Attachments show error signals and RMS channels around the time when IP loop were being engaging.
Both two attachment show the same time. Difference is only the range of Y-axis.
Current time window of RMS calculation is N=1024 (=0.5s for 2kHz model).
Because resonant frequency of tower are much lower than the time window of RMS calculation,
computed RMS is not so different from just an absolute value of the original signal.
Large 'N' for RMS calculation is not so desirable because we need a lot of RMS channels for our purpose.
Hmm...
