Skip to main content
Topic: AAC-LC tuned for 44.1/48 kHz? (Read 5500 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

AAC-LC tuned for 44.1/48 kHz?

I've read that some MP3 encoders (e.g.: LAME) are more or less optimized for 44.1 kHz data. I'm a little less familiar with AAC format.

I'm looking to encode a series of files in AAC-LC VBR (by way of qaac) and the input files are in 16/48. Would there be any advantage to downsampling this (I typically use SoX) to 44.1 kHz? Or does AAC-LC handle 48 kHz just as well as 44.1 kHz at the same bitrate?

AAC-LC tuned for 44.1/48 kHz?

Reply #1
The differences between 44.1 and 48kHz are pretty small from the encoders' perspective. On top of that, non-CD audio, which is getting more common, is more likely 48kHz than 44.1kHz. So it's safe to say AAC encoders receive significant testing at 48kHz. I know for sure we did so at Nero and I see no reason to believe Apple/FhG were different.

So, don't resample.

AAC-LC tuned for 44.1/48 kHz?

Reply #2
The differences between 44.1 and 48kHz are pretty small from the encoders' perspective. On top of that, non-CD audio, which is getting more common, is more likely 48kHz than 44.1kHz. So it's safe to say AAC encoders receive significant testing at 48kHz. I know for sure we did so at Nero and I see no reason to believe Apple/FhG were different.

So, don't resample.


I suppose I should mention that I'll be using lower bitrates (~96 kbps) if that factors into the decision at all.

AAC-LC tuned for 44.1/48 kHz?

Reply #3

So, don't resample.


I suppose I should mention that I'll be using lower bitrates (~96 kbps) if that factors into the decision at all.


I still don't think it necessary to resample 48kHz down to 44.1, and neither do the BBC who stream the majority of their radio output in 48kHz at a range of bitrates; nearly always 48 & 96kbps HE-AAC, and 128 & 320 kbps AAC-LC. AFAIK only their podcasts are resampled to 44.1kHz sampling rate and using MP3 format, presumably for wider compatibility with older devices.

AAC-LC tuned for 44.1/48 kHz?

Reply #4
Some AAC encoders can increase bitrate (somewhat) with higher sampling rates.
As of audio quality it souldn't be an issue. Try to encode at 44.1 and 48 kHz and listen it for yourself. Anyway there is nothing to lose as SoX (very high quality VHQ - best mode) resamples transparently.

AAC-LC tuned for 44.1/48 kHz?

Reply #5
Anyway there is nothing to lose as SoX (very high quality VHQ - best mode) resamples transparently.


Whatever is handling the playback might resample back from 44.1 to 48kHz though, and not necessarily have the same quality as SoX.

AAC-LC tuned for 44.1/48 kHz?

Reply #6
Yes, Windows OS has 48 kHz as default sampling rate while Android (at least some versions) has native sampling rate of 44.1 kHz if I'm not wrong.
It's user who should take care of it and set same sampling rate to avoid unnecessary resampling during playback.

Plus most of economic DACs performs better at 48 kHz.

So, more or less it's the same to use 44.1 or 48.

 

AAC-LC tuned for 44.1/48 kHz?

Reply #7
Several things to note.

(1)

while Android (at least some versions) has native sampling rate of 44.1 kHz if I'm not wrong.


Many recent android devices employ 48k rather than 44.1k as a default sampling rate.
http://audiobuffersize.appspot.com/

I think several vendors recognized that
[quote author=IgorC link=msg=0 date=]Plus most of economic DACs performs better at 48 kHz.[/quote]
I don't know whether this is indeed true or not.

Interestingly, my phone (Samsung SHV-E470S) uses 48k when using wired headphones but 44.1k when using my bluetooth headphone (LG HBS-750) though this receiver supports 48k/16bit as well.
I don't know why Samsung uses such a strange setting. Maybe still Android uses only a single sample rate (cannot changed by users) for A2DP, and Samsung wants to keep compatiblity for old receivers not supporting 48k.


(2) Android 5.0+ clearly has a better resampling algorithm than Kitkat. After upgrading to Android 5.0.1, I never hear any artifacts using a 44.1k sine sweep wave sample, which showed significant aliasing when using 4.4.2.

(3) Previously I had a brief test to compare quality of resamplers. Even though all of testers (4 people) distinguished poorly resampled sine sweep (low quality resampling using Audacity), they couldn't when using typical music samples.
I also failed ABX between SoX resampled and poorly resampled. Of course, I cannot distinguish between the original and SoX-resampled files (twice; 44.1 -> 48 -> 44.1).


 
SimplePortal 1.0.0 RC1 © 2008-2019