I measured transfer functions of IMC servo circuit since large phase delay was measured by exchanging the wideband EOM.

Measured phase delay at 100kHz was 100 degree order. It sounds too big because of bandwidth of current EOM: 1GHz, and new EOM: 10MHz.

Actually measured phase delay at 100kHz in the servo circuit is less than 10degree, even gain is maximized to +31dB and all of ~10 opamps are used. Attached figure show the actual phase delay with 31dB gain at one amp stage. The opamps are AD829, and they are not so slow.

Actual path has two amp stages, and it means phase can be recovered only up to 20 degree.

Now I suspect a phase flip happened in the new EOM path. Or new EOM has a resonance around 800kHz, so the phase delay could be real. Anyway I will try to add a phase flip next week.