HD-DVD TrueHD demux with eac3to: gaps?

Sakura-chan · 19th January 2017, 05:09

Hi,

I have an old HDDVD rip from when the drives for those things were still around (so very old).

The main feature comes split between two EVOs. I want to convert this ancient format to MKV. So I try to extract the tracks with eac3to. The issue is in the split TrueHD track. When outputting to .thd the log talks about a gap that occurs in the middle of the two EVOs (the reported timestamp is the exact duration of the FEATURE_1.EVO)

Code:

eac3to v3.31
command line: eac3to  D:\[...]\HVDVD_TS\FEATURE_1.EVO+D:\[...]\HVDVD_TS\FEATURE_2.EVO 2: D:\newmkv.vc1 3: D:\newmkv.thd 7: D:\newmkv.sup
------------------------------------------------------------------------------
EVO, 1 video track, 4 audio tracks, 2 subtitle tracks, 2:03:31
1: Joined EVO file
2: VC-1, 1080p24 /1.001 (16:9) with pulldown flags
3: TrueHD, 5.1 channels, 48kHz, dialnorm: -27dB
4: E-AC3, 5.1 channels, 384kbps, 48kHz, dialnorm: -27dB
5: AC3, 5.1 channels, 384kbps, 48kHz, 11011ms
6: AC3, 2.0 channels, 224kbps, 48kHz
7: Subtitle (DVD)
8: Subtitle (DVD)
[v02] Extracting video track number 2...
[a03] Extracting audio track number 3...
[v02] Writing new framerate "24fps /1.001" to bitstream.
[a03] Removing TrueHD dialog normalization...
[v02] Removing VC-1 pulldown...
[a03] Creating file "D:\newmkv.thd"...
[v02] Creating file "D:\newmkv.vc1"...
[s07] Extracting subtitle track number 7...
[s07] Creating file "D:\newmkv.sup"...
[a03] Audio has a gap of 96ms at playtime 1:06:14.  <WARNING>
[a03] The audio gaps/overlaps can't be removed from the TrueHD bitstream.  <WARNING>
[a03] In order to remove them you'll have to transcode to another format.  <WARNING>
Video track 2 contains 177704 frames.
Subtitle track 7 contains 1069 captions.
eac3to processing took 8 minutes, 36 seconds.
Done.

It says I need to transcode... do I have to transcode or should I leave that gap as it is?

Transcoding to FLAC the log is like this:

Code:

eac3to v3.31
command line: eac3to  D:\[...]\HVDVD_TS\FEATURE_1.EVO+D:\[...]\HVDVD_TS\FEATURE_2.EVO 3: D:\transcoded.flac
------------------------------------------------------------------------------
EVO, 1 video track, 4 audio tracks, 2 subtitle tracks, 2:03:31
1: Joined EVO file
2: VC-1, 1080p24 /1.001 (16:9) with pulldown flags
3: TrueHD, 5.1 channels, 48kHz, dialnorm: -27dB
4: E-AC3, 5.1 channels, 384kbps, 48kHz, dialnorm: -27dB
5: AC3, 5.1 channels, 384kbps, 48kHz, 11011ms
6: AC3, 2.0 channels, 224kbps, 48kHz
7: Subtitle (DVD)
8: Subtitle (DVD)
[a03] Extracting audio track number 3...
[a03] Removing TrueHD dialog normalization...
[a03] Decoding with libav/ffmpeg...
[a03] Encoding FLAC with libFlac...
[a03] Creating file "D:\transcoded.flac"...
-------------------------------------------[truehd @ 02552a60] Lossless check failed - expected 00, calculated 92.
------------------------------------[truehd @ 02552a60] End of stream indicated.
[a03] Original audio track, L+R+C+LFE: constant bit depth of 16 bits.
[a03] Original audio track, SL+SR: constant bit depth of 24 bits.
[a03] Audio has a gap of 96ms at playtime 1:06:14.  <WARNING>
[a03] Starting 2nd pass...
[a03] Decoding FLAC...
[a03] Realizing RAW/PCM gaps...
[a03] Encoding FLAC with libFlac...
[a03] Creating file "D:\transcoded.flac"...
[a03] Processed audio track, L+R+C+LFE: constant bit depth of 16 bits.
[a03] Processed audio track, SL+SR: constant bit depth of 24 bits.
Video track 2 contains 177704 frames.
eac3to processing took 14 minutes, 13 seconds.
Done.

(the two lines starting with "------------------------" didn't appear in the log, I copied them from the cmd window)

From what I could find the "Lossless check failed" happens because the original is split between two files.

I'm currently muxing the video with the original THD and again with the FLAC to see if the both audios keeps sync with the video.

But in case it's the same or it isn't noticeable, which should I keep? The demuxed THD or the converted FLAC? Which would be the most "proper" audio to archive? I'd like to keep the THD but as it's lossless-lossless I really don't care. Also the FLAC takes less space :P

Thanks

P.S.: Why in the same track are channels with different bit depth? Is that correct?

tebasuna51 · 19th January 2017, 10:42

- Seems than the gap (without sound for 96 ms of video) occurs at end of first part.

eac3to can create extra audio frames, to compensate the gap, for many formats but not for TrueHD.
Then all the second part have the audio advanced 96 ms over the video with the extracted TrueHD track.
Maybe is unnoticeable for you but exist.

The recoded flac don't have this problem, then, if multichannel flac is not a problem for your player, is the "proper" audio to archive.

Maybe there are a workaround to preseve the TrueHD track, if the player don't support multichannel flac but yes TrueHD:

1) Convert each part to mkv (with same video parameters and encoder version)
2) Append the two mkv with MkvToolNix, I hope (not tested) than the gap is reflected in timestamps and can be played without async.

- The bitdepth of 24 in SL+SR channels when L+R+C+LFE have only 16 maybe can be explained with:

The original track have only L+R+C+LFE channels at 16 bits, and the SL+SR channels was created by soft from L+R channels with some effects (delay, etc.).

For me is safe to create the flac with the -down16 parameter.
Only the artificial SL+SR channels was downsampled.

Music Fan · 22nd January 2017, 10:16

You can also try first to remux EVOs in 1 TS, then extract (or convert) audio off this TS.

Sakura-chan · 23rd January 2017, 18:01

Thanks a bunch

Preserving the THD isn't an issue at all, as long as conversions are lossless->lossless (preserving bit depth and freq, i.e. no resampling) I couldn't care less. It's just that I don't bother converting THD or DTS-MA to FLAC when remuxing BDs. Even though I've seen FLAC being more efficient than THD or DTS-MA. In fact, in this case the FLAC ended up being smaller (with no downsampling of the 24bit chs) than the original THD. One wonders why DTS and THD exist when FLAC is free and better :P

Anyway, after trying both THD and FLAC audios, the THD clearly plays wrong. When the time of the gap comes, the video freezes for several frames, the sound disappears, and then the freezed frames play at something like 20x speed. After that the video resumes playing normally but the sound never comes back. This is when muxing the extracted "wrong" THD with the video in MKV, then playing with PotPlayer.

On the other hand when having eac3to convert the THD to FLAC and then muxing with the video, everything is perfect.

Everything is crystal clear then, in these cases of split source files better let eac3to convert the audio to FLAC. Amazing tool really.

Btw, I left everything at 24 bits, just to be 101% sure. I'm not really concerned if the padding in the 16 bit channels make the audio bigger (still, it's smaller than the THD). Since it's a remux the MKV is already big and a bit more doesn't hurt. Still, thanks for the advice. I guess downsampling to 16 bit is the most reasonable option. If I get around to tinker with the HDDVD again (I still have it in my HDD) I'll see what the file size difference is between downsampling or not.

19th January 2017, 10:42	#2 \| Link
tebasuna51 Moderator Join Date: Feb 2005 Location: Spain Posts: 6,915	- Seems than the gap (without sound for 96 ms of video) occurs at end of first part. eac3to can create extra audio frames, to compensate the gap, for many formats but not for TrueHD. Then all the second part have the audio advanced 96 ms over the video with the extracted TrueHD track. Maybe is unnoticeable for you but exist. The recoded flac don't have this problem, then, if multichannel flac is not a problem for your player, is the "proper" audio to archive. Maybe there are a workaround to preseve the TrueHD track, if the player don't support multichannel flac but yes TrueHD: 1) Convert each part to mkv (with same video parameters and encoder version) 2) Append the two mkv with MkvToolNix, I hope (not tested) than the gap is reflected in timestamps and can be played without async. - The bitdepth of 24 in SL+SR channels when L+R+C+LFE have only 16 maybe can be explained with: The original track have only L+R+C+LFE channels at 16 bits, and the SL+SR channels was created by soft from L+R channels with some effects (delay, etc.). For me is safe to create the flac with the -down16 parameter. Only the artificial SL+SR channels was downsampled. __________________ BeHappy, AviSynth audio transcoder.

22nd January 2017, 10:16	#3 \| Link
Music Fan Registered User Join Date: May 2009 Location: Belgium Posts: 1,744	You can also try first to remux EVOs in 1 TS, then extract (or convert) audio off this TS.

23rd January 2017, 18:01	#4 \| Link
Sakura-chan Registered User Join Date: Sep 2010 Posts: 34	Thanks a bunch Preserving the THD isn't an issue at all, as long as conversions are lossless->lossless (preserving bit depth and freq, i.e. no resampling) I couldn't care less. It's just that I don't bother converting THD or DTS-MA to FLAC when remuxing BDs. Even though I've seen FLAC being more efficient than THD or DTS-MA. In fact, in this case the FLAC ended up being smaller (with no downsampling of the 24bit chs) than the original THD. One wonders why DTS and THD exist when FLAC is free and better :P Anyway, after trying both THD and FLAC audios, the THD clearly plays wrong. When the time of the gap comes, the video freezes for several frames, the sound disappears, and then the freezed frames play at something like 20x speed. After that the video resumes playing normally but the sound never comes back. This is when muxing the extracted "wrong" THD with the video in MKV, then playing with PotPlayer. On the other hand when having eac3to convert the THD to FLAC and then muxing with the video, everything is perfect. Everything is crystal clear then, in these cases of split source files better let eac3to convert the audio to FLAC. Amazing tool really. Btw, I left everything at 24 bits, just to be 101% sure. I'm not really concerned if the padding in the 16 bit channels make the audio bigger (still, it's smaller than the THD). Since it's a remux the MKV is already big and a bit more doesn't hurt. Still, thanks for the advice. I guess downsampling to 16 bit is the most reasonable option. If I get around to tinker with the HDDVD again (I still have it in my HDD) I'll see what the file size difference is between downsampling or not.