First we checked nonlinear gain with 25mW green.

 

BAB with green: 210mV

BAB without green: 46mV

Nonlinear gain: 4.6

 

This is more or less consistent with theoretical value (elog1131).

 

Then we found that the TAMA mixer for CC1 has offset. We changed the TAMA mixer to an another mixer (attached picture) and the offset in CC1 error signal disappeared. TAMA mixer for CCFC also has offset, so we need an additional mixer for CCFC.

 

However, when we try to lock CC1 with the new mixer, GRMC and MZ always unlock.