I was encoding some track to do ABX tests with the latest Nero encoder (the August 6th, 2007 one) when I came across a track (the whole album really, but especially this one track) that causes the bit rate to go waaaay up.
Encoding this track at -q 0.54 (the setting I use for my iPod) yields an average bit rate of 228, while -q 0.5 yields an average bit rate of 227. Despite both having the same bitrate, the -q0.5 encode is supposedly of a lower quality, and the spectrogram seems to show this.
I understand that VBR is supposed to use more bits when necessary, but obviously there is something inefficient happening at -q 0.5. I was thinking about switching to 0.5 for my iPod, but I might as well stick with -q 0.54 considering they will be the same size.
The track is quite treble-heavy and rife with clipping, perhaps that could be what's tripping up the encoder?
I've also upload a 30-second clip of the song (The Rainbow, by The Apples in Stereo) for anyone who's interested.
https://download.yousendit.com/EB9C33F77A18C1FC
Full Version: Bitrate anomaly in Nero AAC -q 0.5
Whats the problem here?
One track not representative. You need to check a dozen or so .
I noticed the same.
Fear Factory - Digimortal (15 tracks)
q 0.31 (85 kbps) gets 109 kbit 43.7 Mb
q 0.36 (96 kbps) gets 107 kbit 42.3 Mb
Fear Factory - Digimortal (15 tracks)
q 0.31 (85 kbps) gets 109 kbit 43.7 Mb
q 0.36 (96 kbps) gets 107 kbit 42.3 Mb
QUOTE(shadowking @ Jun 14 2008, 21:14) 
One track not representative. You need to check a dozen or so .
Well it occurs throughout the entire album, that was just the one song I was testing.
In the latest version we use a special mapping to get the bitrates on switches of tuning tables as close together as possible, we use a set of files to do this. On a different set of files, results may differ and higher bitrates might be reached even with lower q value. In older versions of the encoder the steps in bitrate on a switch of tables would be huge, which is also not good. Now the bitrate more or less increases very gradually over the q range, with the disadvantage that you might sometimes get the results described here.
Different tuning tables mean different cutoff frequencies, which is what you notice in your "spectrogram quality assessment" (which is a TOS violation BTW). Do some listening.
Different tuning tables mean different cutoff frequencies, which is what you notice in your "spectrogram quality assessment" (which is a TOS violation BTW). Do some listening.
QUOTE(menno @ Jun 16 2008, 10:10)
In the latest version we use a special mapping to get the bitrates on switches of tuning tables as close together as possible, we use a set of files to do this. On a different set of files, results may differ and higher bitrates might be reached even with lower q value. In older versions of the encoder the steps in bitrate on a switch of tables would be huge, which is also not good. Now the bitrate more or less increases very gradually over the q range, with the disadvantage that you might sometimes get the results described here.
Different tuning tables mean different cutoff frequencies, which is what you notice in your "spectrogram quality assessment" (which is a TOS violation BTW). Do some listening.
I see.
About the spectrogram, I know they are not a foolproof method of assessing subjective audio quality, hence the phrasing "seems to show". However, I don't think they should be excluded entirely. That said, I was unable to confirm any audible difference in an ABX test.
Keep up the good work on the encoder... looking forward to the next version.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.