I, being a notorious ignorant fool, recently found that oversampling makes things sound great. Sort of de-muckifies the high end. Little did I know that it was a form of antialiasing. (I had always accepted aliasing as a fact of life) It's pretty amazing when I go from a 8x oversampled sound back to no oversampling, and how horrendously smeared everything sounds.
So I've got some native plugins that support it, but many don't. The question is, where in the mixing chain are you supposed to apply this?
Theoretically, since it's a way to not introduce aliasing where heavy aliasing could occur, so earlier up stream is better. At least from what I know, once the nyquist mirroring happens, those mirrored harmonics can't be retroactively cleaned up, so better to not have them be there in the first place. At the very least, I think this means that doing some oversampling on the master channel isn't really going to fix things since the signal already contains a lot of aliasing by that point.
But then at the track level, the oversampling would need to happen before any audio comes out of the vsti or whatever sound source is generating above nyquist harmonics.. which is pretty much impossible since that's mean opening up effects and synths.. but from my theoretical understanding, this would be where antialiasing would be effective. If applied after aliasing has occurred, it seems to me that oversampling won't remove any aliasing.
It makes sense to have oversampling in effects like overdrive / distortion where the point is to create a bunch of higher harmonics. Or in a synth, exciters.. I guess maybe realistically there's only a very specific subset of processes that are known to cause aliasing, and so those would be where oversampling should be applied. Summing doesn't generate harmonics so probably doesn't matter as far as aliasing is concerned.
Does that sound about right? If it is, then it sounds to me we're pretty much at the mercy of plugin makers to have built in oversmapling or some sort of antialiasing if the plugin tends to cause lots of aliasing, and once the signal is outside of the plugin, aliasing has already occurred, and cannot be removed. Which kind of sucks. Less aliasing already sounds pretty great.. I'm imagining having no aliasing would sound immaculate. It just sucks you can't just build a mixer (or mixer in a DAW) with antialiasing built in since the fix needs to go into every plugin and alias generating fool in the system..
Which makes me think.. is that a problem that Scope solves internally? Like, is every module made to not cause aliasing? If so, I may need to quickly rebuilt my scope environment back up again.
I guess the simple solution is to keep increasing the project sample rate...
oversampling question
Re: oversampling question
many devices use oversampling, for instance the Minimax filter.
many devices also use very high bit depth, which also tends to limit aliasing. the STM2448 and 4896 are examples.
many devices also use very high bit depth, which also tends to limit aliasing. the STM2448 and 4896 are examples.
- Sounddesigner
- Posts: 1085
- Joined: Sat Jun 02, 2007 11:06 pm
Re: oversampling question
Yep. I also read in Minimax's manual it uses anti-aliasing oscillators to. I read it in either the SCOPE plugin manual or the ASB, but it should be in both. I know many SCOPE plugins oversample and I doubt only Minimax uses anti-alias oscillators. But to my ears the SCOPE plugins that don't oversample usually sound better than the Native that do. O/S as we all know isn't always present in high quality plugins and I've heard some plugins do 16X O/S and still sound like crap, cause simply put the algorithm isn't any good. O/S is extremely common in Native plugins now and used as a marketing buzzword to make alot of mediocre plugins seem better than what they are. I know SSL Duende Native did not oversample when I use to own them but at that time they sounded better than every other comparable Native plugins I tried even the ones that did oversample and used a ton of cpu resources. The duende were also quite cpu efficient in comparison.garyb wrote:many devices use oversampling, for instance the Minimax filter.
many devices also use very high bit depth, which also tends to limit aliasing.
That said I generally dislike Aliasing and view it as a general factor that counters the highest of quality and potential that often can be reached with algorithms (even tho some plugins sound really good with aliasing I believe they'd generally sound even better without) and thus I simply run SCOPE and Native at 96khz project samplerate to deal with the plugins I own that don't oversample (and usually don't have to engage O/S mode for the plugins that do O/S because the higher project samplerate is sufficient). You can avoide added latency from O/S and bad sounding O/S algorithms and unnecessary samplerate conversions from sending a audio signal threw multiple O/S plugins by simply running the whole project at a higher samplerate. All O/S algorithms are not equal.
- kensuguro
- Posts: 4434
- Joined: Sun Jul 08, 2001 4:00 pm
- Location: BPM 60 to somewhere around 150
- Contact:
Re: oversampling question
interesting. I guess the part that may be different from developer to developer is the filtering and maybe how they down sample?
- Sounddesigner
- Posts: 1085
- Joined: Sat Jun 02, 2007 11:06 pm
Re: oversampling question
kensuguro wrote:interesting. I guess the part that may be different from developer to developer is the filtering and maybe how they down sample?
Yes those are the two things I'm aware of. I know in the update note details of some plugins you'll read at times things like "improved oversampling quality" (sonicly speaking). From my own experience and info from developers and others all O/S algo's are definitely not created equal.