thanks for your suggestion!
My DAC chip is included delta-sigma modulator and class D amplifier.
The active filter is connected off chip, so I use an ideal differential op in Spectre to model a filter off chip.
I think an active filter behind a class D amplifier can reduce the size of my class D amplify MOSFET and be low power of my chip.
Voltage of my class D output is 3 level(1.8V, 0.9V, 0V), so I give 0.9V to the filter for common mode voltage. like pic. A.
Is it correct?
I agree with you entirely that a linear filter can hardly worsen SNDR. I will check again my calculate code.
I really appreciate your help.