of DARM and jitters, and pitch jitter would be correlated to DARM around 150-164Hz. In th figure I multipled 1e-4 to the DARM amplitude to match the peak height to those found in the jitter channels. The difficulties for offline subtraction are:

(1) First of all the IMMT1T are not DQ with sufficient speed! the DQ speed for those relevant channels should be increased!

(2) The band width. As shown in the figure, the correlated bandwdith is very narrow (~15Hz) but the noise floor of the jitter sensors are also close to the DARM noise level now. If you simply try to subtract the correlated peaks, the drawback might be increase of the other frequency without appling very sophisticated band pass filtering... It mght be nice if these jitters should be erased at the entrance of the laser itself, but we don't have such actuators after IMC.

... and the jitter couling disappear without doing nothing in the other locking... hmm... (see Fig 2)

In the model k1imcasc, we increased sampling rate of K1:IMC-IMMT1_TRANS_QPDA{1,2}_DC_{PIT, YAW, SUM}_OUT_DQ from 256 to 16384. T he parameters are burted back, and the modified models are svn-ed as committed version of 2887.