THX, the link that you provided cannot be openned in my country for I'm in China whose goverment blocks Google for political reasons.
But I've got your idea that you might think that the reason for the spurious components in audio band is due to limit cycle. However for high order SDM such as MASH 2-1 structure, would limit cycle be so problematic?
When I stimulate the ADC with large signal such as -1dBFS or -3dBFS, the annoying compoents still exist. As a result, I don't think limit cycles give rise to my problem.