MP3 Lame is totally flexible VBR, you could use LR stereo (or MS or IS stereo) even with -V 0. All it would do is, that VBR scales the bitrate higher. There's no quality issues with MP3 why you couldn't use LR stereo even with -V 0. Of course when you use LR stereo, the bitrate will be higher than using MS stereo.

I can repeat this with MPC, MP2 and AAC. You mentioned a
VBR property, not a Ogg Vorbis property.

Note that this posting implies no quality assessment of any codec, only tries to correct your portrayal.

MP3 Lame is not totally flexible VBR. First you have the bitrate limit of 320kbps, which is quite low for max bitrate, secondly you have fixed frame sizes (...128,160,192,...).
Still of course I didn't mean that flexible VBR is only Vorbis property, of course it's MPC's as well, but the original question was about Vorbis, so I used Vorbis as an example.

320 kbps limit: That's right, but there are only some tracks
where this makes trouble (the most famous not artifical is
fatboy). BTW I expect that OoOE of fatboy would be able
to allow transparency of fatboy at 320 kbps.

MPEG-1 has a bitpool of up to 511 bytes. At 44.1 kHz this
can wobble the fixed bitrates by 156.5 kbps. That's even much
more than the 64 kbps difference between 256 kbps and
320 kbps.

You have only some fixed frame sizes, but audio frame sizes
are different from frame sizes.

Ogg Vorbis/AAC/MPC are much more flexible, but for most
encoded files this is not a limitation.

MP3 has problems, but this problem is not in my Top10 MP3 problems. It's on place 11 or 12.

Well that depends on your quality needs. There are even whole music genres that would need clearly over 320kbps with mp3 to sound even close to transparent even to a non-audiophile. The problem is of course pre-echo. Fatboy is even not so big problem...

If 320 kbps MPEG-1 Layer 3 is not enough, I would use
256 kbps MPEG-1 Layer 2 or 128 kbps MPEG-4 AAC.

Above 256 kbps MPEG-1 Layer 3 makes no sense. If you need
such bitrates, the reason for this high bitrate are flaws introduced with Layer 3. Often Layer 2 performs much better
than Layer 3 at the same bitrate. And note that Layer 2
also supports 384 kbps.

For Fatboys also MPEG-1 Layer 1 performs better than Layer 2
at the same bitrate.

For transparent encodings of typical pop music Layer 2 normally
outperforms Layer 3.

OT:

Which layer II implementation are you using Frank?

David.

320kbps mp3 frames make perfect sense. None of the default Lame VBR presets restrict the bitrate to 256kbps frames. Many times MP3 music needs 320kbps frames, many times high quality Lame VBR preset would need to hit even higher than 320kbps but it's impossible. It's absolutely not recommended to restrict the max bitrate of tweaked Lame VBR preset to 256kbps.
It's easy to say that you should use another format then, sure, but many times there are other reasons, like compatibility etc.
You ask me to use mpeg1 layer1 or layer2 instead then. Show me one layer1 and/or layer2 codec which has good, well tweaked psychoacoustics. All the codecs I have tested perform even worse than Lame even at much higher bitrate with many difficult test samples, because their psychoacoustics sucks. And even then I should use insanely high CBR bitrate. Does not make sense.

Are you also saying that it makes no sense for Vorbis to hit very high bitrates then? Do you think its max bitrate should be forcefully restricted to something like 256k or 320kbps? Does not make sense...

MPC2. Is a modified MPC encoder with two modifications and
a MPC->MPEG Layer 2 bitstream transformer.

The fixed block sizes of MP2 sucks due to the missing bit pool
(which you have at Layer 3).

MP3 audio frames which need more than 320 kbps to be transparent have this high bitrate demand due to at least
two flaws introduced with the Layer 3 MDCT.

If there would be a HQ Layer 2 codec it would make sense to use
this HQ Layer 2 codec which outperforms MP3 codecs.

The layer 3 of MPEG-1 is IMHO a hack which works good for
most classical music and which makes things worser for
typical pop music.

MP3 was designed for 128 kbps and it outperforms MP2
at data rates at 192 kbps and below.

Yes, this depends totally on the encoder implementation. It's no use if this is only theoretically true.

Are you going to release MPC2?

actually the 320 kpbs bitrate limit is not correct. lame can create a free format that can go uptp 600 kpbs, its just that only a few players can play it ie mad

Hrmm... this could be very, very interesting. What is the likelyhood of this getting released? I could see many uses for this, since a lot of software/hardware out there is compatible with MP2, and so far there isn't really a "good" MP2 implementation, at least not on the order of what you are speaking of. I, for one, would be very interested in experimenting with this.

If you don't plan to release it, is there any reason why not?

Freeformat is not something you can call true mp3. Freeformat is not MPEG1-layer 3 compliant.

I think a good MP2 encoder will be valuable because VCD uses MP2 for soundtrack encoding.

Freeformat is mp3 compliant. It's part of the standard, and written in the iso docs.

BUT:

mp3 compliant decoders only have to support freeformat up to 320kbps. They can support higher than 320, but it's not mandatory.

Uh, oh. Should have remembered that. Thanks for correcting me.
Anyway, nobody will in practise use freeformat if one wants to create MP3s which are generally playable. Also I think I remember that Freeformat doesn't support VBR, correct me if I'm wrong.

You're right, free format is fixed bitrate (not even bitreservoir in the free format mode we're speaking about)

The only use of free format <320kbps is when you should enforce a custom constant bitrate.

ex: I have a sattellite communication with 162kbps, so I want to use all those 162kbps, because using 160 would loose 2kbps, and 2kbps of satellite bandwidth cost xxxx$

It's just a shame that nobody is working on tooLame. It's been 8 months since the last CVS commit.

MPEG-1 layer II is still the codec of choice for devices that need to allow random access (since unlike MP3, each frame can be decompressed independently), like (S)VCD. Standardization and profilic hardware support are the main reasons, I presume.

so how do mp3DirectCut or mp3Trim work ?
don't they cut out specific frames, decompress them and do things frame based-ly ?

Good question. I assume that they simply look at what frame you want to keep and don't cut out any frames that said frame depends on. With is a maximum dependancy of 9 frames, or 10368 samples for MPEG-1 material (which is only ~0.235 seconds at 44.1 kHz), it's practically imperceptible.

Of course, I'm just guessing. I don't know exactly how the bit reservoir of MP3 works, or how those programs deal with it. For all I know, they could take the bits and put it in a whole new frame.

Frank,

Please, please, please (did I say please?) could
you release MPC2, the"modified MPC encoder with two modifications and a MPC->MPEG Layer 2 bitstream transformer."

I would love to test this... and the VCD/SVCD video community would praise you forever....

Please give it some consideration....

Best,
RD

And could you release the source while you're at it?

When I cut an MP3 with mp3DirectCut, the MAD plugin for WinAmp shows 1 "other error".

Surely this is something with bit reservoir

This is really a nasty hack. It always uses MS Frames, it wastes
on average 20...40 kbps due to MP2 package quantization,
it produces a wrong bitstream if bitrate drops below 112 kbps.
There's some code needed not needed in MPC.

Typical bitrate is 320 kbps and vasting 40 kbps because stream needs 270...290 kbps (where the according MPC stream needs
180...190 kbps).

Maybe SV8 can write a MPEG Layer 2 stream on demand.

Another advantage of MP2 over MP3, AAC, OV, MPC, ... is
that it is much more error resistant. In a typical MP3 stream
you can hear 50% of the bit errors, i.e. you need BER below
10^-7, better 10^-8.

For MP2 this is around 10^-6.

A bit error in MP3 normally damage a whole frame, in MP2
normally a subband sample as long as it not occures in the header bits.

MPEG-1 Layer III has design flaws because of politics. Yes - politics! MPEG commitee had two proposals "on the desk" during the initial phase ("competetive phase"):

- ASPEC (AT&T, FhG, Uni-Hannover, CNET)
- MUSICAM (Philips & Crew)

MUSICAM parts are dominant in Layers I and II, and Layer III is more or less based on ASPEC - BUT - MPEG commitee wanted to keep some kind of compatibility and "joint solutions" between layers (although this compatibility is more or less useless) and they kept "hybrid" filterbank in MP3 as well as frame-based M/S and IS switching (instead of frequency-band wise) - these are most problematic flaws in MP3.

Also, commitee insisted on keeping 1152/384 samples windows in MP3, that are not optimal for MDCT filterbank - optimal values are 2048/256 samples (as used in AAC), etc..

Ivan, you have made me very curious...

You mention politics but maybe there was also a political reason why they wanted to cripple mp3?

Did they not want it to be too good because they wanted AAC (which might have been in early development) to be the best (and expensive)???

I mean if they knew the "hybrid" filterbank was a bad choice, and that 1152/384 samples were not optimal for MDCT filterbank why approve them?

Any enlightenment appreciated....