Ideas about solutions of oversampling

Having trouble with LMMS? Ask about it here.
Hi,

I used LMMS 1.3.0 to make a test, and this is the results in detail: https://1drv.ms/u/s!AurPWHDfJoZolh1WeBw ... N?e=80d9RX

Same thing is happened with LMMS 1.2.x, to the best of my knowledge.

So, with square wave (TripleOscillator plugin) at 33% of amplitude, C5, C6, C7, and C8.

My test consisted to export them at 44.1 kHz, and 88.2 kHz at 16 bits, and oversampled each time in x1 (no oversampling), x2, x4 and x8. I don't change the interpolation.

What is expected: Aliasing decreases when oversampling increases.
Actual behavior: Aliasing doesn't change when oversampling increases.

BUT, when I only change samplerate, the treble sound seems have less aliasing.

I noticed one thing: while we export a song, the antialiasing filter works in same time, but it seems problematic.

So, my proposition is this: For example, if we export a song at 44.1 kHz with x8 oversampling, we could make sure that song is exported with samplerate 352.8 kHz in first. Then, at the end of exportation, LMMS could apply lowpass filter at 22.05 kHz and resamples at 44.1 kHz on audio file. I don't know if you understand my idea.

Maybe it could solve a part of problem (aliasing, audio effects, etc.). By the way, I'm not programmer, it's just to give you an idea.

Thank you.

benbou92
benbou92 wrote:
Fri Dec 04, 2020 11:20 pm
Hi,

I used LMMS 1.3.0 to make a test, and this is the results in detail: https://1drv.ms/u/s!AurPWHDfJoZolh1WeBw ... N?e=80d9RX

Same thing is happened with LMMS 1.2.x, to the best of my knowledge.

So, with square wave (TripleOscillator plugin) at 33% of amplitude, C5, C6, C7, and C8.

My test consisted to export them at 44.1 kHz, and 88.2 kHz at 16 bits, and oversampled each time in x1 (no oversampling), x2, x4 and x8. I don't change the interpolation.

What is expected: Aliasing decreases when oversampling increases.
Actual behavior: Aliasing doesn't change when oversampling increases.

BUT, when I only change samplerate, the treble sound seems have less aliasing.

I noticed one thing: while we export a song, the antialiasing filter works in same time, but it seems problematic.

So, my proposition is this: For example, if we export a song at 44.1 kHz with x8 oversampling, we could make sure that song is exported with samplerate 352.8 kHz in first. Then, at the end of exportation, LMMS could apply lowpass filter at 22.05 kHz and resamples at 44.1 kHz on audio file. I don't know if you understand my idea.

Maybe it could solve a part of problem (aliasing, audio effects, etc.). By the way, I'm not programmer, it's just to give you an idea.

Thank you.

benbou92
Best thing here would be to make a enhancement request.
https://github.com/LMMS/lmms/issues/new ... .md&title=
I dont oversample, so i an not aware of the issues. However i understand that oversampling' mostly is done in order to bring the filesize down, but at the same time, accept a quality loss..
In LMMS Sample-exact and Anti-aliasing oscillators are features which haven't been implemented yet, so i am not sure oversampling really 'works'(?)
musikbear wrote:
Sat Dec 05, 2020 9:26 pm
In LMMS Sample-exact and Anti-aliasing oscillators are features which haven't been implemented yet, so i am not sure oversampling really 'works'(?)
Like this? https://github.com/LMMS/lmms/pull/4397
For me, it's a good idea.

benbou92
benbou92 wrote:
Sat Dec 05, 2020 10:59 pm
musikbear wrote:
Sat Dec 05, 2020 9:26 pm
In LMMS Sample-exact and Anti-aliasing oscillators are features which haven't been implemented yet, so i am not sure oversampling really 'works'(?)
Like this? https://github.com/LMMS/lmms/pull/4397
For me, it's a good idea.

benbou92
If you feel it covers yours issue, the its ok, else you should make a requests