I'm using the analogLib/cmdmprobe to find phase margin in a fully differential opamp (capacitive feedback) using stb analysis. Supposedly it doesn't matter where in the loop you insert it as long as you aren't inside a local loop... however when I put it at the front of the opamp I get a loop gain of 47 dB, whereas if I put it at the output of the opamp I get 45.3 dB. The response after 1GHz (2x my unity gain bandwidth) is also very slightly different. Otherwise the responses roll off at same rate and cross 0dB at the same point so the calculated phase margin is the same

I've looks for parasitics forming a local loop but just can't see any.

My interest is academic (I don't trust the models enough to believe either answer to that precision). Does it actually matter where the cmdmprobe is placed?

Best regards,
Rob