Hi

I was wondering if you guys could help me out with something strange I've noticed with my music files. You see, after reading some of the forum posts around here I've come to the assumption that AAC is generally considered the more modern and efficient codec. I'll be honest, I cant tell the difference between MP3 320 and AAC 320 vbr from a audio standpoint. However, I noticed that file sizes were LARGER in AAC 320 vbr than 320 MP3! For example, I've burned the Halo ODST soundtrack with both WM11 (MP3 320) and iTunes AAC 320 vbr and the file sizes (from the properties tab in windows explorer) for the track 'The Rookie' were:

MP3: 12.9 MB (13,546,204 bytes)
AAC 320 vbr: 17.5 MB (18,411,520 bytes)

Whats up with the difference? The ONLY thing I can think of that could have caused this is that the MP3 was burned to a local hard disk while the AAC file was burned to a networked hard disk. I dont see why that'd make a difference but I'm willing to burn AAC again to the same internal HDD if you guys think thats the reason...

If this is true I'm guessing that it doesnt make much sense to continue ripping my collection into AAC does it? Generally speaking what should be the differance in file size between AAC 320vbr and MP3 320?

Thanks
Fire_Zealot

A CBR 320 mp3 will be EXACTLY 320 kbps. A VBR AAC will vary as needed to meet the requested quality, but should, over a large range of music, be APPROXIMATELY 320 kbps, but will vary quite a lot.

It makes perfect sense that iTunes 320kbit constrained VBR would have higher bitrates than 320.

At those bitrates you won't see any quality difference between the two. In fact a lower bitrate setting of around 200 kb/s would be almost indistinguishable.

If you're going for bitrates that high you should just encode it to lossless.
JXL

That that at you too big difference!
Here I just compared:

MP3 CBR 320 kbps (polyphase lowpass filter disabled)
The size of a file - 12232 KB
FAAC-q 500 322 kbps
The size of a file - 12322 KB
And even in spite of the fact that at MP3 polyphase filter was disabled, AAC all the same had better quality of a spectrum!
At options FAAC-Q500 the spectrogram is indistinguishable from original WAV 1411 kbps!

I didn't realized that we watched out music and didn't listen to it.

You can't use spectrograms to make sound quality claims.

Ahh, where to begin?

First, the size of the CBR file will be the same whatever the lowpass.

Next, you are looking at spectrograms, which tell you nothing about how the files sound. Often the one that "looks" best, sounds worst. In particular, if the FAAC file is similar in size but contains a lot of high frequencies that are not audible, then it is wasting a lot of bits and its sound quality will suffer.

Judging by a spectrum, AAC has kept not only high, but also all frequencies

Please re-familiarize yourself with Hydrogen Audio's Terms of Service to which you had agreed upon registering, paying specific attention to #8.

Also, please notice that the OP is not using FAAC, making your comparison even the more useless.

EDIT: There is no need to for anyone to continue the off-topic discussion concerning UVSM's TOS8 violation in this thread any longer.

Sorry greynol, you beat me to it (I didn't see your post when I wrote mine).

I do have some information for Fire though. The VBR setting in iTunes is actually a constrained VBR setting. iTunes will use QuickTime for AAC encoding. The VBR constrained setting means that files will be encoded at a minimum of the selected bitrate while allowing to increase (within range) when necessary. So a 320kbps VBR AAC file, encoded with iTunes, will have a minimum bitrate of 320kbps. AAC can actually go above a bitrate of 320kbps without breaking compatibility. The Lame mp3 encoder can go above 320kbps, it is called free format mode and breaks compatibility with almost everything out there.

So there you have it. The Lame mp3 encoder is using a constant bitrate of 320kbps and won't go any higher. The iTunes AAC encoder is using 320kbps as a minimum and will go higher.

However, your best option would be to conduct a series of blind ABX test to determine which setting is right for you. That is not what you asked about so I will stop there. Just know that 320kbps (whether it is ABR, VBR, or VBR constrained) is often overkill for a good majority of people with most non-killer samples.

actually, unless this has changed since the last time I did any experimenting with Constrained VBR (C-VBR) and QuickTime AAC, the minimum bitrate can drop well under 320kbps, but the average bitrate of C-VBR-encoded files will go no lower than 320kbps (unless, of course, there is a significant amount of silence in a track)... I know what you meant, I just thought I'd throw this out there.

also, to the OP, something to keep in mind is that if you have imported tracks in iTunes using C-VBR, iTunes will display whatever bitrate you chose in your settings, not the actual bitrate. for instance, files encoded @ 320kbps C-VBR will usually have bitrates significantly higher than 320kbps (I've seen average bitrates upwards of 360-370kbps using 320kbps C-VBR, which would explain the difference you see in your filesizes), but iTunes would still display 320kbps as the bitrate- my guess is that this is to avoid confusing the general public who would otherwise see these much higher bitrates and likely think that there was something wrong with iTunes or the settings they chose.

I don't know about iTunes bitrate deviations, but what about obvious things that affect file size such as lots of artwork in the tags

The most obvious thing is that iTunes uses constrained VBR (and we are talking about iTunes here, which is different from QuickTime); besides, the native way that iTunes handles artwork is to put it in a location external to the audio files.

Yes, you are right. The bitrate can drastically drop for quiet parts of songs and for more simplistic music (such as techno). However, generally speaking, the bitrate won't go below the selected value. Then again, my experience is mostly with hard rock and lots of metal. The bitrate only ever drops during periods of silence for me. I ripped a CD for a friend at 128kbps VBR (constrained) and the overall average bitrate of each song was down to about 112kbps. This was a techno/electronic/whatever CD.

I find The Lonely Island's Jizz In My Pants, to be a good and humorous example of how VBR works.

Bitrates can dip very low on mono recordings such as old Beatles records. I've seen bitrates to stay low with Rap, Hip-Hop, and certain electronic artists such as Tubeway Army, Gary Numan and Vangelis. But for Rock and Metal music its a different storey especially with Death, Doom, and Metalcore metal music; the only metal music i've seen that produces low bitrates is mostly tracks by Tool.