To add a bit more detail on noise modulation, here's a snippet from the AKM4399 datasheet to illustrate :
View attachment 7468
I'm able to decode these figures to show noise modulation because they also include a helpful FFT plot right at the back of the datasheet :
View attachment 7469
The FFT is showing the THD for a full scale 1kHz sine stimulus. The highest harmonic shown is -109dB for the 3rd. Armed with this information go back to the first snippet and note the THD+N for full-scale 1kHz sine stimulus is shown as -105dB. This tells me that the THD+N figure is not determined by the THD component, rather the 'N' component. With full-scale sine stimulus, the DAC is putting out -105dB white noise.
The noise is not constant notice because the -60dBFS stimulus figure shows a THD+N of -60dB. Therefore the noise has fallen 15dB in level to -120dB.
Note 5 tells us the figures were obtained in averaging mode on the AP2. Thus its safe to assume the peak noise modulation is considerably worse than 15dB.