If i were to encode my mp3s at 44khz or 48khz, what would the difference be and would i be able to distinguish a difference when listening?

The difference would be in the resampling, or lack thereof.

Don't resample. Encode your CDs at 44.1kHz, and encode your DVD LPCM rips at 48kHz. Even if you can't hear it, it does no good. Avoid resampling altogether.

If you have a not-so-good soundcard, don't bother to resample your files to 48kHz; just resample them on playback. Better yet, get a better soundcard.

Just say no to resampling.

There won't be that a big difference...
Though, for lower bitrates lower sampling rates seem to perform better. (less side-info overhead due to the lower frame rate & higher spectral resolution / decorrelation possible).

SebastianG

Thank you.

Just use the original sample rate. Resampling causes artifacts.

In other words don't resample unless you really really need to!!!

If you do use SSRC!!! (Does that still hold true? Haven't been around to be in the know)


Regards

what about the quality of the resulting 48khz file?

for what i know is that lame, musepack or whatever is optimized fo 44.1?!

"Theoretically," if the source is 44.1(cd audio), the quality could be worse at 48kHz due to quantization. Just stick with the original sampling rate.

Another side effect would also be the need to downsample to burn a cd.

You should encode them at whatever sample rate the wav file is at.
Some players may not like 48khz in which case you should first downsample the wav file using something like sox to 44.1 and then encode the downsampled wav file.

Interesting topic.

Let's say that I need to burn some 48 kHz LPCM extracted from a DVD to CD. With what separate app should I resample it? Or should I just leave that to Nero 6 that would resample it anyway?

I always thought that the --alt-presets in LAME were strictly optimized for 44,1 kHz? My 48->44,1 khz resampled --alt-preset standard files actually sound better (yes, I've ABXed) than those which are left as 48 kHz before encoding.

Some links with good reading about the badness of resampling would be interesting too.

If you need to downsample something from 48 kHz to 44.1 kHz, you should use an application that apply proper dither. As far as I know, Nero does not do that when source and destination are at 16 bits. Use ssrc instead. Which preset exactly is subject to discussion, but one should apply proper dither to avoid (hearable) distortion. Especially in the low amplitude range.

My personal way of working is with the GUI-version of ssrc and than select ATH based noise shaping and the pdf at rectangual or triangular. According to the theory one should apply triangular but if this CAN introduce too much noise.

Regards,
Jacco