by
diiz » Thu Apr 03, 2014 2:52 am
Firstly, it's always best to import as wav, not ogg. Mainly because you may need to do some post-processing steps, like normalization. And also because, well, I'm not sure if the same applies to ogg, but at least mp3 compression causes distortion at near-0dBv levels, so it's best to normalize to a bit less than 0dB, like -0.2dB. If you export directly to ogg, you can't edit the resulting sound anymore (or you can but not losslessly, meaning the lossiness cumulates: it's kind of like taking a photo of a jpeg and then saving it again as a jpeg). With both audio and image editing, the principle is the same: always use lossless formats for editing and only convert to lossy as the very last step.
So I always import as wav, do the finishing touches in Audacity (may need some extra limiting/compression, as well as some slight equalization, in addition to normalization). Audacity is very good for this kind of post-processing.
Samplerate is the "resolution" of the sound file. It's how many times per second the sound is sampled to produce the waveform. The higher the sample rate, the better the quality. Generally, 96khz is a pretty good sample rate to export, even if you later downsample it to a lower rate... when you render in a higher sample rate, the calculations used for rendering the sound are more accurate, resulting in a better sound quality.
Bitrate only applies to ogg files, so we can ignore it. But briefly, the same applies here: bigger is better.
Interpolation is hard to explain, it's a mathematical algorithm used for resampling sound. I think it currently only affects samples in the AFP (not sure though, will have to check). In any case it's pretty safe to use the medium setting, as recommended.
Oversampling does what it says: it renders in a higher sample rate than specified, then resamples it down to the specified sample rate. Basically, using 44100 hz samplerate with 2x oversampling is practically exactly the same as using 88200 hz samplerate with 1x oversampling (except, in the second case, the resulting sound file will be 2x larger). If you plan on doing any post-processing, it's better to not oversample and just render in the higher samplerate instead, then do the downsampling manually later (same principle applies as with lossy/lossless formats, as explained above).
Sample-exact and Anti-aliasing oscillators are features which haven't been implemented yet, so their presence is a bit confusing... you can pretty much just ignore them at this point, they will be useful in later versions though.