Skip to main content
Topic: Do HE extensions hurt quality above X kbps? (Read 1914 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Do HE extensions hurt quality above X kbps?

(I apologize if this has already been discussed before, but I've done my best searching the web and forums for answers and couldn't find one.)

If we know that HEv2 (SBR+PS) is most effective at bitrates <= 48 kbps, and that HEv1 (SBR) is most effective at bitrates <= 96 kbps - is there any downside to using those HE extensions at bitrates higher than their ideal range?

For example, does HE-AAC v2 audio at 192 kbps sound worse than AAC-LC at 192 kbps?

I've looked at a lot of audio blind tests but couldn't find anything that addresses that particular question. Most tests seem more concerned about comparing various codecs, but I haven't found anything that compares various AAC profiles at a range of bitrates.



Do HE extensions hurt quality above X kbps?

Reply #1
As I understand it, there is not much point in using SBR at 80kbps and above due to the (supposed) ability to handle the higher frequency range(s) at those bitrates (although at 80kpbs the lowpass is still set well below 16kHz IIRC).
"Something bothering you, Mister Spock?"

Do HE extensions hurt quality above X kbps?

Reply #2
As I understand it, there is not much point in using SBR at 80kbps and above due to the (supposed) ability to handle the higher frequency range(s) at those bitrates (although at 80kpbs the lowpass is still set well below 16kHz IIRC).
 

  I think we can probably all agree that using SBR or PS above certain bitrates is pointless. But the question I'm asking is: does it in fact hurt quality (compared to using plain LC at the same bitrate)?


Do HE extensions hurt quality above X kbps?

Reply #3
Oh, sorry. I forgot that part.

I guess I should have added: due to SBR being merely a reproduction of HF based on mathematics and such, that it does well for low bitrate situations; whereas it does have shortcomings at higher bitrates since psychoacoustic modeling has a more accurate HF perceptibility calculation compared to the "phony HF" SBR reconstruction.

I didn't find much on listening tests, it seemed logical to go with a straight-up decent psychoacoustic model across the spectrum rather than add on the extra reconstruction phase (which I believe SBR takes more of a hit on battery than LC on some portables).

Other than that, are some threads here on HA about SBR vs. LC but no one I read had access to an encoder that unlocked SBR at higher bitrates so listening tests are utterly lacking so far.
"Something bothering you, Mister Spock?"

Do HE extensions hurt quality above X kbps?

Reply #4
Other than that, are some threads here on HA about SBR vs. LC but no one I read had access to an encoder that unlocked SBR at higher bitrates so listening tests are utterly lacking so far.


Doesn't Nero AAC encoder allow forcing HEv1 and HEv2 modes at any bitrate? When I tried this a while ago I was able to see SBR and PS metadata in MediaInfo, confirming that Nero was actually respecting the forced parameters.


Do HE extensions hurt quality above X kbps?

Reply #5
As I understand it, there is not much point in using SBR at 80kbps and above due to the (supposed) ability to handle the higher frequency range(s) at those bitrates (although at 80kpbs the lowpass is still set well below 16kHz IIRC).
 

  I think we can probably all agree that using SBR or PS above certain bitrates is pointless. But the question I'm asking is: does it in fact hurt quality (compared to using plain LC at the same bitrate)?


AAC-HE lowpasses at 11kHz for 44.1khz audio and then fills in the missing spectrum by mirroring from lower frequencies, so yes, its probably a lot less accurate than LC at higher bitrates.  I would not use it for any bitrate high enough to encode those frequencies directly.

Do HE extensions hurt quality above X kbps?

Reply #6
AAC-HE lowpasses at 11kHz for 44.1khz audio and then fills in the missing spectrum by mirroring from lower frequencies, so yes, its probably a lot less accurate than LC at higher bitrates.  I would not use it for any bitrate high enough to encode those frequencies directly.


That was my assumption too. Here's the hard part: has anyone done ABX testing to prove this conclusively? It'd be interesting to find out at which bitrates the SBR and PS extensions pass the point of diminishing returns and then start hurting instead of helping quality.


Do HE extensions hurt quality above X kbps?

Reply #7
There are various listening tests from many years ago that were done when AAC-HE was standardized.  You can check the wiki, and probably find even more individual ABX tests on the forums.  You could also try it yourself.  Probably samples with a lot of high frequency content will be most obvious.  There was recently someone on the forums asking why AAC-HE sounded bad with chiptunes for instance.

Do HE extensions hurt quality above X kbps?

Reply #8
I've done some personal testing few years ago, and was always able to distinguish SBR encoded file from original even at higher bitrates, because SBR adds some unnatural hiss at hihats, for example. But I haven't forced it at bitrates higher than 128 kbit - I saw there wasn't a point. You are welcome to try, however.

Do HE extensions hurt quality above X kbps?

Reply #9
I think we can probably all agree that using SBR or PS above certain bitrates is pointless. But the question I'm asking is: does it in fact hurt quality (compared to using plain LC at the same bitrate)?

Yes, it does. SBR and especially Parametric Stereo of HE-AAC v2 are parametric coding tools which cannot preserve the exact waveforms of the input signals (except maybe for very trivial/special content).

In principle, the SBR start frequency could be set higher than 11 kHz by increasing the codec sample rate, but as others said,
at high bit-rates AAC (LC) works well enough without SBR.

Chris
If I don't reply to your reply, it means I agree with you.

 
SimplePortal 1.0.0 RC1 © 2008-2019