Why common-mode gain tends to follow a U-shape curve?

[checkmate]

Can anyone enlighten me on why common-mode gain tends to follow a U-shape curve? What is unexpected is that simulation results almost always exhibit a CM gain degradation at low frequencies, with the lowest gain surprisingly not at near DC frequencies.

My inputs are isolated via DCFEED inductors and DCCUT caps, hence already forms an ideal filter at DC. This is verified simply by observing perfect 0dB input waveforms, showing no feedback contribution from the isolation filter. Yet, the output node still exhibits a CM gain degradation at low frequencies. Anyone knows why? Or if this is just one of the quirky limitations of the simulator?

 

[windknows]

Re: Common-mode Gain

i dont know how you build the cm testbench,
but for the LC lpf method to test the open loop gain, actually there will be a inverse-U shape, i recall this from the book of 'The Designer's Guide to Spice and Spectre' .
if you change your opamp to a simple one-pole system, with some calculation, you will find the LC will add additon poles and zeros close to 0 frequency, with zeros at lower frequency than poles.

 

[checkmate]

Re: Common-mode Gain

Thanks for the advice. Your reason that the LC forms a low-pass filter feeding some low-frequency feedback into the inverting input does not hold as mentioned as I have already used an ideal filter, ie dc-open and ac-short. Probing the opamp inputs shows a perfect 0-db frequency independent common-mode injection.

 

[windknows]

Common-mode Gain

not quite understand what you mean,
but 0dB does not means 0V