Intermittent glitch on one frame on DCP

[Kirk Futrell]

We were alerted to a potential visual glitch during a screening of RAGING BULL - it was described to us as 'a frame of colored squares' during a particular scene. When we went to try to replicate the issue, we were immediately able to. However, when we tried to replicate it a second time, we couldn't; nor the third or fourth. Eventually we saw it again, but it wasn't quite the same pattern of squares. We paused the film and tried to advance frame by frame, and did see the artifact, but it flashed up and disappeared without us advancing to the next frame.

By recording about 2 minutes worth of the same 10s of the film, we were able to capture an instance of this artifact appearing:


The setup is a Doremi Showvault with IMB and a Barco 4K 23B. We had occasional quick glitches on screen with another IMB in another hall which replacing the IMB seemed to correct. The question is, because this happens in the same spot when it happens, is this the DCP or the Doremi system or both? We tried the same DCP on an IMS 3000 in another hall trying to get the glitch to show up and it never did.

Also as a side note and this is probably for another thread, but we were provided with a 5.1 and 2.0 DCP of Raging Bull. After much research and testing we confirmed it is a 2.0 surround from the original LT RT 35mm source. So, don't play it as normal. I told Park Circus they should probably let people know that, but they didn't reply. The 5.1 seems to be a matrix decoded mix put in a discrete package since there is bleed in the surrounds.

[John Thomas]

Could be an issue with the actual DCP that only crops up on certain hardware. It could also be a bad ingest or bad drive. So you could delete and re-ingest the DCP from the same drive, or a different drive if you can get them to send another. Reseating the IMB can't hurt either.

[James Gardiner]

J2K decoder is a complex algorithm, especially when implemented on embedded systems like the cards in the older units. It's most likely a small bug in the decoder used on that old card. A small memory leak or something that never really showed up but under the most unique circumstances. That's how it can come and go as it's connected to previously decoded frames, resulting in memory usage on the embedded decoder that causes the glitch.

From what I hear, there is a more serious version of this type of error causing random crashes. A certain film/DCP on a certain decoder... And unreliable in attempting to repeat. They don't really know what's causing the issue, but they are pretty sure its a J2K decoding edge case.

Its a type of problem, not even the big content distribution entities can foresee due to the difficulty is putting the finger on the exact cause.. And as such, they have yet to be able to create a certification test to detect the possibility, But realistically, considering how many screenings happen per day and how rare it is...
I am not sure it will ever be tied down.

[Carsten Kurz]

The Doremi Dolphin decoders are notorious for these glitches when the datarate / components approach certain values. I have seen Dolphin systems fail well below the 'official' 250MBit/s spec. The IMBs also have issues when playing back complex 4k streams. They usually don't crash, but fall back to 2k decoding.
Good/well established mastering companies know these limits and keep their files within constraints.

Not saying that this MUST be one of these Dolphin weaknesses. I have a professional DCP in my backup that shows a single green flash frame somewhere in the middle. Hash check without complains. In this case, the bad frame obviously had been encoded properly to j2K, so turned up before/during the J2K encoding.

The different outcomes during your testing may be the result of data rate issues in combination with buffering. The behaviour of the decoder pipeline may be different when you play through the whole feature vs. cueing to it.

You could try to run that DCP through DCP-o-matic player's verification tool. Maybe something interesting shows up there.

[Marcel Birgelen]

Do you have another room with the same server/IMB/projector setup? Would be interesting to see if it also happens there, because then it would be a compatibility issue with the media block.

Whether it's an encoding or decoding error is often difficult to say without being able to do a bit-by-bit analysis of the source material. Also, JPEG2K is a difficult enough specification that leaves some edge cases that can be interpreted differently.

You should report those issues to the distributor though, so they know the DCP may be problematic on some equipment. If they receive multiple reports, they may actually want to look into it and fix it.

[William Kucharski]

Some of these issues can be so complex.

I recall when seeing The Unbearable Weight of Massive Talent at my local AMC, the end credits were glitching all over the place.

The projectionist (he was actually there at the time, and it was the last showing of the day) went back and played the DCP starting in the middle of the end credits and they played perfectly, so of course he thought I was nuts.

But then for some reason he restarted from the beginning of the end credits, and they glitched all over the place just as I saw.

No idea what the issue was, but since it was the last night of the run on that screen, he wasn't all that concerned; I was just glad he no longer thought I was seeing things.

[Kirk Futrell]

Thanks everyone!

Carsten, I recall you and I trying to figure out if the IMB could do complex 4K. Lots of test patterns. Fun times!

James, would the DCP you are talking about happen to be RRR? If so we had that happen to us. It would crash on both of our IMB systems at the same spot. It was created by Deluxe and they almost didn't believe us.

Marcel, we just switched one of our other hall from an IMB to an IMS3000, due to the HDMI failing on it and occasional momentary rainbow line glitches during films. When we ran this DCP on the new IMS3000 we couldn't replicate the issue. Also, we have reported so many issues lately to distributors, I think they either think we don't know what we are doing or are just pests.

William, it is funny how you can start to feel gaslit or crazy when you are the only one who keeps noticing the issue.