CIE Chromaticity shows data out of the Rec 2020 space

Greetings,

I'm using "Davinci YRGB Color Managed" as my Color Science, and
Timeline and Output Color Space as Rec. 2020 (Scene).
The Input Color space is Cineon Film Log.

When I look at the CIE Chromaticity of my images, I find that the data appears to be outside the Rec 2020 space... even though I specified Limit Output Gamut To to be Output Color Space. Any idea why this discrepancy?

I've attached Screen grabs of the CIE Chromaticity as well as my Color Space & Transforms page. Thanks in advance for your response.

Regards,
Rajnesh

CIE Chromaticity:

CIE Chromaticity of Image

CIE Chromaticity.jpg (458.96 KiB) Viewed 2629 times

Color Science:

Color Science

Color Science.jpg (128.35 KiB) Viewed 2629 times

If I remember correctly there was a bug in gpu scopes which was supposedly fixed in one of the latest Resolve versions. Try disabling gpu scopes if you use them or check your R version.

What is the gamut of your original footage by the way. Cineon log curve specifies just the transfer function, but there is some kind of implicit interpretation of source gamut. You can force separate gamma and set it to what it should be, then check the scopes again.

Why gamut limiting does not work I don't know.

Greetings Hendrik,

Thank you for replying. I was given Log files from the Lab, and I'm not sure what Gamut they used.

I am also new to color correction. If I want to learn about the concepts you've referring to in you message, could you suggest a good starting point - something that I can read and gain a better understanding.

Thanks again.
Rajnesh

I am also new to color correction. If I want to learn about the concepts you've referring to in you message, could you suggest a good starting point - something that I can read and gain a better understanding.

If you're a beginner, start the session over and live with the default DaVinci YRGB settings and leave it all in 2.4, which is also the default. Don't make this harder than it needs to be. Just correct it all in Rec709, calibrate a display for Rec709, and use a Blackmagic output card if possible for monitoring.

Unless you have a Rec2020 monitor (and those are rare and expensive), I would avoid that and just keep everything simple.

You say:

I was given Log files from the Lab, and I'm not sure what Gamut they used.

It could be anything. Call them on the phone (or email them) and ask them what files they are. At worst, go to the media page and see what Resolve says about the files. They literally could be anything -- there are at least 100 file formats and a huge number of potential gamma/color spaces available.

Thanks for your reply, Marc.

I think I will switch to the Davinci YRGB and leave in the other defaults - with one exception that is forced on me.

The film aggregator, I'm using, wants the Color Space to be Rec 2020. The rest of the specs say 4K and ProRes 4444 XQ, so I'm keeping my fingers crossed and hoping that Davinci's UpRes from 2K will look good enough (we shot the film on Super 16, and scanned into 2K).

As for equipment, I have a Macbook Pro and a Dell U2720Q monitor that I use to view the footage. It has 8bit + FRC and shows about 67% Rec 2020 (not great, I know, but it's the best I can do).

Any other recommendations would be greatly appreciated.

Thanks,
Rajnesh

I'm resurrecting this issue with some of my own troubles.

I'm currently working on a project where we'll have a few abstract sequences that we want to be extra strong in red and we'd like to use the P3 red primary to that effect.

I'm most comfortable working in YRGB non-managed (monitor) color space so I set the project to P3 and gamma 2.6. We will do the final grading on a calibrated NEC PA311D that has DCI P3 preset LUT, but in the meantime I'm doing tests on a monitor that covers 85% DCI P3 and is calibrated to 2.6 gamma. Thus, to achieve the effect, I was counting on eyeballing the new CIE diagram and pushing the controls so that that red pixels just start touching the red primary point of the P3 triangle. However, when I do that, the actual color on screen looks really pale, not close even Rec709 red, so I start clipping against the edge of the triangle to achieve the sufficient red color deepness. I then made a test DCP which we run in a cinema and what should have been a P3 red turned out YELLOW! That's possibly because colors are not clipped to P3 but start spilling around since the color space of a DCP is XYZ, so much wider.

So the question is: even with the fully DCI P3-capable NEC and the CIE diagram that currently looks like it's of no use, how do I know when to stop pushing so that my end result reflects what I see on the monitor and doesn't flip over to unexpected outcomes?

Thanx.

If you handle your footage as P3 in all subsequent conversions, you exceed the gamut only when producing negative values. So clip to zero. It kind of sounds like problem is somewhere else though... can you share a still of your supposed P3 image as a tif or some other file?

Probe your image values, if you push to one primary, red for example, RGB values should be close to 1.0, 0.0, 0.0.

If you handle your footage as P3 in all subsequent conversions, you exceed the gamut only when producing negative values. So clip to zero. It kind of sounds like problem is somewhere else though... can you share a still of your supposed P3 image as a tif or some other file?

Probe your image values, if you push to one primary, red for example, RGB values should be close to 1.0, 0.0, 0.0.

When I probe the grade it's 1024,0000,0000, when I probe the reimported DCP I get 1024, 0955, 0000 whis IS yellow. I also check the DCP export in easyDCP player and it's same there as is on the projector.

No grade:
https://drive.google.com/file/d/1GJowWr ... sp=sharing

Grade to maximize red:
https://drive.google.com/file/d/1kbsdhZ ... sp=sharing

Reimported DCP:
https://drive.google.com/file/d/1t7xMLq ... sp=sharing

Further testing suggests this happens when I push the color wheels gain above 1 or hue vs saturation to 2...but I can't get the red effect i.e get rid of orangey bits if i don't push those controls.

But the most important thing for is to confirm whether I can trust the CIE to actually show that the colors are clipped at the P3 triangle boundary and this problem is just some DCP conversion bug.