Oversampling filter design

[nbdy]

I would like to start a thread about audio DAC oversampling filter design, which is believed to have something to do with DAC sound.

Since current FPGAs offer some serious computational power for building such filters (they also low cost and power consumption is not an issue here), it's now quite easy to implement practically any
filter neccesary for a better DAC. The missing part (of my knowledge) is which oversampling filter parameters are important for the resulting sound?

E.g., what makes PMD100/200-based designs sound good and what makes PCM1792 sound (as believed by some people) weak compared to its impressive specs?

Passband ripple and stopband attenuation seem to be unlikely candidates since almost every decent DAC/filter is good in this part. Phase linearity seems like a possible answer, but I have hard time

believing that PCM1792 for example has non-phase linear filter (however, its datasheet doesn't state the opposite while datasheets for other TI DACs sometimes do specifically mention phase-linear filters). Maybe round-off errors?

More, AK4396/5 is believed to be sonically superior to matching TI/Crystal offerings - how this could be explained? Does having less quantization noise above the audio band (as suggested by the
manufacturer) or maybe a better oversampling filter help here?

All of this may be also applied to the PCM-PWM modulator design (e.g. TI PurePath modulator analog with more flexibility wrt say dead-time adjustment or digital domain PWM stage error correction ala John Westlake designs)

Regards,