Now forgive me for being total newbie here, but I (at least not yet) haven't stumbled on any informational, non-commercial and easy-to-understand site about AAC in general. Especially one which would concentrate on archival quality, not internet streaming!

With just the slightest bit of trouble from experts, I would be much happier . Questions follow:

1) Is AAC the same thing as DVD's AC-3?

2) Wasn't AAC so very strongly patented and licenced that freeware encoding/decoding tools are nonexistent, rendering it useless to end-user (like me..) who won't pay 249$ for 2GB monster-multimedia-encoding-package?

3) Come on, is AAC REALLY that superior to current quality standards, like MPC?

4) Back in LAME 3.87, I was asking very similar questions about MP3 quality. Luckily "mp3 quality" google search pointed me to r3mix.net which answered my questions more than adequatly. Is there this kind of page for AAC?

Thanks for your trouble!

niktheblak

Hi,

You will find some of your answers on:



The site is not complete yet, but be patient!

AC-3 is Dolby's proprietary multi-channel codec, Mpeg4 is an international standard.

I am no expert, but maybe that helps?

No.



Yes, AAC does have some pretty stringent legal issues surrounding it. That doesn't render it useless, but it does make it a harder format to use for personal encoding.



AAC isn't superior to MPC, but then that isn't really its goal either. Its goal is mainly to be superior to MP3 and to provide higher quality at lower bitrates. For this it does succeed. However, a good AAC implementation does come pretty close to MPC in quality, though overall MPC still has the edge in terms of quality alone.

If I were to give a ranking of current overall codec quality and where AAC fell within that scale it would be something like this:

1. MPC
2. AAC
3. Vorbis
4. MP3

This is of course taking into consideration VBR and reasonable bitrates (160-192kbps). As you go below 128kbps I think the clear winners would be AAC and Vorbis.



Well the idea behind this site is to eventually provide general information about all the codecs represented here, so hopefully this site would fit that need. Currently the FAQs are not up yet though, and I don't really know of any other AAC sites geared towards archival quality. Honestly though, the situation with AAC isn't nearly the same as with MP3. With MP3 there is a very wide variation in quality among encoders, while with AAC I don't think this is so much the case. That isn't to say there aren't differences (Ivan's test shows that there are) but the situation isn't really the same. With probably the most widely used ISO AAC compatbile encoder (PsyTEL) the presets are already tuned very much for quality, so there isn't much to talk about there either like there would be with LAME.

Hmm. Somehow I seemed to have mixed these two things together for a pretty long time. So let me see if I got things straight this time:

AC-3 is MPEG-2/4-based multichannel audio licenced by Dolby.
AAC is MPEG-2/4-based multichannel audio licenced by Dolby and others.

How on earth have I failed to build a clear distinction between these two



Good to know. I somehow seem to express severe dose of healthy scepticism towards everything that is advertised to be "revolutionary" or "superior to any current product". And in case of AAC, if that claim is backed up at all, is backed up with some comparisons between 64kb/s AAC and MP3.



My spoiled eardrums would begin to pulsate painfully if I even considered using less than VBR ~180kb/s for audio encoding



And undoubdly it will when the FAQs come out and when the forum gets a little more messages in it. Unfortunately the net knowledge on this site seems a bit thin at the moment.

niktheblak

I'm not actually sure if AC-3 actually has anything to do with MPEG-4 or not. I believe it was standardized before MPEG-4 was even being really discussed though I could be wrong. AC3 is mostly used on dvd's though and its use is limited at lower bitrates because it is outperformed by other codecs such as AAC.



Eh.. not quite. AAC is without a doubt superior to MP3. There is no question about that at all. AAC in a way is kind of like the successor to MP3, it is much more technically advanced and it addresses a lot of the design flaws in MP3 (such as the lack of a scalefactor for the last scalefactor band, or basically the whole > 16khz issue). AAC has been shown to be superior to MP3 across the entire board in pretty much every situation.



This is part of the advantage of AAC over MP3. With MP3 you need vbr bitrates near 256kbps to really acheive transparency across the board in most situations I think and even then this is not always possible, but with AAC this is much more attainable at bitrates near 180kbps.



True, however I don't think you will find another site out there with so many knowledgable people gathered in one spot involved in a community like this.. so at least that helps for the time being

Yes, I trust you and everyone else saying this. I'm just saying that I would be using RealMedia or WMA @ 64kb/s if I immediately bought all this commercial hype. Now when AAC's superiority (at least over MP3) has been verified by sources I trust I can move forward to test-listening period



Now that's also true. Heck of a job you did back here, Dibrom! Magnificient idea, having quality-oriented audio encoding site with most of current developers and experts as moderators or at least regular readers. Simply superb!

niktheblak

AC-3 v AAC:

If you play with Liquid Audio, 192 AC-3 seems to be similar to 128 AAC.

AC-3 is pretty inferior to AAC - no matter what Dolby says.

However, AC-3 is the de-facto standard in the USA for high bitrate coding (DVD, HDTV, etc...)

Try: www.mp3-tech.org

It explains a little about the different formats architecture....Including a "what is AC-3?" part.

Hope that is of some use,

Cheers,
-Nic

AC-3 has its advantages - it is very low complexity, and at the time it was invented AAC didn't exist. DSPs able to encode MP3 were much more expensive than today.

And, while the MP3 was basically European invention, AC-3 is pure USA product of the USA company - and the most of the USA standards are being standardized by USA companies without goverment funding and support involved (unlike Europe), so USA companies were in position to choose their coding systems in HDTV and DVD standards.

"Dolby states quite clearly that CD-equivalent results begin at 192 kb/s with AC-3 (the codec at the heart of Dolby Digital)."

HiFi News May 2000

Ruse

That would be in the same way that CD-equivalent results start at 128kbps with mp3.

no it doesn't silly. it starts at 64kbps with wma!

The point I was trying to make was that AC3 was not designed as a low bitrate codec!

As I stated quite a long time ago I had to do some serious listening tests with AAC and I'm quite happy with my results.

For the encoder I user PsyTEL's aacenc 1.2 with the preset commanlines -normal and -extreme. They gave my test albums average bitrates of ~195kb/s and ~202kb/s for -extreme and they sound VERY good. One of best sounding (if not THE best sounding) lossy codec I've seen up to date!

This would be very ideal solution for me since I dislike MPC for it's unnerving behaviour with middle-range energies (I've brought this issue up rather long time ago but won't go further into it, not without ABX results...) but few problems remain.

1. Legality. I'm not quite sure if it's even legal to use PsyTEL's aacenc (or any AAC encoder whatsoever) even for personal use. Since I don't want to violate the rights for the copyright owner I will switch to a freeware solution (*COUGH* *OGG* *COUGH* ) if I'm required to do so.

2. Personal lack of experience for AAC and AAC encoding. I have absolutely no idea what the parameters do. I.e. what command line I need in order to make a good 192kb/s CBR etc.

3. This is a minor problem, but a problem anyway. Speed. "aacenc -normal" encodes about 0.9x of realtime on my Duron 650. That's very slow compared to MPC and quite slow compared to OGG. I hear that there would be some kind of "fastenc" for AAC, can anyone provide some insights on this?

So can you do some abx tests? I don't remember seeing this before. But it would of course help to know more, I haven't noticed any middle-range energy problems with MPC.

CD-rw says at r3mix forum that MPC has exaggerated highs, some other says at MP+ forum that MPC has no highs. And nobody ever shows ABX results... I wonder why.

Niktheblak and others: I splitted the thread so MPC quality discussion continues here:
http://www.hydrogenaudio.org/forums/showth...s=&threadid=325

In spite of that MPC spinoff conversation, I still would like to get some information about AAC in general.

In addition to my previous questions, I would like to ask about Liquid Audio (uses AAC I hear):

I see very many people stating that Liquid Audio is currently the best AAC encoder available, however few points remain unclear;

Can Liquid Audio do VBR (or multichannel, higher bitrates) at all? There's absolutely no technical information in the homepage.

And how about it's quality? According to LA homepage "it encodes up to 160kb/s" and I assume that it's CBR. So, is LA that good or is it just designed for the 128kb/s market?

If I want archival quality should I go for PsyTEL instead?

I don't know about multichannel, but it can do VBR (I'm referring to Liquifier 5.0 throughout). The presets are rather limited: 128 CBR, 128 VBR, and 192 VBR, but you can specify the bitrate and mode to your liking.



I guess Liquifier 6.0 must have different presets. But even just 128 VBR in Liquifier 5.0 is excellent quality.



Psytel sounds very good as well in its archive setting. At 128 CBR, I think Liquifier is superior (and Ivan's test shows that as well on castanets.wav).

ff123

Liquid Audio is my favorite for the rates 128Kbps and lower. I've tried both Professional Liquifier, and the freebee encoder which comes with the latest Liquied Audio Player (rates up to 78Kbps, but can be upgraded for the higher rates for only $20.00 or so).

Having said that, it appears that the future of Liquid Audio looks rather bleak. They are not doing that well, layed off part of the staff, they apparently have a fairly large Mac user base, but their Mac-based software is really buggy. Very unfortunate I would say, since at this point they are the only one who provide free encoder which is perfectly legal to use. (just download their latest player from their site, and encoder is included).

And so it appears that the safest bet is use LAME, or OGG, or MPC whose existance doesn't depend on market conditions and a good business plan (well, I"m not sure about OGG or MPC, but at least this is the case for LAME).

My 2 cents.

Yes, for commercial products there is always the risk of inverstors taking their money elsewhere.

But since it seems quite sure that AAC itself isn't going anywhere in some time, how about open-source developement?

Remember, when MP3 emerged, commercial products (FhG) were much better than freeware implementations (early Blade and LAME). Nowadays LAME outperforms commercial FhG and Xing just about anywhere.

In AAC scene we now have commercial Liquid Audio outperforming open-source FAAC but if history repeats itself the future looks bright

Only thing to worry about is patent differences between AAC and MP3 format. I see however that AAC encoder/decoder source code can be distributed "for educational purposes" just like in LAME.

MP3 and AAC share a lot of same patents, however the difference is in the licensing authority (Thomson for MP3 and Dolby for AAC, in the future MPEG-LA LLC will be responsible for the MPEG-4 patents)

Another difference is the patent licensing policy - Dolby&Crew agreeed not to allow non-SDMI AAC implementations for end-user market. That is very bad since we deal with industry trying to forbid access to the high-technology to the ordinary user.

And, finally - making a good AAC implementation is little bit tricky since many encoding tools in AAC are badly explained (if explained at all!) in ISO docs. Few most important things are totally non documented, probably on purpose.

FAAC is a great project, encoder lacks high quality at the moment, but decoder is really a state of the art. Menno is now rewritting FAAD completely and I expect new library to be a lot faster, too. We will try to integrate some advanced Intel optimizations with new Intel IPP library.