[Mp4-tech] public AAC at 128kbps listening test open

bond b-o-n-d gmx.net
Sat Feb 21 11:36:41 EST 2004


forwarding the reply of Robert Amorim:
> ----- Original Message ----- 
> From: ejenn2000-mp4 yahoo.com 
> To: mp4-tech lists.mpegif.org 
> Sent: Friday, February 20, 2024 9:28 AM
> Subject: Re: [Mp4-tech] public AAC at 128kbps listening test open
> 
> 
> Glad to see this is happening. However, I am sorry (and hate) to say
> that the result may not have any value because it is comparing apples
> with oranges. The average bitrate for Ahead/Nero AACenc (a VBR codec)
> to encode these test samples is as high as 141 kb/s!!!!! However, CBR
> codecs like iTunes and Real are only at 128 kb/s for all these test
> samples. Any experienced AAC developer will tell you that at this high
> bitrate 7-8 kb more will significantly improve the resulting sound
> quality. It would really be a surprise if iTunes or Real can beat Nero
> with such a lower bitrate. This is just not a test at 128 kb/s as the
> title called. It is comparing samples encoded at 141 kb/s and at 128
> kb/s. I think the winner is obvious without going thru the test...

There are several problems the listening test conducer needs to cope
with:
- Usefulness: It won't help if a different encoder setting is used
for each sample (I.E, to hand-tweak until the 128kbps mark is
reached). Because that doesn't reflect the dad-to-day usage of codecs - 
users don't re-encode their album tracks with several different quality 
settings to reach exactly 128kbps at each one of them, they choose a
setting and encode the entire collection with it, in the expectative
that it will reach a desired average bitrate in the end.
- Few VBR profiles: Ahead/Nero has the -radio profile, which average 
bitrate for the test samples would stay under 100kbps, and the -internet
profile, that was the one used and ended up at 141kbps. My choices,
considering those results, were either go with CBR encoding at 128kbps
 - again, a setting that wouldn't reflect day-to-day usage for the
target audience of my tests that prefers VBR -, to give up testing
Nero - which would also make the test audience/participants unhappy - 
and go with a bitrate slightly higher than the others. That was my 
choice. I might repent of it later, but it seemed the most sensible at 
the time I was deciding about codecs and settings.
Last but not least, it's worth mentioning that on large sample suites, 
Nero does indeed come close to 128kbps. References:
http://www.hydrogenaudio.org/index.php?showtopic=18474&view=findpost&p=185868
http://www.hydrogenaudio.org/index.php?showtopic=18474&view=findpost&p=185992
Regards;
Roberto.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20040221/5077d655/attachment.html


More information about the Mp4-tech mailing list