As pointed out by Craig and myself, in all lock stretches, regardless of what CHARD_Y controller we use, there is a rising 3.4 Hz line, that ends up dominating the RMS of most of arm ASC signals (in the trend plot: first lock is nominal CHARD_Y, second lock is new CHARD_Y, third and fourth locks are nominal again). The new CHARD_Y controller doesn't necesasrily make the 3.4 Hz worse.
Yesterday we switched off the calibration lines, and noticed an improvement in DARM above their frequency.
It turns out that the lines we see between 18 and 21 Hz are sidebands of the calibration lines, shifted up by the 3.4 Hz motion.
Therefore it is important that we understand how to reduce this 3.4 Hz motion