In what is AAC better than AC3?

[Tom|420]

I have played with AC3 just enough to playback movies already encoded with that (never encoded any file with AC3), but yet never experimented with AAC. All I know is that AAC is supposedly better than AC3.

My question is: In what is AAC better than AC3?

Is is sound quality? Is is file size? Is is open source?

And while we are here, is it better than Ogg, and is what is it better (or worse)?

[sh0dan]

Originally posted by Tom|420
My question is: In what is AAC better than AC3?

Is is sound quality? Is is file size? Is is open source?

In quality vs. filesize: Have a look at the 128kbit/second listening test (http://audio.ciara.us/test/128extension/results.html).

I think you'll also be able to find good answers at hydrogenaudio (http://www.hydrogenaudio.org/).

[bilu]

AAC is supposed (never tried myself) much better at low bitrates than any other, including Vorbis (which comes 2nd if I remember well).

Bilu

[tiki4]

I guess the guy is interested in 5.1 encoding. In that case AAC should be much more efficient in terms of file size and quality than AC3 (VBR, HE profile, more andvanced psy-model). The problem is, that most people here transcode from AC3 source and thus a real comparison is at least very difficult...

Otherwise, AAC isn't open source*, it's a heavy patented format and part of the MPEG2 and MPEG4 standards. It's the supposed successor of the MPEG1 layer-3 (MP3) and was developed by Dolby, Fraunhofer and others. I terms of compression it's much more efficient for 5.1 sources than Ogg Vorbis, because Vorbis still hasn't implemented proper channel coupling, so the bitrates you get are much higher.

Cheers,

tiki4

* Just to clarify this: There is an open source implementation called faac (http://www.sourceforge.net/projects/faac), though binaries maybe illegal in some countries.

[The Edge]

I converted a 5.1 AC3 which was 370mb to AAC 5.1 LC at just a little over 150mb. Sounds perfect to me on playback. Just changed it from 48->44.1 to be on the safe side. Ogg just didn't give me the same quality at the same filesize so AAC wins for me :)

Edge

[bond]

Originally posted by Tom|420
My question is: In what is AAC better than AC3?as the others already wrote aac is much more efficient than ac3 (or mp3) and also usable for multichannel

Is is sound quality? Is is file size?its both :D
better quality at lower filesizes (than ac3 or mp3 for example)

Is is open source?the compression format itself can never be open source, only the implementations!

aac is an open standard (part of the mpeg-4 standard), which means the specifications are open to everyone and everyone can write its own aac codec (the same goes for mp3 for example and there already exist a lot of aac implementations/codecs with different quality: quality comparison (http://audio.ciara.us/test/aac128test/results.html))

wma is for example not an open standard, which means that only microsoft can publish a wma codec (and only ms can earn money with it ;) )

now if you compare aac with vorbis, they are both open standard, which means that everyone can publish its own vorbis codec too (like garf did)
the difference to vorbis is that the people who developed the aac technologies (and hold the patents on these) want to charge money if someone uses these technologies whereas vorbis is said to be patent-free and you dont have to pay for the technologies vorbis uses

as i said the aac implementations/codecs themselves can be opensource as for example the faac aac codec already is: http://sourceforge.net/projects/faac

And while we are here, is it better than Ogg, and is what is it better (or worse)?we have to differentiate here too:
there are mainly two aac profiles: he-aac and lc-aac
1) lc-aac provides ~ the same great quality as vorbis does on medium bitrates (96/128kbps)
2) he-aac was developed especially for low bitrates (64-kbps) and multichannel and is also better in these ranges than vorbis (http://audio.ciara.us/test/64test/results.html) the only available he-aac implementation i know atm is from ahead/nero

[unmei]

AAC is probably unmatched at eating CPU cycles on decode. I love this codec, but decoding a dual-audio HE-AAC Xvid is a bit much for my ol' P3-600.

[Tom|420]

Thanks to all for the answers. Now I have a better idea of how to rank it.

From your experience/opinion, how will it compare to Ogg once multi-channel coupling is done?

Ogg is my favorite at the moment. I find it way better than MP3, and about equal to WMA, but the fact that it's 100% free gives it several extra points in my heart.

tiki4: No I am not much interrested into 5.1. My sound system is a 2.1 and I have no plan into upgrading. However I have some interrest in audio/video in general and I'm always happy to hear more about codecs/format eventhough I can't see the full quality in all of them

[bond]

Originally posted by Tom|420
From your experience/opinion, how will it compare to Ogg once multi-channel coupling is done?vorbis (ogg is only the container not the codec) is not tuned for multichannel encoding atm (lets hope that vorbis 1.1 fixes this), i wouldnt use it for that

[Latexxx]

Originally posted by bond
2) he-aac was developed especially for low bitrates (64-kbps) and multichannel and is also better in these ranges than vorbis (http://audio.ciara.us/test/64test/results.html) the only available he-aac implementation i know atm is from ahead/nero

HE AAC is similar to mp3 pro. It includes sbr which makes it pretty good for low bitrates but doesn't help at bitrates over 90. There is also another implementation of HE AAC. It is made by Coding Technologies and they use a commercial name "aacPlus". AacPlus is the original thing. It was preferred by the standardisation committee when Dolby AAC and aacPlus where fighting for the standard position at low bitrates.

[bond]

yes, aacplus is just another name for he-aac (which is the official name in the mpeg-4 standard)
he-aac = aacplus = aac+sbr = aac+

[SeeMoreDigital]

SBR = Spectral Band Replacement
HE = Head Extension
HE is a less technical term SBR.

From what I can see from the SBR specs it works in a very similar way to an old analog compression technique, which was called DBX.

DBX was quite an incredible innovation for it's time and still holds up today. In fact it was able to offer a dynamic range (the difference in volume, measured in decibels, between the loudest and quietest sections of a musical passage) that still connot be beaten by many of todays 'digital' audio CD discs/players!

Cheers

[Latexxx]

Originally posted by SeeMoreDigital
HE = Head Extension
SBR = Spectral Band Replacement
HE is a less technical term SBR.


SBR means "Spectral Band Replication" http://www.audiocoding.com/wiki/index.php?page=SBR

"High-Efficiency AAC" is the official MPEG name for the combination of AAC and SBR since becoming an ISO/IEC International Standard in May 2003 http://www.audiocoding.com/wiki/index.php?page=HE+AAC

HE doesn't directly mean SBR.

Don't post BS, please.

[SeeMoreDigital]

Originally posted by Latexxx
Don't post BS, please.

Thanks for putting me straight.

I could have sworn SBR stood for Spectral Band Replacement and not Spectral Band Replication!

It's a good job this was not the final question on... "Who Want's to Be a Millionaire"

Cheers

[Neo Neko]

Replacement replication they are both applicable. The replication is a replacement to keeping the actual high frequencies. :p

[Atamido]

Originally posted by bond
aac is an open standard (part of the mpeg-4 standard), which means the specifications are open to everyone and everyone can write its own aac codec I was under the impression that to get a copy of the AAC specs, you had to pay the ISO a fee. Is this incorrect? I am also sure that to distribute an AAC codec you have to pay licensing fees, even it was made by you. (I live in the US, so my statements apply specificaly to there.)Originally posted by bond
vorbis is not tuned for multichannel encoding atm (lets hope that vorbis 1.1 fixes this)It was stated on the #vorbis channel that they intended to fix channel coupling before Vorbis 2.0. So, 1.1 is probably accurate. Timeline though is, "It will be done when it is right".

[bond]

Originally posted by Pamel
[B]I was under the impression that to get a copy of the AAC specs, you had to pay the ISO a fee. Is this incorrect?hm good question...
afaik there are offers like you can get 5 iso documents for free or so... but if you have to pay for the standard specs, you pay these fees to the iso organisation and not to the mpeg-4 standard creators, but i am sure you can find the specs also somewhere else

I am also sure that to distribute an AAC codec you have to pay licensing fees, even it was made by you.well the codec was made by you, but the coding technologies you use were developed by someone else, and these guys hold the patents on these technologies and they want money for it

It was stated on the #vorbis channel that they intended to fix channel coupling before Vorbis 2.0. So, 1.1 is probably accurate. Timeline though is, "It will be done when it is right".great news!
now i hope that it will not take a year or so till we see vorbis 1.1 ;)

[Tom|420]

We waited over 2 years to see Vorbis 1.0, so one more year for 1.1 isn't that bad :) Of couse I would like to have a complete Vorbis yet, as this is my favorite Audio codec ATM

I believe even Stereo (2 channels) isn't done yet, resulting in the 2 channels being encoded seperatelly as 2 streams. Am I right or as they completed that step already?

Also I am not quite sure about some semantics. When we hear of "5.1" are there 5 or 6 channels encoded? That is, is the sub channel actually encoded or is it created by extracting low-frequency sounds from the other channels?

[Neo Neko]

Originally posted by Tom|420
I believe even Stereo (2 channels) isn't done yet, resulting in the 2 channels being encoded seperatelly as 2 streams. Am I right or as they completed that step already?


It's done. Below -q 5~6 you have sterio channel coupling. The lower quality you set the less dynamic the channels get. Resulting in something sounding a bit closer to mono. Though not quite. It depends on audio complexity.

[Happygolucky]

Originally posted by SeeMoreDigital
SBR = Spectral Band Replacement
HE = Head Extension
HE is a less technical term SBR.

From what I can see from the SBR specs it works in a very similar way to an old analog compression technique, which was called DBX.

DBX was quite an incredible innovation for it's time and still holds up today. In fact it was able to offer a dynamic range (the difference in volume, measured in decibels, between the loudest and quietest sections of a musical passage) that still connot be beaten by many of todays 'digital' audio CD discs/players!

Cheers

I had forgotten all about DBX. I have (somewhere, don't know if it still works) an old Technics cassette deck that has DBX. It was a very advanced deck for its time, has Dolby B and C as well as DBX, and supported the high-end metal tapes. I always had excellent results on that deck.

[Wolfman]

dbx = companding?? less hiss on cassette tapes
or EL-cassette (remember those?)