[FFmpeg-devel] MLP decoder problems

[Ian Caulfield]

Would it be possible to get a sample of the AOB? The first 100M or so
of the AOB containing those first two tracks should do...

Ian

On Nov 8, 2007 8:45 AM, Olev <hannula at gmail.com> wrote:
> I tried ffmpeg build ffmpeg_eac3_mlp_rev10905 from
> http://kurtnoise.free.fr/index.php?dir=misc/ and tried to decode the
> infamous bootleg DVD-A release of Pink Floyd's - Dark Side of the
> Moon. I used DVDAExplorer_a7 to get MLP files from the .AOB container
> (non encrypted, this DVD-A has no content protection). When decoding
> the .mlp files I get strange errors:
> [mlp @ 0092B3E0]More matrices specified than channels
> Error while decoding stream #0.0
> [mlp @ 0092B3E0]Lossless check failed - expected 15, calculated 0
> [mlp @ 0092B3E0]More matrices specified than channels
> Error while decoding stream #0.0
> [mlp @ 0092B3E0]Lossless check failed - expected 98, calculated 0
> Error while decoding stream #0.0
> ...
>
> Next I joined all the MLP files to get a general view of things, seems
> to work, can anyone comment if MLP is not join files safe? I hear no
> clicks or pops or silence between the track after just joining them.
> Ffmpeg decoded the one large MLP file without no new errors, only the
> Lossless check failed that it shows on single files too. I used
> -acodec pcm_s24le since the MLP is 4.1 (left, right, lfe, surround
> left, surround right) channel 24bit 96kHz. I decoded the same large
> .MLP file with Surcode MLP (v1.0.29) "replace file before verify"
> method. What suprises me is that a lossless codec is decoded
> differently, heres an avisynth SoundOut() analysis of the two
> decoders:
> FFmpeg:
> [Ch 0] Maximum:-0.57dB. Average:-22.41dB. RMS:-18.91dB.
> [Ch 1] Maximum:-0.28dB. Average:-22.35dB. RMS:-18.82dB.
> [Ch 2] Maximum:-11.44dB. Average:-35.10dB. RMS:-30.92dB.
> [Ch 3] Maximum:-0.86dB. Average:-25.83dB. RMS:-22.10dB.
> [Ch 4] Maximum:-2.10dB. Average:-25.58dB. RMS:-21.78dB.
> [All channels] Maximum:-0.28dB. Average:-25.24dB. RMS:-21.01dB
>
> Surcode MLP:
> [Ch 0] Maximum:-0.57dB. Average:-22.36dB. RMS:-18.86dB.
> [Ch 1] Maximum:-0.28dB. Average:-22.29dB. RMS:-18.76dB.
> [Ch 2] Maximum:-1.#JdB. Average:-1.#JdB. RMS:-100.00dB.
> [Ch 3] Maximum:-11.44dB. Average:-35.04dB. RMS:-30.86dB.
> [Ch 4] Maximum:-0.86dB. Average:-25.75dB. RMS:-22.00dB.
> [Ch 5] Maximum:-1.64dB. Average:-25.48dB. RMS:-21.67dB.
> [All channels] Maximum:-0.28dB. Average:-26.76dB. RMS:-21.73dB
>
> Surcode MLP has one extra channel because it was easier to join the 3
> files it produces and it produces a channel of silence so that all
> it's extracted files are stereo. And the result is different length
> too, FFmpeg is 42:55, Surcode MLP is 43:10.
>
> Who does not conform to standards, extracting from a lossless source
> should be bit identical. I tried to extract MLP directly from the
> non-encrypted .AOB files but that failed - but this is a different
> issue, MLP support without reading it out of AOB is only half the
> things needed to decode DVD-A, FFmpeg reads something out of .AOB but
> it does not see it as being MLP and produces noise filled waveform.
>
> But I sure think this is the best support for MLP it has ever had, GREAT WORK!
>
> Regards,
> Olev
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> http://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel
>