Skip to main content

Topic: Dumbass Question (Read 2464 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • krsna77
  • [*][*]
Dumbass Question
Ok, first of all, lemme say that I've been playing with Garf's tuned encoder, -b 999, and it sounds in-f@&king-credible! Fantastic!

Anyhow, I was wondering 'bout something kinda stoopid (remember, I'm no guru - that's why I ask you guys). On the r3mix forums awhile back, there was a discussion about resampling files from 44100 Hz to 48000 Hz for MP3 encoding. Obviously, merely upsampling the files cannot really increase resolution. But, for reasons I haven't fully grokked yet, this technique was supposed to better mask certain MP3 artifacts.

Can't say I've played with it much, but I was wondering if this is indeed worthwhile for MP3, and if it would be worthwhile at all to upsample a file for Ogg Vorbis as well. Would this also mask artifacts - or just be a silly waste of time?

(not that I have a big problem with artifacts with Vorbis, mind you...

Just curious.

  • Garf
  • [*][*][*][*][*]
  • Developer (Donating)
Dumbass Question
Reply #1
Quote
Originally posted by krsna77
But, for reasons I haven't fully grokked yet, this technique was supposed to better mask certain MP3 artifacts.


The reason for this lies in the root of the artifact itself.

Preecho occurs because encoding errors are spread out over a whole sample window. The sample windows are fixed in size (256 for Ogg small blocks, it's possible to use other values though). The time that one sample block takes is 256/44100 = 6ms. Upsampling makes this 256/48000 = 5ms. The range over which the error (the preecho) happens is thus about 1ms smaller. This isn't much, but it's enough to appearently give an audible improvement.

Quote
Can't say I've played with it much, but I was wondering if this is indeed worthwhile for MP3, and if it would be worthwhile at all to upsample a file for Ogg Vorbis as well. Would this also mask artifacts - or just be a silly waste of time?
[/b]

The same technique would work with Vorbis. But there is probably little use doing so. Vorbis is more flexible in bit allocation and preecho can be improved by other means. My tuned modes do just that.

--
GCP

  • krsna77
  • [*][*]
Dumbass Question
Reply #2
Awesome! My understanding is increased. Thanks for the clear, nutshell explanation!