Howdy!

Here's a relative newbie question for you all:

At this point, when designing a commercial portable digital audio player, is there any reason to support the AAC-Main profile on top of AAC-LC?

I understand that AAC-Main achieves better subjective scores than AAC-LC, but by margins that usually warrant the significant additional computational and memory complexity. A bit of Googling has me concluding that most all widely-used encoders/decoders use AAC-LC rather than AAC-Main. Also, doesn't iTunes still use AAC-LC over AAC-Main? In general, is there any popular encoder/decoder that regularly uses the Main profile?

Thanks for any insight!

Cheers,
~dave

Not at all.



Do _not_ warrant. But I think that's what you meant?



Main profile is dead. Even the improvement over it introduced in MPEG-4 (LTP profile) is just as dead.

HE-AAC and HE-AACv2 are meant to be used with LC-AAC, not Main-AAC or LTP-AAC. So clearly, the Main and LTP profiles aren't just dead but already buried as well.

> Do _not_ warrant. But I think that's what you meant?

Indeed, I missed that key work, "not"...

> Main profile is dead. Even the improvement over it
> introduced in MPEG-4 (LTP profile) is just as dead.

> HE-AAC and HE-AACv2 are meant to be used with
> LC-AAC, not Main-AAC or LTP-AAC. So clearly, the
> Main and LTP profiles aren't just dead but already
> buried as well.

I see... Thanks for the info!
~dave

It is my understanding that one of the main, perhaps the major, contributor to MPEG-2 AAC referred to "main profile" as the "High Complexity" profile. While that is an informal opinion, I suspect that one could investigate with the fellow who said that and ask for more information, if he wishes to actually comment on AAC at all at this time.

My own understanding is that it takes a lot of processing, and might have some effect on difficult material with stationary tones, and that about 7 or 8% of the time.

Well, HE-AAC is far from being transparent.. compared to LC or Main or LTP mode..

I'm not sure what this is supposed to mean. HE == LC profile for the sake of this discussion. They're compatible, and HE is an extension on top of LC (and not Main or LTP).

Do you have a Main or LTP encoder that is competitive with the current top LC encoders at high bitrates or with HE-AAC at low bitrates? I certainly haven't seen such a thing.

For hardware portable audio designers point of view, transparency means close to CD / Hi-Fi quality at whatever bitrates.

I am sure HE-AAC isn't designed for this purpose in mind. HE-AAC is a form of parametric audio coding useful only for low quality audio transmissions such as FM radio broadcasts, DVB etc-etc.

Well, HE-AAC can sound transparent to some people, especially those with low end speakers. Also, noisy environments (common with DAPs), and cheap head/earphones help. It is quite suitable for usage in DAPs, especially since it saves space. It's not like the de facto standard of 128kbps CBR sounds transparent in most cases.

For the second time, what has this to do with anything? This is about Main-AAC, LTP-AAC and LC-AAC. You are _completely_ missing the point.

HE-AAC is an extension on LC-AAC and LC-AAC can definitely be a "transparent" coder. An HE-AAC decoder is also a LC-AAC decoder. Not the same for Main and LTP profile. If you design an HE-AAC decoding portable you can get transparent quality decoding too, because you will have made an LC-AAC decoder as well. If you design an LC-AAC decoder, you will still be able to play back HE-AAC but at reduced quality. If you design a Main or LTP decoder, you aren't going to play back j*ck sh*t!

There is *zero* reason to design anything around Main or LTP AAC right now.

Yes.. but even LC-AAC requires at least 192 kbps to be near CD / Hi-Fi like. Signals with lots of closely spaced tones or contains a lot of pitch information cannot easily be coded using LC mode. That is why the LTP and Main modes come in. At 128 kbps, I'm sure LTP or Main profiles outperforms LC (including HE-AAC)



Oh.. I am sure there must be some reason why the MPEG committee included those profiles into the standard.

Hmm - lessee:

In the recent 128 kbps listening test AAC scored quite near CD / Hi-Fi like:

http://www.hydrogenaudio.org/forums/index....showtopic=40607

Even taking the lower end of the confidence interval, LC AAC at 128 kbps scored at least 4.60 (worst score) - well beyond the ITU-R/EBU's criteria for "indistinguishable" quality. So much for the "CD Like"

Regarding the MAIN/LTP, even if we try to forget the TOS8 and think of what could happen if we include LTP/MAIN into modern AAC codecs of today (such as Nero and Apple) I don't think that you would gain anything even remotely close to claim any statistical significance.

In 1998 - the LC and MAIN codecs scored like this:

http://www.tnt.uni-hannover.de/project/mpe...ublic/w2006.pdf

As you can see, AAC MAIN and AAC LC are statistically tied - I have no reasons to believe it would be different today - they would most likely be even more tied as LC core got better.

The burden of added complexity completely overturns any gain made from these profiles - it is simply not enough.



Yes, there must be a reason but I think it falls out of scope of this forum

I'm not so sure, which is why I asked you to provide evidence earlier in the thread. In the latest 128kbps test, LC AAC was transparent for like 80% of the listeners/samples, and this was with known difficult material.



The same one that caused MP3 to get a hybrid filterbank, perhaps?

Note that LC AAC is the de facto standard now, so reasons matter little.

How about AAC SSR and AAC Scalable?
Are they dead too?

Dead for a long time.. I have not seen an implementation using the SSR mode.