5.1 Audio goes out of sync over time, after DCP is made

[ryanwellence]

Deliverable specs:
4K Scope
6 channel 5.1 Surround Sound
24fps
SMPTE

The DCM I received were:
DNxHR HQX 10bit 4K Scope 24fps (~19 minutes)
Linear PCM 48000Hz C mono
Linear PCM 48000Hz L mono
Linear PCM 48000Hz LFE mono
Linear PCM 48000Hz Ls mono
Linear PCM 48000Hz R mono
Linear PCM 48000Hz Rs mono

Additional notes:
The audio files have a 2-pop reference tone in the beginning and end, with a trim from the beginning of 2 seconds and 2 frames, and a trim from the end of 2 seconds and 22 frames. After trimming, the audio and video files have the same duration.

The [video range] settings are set to [video].

I am running DCP-o-matic v2.16.35 git 99cb0937e5 on my Microsoft Windows 10 Home v10.0.19045 Build 19045

THE PROBLEM:
Sound begins just before video, then lines up about 1 minute into the film, then gradually gets behind video. My [video framerate] settings are set to [24] like the video DCM, I've picked the correct [audio channel] settings of [6 5.1], and all audio files are routed to the correct channels.

We've tested the DCM files back in both Premiere and Resolve and both show that the files are in sync. We've testing the playback multiple times in theater. The theater has had no issues with sync over the last 3 days of 8 hour screenings.

If anyone has any idea why this is happening, please let me know as soon as possible.
-Ryan

P.S.
I have asked for the stereo mix as a backup (using the recommended 6 5.1 settings, only routing to the L and R channels), and made a DCP with that and will test later tonight as a backup.

[carl]

Is the DCP available anywhere to download?

Could you attach the log, metadata.xml and ffprobe.log files from inside your DCP-o-matic project? (or email them to carl@dcpomatic.com)

[ryanwellence]

Ah, shoot. Sorry, but I deleted the old DCP for space reasons. And, I don't have it up for download right now. I could render it again now (it'll take about 3:30 hours to complete) and attach the log file.

If it's any thought though, I can send the stereo DCP's log? It's virtually identical in DCP file settings, except the only DCM audio files are just 2 mono L and R files.

[carl]

The log would be interesting. Also, if you can make the original files available somewhere I can give it a try. I can't immediately see anything you are doing wrong, so I would suspect either a bug in DCP-o-matic or some strange property of your source file that DCP-o-matic is misinterpreting.

[ryanwellence]

I'd like to, but I don't think it's mine to give out right now. But, here is also the log file: https://drive.google.com/file/d/1LR0Z4H ... sp=sharing

for some reason I can't upload it to the forum's, invalid file type, or something.

I'll also go ahead and render the 5.1 mix again.

[carl]

Thanks. There's a lot of errors in there coming from the library that decodes the source video. They could easily cause your sync problems.

I think it would be worth trying an encode with the test version of DCP-o-matic. This includes an updated version of this decoding library.

If you can't share the whole file, a short clip might be enough to trigger the problem so I can take a look. I understand if that's not possible, though. Also, you can share through carl@dcpomatic.com rather than on the forum if you prefer.

[Carsten]

Oh wow yes, loads of codec related errors. Do you know which application created this HQX file?

[ryanwellence]

Do you know which application created this HQX file?

Yes, Premiere Pro, but I'm not sure which version number. I'll have to ask the editor.

Also, as an update for anyone else reading this problem:

On the test version 2.17.4 the render did fix the audio issue, but now there is a graphical error where a black frame appears every 2-4 minutes in the video. So, not release ready for these variables.

Here is the log file from the new render in the test version: https://drive.google.com/file/d/1SRVkgh ... sp=sharing

[carl]

Looking at those logs again, 2.17.x had the same problems decoding the video but I guess the improved timing code in that version resulted in video drop-outs rather than sync errors.

At this point it's unfortunately going to be difficult to debug any further without a video file that triggers the problem. If it's still impossible to obtain any of the original, perhaps a test video exported using the same settings would be enough. So long as it triggers those errors in the log, it will be enough to investigate further.