I made a stupid mistake in the calculation of the cut-off frequency by the coild inductance.

I put R/L in a square root when calculating the cut off frequency, which is not valid. This is not an LC circuit. Therefore, the cut off frequency by the coil inductance should be around 1MHz: irrelevant for our calibration.

Now a big problem arise. Without this klow pass filter around 550Hz (as explained in the updated note below), we cannot explain the measured open loop gain with the model. 

https://nbviewer.jupyter.org/urls/dl.dropbox.com/s/q94boclxu2mgzll/MICH%20Calibration.ipynb

We need to find where this LPF come from. I could not find any in the digital model at least from the MEDM screens.

If this LPF exists somewhere in the input chain (RFPD or I-Q demodulator, AI filter, etc), we need to take this into account in the generation of h(t) because the effective optical gain is no longer flat, which we assume so far.

Actually, by comparing the transfer function measurements of DC and RF MICH signals, we can see that the LPF exists in both cases. This excludes RFPD and I-Q demodulator from the culprit candidates.