I was just wondering:

when does it stop being useful at throwing more bitrate at recoding an already lossy file?

For example if I have a 192kbit mp3, it would (duh) make no sense recoding it to 256. But what about an 192kbit mp3 to Ogg Vorbis? Im sure the difference in encoding systems will mean that ogg will introduce errors where mp3 still kept 'full integrity' and some spots where ogg will keep the mistakes already introduced by mp3. But will these differences warrant encoding at high bitrates?

As I said, recoding 192kbit mp3 to 256kbit mp3 is silly, so would mp3-192 -> ogg 256 be equally silly? As a rule of thumb ogg-128 sounds about the same as mp3-192 from the same source, but if we're transcoding would it make sense to use ogg-160?

I know my question is a bit strange but I hope people will understand what Im referring to

Uh, if you mean to turn a 192kbps mp3 DIRECTLY into a Vorbis, that's a big no-no. That's called transcoding.

Here's a link for why NOT to do that.

Transcode it if you REALLY want to, but otherwise, the general recommendation is to keep it as is.

Eeeeek indeed you're right I've been mixing the words 'transcoding' and 'recoding' up a bit. >_<

But my question especially is for transcoding. I -know- that a trancoded ogg-128 will sound worse than an ogg-128 encoded from the CD source, but sometimes an mp3-192 is all you have, which is exactly why Im posting this topic.

Then keep the mp3 files as they are. Why do you need to transcode them? What does ogg vorbis offer that mp3 does not?

Well actually Ogg was an example, AAC is more my reality. I have a portable player capable of playing AAC and MP3 (iPod) and I am toying with the idea of re-ripping all CD's to 128-AAC and then I happened upon the mp3's I don't have the source of....

Point is: if I can have a file of 2/3 the size of the original with 'comparable quality' (especially given the outside/noisy nature of listening to a mobile player) then I can fit more music on my device, or same amount with more free space

When transcoding, I usually aim to halve the file-size.. so a 192kbs mp3 I would output as a 96kb AAC/OGG.. I think after that much reduction it is borderline if I will notice the difference between (original-->aac) and (mp3-->aac).

I don't think this is a neat idea in general. The space difference is negligible unless you have a 64MB portable device or something. But it's always up to you to decide what to do with your music. I'd say keep the 192 MP3s as they are.

Heh, especially with an iPod, even supposing it was the 'just' 10 GB version, the difference between 192 kbps to 96 kpbs is relatively small. Of course that depends on how much of your collection consist of these kind of 192 kbps MP3s (and how big it is, of course). Anyway, I support the suggestion to leave them as they are, although the final decision is of course yours.

Yes, the difference is 96kbps. That means you can fit twice as much music on your device, or 100% more. Considering that computer people consider a jump from 3Ghz to 3.2Ghz (6.25%) a BIG jump in CPU speed I wonder why you say 'small'.

And until the new windows pod firmware came out (2 days ago) I only had 192kbps mp3's on my ipod.

Soo the question stands..

If a 96kbps AAC file encoded direct off the CD would be "100% quality" (for that bitrate) how much would an 96K-AAC off 192kbps MP3 be? And 96K-AAC off OGG-192?

99? 95? 90? 85? 80? 75? 70? 65? 50? 45?

And secondly how much more bitrate than 96 is needed to get quality over 95% again? I know I know its hard to present hard facts but Im looking for an educated guess by someone who actually knows (in detail) how lossy encoders work

This question is ridiculous. There's no such per cent quality scale for lossy codecs and there couldn't ever be. Similarly comparing transcoded music the non-transcoded lower bitrate music is very crude and illogical. You'll lose a good amount of quality, that's for certain.

No it isn't, in fact it is a VERY complex question to answer, but not impossible. You're just unwilling to give it some thought.

Factors that come into play are:

1/ How much do 192kbit mp3's deviate from the original in certain frequency bands
2/ How closely does the AAC/Ogg codec at that bitrate look at an 'affected' frequency band.
3/ How heavily will an artifact created by 'mp3'ing to 192' be amplified by AAC'ing the result.
4/ Will mp3 artifacts have a certain quality about them that will cause them to propagate significantly harder to the AAC encoded output than actual sound?

And so on. A simple example would be suppose that MP3 would only have artifacts in the 18000+ frequency range. And suppose 128kbit ogg encoding setting would cut off at 17000Hz. In that case transcoding from the mp3 instead of the source would have 0 impact on the quality of the resulting file.

To answer all this correctly will require quite a few listening tests and so forth, but a certain value can be given to the perceived quality of the transcoded file compared to the directly encoded file, and yes, that can be indicated in %'s. And all I was hoping for is some 'guru' to just take a little stab at how much this would be.

I'm not a guru, but transform codecs mess with sound in deep creative ways that are not limited to particular frequencies. The interaction of two codecs in series is impossible to predict. Given that the accepted wisdom is that transcoding is bad, you will have a hard time convincing anybody to research that.