View Single Post
Old 16th May 2010, 02:47   #9995  |  Link
Abradoks
Registered User
 
Join Date: Mar 2008
Posts: 71
Quote:
Originally Posted by jruggle View Post
I haven't ever seen anything like this reported before. Do you have a sample? And I'm assuming you mean using eac3to with Nero rather than libavcodec for AC-3 decoding...

If you do have a sample, it would be great to have the original AC-3 snippet, along with the eac3to wav or flac output.
Hi, jruggle. It's nice to see you here. I thought I should better investigate the problem before reporting it to ffmpeg, but was too lazy recently.
Here are some samples:
1) 5amctest.rar (mirror) : test.wav — mono source; test.ac3 — 2.0 ac3 produced from source (both channels are same) by Dolby encoder. Man who originally reported the problem has made this graph, where you can see attenuation in the high frequencies when using eac3to with libav. ffmpeg (SVN-r23056) has the same problem but attenuation starts on higher frequency. The funny thing is that a52dec hasn't such issue.
2) qs.7z (mirror) (14.5 MB) — two another samples and source file. 5.1, Dolby encoder. For 640.ac3 eac3to (with libav) shows 27.809 dB SNR, while ffmpeg 11.547 dB. liba52 (through foobar2000 plugin) gives 27.537 dB. When comparing average sound level graphs, it looks like eac3to and ffmpeg start to attenuate at ~16.8 kHZ. And ffmpeg is always ~2 dB quieter than eac3to. For 224.ac3 attenuation starts near 4.5 kHZ but only for eac3to. ffmpeg is just always ~2 dB quieter than source.
Also in all 3 samples ffmpeg has average sound level about -140 dB after frequency cutoff. Is it ok?
Graphs:

Last edited by Abradoks; 7th July 2010 at 11:48. Reason: links
Abradoks is offline