Oversampling Question

[Snarf]

I'm almost done with another song, and I exported the (nearly) final mix so I could listen to it. Like usual, I did a bit rate of 192, medium speed sinc, and 2x oversampling. I'm used to having the oversampling make the song sound a little different (almost always in a good way) but this time, it really affected the sound a lot. I used a ton of lowpass filters (on almost every track at certain parts) and the 2x oversampling made these filters way overkill (everything sounded very muted). Of course, everything sounded nice and smooth and less edgy, but a lot of the subtleties and dynamics of the song were lost. The character of the sound was completely changed, and not for the better this time. Just to make sure, I exported it again with the same settings minus the oversampling, and it sounded just like how it did while editing it.

My question for you guys is: do you use oversampling? If you do, how do you make it sound the way you originally intended? It's pretty hard to mix and master a song when the export process completely changes the dynamic balance...

Thanks

[musikbear]

The best post written is by diiz, who also ofcause is main engineer, and highly knowlegable in these matters:
http://lmms.io/forum/viewtopic.php?f=4&t=1174

[Snarf]

I solved the problem by using the settings suggested in that article. I also had to raise the cutoff frequency on every filter in the song. It was annoying but now it sounds much better.

[Sti-Jay]

Never use oversampling. Don't understand what it does and if it changes my sound, then it is not good, cause I make songs so they sound good (I believe) and don't want them to be exported into something different.

I use 32 bit, 320 kbit/s, 44100 Hz and sinc best. I could probably render with 48kHz but I have heard rumors about some plugins/effects (even native LMMS) not supporting it and thus there is not point. I kind of regret that I haven't exported in 48kHz the last year. Oh well, only CD quality on my songs then

[Snarf]

Ok oops...96k Hz song incoming...

[HenryDeath]

I'm used to having the oversampling make the song sound a little different (almost always in a good way) but this time, it really affected the sound a lot.

Not sure if it’s useful to answer this post after so many year but let’s go.

I too find the oversampling makes the song sound a bit different (even at x2). The fact it’s different is not a good thing to me, sometime, on some parts, it even produces sounds I don’t like. But I admit it quite always, globally, sounds “brighter” (not sure if it’s the right adjective, but hi-end often sounds better IMO). So… what it leads me to do is mixing (in Audacity), a non oversampled export with an oversampled one, in variable proportion. For example: reducing the volume of the oversampled export by 7 dB and reducing the volume of the non-oversampled one by 3 dB, then mixing the twos.

When I do this I find the final result more faithful to the original creation, but still with a richer, nicer sound.

[musikbear]

but let’s go.

well the sad truth is that oversampling in LMMS does not work well, there are artefacts in well over 50 of renders.
One way to to oversample anywitch is to render in hiQ wave and the resample in audacity or an other similar prog.

[HenryDeath]

well the sad truth is that oversampling in LMMS does not work well, there are artefacts in well over 50 of renders.
One way to to oversample anywitch is to render in hiQ wave and the resample in audacity or an other similar prog.

I can confirm, it’s only interesting some times, and some other times it’s just awful. Today I exported a project which has a SF2 player plugin which use the "Cello" instrument from the default “General MIDI”: The oversampled rendering (x2 or x4 I don’t remember) has the cello playing maybe an octave lower… I don’t think oversampling should do that in any case. But I may be wrong.

Personally I care less about oversampling than about the impossibility to export a 24bit FLAC file from command line, or the fact that rendering to a file doesn’t honor the MIDI connections from an instrument to another (ie: connections within LMMS itself).

I also have questions about the “tracks rendering” functionality. I will probably make a post for that.

Although all the problems LMMS still has, I like it very much.

[OroBit]

Never use oversampling. Don't understand what it does and if it changes my sound, then it is not good, cause I make songs so they sound good (I believe) and don't want them to be exported into something different.

I use 32 bit, 320 kbit/s, 44100 Hz and sinc best. I could probably render with 48kHz but I have heard rumors about some plugins/effects (even native LMMS) not supporting it and thus there is not point. I kind of regret that I haven't exported in 48kHz the last year. Oh well, only CD quality on my songs then

I also use 32 bit, 320 kbit/s, 44100 Hz