View Bug Details

Jump to NotesJump to History
IDProjectCategoryView StatusDate SubmittedLast Update
0001267DCP-o-maticBugspublic2018-04-05 16:452023-12-22 22:32
Reportercarl Assigned To 
PrioritynormalSeverityminorReproducibilityhave not tried
Status acknowledgedResolutionopen 
Summary0001267: Investigate problems around 250Mbit/s
Description

Not sure if I've checked that DCP-o-matic doesn't go over 250Mbit when you ask it for 250... worth checking if we can legitimately blame servers for problems at high bit rates.

TagsNo tags attached.
Branch
Estimated weeks required
Estimated work requiredUnknown

Relationships

related to 0001664 acknowledged Check that J2K frames are not too large 
related to 0002450 acknowledgedcarl Check frame size and color component sizes of frames in verifier 

Activities

Carsten

2018-04-16 15:33

manager   ~0002374

Last edited: 2018-04-16 20:01

There is probably something to investigate - in 'earlier times' I did some compression tests with 2k and 4k noise patterns. I just recreated such a 4k noise pattern in Photoshop, and did some compression tests with DCP-o-matic 2.12.4. This pattern is very complex and brings every common compressor to it's knees.
Created a 10s still DCP from it. These numbers are for the J2K MXF file:

2k - 125Mbit/s-24fps: 156MByte/124MBIt/s
2k - 230Mbit/s-24fps: 287MByte/229MBit/s
2k - 240Mbit/s-24fps: 300MByte/240MBIt/s
2k - 250Mbit/s-24fps: 312MByte/249MBit/s
2k - 250Mbit/s-48fps: 312MByte/249MBit/s
4k - 250MBit/s-24fps: 250MByte/200MBit/s
4k - 250MBit/s-48fps: 250MByte/200MBit/s
4k - 500MBit/s-24fps: 250MByte/200MBit/s
4k - 500MBit/s-48fps: 500MByte /400MBit/s

The standard size of 312MByte at 250MBit/s/24fps/2k computes to 249MBit/s - nice one.

It's a bit strange (though at least safe) that the 4k/24/500MBit/s doesn't nearly reach the target size, but the 48fps version does.
It's possible, though, that my test image does not really allow to saturate the 500MBit/s at 24fps, I may need to investigate this. It's just my assumption that this noise pattern is worst case. I tried to replicate this by using a high-res nature image, and applying noise to it. That created a more widespread color spectrum in video analysis, but still gave the same result at around 200MBit/s.
Maybe this is an OpenJPEG implementation issue with the cinema profiles. Is it at all possible they may not support higher Bitrates?

Hmm, interesting, I just tried it with the same image in DCP-o-matic 1.83.0 (YES!!!!), and it creates a 625MB file (10s-4k/24fps/500MBit). That computes to...
500MBit/s ... hmmm... maybe the 4k modes need some tweaking.

  • Carsten

Bug History

Date Modified Username Field Change
2018-04-05 16:45 carl New Bug
2018-04-16 15:33 Carsten Note Added: 0002374
2018-04-16 16:15 Carsten Note Edited: 0002374
2018-04-16 16:24 Carsten Note Edited: 0002374
2018-04-16 20:01 Carsten Note Edited: 0002374
2019-11-20 11:58 carl Relationship added related to 0001664
2023-04-15 22:37 carl Relationship added related to 0002450
2023-12-22 22:32 carl Status new => acknowledged