Black-and-white film, Resolve P3-DCI,

Discussions - Using ACES Post (DI, Edit, Mastering)
1

First of all I’m sorry that My English is not so good.

I use davinci resolve 15 linux version.
I’ll grade 4K scanned dpx sequences.(scanned by Scanity)
Display is a 4K sony dci theater projector. calibrated x-0.314 y-0351
I want to use ACES color science in Resolve 15.

*Project settings
color science :ACEScc
ACES version : 1.1
IDT : ADX10
ODT : P3D60? or P3-DCI(D60 or D65 sim)?

It’s a black-and-white film, so white point is very important.
It has to be seen in monochrome.
The main output is DCP.

When I set P3-DCI(D65 sim)ODT, resolve internal vector scope shows magenta.
Projection screen is fine, I think.(it looks like rec709 d65 white)

When I set P3-D60 ODT, resolve internal vector scope shows nearly no saturation.
My projection screen shows a bit greenish.

I render DPX sequences to make DCP with clipster.

Which ODT should I use? Is DCI WP a bit greenish or just white?
It really confuses me.

2

Yes DCI white is greenish.

Using P3D60 for cinema screen would be fine. Adjustment to make the white a little warmer grey may be necessary. Because DCI projectors vary on different sides of the screen between green and magenta, it is better to have the overall balance tilted to one side for black and white. It depends on content and how warm or cool it should appear.

Jim

3

Ok, what you’ve said makes sense. Let me try to explain each one and then give you a recommendation:

Ok, this makes sense. The projector is calibrated to x=0.314, y=0.351 (from here on out = “DCI white”). Resolve’s vector scope is showing you the code values being sent to the projector in RGB mode. The reason it looks “magenta” on the vector scope but ok on the screen is because in order to make a D65 white (x=0.3127 y=0.329) appear on the screen, the ODT is outputting unequal code values (magenta-ish) to counteract the green tint of the DCI white.

You can also think of it this way, “neutral” on the vector scopes shows that the code values are equal, but the vector scope doesn’t know what your projector is calibrated to. If you send equal code values to the screen when it’s calibrated to DCI white, you will get DCI white. But the ODT is labeled as P3-DCI (D65 sim). What that means is that it’s designed for the use case where you have the projector set to P3 primaries, DCI white, but want to have “neutral” ACES values appear as D65 white.

With this ODT, no compensation is done to make the white appear different. But it assumes that you have your projector set to the ACES white point (“D60”). Most people don’t calibrate their projectors to this (although we do!), so this is probably not the ODT you will want to use. You say it looks greenish which makes sense because if you actually measured the white on the screen using this ODT, it would be DCI white (0.314, 0.351) in your setup. (It shouldn’t normally be DCI white, but it will be in this instance because this is the “wrong” ODT for your display configuration)

You render DPX sequences into a DCP? If using ACES I would recommend using Resolve to export DCDM encoded TIFFs and then package those into a DCP.

DCI white is greenish. I would not recommend using that as “neutral”. I would recommend using P3-DCI (D60 sim) ODT for viewing on your projector. Again however, on the vectorscope, the neutrals will look magenta.

In ACES, a D60 white is considered ideal, but many people use D65 because it is the standard for video deliverables. Use either P3-DCI (D60 sim) or P3-DCI (D65 sim).

Before export, I would then recommend switching your output to DCDM with either (D60 limited) or (D65 limited) - corresponding to whichever you chose to grade with. This will output properly encoded X’Y’Z’ data and clip to the gamut that you used to master so that you don’t get any unexpected colors that you couldn’t see when grading. The Deliver tab can be confusing though. You don’t want to “double convert” from RGB to XYZ. This is also a danger in Clipster if you don’t have your settings right. It depends where you want to do the conversion:

  1. let the ODT handle it
  2. trust Resolve’s conversion in the “Codec” dropdown menu
  3. trust Clipster’s conversion
    Any can work but it might seem counterintuitive based on what you’re input/output are. But that’s not an ACES issue - that’s a common and general RGB-to-XYZ conversion issue…

I hope some of what I’ve said makes sense and helps you out.

1 Like
4

Thank you.

I understand what should I do.

And I figured out scopes and UI viewer show magenta with P3-DCI(D65 sim) odt. I generate 3D LUT(only color space transform from P3D65 to P3DCI).
I put this 3D LUT to viwer lut and scope lut I can see the neutral.

Clipster converts P3 to XYZ well but I’ll test p3dci(d60 and d65 sim)odt dpx seqs and DCDM odt tiff xyz seqs for DCP.
I know DCDM xyz tiff is more proper format for DCP. But my client wants DCDM as RGB dpx seqs. I think it’s wierd.

5

I use ACES 1.1 cct, calibrated the projector white point to DCI white, Davinci Reslove ODT set to P3DCI (D65 sim), when the color grade was done, I rendered a ProRes MOV, if I drag this mov to timeline, what IDT should I choose for this MOV? I found if I chose IDT P3DCI(D65 sim), it was not the same as color graded clip, look little green …

6

Does anyone have any issues when outputting BW to the DCP? When packaging to j2c files (the main was DPX RGB) I have strange color artifacts, green and magenta, on several instances of the film. Tried to pack this j2c at a QUBE station and from Resolve. Both generate colored artifacts. Is this a j2c compression issue?

7

How are you viewing the exported files? Through some software? On a calibrated projector or on a computer display?

8

All of the above… Opening the DCP on Resolve results on the vectorscope bouncing all over the place. Visually, on a DCI projector (Chrisite) I can see magenta and green spots, though they seem fixed, rather then randomized. This appears on any theatrical exhibition of this project, so it’s not our projectors fault. The spots are magnified on the big screen and they seem very discreet on a 31" calibrated OLED monitor. The DPX and all other masters (I must say that there’s an exception on heavily compressed files, such as h264) are pure bw, no noticible artifacts and the vectorscope is dead on BW. Since j2c files are only intraframe compressed, i would not expect this kind of behaviour. But since they are not "mono"files (carry RGB distinct yet equal values on BW) maybe the vary form frame to frame on these values. Im out of resources on undertanding why this is happening.

9

Good morning,
I’m re-opening the issue.
Indeed, I restored a feature film in B&W but I cannot correctly encode the 4K DCP from Doremi Cine Asset without generating green and red artifacts…
The calibration was done from Resolve and exported to DPX. Has anyone already encountered the problem?
Thanks in advance.

10

I had the issue in the past with complex Images, the J2K compression introduced colour artefacts.

To verify that it is compression related you could test rendering out a short clip with a strong blur applied to the shot (lowering entropy). If that blurred image is ok then, your artefacts are probably compression related.

1 Like
11

Hello,
Thank you very much for taking the time to answer my question.
Unfortunately, the Gaussian blur did not solve my problem.
The only thing that has improved (a bit) the artifacts is the output directly in JPEG2000 from Resolve. Then create the DCP in CineAsset without converting color space.
Otherwise, I noticed a satisfactory result with a srgb LUT to DCI XYZ then an export to TIFF RGB 8bits.
But for the moment it is only tracks that do not completely remove the problem. Then create the DCP in CineAsset without converting color space.
Julien

12

If you use a LUT to Concert to XYZ, you need to be careful, it is not easy to make a LUT so that the lattice points in XYZ fall on the choose neutral axis.
And if your neutral axis is not directly on a slice of your cube you get LUT interpolation errors.

If you send a neutral ramp through your pipeline and you get “sick sack” waves on the output you probably have a LUT interpolation issue.

You could try to convert to XYZ using only formulas.

Generally speaking try to avoid any LUT for the output pipeline.

13

Hello,
Thank you very much for your thoughts.
Indeed, my workflow was extremely simple when compared to your examples.
I’ve never talked about chromaticity coordinate conversion. I’ll try to find articles to understand how to do that.
Maybe if you have an idea where I could start because I don’t know how to go about it with the chromaticity coordinates.
Thank you for your answers anyway.
Julien