The GRMC servo was modified and repaired.
Firstly after the new measurement of the mechanical transfer funtion of GRMC, the notch filter 2 was modified as below:
The center frequency of the notch filter was tune to 9,47kHz without changing the Q factor (1 as previously).
C60; C63; C61; C62; : 560 pF (unchanged)
R79; R80; R81; R82: 30 kOhm (instead of 24 kOhm previously)
We didn't succeed to lock the GRMC, so I checked completely the filter section of the servo: it appears after measurement of the transfer function of the 1/f filter, that the corner frequency was displaced from 2,2 kHz to 9 kHz.
It appears that the C38 capacitor was damaged. The C38 capacitor was replaced by a 33nF capacitor and the transfers function was restablished.
We always didn't succeed to lock the GRMC. It appears that the error signal in scan mode has an amplitude of -20mv / + 30mV which is weak for the current design of the servo. So the hysteresis of the detection comparator (U23 - LM311DR) was reduced drastically from 10mV to 0.5mV (largely less than the amplitude of the error signal in scan mode).
The R108 resistor was changed to a 10 MOhm resistor (instead of 470 Kohm previously).
After this modification, the GRMC locking was succesfull and robust.
We tuned the gain (position 3 of the gain potentiometer) in order to have a gain margin of 10dB (at phase 0 deg) and a unity gain frequency of 1.6kHz (more than 50 degree of phase margin).
So the GRMC servo is now functional and we can go on the modification of GRMC and MZ servos for synchronized locking.