Huh? No, normal HE-AAC is meant to be used over a reliable channel. There are some versions with error resiliency (as used by DRM), but the entire solution is hardly designed up front for that, and those variations of AAC are not generally used. (The same applies to MP3)

Since HE-AAC incorporates LC-AAC, a HE-AAC solution will scale from the lowest to the highest bitrates, and is extremely efficient all over that range...

How's aacPlus supposed to be at higher bitrates? I know it's regarded as useless or wasting but is it as good as plain vanilla aac for example or worse?

The reason I ask this is that my favourite internet radio Radio Free Colorado recently experimented with 96kbps aacPlus (in addition to 320kbps MP3!). Didn't hear it though. Normally they stream aacPlus at 64kbps.

In the general case it will be worse than normal AAC.

Do all currently available aac-capable mobile phones
only support LC AAC?

I'm particularly interested in the Sony Ericsson W800...

I've read W800's white paper, and it says LC-AAC only.

Anyway, what's the difference between Main Profile (MP), Low Complexity (LC), and Scalable Sampling Rate (SSR)...?
And how do you encode an AAC w/ any of those profiles aside from LC?
I just came across those from reading W800's white paper.
Well, I know LC, but I don't know what's in it (as in why is it called LC).

if i didnt get it wrong lc devices such as the k750i/w800i can play he streams, the extra info just wont be processed. and its my understanding that the so-called aacplus v2 is he-aac with ps. is that correct? has anyone tried playing back an aacplus v2 48kbps ps stream on one of those two se phones?

You Don't Know To Know And You Don't Want To Use Anything Besides LC

(Note that HE-AAC/aacPlus is a LC-compatible)

If you do want to know, use the search, audiocoding wiki, or google

Please read post #8 in this very same thread. Sometimes it is annoying to write twice the same.

SSR (Scalable Sample Rate) is designed for streaming through variable bandwith channel. Most of us will never use this profile, including you.

HE-AAC v2 (including HE-AAC and AAC LC) is the standard for 3GPP release 6 compliant mobile devices, but it is not yet a consolidated standard in the market. Give it some time.

Regards,
Oki

So the difference between LC and MP is LTP (Long Term Prediction).

Why would you not want to use this feature, besides hardware compatibility reasons?
Is there no encoder that implements it properly? Is there very little to gain from it?

I mean, surely it exists for a reason.

Main profile is implemented in all encoders IIRC unless otherwise especified. LTP adds complexity and the compression ratio is not improved that much. This is the main reason why AAC LC profile is used instead AAC Main.

IMHO, with a good PAM controlling the configuration of the tools used in compression, LTP could be very useful for heavy tonal parts of the compressed sound, allowing a huge gain in quality/bitrate. Same states for TNS with sounds with high dynamic range. Maybe commercial AAC encoder developers could give more details about this, based on real test.

Regards,
Oki

No, no, no, no!

LTP is a *different* profile from LC and Main.

AAC LC = basic AAC
AAC Main = AAC with predictors
AAC LTP = AAC with better, different predictors (only MPEG4)

Egad! I noticed I didn't post any controversial opinion in this thread.

This will not do!


It is my opinion that all these low bitrate enhancements are a cruelty of fate, because they are being developed precisely at a time when they are not needed anymore.

If these enhancements were released in the late nineties or early 2000s, they wouldn't be anything short of groundbreaking. Back then, everyone was still on dial-up (great streaming possibilities!) and the most expensive flash-based DAPs had 128~256MB of storage.

These days, most people are on broadband, making streaming a moot point, and you can hardly find a DAP with less than 512Mb. Microdrives handling several gigabytes are being added to cell phones and optical storage is breaking the 50Gb/unit barrier.

Likewise, "spatial audio" (the multichannel version of parametric stereo, another novelty that came too late) would have been great for 1-cd DVD rips in 2003. Nowadays, most people don't even bother to rip DVDs, instead just copying it verbatim to another DVD.


I'm confident these fancy new buzzwords will find their place. SBR and PS are very useful as AM replacement in DRM, and some people more space-conscious than quality-conscious will want them on their DAPs. But for most people and purposes, they are too late, too little.

Regards;

Roberto.

By 'only supports LC-AAC', I meant it'd only play LC-AAC properly.
And HE-AAC'd only get the same treatment as LC-AAC and makes no sense creating them specifically for those SE phones use.


About the LC bit, I actually have read it before; I just forgot I did.


'Variable bandwidth channel'?
Like as in what?



Hmm...this and w/ Oki's post some times ago.

AAC LC = MPEG2 AAC + PNS --> (???)
AAC Main ~= MP3 + TNS + TP or MPEG2 AAC --> (???)
AAC LTP (or MPEG4 AAC) = MPEG2 AAC + LTP + PNS

What is 'basic AAC' anyway?

Btw, from what I've read on Wikipedia (not HA's own Wiki), SSR is introduced by Sony.
That it's somehow similar to its own ATRAC and ATRAC-3 formats.
And that its quality is below the other AAC profile on the same bitrate.
Anyone can confirm this?

Sorry if I'm asking too much...
I just still feel pretty mixed up w/ all these naming conventions...
And much thanks for the replies!

I am sorry, The information about MPEG4 I provided in post #8 was a little misleading. My post did not cover all the technologies in AAC like CELP, TwinVQ, etc because those are not used that much. I tryed to condensate my post as much as possible but explaining the difference between different AAC profiles was not the main target. I was more focused to the tools.

The profiles are described in ISO/IEC 14496-3, Subpart 1, paragraph 2.1. It is not easy to explain all the different profiles since, for instance, MPEG4 AAC Main is like MPEG2 AAC Main but supports also PNS. Same for MPEG4 AAC LC. It is like MPEG2 AAC LC but also pupports PNS. There are more than 15 different profiles in the original MPEG4 AAC. After 6 admendments there are more than 29 (SBR, PS, etc...) IIRC. Prese refer to the ISO/IEC 14496-3 and its admendments for a complete explanation.

Regards,
Oki

It is an hierachical scheme of streams inside one AAC stream. the same stream provides different qualities for different targets. One stream can contain up to 3 streams for 11.025 kHz (Stream 1), 22.05 (Stream 1 and 2) and 44.1 kHz (all 3 Streams) sampling frequency.

The basic AAC specification has several profiles too. The basic AAC is described in IEC/ISO 13818-3. I do not have that specification since you have to pay for it. I would like to have it lately.

You sacrifice quality for scalability. Main profile's quality/bitrate ratio is higher than SSR's.

Regards,
Oki

You are completely wrong on all aspects.

Firstly, the target of these formats is largely broadcasting, especially so for spatial audio. For a broadcaster, the equation is simply less bitrate == more channels or less bitrate == less costs, and it will always be the case, even if bandwidth gets cheaper, less is still more. Just look at what happened with DAB in the UK. Or internet based broadcasting. More quality in smaller spaces is always welcome. Unlimited quota broadband is not something that exists in 99% of the world.

If I read your post it's like the only application for compressed audio is people ripping music on their digital players, or making DVD backups. That is false. These formats are now being used for example to push digital television further along.

Don't forget what the M in MPEG standard stands for.

Secondly, no matter how big your mobile player storage is, 32kbps PS AAC will still fit four times as many songs as 128kbps, and it will also sound acceptably good. Don't tell me this is useless if the storage is so big. I would expect most people to be always craving for more space, or for better quality in the same size.

I'm wondering why you left out of the quote precisely the paraghraph where I wrote the caveat about these enhancement techniques not being useless - and I specially mentioned broadcasting

Let me quote it then:


My point is that a window of opportunity for these formats has pretty much closed now. That's the "too late" part in my post.

Can you deny these technologies would have been much more useful years ago?



Well, I wouldn't expect an AAC vendor to do otherwise, but come on, Garf... 32 kbps acceptable? Now you're sounding like Microsoft

Besides, that's precisely what I said in the part you forgot to quote. "some people more space-conscious than quality-conscious will want them on their DAPs." As DAPs are coming with 60Gb these days and storage will only get bigger and bigger, fitting more songs on the player won't make a difference because most people simply won't have such a big music collection anyway.


So, no, I'm not completely wrong on all aspects, although I understand that, from a professional point of view, you wish I was.

I reread the post I was replying to, and this wasn't stated anywhere.

You only mentioned DRM. I was thinking about *quite a bit more* than that. Broadcasting is way bigger than just one shortwave standard.



I don't see how, and I don't see any argument why. Because some players have 60g disks? The smaller ones are still cheaper and handier, you know.



Their effect is the same as their effect is today. You can do with significantly less storage, which means less costs.

So, no, I don't think so.

I guess putting todays technology into the world of 10 years ago would be interesting, but the fact is that it never happened that way.



32kbps PS AAC is very acceptable for portable playback. This isn't marketing, it's a fact, supported by your own listening tests.

32kbps AAC was rated above or equal to MP3 at 128kbps encoded by FhG, Gogo, iTunes, and Xing.

If that is not acceptable, then why are or where so many people encoding to those formats?

I'm not saying it's great or that it's transparent, but it's usable.



Is trolling really only the only thing you are capable of?

DRM was an example. I could post several more, like Tuner2 and the like.



Even "smaller ones" are going past the 1Gb capacity, and getting bigger and bigger as time goes by.



Giving up on quality, so it's not the same thing as just comparing storage prices. If SBR was lossless, I would be all for it. But it's very far from it.



Whoa. You know more about my listening listening tests than I do.

Is it just me, or are you trying to extrapolate results across tests? As far as I remember, that's highly unrecommended around here. Specially with two tests conduced at such different bitrates.

Hehe. Ok, ok... I would say that is rather your opinion than a fact, but what is the point of arguing with you?



Your definition of trolling is too broad, Gian-Carlo. Why you get so touchy when your professional interests are mentioned? It's perfectly natural to give some priority to professional interests.

"Giving up on quality" is subjective. The question is if most users really feel like they are doing that, or even whether the tradeoff is worthwhile or not.



You are right, the 32kbps was anchored twice, but the 128kbps one was not. My bad, conclusion must not be taken -yet-.

Shall we do a new test?



I definitely see some signs of trolling:
Sign 1: turning it into a personal matter.
Sign 2: trying to undermine what I say by attempting to attack my credibility

Please, be calm.

Garf, I understand that if you have acceptable quality for a portable DAP at 32Kbps HE-AAC v2 then no one will prefer 128Kbps MP3. It is the difference between 3 minutes and 12minutes uploading to the portable device. It means a lot for commercial purposes too, most DAB content and its rips will be done in HE-AAC v2 as well as DVB's. I have no doubt that the future is for AAC and its versions. MP3 will be supported only as legacy.

Roberto, I also understand that for audiophiles, it is easier to obtain lossless quality cheaper and cheaper as the time goes by. The time will bring cheaper and with bigger capacity portable DAP, more Internet bandwidth, cheaper DVD-R, more capacity and features in our mobile phones...
It is true that AAC would have been a revolution 10 years ago, but also AVC (h264), wifi, perpendicular recording, PCI-express, dual cores, etc... Does it means that evolutions are not valid, only revolutions?

Roberto, I am more aligned with Garf, Remember that most people out there can not identify 32Kbps HE-AAC v2 and 128Kbps MP3, even most people think that iPod is a "MP3 player". They only understand battery life, uploading time, recharge time and sound effects (usually more bass). But I agree with you that 10 years ago HE-AAC v2 would have been a big fat revolution.

Regards,
Oki

Of course. I've been awaiting a successor for more than an year. Noone felt up to the challenge so far, it seems. Whenever someone does, I'll be extremely happy to help him/her in every way I can.



I never tried in any fashion to turn this discussion to the personal side, even though it's a well known fact around here that I have no sympathy for you. All my opinions in this thread are based on my view of what happened in the past and what might happen in the future of the audio compression arena.



I'm not trying to attack it. I'm just pointing to an obvious reason for you to want HE AAC to succeed. It's your job at stake. Anyone looking at your member title can come to the same conclusion.

Considering that HE-AAC v2 is pretty new, and that there haven't been any large scale public tests, and considering the difficulty to prove such a generalizing statement about "most people" being not able to distinguish them using listening tests, I wonder what makes you think that this could be established as a fact?

My girl friend has an ipod. One day I told her, ipod plays AAC and AAC gives more compression and suggested her to use AAC instead of MP3. She said to me, "my ipod is never full". This is a good example showing that AAC brings less impact to our daily life than it could several years ago.



As my personal feeling, when using ear phones, HE-AAC-PS @ 32kbps does not always sound as good as MP3 @ 128kpbs VBR, maybe for some songs it really does, but just not always. If one day HE-AAC-PS-VBR gets widely available, things might be better.

Acceptable, yes.


???
Am I missing some test results?
I do not doubt that in internal tests you found some tracks where AAC-SBR/PS @32 was better than some MP3 @128, but a generalization seems highly risky.

note: if for whatever reason one day you fall into pure marketing, I'd suggest you to compare against Shine, this way you could have AAC-SBR/PS @32 superior to MP3 @192.

It is my experience and you can try for yourself. Ask your neighbor, sister, mother, etc... as random as possible to listen something @ 32Kbps HE-AAC v2. Please do not interfere with their hearing, mood or environment if you want a fair result. You will be surprised how most people can not identify what is a good audio quality.

Even more, FM radio is considered acceptable even for hearphonesby most listeners when actually its quality is far from 32Kbps He-AAC v2 and XM Satellite Radio is considered as CD quality by the general audience over 128Kbps MP3. XM Satellite Radio is broadcasted @ 64Kbps HE-AAC, very close in quality to 48Kbps HE-AAC v2.

Anyway audio quality is always subjective, and we can talk about preferences for ages.

Regards,
Oki

Could you give some links? I never see such tests. I'm very interested, thanks

Could you back up your claim please? What's the next step: that AAC with PS at 48 kbps is undistinguishable from MPC --insane or LAME -V0?


We're on HA.org. Several users on this board don't care about untrained opinions.

Read post #76. Thanks

Read TOS#8. Claiming that 32 kbps is as good as 128 kbps need ABX results. Results. Not a vague assumption.

More like wishful speculation in that case.

This turns out very interesting, especially the '32kbps HE-AACv2 quality' bit.
Personally, while it might be satisfiable in noisy environs, it won't be in quiet places.
So, 32kbps HE-AACv2 might sound fine on certain situations, but from what I've experienced so far, it doesn't sound as good as 128kbps MP3 LAME encodes when using earphones.

But that's just me.

Anyway, I partially agree in some way to Roberto's thinking.
Especially the 'storage is getting cheaper' part.
It's true that most flash-based DAPs nowadays are capable of sporting more than 512MB, or even 1GB, of storage.
This leads to virtually unlimited music on a small, light-weight, DAP.
(well, personally, I don't hear music on my DAP more than 50 tracks a day, thank you)

But this doesn't mean 'greater compression at an acceptable quality' techniques'd get obsolete.

Who wouldn't want smaller files that sounds almost indistinguishable from its larger counterparts?
It doesn't matter how much storage you have; some people just want more and there's a market for it like DABs.

@Oki
Btw, thanks for the explanation before.
Though I still have this muddled view of AAC naming standards, but I really appreciate it. =)
Will probably try to figure it out by myself and hoping it's not so far from the truth.

I guess, it's a dirty truth but one of the most popular uses for HE AAC will be 5.1-channel movie soundtrack compression for flicks shared through P2P networks. For those folks it cannot ever be quick enough and hence files cannot ever be small enough. Also it is used for homebrew DVD backups through the Nero Digital Package.

Gabriel and guruboolez: I pointed out which test I was referring to. (It wasn't anchored - so the comparison is not valid.)

I would really like such an anchored comparison (*new* Nero HE-AAC @ 32, 48 and 64 kbps, together with CT aacPlus and perhaps Vorbis against MP3@128kbps, in a *large* listening test), the results would surely amaze a few people. But I cannot conduct this publicly myself, since I'm hardly a neutral party.

The last such comparison was in 2003, and then HE-AAC at 64kbps scored 3,7 versus 4,3 for LAME.

I think a new test would show in what direction HE-AAC can go, as was the original subject here

Not yet. Parametric tools run into their limitations at the highest end. They were not designed for a transparency-craving HA crowd. But they will perform fine for everyone else.



That doesn't unvalidate his statement. Not everybody is a trained listener. I would say most people aren't.

Neverthless, he did a claim that begs for factual proof, like a listening test, and he didn't provide any.

"most people out there can not identify 32Kbps HE-AAC v2 and 128Kbps MP3"

Clear TOS 8 violation there.

I made a statement and I also gave the method to duplicate it, according to TOS #8. It is very easy to implement, do it yourself. It is like asking if someone can identify a diesel engine from a gas engine. Most of the people (mostly in the States) will not be able to differenciate the sound. This last statement was also based on my experience, just asking others.

You could be very surprised if you take ito account the experience from others, ramdom others, even your grandfather or your daughter, with their resources since most people has a medium quality sound system and the quality difference is always masked. For some people, even a CD will sound like 64Kbps WMA and this statement does not mean that 64Kbps WMA has the same quality than a CD.

I can say also that most of the people I know (again from my experience) in Texas, UK and Spain listens WMA @ 64Kbps in theirs portable and car DAP because it is enough and it is the fastest way to download and upload music.

I am sorry if my english is too limited and the real meaning of my sentences are wrongly understood.

It you are reading this, please, test this stream using foobar2000 or Winamp 5.1 with your family in their usual listening environment and write if they think if this quality is acceptable or not. Not an ABX test but enough to see if 24Kbps HE-AAC v1 quality is acceptable or not.

Regards,
Oki

Acceptable is not the same as transparency. I found it acceptable. But I cant claim its "better than mp3 in 128" in all situations.

I do think a lot of people cant hear a difference between an original and a 24kbps AAC+ encoded file. But many of them just think "man, that's a really good sound for such a low bitrate!!". But it doesnt mean its better than all other files in LAME 128 VBR fo example. Just acceptable for that bitrate range.

I know you're not talking just about transparency but I'm just saying IMO acceptable sound and tranparent sound are different things and many people claim as if those things had the same meaning.

I suspect you didn't understand TOS 8 at all. Let me quote it here:



The part I emphasized means YOU have the burden of proof - that is, providing the hard data supporting your claim. The thing about posting a sample (method) so that others can reproduce (duplicate) the findings doesn't free you from proving that the majority of people (that's what "most" means) can not identify HE AAC+PS at 32 from MP3 at 128.

If just posting methods was enough, everyone here would go around claiming crazy shit and just reply to anyone contesting it "oh, just follow these steps, you should hopefully come to the same conclusion"



Nope, you should do it and provide the results data. As I said, the burden of proof lies on you and you alone.



So, your method was asking others. Care to clarify? How many people did you ask? Are they a representative group? In what conditions did the test happen? Was it a blind test?

But this situation is a bit different. Oki is claiming that people *can't* hear a difference, which is an impossible thing to prove, even with ABX results, as opposed to saying you CAN hear a difference, which a positive ABX result proves. I suppose he could fire up foobar's ABX utility and randomly click on buttons without even listening to the samples and come out with a failed ABX. That still wouldn't prove that the samples sound the same.

It seems to me a gray area RE the TOS, because it is making claims concerning subjective sound quality, but there is no proof I know of that can prove its truth.

I suspect you didn't understand my statement:

"Remember that most people out there can not identify 32Kbps HE-AAC v2 and 128Kbps MP3"

This statement regards the ability of most people to differenciate two levels of quality, not the quality of the codecs. The rest of you dissertation is futile.

By the way, are you suggesting that my statement is a crazy shit?

Regards,
Oki

Hmm. Where there? Don't you want to be more specific?

I don't see any gray area or impossibility here.

Set up a test with random people, ABX-ing HE-AAC@32kbps vs MP3@128kbps vs originals.

If you can't make a conclusion from that that MP3@128kbps > HE-AAC@32kbps, then
that may not prove they are "equal", but you *have* shown that an average person can't differentiate them with any kind of significance. And that *is* the point here.

No that is not the point. Right now we can't make any assumptions if "most people" can or can't hear the difference. I doubt "most people" have ever heard of HE-AAC let alone tested or listened to it.

Oki made an assumption without providing any sort of test results to back them up and then asked others to prove him wrong. That is cleary not in the spirit of TOS#8, as we can quote from it:

Now where did Oki provide objective support for his claims? I am sorry if I missed it, but I only saw him talk about his "experience", which is rather subjective. It is not any more difficult to claim that most people I know can tell the difference.

So now we know nothing more than we knew before. Some people can tell the difference, some people can't. Obviously that is almost always true if we talk about lossy encoders.

It is strictly a TOS 8 violation, just like my post is. You must provide the evidence for your claims.

I am willing (and would be happy) to set up such a test. But would anyone believe a good result from a test conducted by a "Nero MP4/AAC developer"? I am sure I would just be wasting my time, since people in this thread have already pointed out I am not to be trusted.

So I will STFU now, and we will try to get the new PS encoder out the door

BTW, I fail to understand how "the level of quality of 32kbps HE-AAC and 128kbps MP3" and "the quality of the codecs" is any different. What do you mean?

Exactly. I suppose the correct way to prove that point would be conducing a blind test with a handful of representative samples and maybe 20 listeners. If more than 50% of them can't listen to the difference, I suppose that's enough proof to claim "most people out there can not identify..."

That's what statistics is all about. When people claim "most americans have a television at home", they didn't ask every single american if he/she has a TV set at home. They took a small, representative group and extrapolated the research results.

It is impossible to do that. you can not ABX with a radio receiver, and it is really complicated to do that with a car stereo while you are driving... It is easier just asking if they think if this is acceptable or not. Most people are not doing an ABX test when they are buying a sound system, it they think it sounds good, they buy them. I am not talking about the quality of a codec, but the quality of the sound systems, environments and requirements of the people. If you want to extrapolate the philosophy of an ABX test to the rest of your life then it is better to do a blind taste before selecting a meal from the menu. From my personal point of view this issue has nothing to do with TOS #8.
GREEEAT!!!!

It is the same, and has nothing to do with the codec quality.

The meaning is: Most people has a medium quality sound system with a non optimal listening environment (car stereo, the noise of their children in a home, The working noise in the office and the traffic noise while walking on the street). This happens most of the time to most of the people and they can't hear the difference in their normal environment. An ABX test is artificial and valid only for backing up some hypothesis, trying to give a more objective view of a subjective perception under a unique controlled environment.

I am not saying 32 @ HE-AAC v2 is the same in quality than 128Kbps MP3.

Sorry for my limited english.

Regards,
Oki

I'd be very interested in the results from such a test and happily accept the results even when it is conducted by a part who has economic interest in the outcome, as long as the testing methodology is fair and transparent.

Two requierements on such test:
* don't cripple mp3 by using anything but the best encoder
* don't use a not-yet-released version of any encoder (specifically Nero)

So, continue to work on the new PS encoder, but after it's ready for release, please come back and start that listening test.

These kind of things are an issue, because there is *always* someone popping up to point out that whatever you used is (in their eyes) not optimal, hence your test is invalid, whine whine...

I agree with you on that, but how can I set up a blind test while driving, while at the office, while I am playing with my daughter...? You really need to ABX several persons but what about the testing environment? A blind test requires the same environment and my point is that environment masks the quality of the codec.

I would like to see a blind test under real conditions since all the test described in HA were performed under the best fidelity achieveable.

Regards,
Oki