HE-AAC Nero!

[kotuwa]

A question regarding HE-AAC with Nero (neroAacEnc).
Asking from people who can test it... And also from ppl who already have tested, or seen tests :]

Usually it is recommended to use HE for low bitrates and LC for high bitrates.
What i'm asking is, will HE give better quality compared to LC in high bitrates?

What are Theoretical/technical and Practically Tested answers?

1. -q 0.40 (~132 kbps 2.0)
Will -q 0.40 -he give better quality than -q 0.40 -lc ?

2. -q 0.70 (~240 kbps 2.0)
Will -q 0.70 -he give better quality than -q 0.70 -lc ?

!?

[Sharc]

http://forum.doom9.org/showthread.php?t=157368

I wouldn't suggest to use HE for music. For low bitrate voice HE is good, or when small file size is your primary goal. Otherwise use LC. But try yourself what is good for your ears.

[tebasuna51]

HE-AAC is not recommended over 96 Kb/s (for stereo).

Let Nero defaults when use -q parameter, don't force -he encode.

Some test:
Dark blue -> HE
Green -> LC

[kotuwa]

Thanks.
In above graph, if they used nero, they could have go at least for 96kbps limit in HE. Then it would be more clear.
Apple AAC limit HE at 80kbps, thought nero with higher bitrate will do better... at least at -q 0.35.

Regarding above test in graph, it a vocal / dialogue based test?
That USAC looks like the best! All rounder for low and high bitrates! :)

Also I remember a comment which said something like this... ~"LC part of a 80kbps HE sounded better than 112kbps LC's" !!!
Can not remember it from where.... How about it's validity !?

[tebasuna51]

The test was made for many kind of sources, maybe for a specific source that can be true if the ears/audio equipament of that guy can't listen/play high frequencies.

If we have a source 44100 Hz than the HE 80 Kb/s method uses 70 Kb/s to encode the frequencies until 22050 and 10 Kb/s to encode the SBR part (22050 Hz to 44100), and LC 112 Kb/s need 60 Kb/s to encode the high frequencies, then the low frequencies part is better in HE.

But if ears/equipament can't listen/play high frequencies the best option is downsample 44100 to 22050 and encode like LC.

[kotuwa]

tebasuna51, I can get the rough idea what you said above...

But I don't get the point of down-sampling thing!
And are those frequency values you used correct?
Aren't the sampling rate and audio frequency values are 2 different things?

I mean, in 44.1 kHz sampling rate, isn't the max frequency it supports is 20 kHz? The the upper limit of human ears! (in 48, it is 21.8)
I thought it like that... What I thought is the range is divided somewhere in the middle..
(like 10 kHz, probably not) and then the lower part being encoded with 44.1 kHz sampling rate and the higher part is encoded with its half, 22.05 kHz
!?
:confused:

[tebasuna51]

Aren't the sampling rate and audio frequency values are 2 different things?

Yep, to recover a analog audio frequency of F you need a digital samplerate of 2F. See https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem

Then with a samplerate of 44100 Hz you can recover a analog audio frequency of 22500 Hz, enough for normal human ears limit.

Some damaged or old ears (like mine) can't listen over 11000 Hz then a samplerate of 22500 Hz can be enough.
But there are golden ears than affirm listen differences between samplerates of 48000 Hz and 96000 Hz, congratulations for them.