If SBR works well at low bitrates, shouldn't it also be good for bitrates such as 128 or even higher?
For stereo 48 ABR 128, would I be better off using HE or LC?
These are mainly audio of documentaries which naturally are mostly narration with the odd background music and effects.
The point of HE-AAC is that you can go down to lower bitrates while still keeping audible fidelity intact. Since you aim at high bitrates, and decoder support for LC-AAC is much better, I'd just use LC-AAC. I expect both HE and LC to be transparent at that bitrate. As always, you should/could decide for yourself by performing a double blind ABX test (http://www.hydrogenaudio.org/forums/index.php?showtopic=16295).
I expect both HE and LC to be transparent at that bitrate.
Transparent SBR? Hmm..
The point of HE-AAC is that you can go down to lower bitrates while still keeping audible fidelity intact.
Does it keep?
I expect both HE and LC to be transparent at that bitrate. [/url].
...
Interesting point the OP is pointing out. Let me put the question in other words: could at around 128 kbps AAC-HE be more transparent to more people than AAC-LC? If not, what would limit it technically speaking?
The point of HE-AAC is that you can go down to lower bitrates while still keeping audible fidelity intact.
Does it keep?
I expect both HE and LC to be transparent at that bitrate. [/url].
...
Maybe doing an ABX test at work is suboptimal, but I wasn't able to confidently ABX HE-AAC at ~92 kbps vbr (qaac -v 128 --he) (9/13 trials, 13.3%). "Transparent" was likely overzealous, but it doesn't obviously fall apart, like you seem to imply. For mobile use it sure sounds fine enough. I'd expect I could be happy with even lower bitrate while biking or riding the bus.
with regard to the original question, I am under the impression that AAC-HE can only be used up to 80 kbps (variable would obviously give approximately 80kbps at the highest setting, but constant would be only 80kbps at the highest). I checked with qaac.exe (2.18) with the --formats option. only goes up to 80 kbps for AAC-HE with 48/stereo (and 44.1/stereo). also, encoding with "qaac.exe -v160 --he track04.cdda.wav" said on the console window that it was encoding to CVBR 80.
maybe someone more familiar with this can verify?
if I'm right (which is unusual in this forum), the mere fact that it is impossible to truly achieve approximately 128 kbps using AAC-HE should mean that AAC-LC would be the choice to go with, if the OP wants 128kbps in the first place.
with regard to the original question, I am under the impression that AAC-HE can only be used up to 80 kbps (variable would obviously give approximately 80kbps at the highest setting, but constant would be only 80kbps at the highest). I checked with qaac.exe (2.18) with the --formats option. only goes up to 80 kbps for AAC-HE with 48/stereo (and 44.1/stereo). also, encoding with "qaac.exe -v160 --he track04.cdda.wav" said on the console window that it was encoding to CVBR 80.
I estimated my ~92kbps by watching and mentally averaging the reported bitrate by foobar2000 during playback, the file metadata says "CVBR 80kbps" indeed. Since I used foobar2000 for conversion, too, I never saw the console output.
if I'm right (which is unusual in this forum), the mere fact that it is impossible to truly achieve approximately 128 kbps using AAC-HE should mean that AAC-LC would be the choice to go with, if the OP wants 128kbps in the first place.
My impression is that this bitrate was arbitrarily chosen. AAC-LC is the choice to go with because of software and hardware support.
If not, what would limit it technically speaking?
Uh, the fact that it's not coding half the bandwidth at all? It just guesses at higher frequencies using the lower-bandwidth content and a little sideband info (1-3kbps sideband info IIRC). It may be better than nothing, but you can't possibly rely on this kind of guessing to give you transparency.
At low bitrates, the low frequencies are sufficiently starved for bits that it makes sense to rob all the bits from the less-audible higher frequencies. At >64kbps, not so much.
Not having read the standard, I don't know whether it's possible to make an encoder that does SBR at >80kbps. But encoder authors have no reason to do so; listening tests seem to indicate that already at 80kbps LC may be better. (this test (http://www.hydrogenaudio.org/forums/index.php?showtopic=74781&hl=) was the first one I found in a search, but I'm pretty sure I've seen similar results elsewhere.) Definitely no point to using SBR for 96kbps and up.
This thread is a good example why you shouldn't post into threads if you have no clue about the subject matter.
*silently sneaks away*
OK, let's see...
For stereo 48 ABR 128, would I be better off using HE or LC?
These are mainly audio of documentaries which naturally are mostly narration with the odd background music and effects.
I suggest you use AAC LC. HE-AAC will give you no advantage at this bitrate. Speaking of compatibility, it would be nice if Kohlrabi could point out some hard-/software which still has problems with HE-AAC decoding, so we can contact the developers (btw, have you always been Super Moderator, Kohlrabi?)
with regard to the original question, I am under the impression that AAC-HE can only be used up to 80 kbps...
Not having read the standard, I don't know whether it's possible to make an encoder that does SBR at >80kbps...
There is absolutely no such restriction in the AAC
standard. In fact, IIRC Winamp 5.63 (and higher) can do up to 128-kbps HE-AAC in CBR mode. Maybe someone can try it out and report. This brings me to the last point:
"If not, what would limit it technically speaking?" Uh, the fact that it's not coding half the bandwidth at all?
Again, this is not a limitation of the standard, it's an encoder design decision. I thought I already made this clear in another thread once, but I'm happy to repeat: there's a mode called "downsampled SBR", in which you can move the SBR start frequency above half the input signal bandwidth. For example, with 44.1- or 48-kHz audio input, you could let SBR code only the frequencies above 16 kHz or so. Such a setting will be transparent for many people (assuming the core bit-rate is high enough). Fraunhofer's encoder supports downsampled SBR, but that mode might not (yet) be available in Winamp, I don't remember.
Chris
SBR is great for speech. You can cram much more audiobook material that way.