P3 color conversion to XYZ

[VitorC]

Hi, I’ve been searching for this topic and I didn’t seem to find the exact question I am in need of an answer for.

I have recently graded a short film in a color grading suit that had a projection in P3-D65. I am now preparing a DCP for a test screening in a small multiplex theatre, and I am left in doubt within DCP-o-matic’s color conversion to XYZ options, as the “P3” input option does not mention if it is P3-DCI (which I assume it is), P3-D65 (what I graded for) or anything else.

My thought was to change my output color space in Davinci resolve color management before exporting, so that I would have my export in P3-DCI and feed that into DCP-o-matic. I ended up doing that AS WELL AS exporting the film in P3-D65 and making another DCP.

If I import the DCPs into Davinci Resolve, I see no difference between the two, and they look as my film does.

If I import the DCPs into DCP-o-matic Player OR easyDCP player, I see no difference between the two, but they both look slightly green (white point greener?).

So my questions are
- what white point does DCP-o-matic assume with color conversion from P3?
- why does my DCP look slightly greener in DCP players and looks fine in Davinci Resolve?
- am I missing something?

Thank you,
Vitor

[VitorC]

Hi again - I wonder if anybody know the internal build of dcp-o-matic and can answer the above question? How can I feed two files with different white points and do the same xyz conversion and see no difference between the two? Can anyone elucidate me on this? Does white point not matter at all? Why do the dcps seem to have a green white point on dcp players and seem fine in resolve?

Sorry to post again the same question but it seems unlikely this has never been an issue before.

Thank you!

[Carsten]

I can only say that PC based players can not offer a realistic color reproduction.

You may try some things out using the custom color conversion dialog - it has options for white point. Start with the P3 preset.

As far as I know, Carl is on holiday right now, of course he does know about the inner workings of the color conversion in DCP-o-matic.

[VitorC]

Hi, thanks for the reply.

Yea I assume the reproduction is not realistic but since I’m comparing two different DCPs that I expected to have some slight difference, and I see none, I’m left without knowing which one is done correctly.

In the P3 custom menu I believe the white point values are the ones of P3-DCI, but I just wanted a solid confirmation on this and the above questions. I will wait, gladly!

And of course I’m testing the DCP in a commercial theatre, on Monday btw, but there are only so many times I can go test in a theatre environment, so I’d like to minimise the potential issues the DCP might have, and use the test to look carefully at color rather than feel like the DCP is made wrong.

Anyway, I will wait for Carl’s input, pay very close attention to the two versions of the DCP screening on Monday, and eventually some of these questions will have answers.

Thanks again!

[Carsten]

DCI is x'y'z', which is the DCP storage colorspace, with DCI xyz having it's own whitepoint, around 6300K I think.
And P3 being a display colorspace, it obviously has a different whitepoint. In the case of P3-D65 - 6500K.

https://en.wikipedia.org/wiki/DCI-P3

https://www.liftgammagain.com/forum/ind ... nts.11109/

As you can see, there is a bit of confusion about wether xyz actually defines a whitepoint. As I understand it, xyz does not have an actual whitepoint per definition (xyz exists beyond DCI or other color spaces), but within DCI specs, there is a 'DCI whitepoint' defined within xyz, which originally was based on cinema projector xenon maximum brightness white - which is somewhere around 6300K. The difference between 6300K and 6500K may be what you are seeing.

[NominalFilm]

Hello everyone, I can confirm Vitor's observation. I tested a Tiff sequence (DCI P3 color space, gamma of 2.6) in cinema and the result had a green cast too. The color space input in DCP-o-matic was set to P3. Does DCP-o-matic convert from DCI P3 to XYZ?

In my opinion, it cannot be due to the Kelvin since 6300K should have a little less blue content (i.e. more yellow content) than 6500K, but should not have any effect on the magenta-green axis.

@ Vitor: have you made any progress with your tests? Have you seen your test in the cinema by now?

Best wishes, Max

[Carsten]

>I tested a Tiff sequence (DCI P3 color space, gamma of 2.6) in cinema and the result had a green cast too.


Can you explain in more details how you did that? Technically, you can not test a P3 color space TIFF sequence in a cinema - it has to be converted to X'Y'Z' before. I guess you know that, but, just to make sure we are talking about the proper procedures.

DCP-o-matic will do all necessary conversions. The main question in most of these discussions is - do we have a 'slight' color difference, or is it 'green tint'. If your image is actually 'green' at the cinema, the most likely explanation is that NO color conversion has been performed at all.

P3 in a cinema is just an intermediate display color space that is used between server and projector, or, within the projector. You would not approach a DCI projection cinema system yourself with P3 files. You may of course use P3 for grading, but, when aiming for a DCI projection system, you need to go through X'Y'Z' as the final stage.


In DCP-o-matic, you can apply preset color conversions from all common color spaces, create your custom conversion, or disable conversion completely. The latter should only be chosen if you know for sure that your source is X'Y'Z' (which is XYZ with an inverted 2.6 gamma curve).

If you use P3 input files, it is indeed important to know or match the whitepoint. The trouble for users may be, that DCP-o-matic uses an xy pair for whitepoint interpretation, not the common Kelvin single parameter. You can find common values here and may play with them:

https://en.wikipedia.org/wiki/DCI-P3


Easy to create a few short clips with different conversion parameters and try them at the cinema. When you compare DCI and P3-D65 whitepoints in the CIE1931 table here, you will notice that DCI 6300 is not just 'warmer' than P3-D65, but shifted slightly towards green/yellow:

https://upload.wikimedia.org/wikipedia/ ... 20.svg.png

As this DCI whitepoint is not on the planckian locus/planckian white point path, the typical 'lower Kelvin - less blue/more red' simplification does not apply here. Therefore, it is wrong to call it 'D63'.

[Carsten]

Hmm. I did some numberchecking and it may indeed be that the default P3->X'Y'Z' conversion preset in DCP-o-matic is not practical. It seems to assume the DCI whitepoint for P3 sources, while, when considering typical grading/editing workflows, it would be more useful to use the Display P3-D65 whitepoint for source interpretation (because that is what is typically used with P3 calibrated grading systems), and then convert it towards DCI whitepoint (0.314/0.351) for X'Y'Z'.

Now, I suggest you create your custom conversion preset using the default P3 preset, and alter the RGB conversion matrix to:

Bildschirmfoto 2023-08-26 um 16.17.38.png

Please try this, compare at the cinema, and then come back here and tell us your results. Best is to prepare two identical clips, one converted using default P3 preset, one using the modified preset, and then play them next to each other at the cinema.

I think I don't have any dependable P3 source material here right now to check this on our projection systems.


If it all comes together like that, I guess it would be wise to change DCP-o-matic 's default P3 conversion preset to Display P3 whitepoint/D65.

[RAY-NUE]

P3 must be viewed with HDR10 compatible devices.

[Carsten]

That's not true, P3 was there long before any HDR technique was available. And essentially all DCI projectors use P3 and are not HDR10 capable (with some rare exceptions).