note that its now called nctu-aac and not nctu-faac, as it was named till now
well i wonder if this now should make people think that it isnt based on faac anymore

"NCTU-AAC is a new generation of AAC encoder to release the limit of LGLP license problem compared to its last generation NCTU-FAAC coder."

anyways there is no sourcecode available

http://psplab.csie.nctu.edu.tw/projects/in...c-download.html

I dl'ed this. Options are not all that clear. there does not seem to be any provision for tagging or the mp4 container, so the output will need additional processing. Of course their tests say it sounds as good as Quicktime at 128.

Have a look at their great comparison:

The only reason why QT6.3 wins in their test and not NCTU-AAC is that they couldn´t find an old version or a totally bad commandline for QT!

Wait a minute.. this is part of my copyrighted, optimized, geniously designed, fetish command-line

You are right!! LAME3.88 & LAME3.93 testing are not their "best" setting. They should be changed..

I just want to say one thing again:

Don't trust the ITU-R BS.1387 Peaq Basic -model's ODG figures they are using here and claims to be near QT 6.3.
It doesn't necessarely say much anything correct about the true audible quality, and it's worthless for measuring temporal accuracy.

Read here Alexander Lerch's (Zplane development) comment about his EAQUAL - Peaq Basic tool.
The most important point he says is:


Listening testings are extremely important, even more so, when all you have is Peaq Basic, which is practically deaf to temporal accuracy.

You don't see any other codec developer presenting ODG numbers even from Advanced Peaq model (even less from Peaq Basic -model NCTU is using). That should tell everybody even without technical knowledge how much value it should be given especially in codec vs codec comparison, for what Peaq Basic is totally worthless (and the exact use NCTU is using it).

Peaq Basic has some use in codec development, when you measure the same codec's relative quality development (in non temporal cases). But when using the Peaq Basic for "codec vs codec"-testing, it doesn't give any reliable results, because the "offset" can be different due to differences in psychoacoustic models which means you can't take the figures from many codecs and put them to the same absolute scale against each others. (But NCTU still does this)

Someone might remember the Peaq basic model testings we did here at HA long time ago. In some cases for example Vorbis samples got positive ODG, which means the model thought the encoded samples are better audible quality than the original.....

Peaq ODG number is not allowed evidence here at HA of codec quality, and for very good reasons. It has been tested here, and found to be inappropriate for this purpose.
Lets wait for the AAC blind group listening test in February and we will see how NCTU-AAC really ranks.

Before start praising ODG results as "better than QT 6.3" I'd suggest people involved in development to check the subjective quality first.

For example, this file (bassdrum.wav) sounds absolutely horrible, worse than FAAC by any means - and this is only one in a bunch.

I attached the NCTU AAC file (encoded by default 128 kb/s settings)

IMHO, a codec that targets to be better than state-of-the-art must not have failures like this, since they are very likely to occur in other music passages as well.

Right. NCTU-AAC has serious short block issues. And no, I wouldn't rank it near QT 6.3, rather probably lower than latest FAAC...
Pretty bad I would say anyway.

Here's another commonly used example sample:
http://www.ff123.net/samples/florida_seq.flac

NCTU-AAC 128kbps recommended encode attached.

Follow the exchange between Hydrogenaudio members and PSPLab here:

PSPLab Forum Discussion

That's an old thing now, since according to PSPLab they have removed all the FAAC code from their encoder.
I'd leave that behind for now, unless someone does some debugging and finds identical code from NCTU-AAC and FAAC.

Only thing which is bugging me anymore what comes to NCTU PSPLab (though it bugs me a lot) is their praising of the ODG comparison "results", and wild claims based on those. Claims that according to real life tests are very very questional.

Well, I don't know if you are talking about my listening test or another one, but I'm not so interested anymore in NCTU codecs. First, because Winamp is much more widely spread, so comparing Winamp/Dolby to others is much more interesting (In case you don't remember, I was considering either Winamp or NCTU as 5th codec).

Second, because I personally don't like NCTU, for some reasons: their behaviour when confronted with claims of breaking the LGPL (insisting they were on the right, heh), and their insistence on taking PEAQ results as definitive assestments of quality. IMHO, this seriously damages their credibility and therefore I am not inclined to take them seriously.

Regards;

Roberto.

Well yeah, I meant the test you will arrange. But there's indeed some point what you are saying.
Also there's absolutely no doubt that NCTU-AAC is very bad on temporal accuracy currently, although they want to give the impression that they are on par or very close to QT 6.3 on their forums and project pages. Everybody can pretty easily verify their short block encoding quality themself quite easily using for example clips like castanets,fatboy,bassdrum,florida_seq,death2,c44 etc etc..
Though they have lots of other problems too, but their temporal accuracy problems are very bad indeed.

I don't know, you of course decide...

When I said "Of course..." I was being a bit sarcastic. Little did I know that thier testing methods were an issue around here.

Not really.

I'll start discussion about this test in some weeks, and then I'll ask the users what they prefer. I really don't care what codecs are being tested because I'm not making these tests (only) for myself, they are mostly for the users.

Also, even though I think Winamp would be more interesting, I reckon the test would be a good opportunity to expose how flawed PEAQ is for quality comparisions.

As the purpose for academic contribution, some wording in the NCTU-AAC demostration link such as "perfect speedup" seems too business-oriented. One can easily grab the relationship between NCTU CISE PSP Lab. and Intervideo when trying to understand the "perfect speedup" by reading their patent. It seems to lack some research coverage when comparing with Nero and Quicktime by ignoring compaact .. and etc. Much focus on emphasizing comparable with Apple, superior to Nero bases on a quite not fair scores... It is hard not to remind people that all are about advertisements.

Yes, but at least Apple, zPlane and Ahead aren't trying to compare their codec to others, nor they are using a flawed method to claim their superior quality.

Now, I would be fairly happy if they made clear that the results they show were obtained using a software simulator that is prone to have flaws, and no human input (not prone to flaws inherent of software like bad modelling) was considered. The way the results are presented now is misleading and unfair.

Not to mention the "custom" settings they use for each encoder are laughable, and clearly show their deliberate interest in misleading visitors of their page.

Yes, this is the big difference why the NCTU PSPLab guys probably feel they face very critique filled response. They are not using commonly accepted methods when trying to compare (or what ever they are trying to do) codecs.

Their problem is simply what seems to be an absolute belief into Peaq Basic ODG results, belief that ODG would somehow produce an objective average results which are directly and correctly comparable to "average" human hearing, and failure to recognize its limitations and proper usage. They seem to think that their results are ok, even though no other developer or even official organizations like MPEG and ITU use ODG for their official decisions of codec qualities.

Whether this is deliberate or not, I really don't know. I hope it's not deliberate, and this is just something that will change when they gather more experience.
But if they present themselves with "Look at our results, we are the best, we have the best tech and our results prove it" -attitude in their project pages and their public forum, the response will also come accordingly (especially because their claims are obviously false and anybody with some experience can verify it).
With more humble attitude, maybe asking people to help testing their encoder and showing commitment to develope it according to feedback and such, they'd get much more positive and encouraging response also.

Actually, I would like to see NCTU AAC in some subjective test - it would be a perfect opportunity to correlate ODG of all codecs vs. SDG gathered by experienced listeners.

That way, we could make a decent conclusion about PEAQ usefullness in multiple codec testing.

Howabout if we use NCTU-AAC as anchor from AAC test.