Delivering at 96kHz or 192kHz isn't going to help anybody. It's not even helpful to run a session with a sample rate that high. I have a pretty good logic on this. Here we go:
- Oversampling per plugin is less wasteful than running your whole project at a high rate. Running everything, even linear math, at an oversampled rate is a waste of cpu, because linear math is no better at a higher sample rate.
- Unless you put a low pass filter between non-linear plugins, your harmonic + Intermodulation distortion (IMD) will climb each time: the higher frequencies from each plugin will be pushed to the next, which will push them up further. Eventually that will alias. If you oversample only in the plugins that need it, the downsampling anti-aliasing stage will clear this by virtue of how it works.
- All those extra frequencies in the high range technically just add more IMD for every nonlinear process, again unless you low pass filter.
- If you have a real-world playback system at 192k, assuming the actual output translates to analog with that frequency information, real-world systems are anything but linear: more IMD.
- If you low-pass filter to solve 2 + 3 + 4, then there was no reason to use 192k in the first place, because you're removing the high frequency information anyway.
- For those people that say the IMD will be so quiet that you can't hear it, because the high frequency content is so quiet... well then you can't hear the high frequency anyway. So doesn't that make the whole exercise moot?
If you say "it sounds better," then fine, there's no logical argument for subjectivity! But, given all evidence to the contrary, and my own listening experiments, it's a hill I'm willing to die on.
