Has anyone ever published spectra vs level to show the details of the hump?
Are we talking IMD or THD hump?
For THD (measured spectrally) I have seen with various chips some strange ghost frequency patterns in a certain limited level range which may or may not be related to "the IMD hump" -- I think this could be the case, though I haven't seen the phenomenon when using more than one frequency (but haven't investigated thoroughly).
For example, for the AK4490 this range centered around -40dBFS and spanning 20dB (see
here), for the AK4493 things start 10...15dB earlier. And the AK4499 is reported to have that as well from reliable sources.
I now have finally understood what causes it (at least I think so), the
Data Weighting Algorithm used to linearize the final DAC proper (after the multibit Delta-Sigma modulator). And from that it follows it is completely irrelevant for music playback, these limit cycles do never ever appear with music signals.
As for trimming THD/IMD, digital DC offset does a lot of things, as does sourcing/sinking DC current at the Vcm pins of AK4490/93.