Topic: Getting the most out of your MP3s

[Lunar_flare]

FL uses LAME, matteroffact. If you don't change FL's bitrate slider from 128kbps CBR, though, you're not gonna get good results.

so what SHOULD we put it on?

after sifting through this technical discussion one question still remains:

WHAT THE EFF (David Blaine)?

[Toxikator]

so what SHOULD we put it on?

I suppose if it's ABSOLUTELY ESSENTIAL that you export an mp3 from FL, set it to 256kbps.

But the smart way to do it would be to export .wavs from FL and convert them using a better converter.

[^sPikE-]

I also am frustrated because my mp3's dont sound the same as in FL.

So if im getting this right, exporting as wav and then encoding to mp3 in foobar would make the mp3 sound better?

[Toxikator]

No. FL can export a 256kbps mp3 (by lame)... that SOUNDS as good as a v0, it's just much larger (and FL won't write gapless tags so it's doubly a bad idea to use FL). If your mp3s don't sound good coming out of FL it's probably because your bitrate slider is too low.

[^sPikE-]

I have the bitrate set to max (i think tis 450 now), but today i found our i was using 6point hermite interpolation(what the heck is that anyway), so that probably made my mp3 sound different.

And also, what does Dithering do?

[The Colorado Kid]

                      http://www.earlevel.com/Digital%20Audio/Dither.html

[Toxikator]

I have the bitrate set to max (i think tis 450 now)

That's 320 still, but FL uses the same bitrate slider for oggs too so it goes higher for an ogg export.

but today i found our i was using 6-point hermite interpolation(what the heck is that anyway)

It only has to do with the way samples are repitched or upsampled.

so that probably made my mp3 sound different.

Probably not, unless you're doing extreme repitching (what it usually does is causes some differences in the high-frequency content)

And also, what does Dithering do?

Adds low-level (I mean, REALLY low level, totally inaudible over even the best gear's cable hiss) noise to the signal. This helps to "randomize" rounding. If a sample's volume falls between two possible values that it can be stored in, it must be rounded. Dithering prevents all of these roundings from going the same way, thus changing the period quality of the audio being sampled and adding overtones (called harmonic distortion); by instead replacing it with noise of equal value (the idea being that noise is naturally less-offensive to the ears). Of course, gol includes it basically as a gimme for audiophiles and tech-heads because it's pretty widely accepting (by audio-objectivists) that no one can hear either the harmonic distortion or the dither noise in a 16 bit file; it's too low. It's more a leftover of 8-bit audio, when it WAS an important part of the sound.

[brstma]

You only need to dither when your samples are at a higher bit depth than the render (that's what I learned from Recording magazine anyway) I just leave it on so I don't have to check all the samples

[Toxikator]

but even then, you only need it on when going to something lower than 16-bits (probably around 8-10 you'd start to hear it) since any harmonic distortion introduced would be inaudible.

[brstma]

so it's one of those myths?   

The article I read said to dither any time you go from a higher to lower bit depth.  This is why so many people are confused 

It also said it's bad to render a file to a higher bit depth than original.  I assume it's not true either 

[WhiskeyPriest]

I don't think it makes any difference to save at a higher bit depth than the original. Maybe you meant sample rate? Saving at a higher sample rate than the original file introduces some noise.

[Toxikator]

saving at a higher sample rate OR at a higher bit depth doesn't introduce any noise and if it does you're using a shitty interpolator for resampling and/or a shitty dithering algo.

Upsampling is only bad if you use a linear (or some other very lazy) interpolator as it introduces high-frequency noise. increasing the word length (bit depth) is only bad if your dithering algo sucks and it produces audible noise.

In neither case is increasing the file's resolution a bad thing. It's just a waste of disk-space and resources, it won't sound WORSE.

[brstma]

Can you explain,  or recommend some reading about what exactly interpolation is?


Does it refer to how the software decides to round the data?

[Toxikator]

Interpolation is how the software (or hardware in the case of DACs) determines how the "line" between two samples should be drawn when reconstructing a sinewave.

consider the picture attached: if you were to draw a straight line between those points, rather than interpolate the appropriate curves, you'd end up with a LOT of high-frequency content not present in the original signal. This means that when you doubled the sampling rate, instead of adding in points on the line as we see it now, you'd add points in the whrong places, effectively GUARANTEEING that the unwanted high-frequency content was present. that's why, when rendering resampled audio, it's a good idea to select a good interpolation algorithm (the better the algo, the closer to the "true" data the signal will be; choose a good enough algo and the extraneous data will be so low in volume as to be nonexistant when you listen).

[brstma]

Thanks,  pictures always help