As in the header, is Itunes VBR option a pure VBR or ABR?

Thanks

I think the truth is somewhere in the middle, it is neither a true ABR or VBR mode, it is I guess VBR with a bitrate floor of the nominal bitrate setting.

By this I mean it seems that the frame size can increase above the nominal setting, but it will not decrease below the nominal setting.

The only exception to this is for digital silence which is encoded at 32 Kbps.

In my tests it seems that 128 Kbps VBR can go as high as 145 Kbps VBR, the 160 VBR setting can go as high as 180 Kbps.

They are the extremes that I have seen, others may have files that are even higher.

I use 128 Kbps VBR for my iPod, and think the quality is outstanding at that setting.

Actually, I think it goes as low as 2kbps on digital silence

iTunes AAC encoder operates using an ABR mode by default. If you enable the "VBR encoding" option it will allow the bit rate to flucuate even more on encodes but will still keep the bit rate around the selected the bit rate.

And yes, I can also confirm that digital silence is encoded at 2 kbps with the AAC encoder.

iTunes AAC encoder is CBR by default, not ABR.
Its "vbr" encoding mode is a "relaxed" CBR, where psychoacoustic model can cause some moderate bitrate increase, but the encoding is still not fully driven by psychoacoustic threshold.

If you encode without the VBR option enabled with the iTunes AAC encode you will get an ABR file. This is verified by Foobar as the bit rate flucuates around 128 kbps. iTunes 4.5 and above feature an ABR encoder while versions before that offered a straight CBR encoder.

No, this is CBR, as defined in the MPEG audio standard regarding AAC. CBR is constant over a specified buffer, so instantaneous "bitrate" can fluctuate.

Oh, I did not know that, thank you for that information Gabriel.

I didn't know that CBR can perform like ABR.

Well, in common language terms (not going into dev terms), the iTunes CBR AAC encoding method is quasi-ABR as the bitrate can fluctuate over a period of a few frames. Enabling the VBR option will allow the encoder vary the bitrate even more than the CBR option.

Is it as VBR as Lame mp3? Not really. I have songs encoded at -V 2 and 1/3 of the frames can be encoded at 128kbps, the other 1/3 can be encoded at 256kbps, and the remaining can be split betwen 192 and 320kbps. However, when encoding at 192kbps VBR mpeg-4 AAC with iTunes, the bitrate can go down to 180kbps and can go up to 224kbps, that is about it.

I think Apple was going for a known bitrate instead of a known quality. This means that if you select 128kbps VBR, the resulting song will have a bitrate around 128kbps. With lame, if you pick -V 5, you get that quality and the bitrate can be anywhere between 145kbps (averaged) or 112kbps.

So do you mean it is like a more advanced version of the MP3 bit resivour?


Does it really go down that far? I got the impression that it only goes lower for digital silence, because most files that are under 128 Kbps are only under by a tiny amount, like 126 KBps or 127 Kbps, if that is the average it can't be going down by very much, or very often, which makes me think it is just a bit of digital silence at the start and end of tracks.

An exception of course are songs with hidden tracks, where there is a long passage of silence. These frequently show up as close to 100 Kbps, but they are a very rare exception.

I think the bitrate flucuation is more exaggerated than that. A lot of recently remastered Jazz CDs I have regularly reach into the 140s using the 128 Kbps VBR setting. I haven't seen a file which is over 145 Kbps though, so it may be there is a ceiling as well, or an ABR like scheme that makes sure the bitrate doesn't exceed a certain level.

Is there any tool to check for different frame sizes?

Itunes *MP3* vbr has changed quite a bit within the last year. This runs counter to the traditional "itunes MP3 is awful and hasn't changed much since the year 2000" wisdom, but as usual, the traditional wisdom is wrong.

Itunes MP3 always has used the chosen vbr bitrate as average floor. For example, vbr MP3 at 160 kbps will never averages lower than 160 kbps for the chosen sample. (Actually, I've run across samples maybe averaging at 159 or 158 for really easy material, but this is a quibble.)

Recent versions of Itunes have changed the vbr mp3 implementation so it is MUCH more sensitive to the complexity of the source material. For example, vbr highest quality at 160 kbps would typcially encode at at about 180 kbps for a complex acoustic jazz passage in the older versions of itunes; it now would typically encode at around 210 -- 220 kbps for highly complex material, in my experience. Siimple program material will stick right around the chosen floor value (e.g., 160 kbps). So I'd guess this is a vbr/abr hybrid, no?

I have posted about this change in Itunes on other boards but no one seems to care or to believe me. I use Itunes MP3 as my main ripper, tying to average around 192 kbps using vbr, so I'm pretty familiar with it. (I have experience using LAME but I find Itunes to be much more convenient). I have a lot more confidence in my mp3 rips with Itunes since the recent changes.

Not true - I believe you! I posted something about this a while ago as well. I do think it's true that no one seems to care. It seems there are solutions on both OS X and Windows to make using LAME with iTunes fairly convenient, so most MP3 afficianados are using LAME.

I think another problem with using the iTunes encoder, even if there were evidence of increased quality, is that we have become accustomed to the transparency of the LAME development process (I'm not referring to sound quality) and no one knows what's going on with the encoders in iTunes.

ShowsOn, sorry about that, I ment to say 190kbps and not 180kbps. The bitrate of a song may drop below the chosen VBR bitrate (ie 128kbps VBR) but it will only drop by a few kbps, I have yet to see it drop down by 5kbps or more except during portions of digital silence.

I can attest that I don't know what is going on with the iTunes mp3 encoder. I didn't know that it has been updated. The last time I checked, it hadn't had a major sound quality update since 2000 and it had some bug fixes in 2001. I still doubt that it can attain the quality of Lame but at least its nice to know that they have updated it.

So after this conversation about iTunes's MP3 encoder being more responsive with VBR, I started playing with it. At work, on a windows laptop, using the settings <128 VBR (Highest), joint stereo, smart encoding> I got bitrates ranging from the 140's to 185. At home, on a Mac, the same CD with the same settings resulted in nothing but bitrates in the low 130's. Thus concludes my fun with the iTunes MP3 encoder.

Woohoo! Someone at least acknowledged my post about the updating of the Itunes MP3 encoder!!

From what are saying, then (from your post directly above), the MP3 vbr implementation in Windows Itunes appears to be more responsive than that in the Mac version of Itunes? I assume you have the latest version of Itunes available for a Mac?

I've also seen some indications that the Windows Itunes MP3 vbr ripping takes some account of what type of processor you have. On my two Windows computers, anyway, Itunes indicates it has optimized for the type of processor in the computer.

I would find it hard to believe that the Windows and OS X versions of iTunes have different MP3 encoding engines, and as for the processor my Mac is an Intel Core Duo MacBook, so I wouldn't think that would be the problem either. Something is up with the settings because the first few times I chose 128 as the bitrate (with all the other settings I listed), it displayed 320 as the chosen bitrate in the box that shows up on the preference tab for importing. I think something is wrong with the preferences for the MP3 encoder, and I tried to delete them and try again but it was the same result.