P3DCI_D65sim_48nits and DCDM_P3D65limited - an optimum bundle?

Dear all,

The question:

Does ODT.Academy.DCDM_P3D65limited.a1.1.0 include the identical (pre)processing of R, G and B (before conversion to XYZ) as found in ODT.Academy.P3DCI_D65sim_48nits.a1.1.0 regarding “compensation for different white point being darker” and “roll off of highlights to avoid need for as much scaling” described in the code-text at https://github.com/ampas/aces-dev/blob/master/transforms/ctl/odt/p3/ODT.Academy.P3DCI_D65sim_48nits.ctl?

The context:

The colourist’s projector is calibrated with X’Y’Z’ based test-patches according to SMPTE RP 431-2.
In the normative Table A.4: The patch called “White-2” (DCI’s D63-based projector calibration), with 12-bit integer X’Y’Z’ values 3794/3960/3890, is possible to create mathematically based on initial R’G’B’ 4095/4095/4095.
However, patch “White-1” (D65) and patch “White-3” (D55, legacy Xenon projector lamps) seems “synthetic/wrong” as they cannot be created mathematically based on initial R’G’B’ code values in the 0-4095 range – e.g. the D65-patch would need impossible 4097/3896/4179 as initial R’G’B’ in order to eventually display 48 nits in luminance.

This issue is described in the code-text at https://github.com/ampas/aces-dev/blob/master/transforms/ctl/odt/p3/ODT.Academy.P3DCI_D65sim_48nits.ctl.
The issue does not seem to be described in the code-text at https://github.com/ampas/aces-dev/blob/master/transforms/ctl/odt/dcdm/ODT.Academy.DCDM_P3D65limited.ctl.

Hence, there is a worry that the colourist will enjoy the “roll off” in his viewing-ODT (to avoid the viewing of clipped B’ causing nuance shift in the highlights) – while audience in cinemas may experience clipped B’ feeding the projector lamp, giving an overall nuance shift in the highlights, via the DCDM-based DCP.

(Note: The mastering of the DCDM, and the packaging of the DCP, is not performed by the colourist. Mastering and QC is performed subsequently by another colleague in another viewing environment).

Stockholm, November 18, 2020

It is an interesting question, and something that has occurred to me before. I believe you are correct that the DCDM ODT does not include the highlight compression that is included in the P3 ODTs.

However I do not believe that this will result in clipping. The DCI spec includes the normalising factor of 48 / 52.37 for this specific reason. A DCI projector is set up such that X’Y’Z’ input of [4095, 4095, 4095] would (in theory) produce equal energy white (CIE Illuminant E) at 52.37 nits. This headroom makes it possible to produce D65 white at 48 nits without clipping.

It would be the case that a DCDM produced directly with the DCDM ODT will not be identical to one created from a P3 intermediate, as the latter would include the highlight roll-off and the former would not. The question then is whether the two are noticeably aesthetically different. I have not personally made a comparison, but plenty of movies have been graded on P3 projectors and then mastered using the DCDM ODT without complaints

1 Like

Dear Nick, all,

Thanks for your reply regarding “highlight roll off”. A few additional questions, please.

Question 1) Please, where can I download the very latest draft/final version of “Technical Bulletin – ACES Output Transform Details”?

Question 2) Please, where can I download the nine test patches (N1, N2, N3, R, G, B, C, M, Y) that the Technical Bulletin describes? (I.e. ACES-based OpenEXRs).

Question 3) The code-text at https://github.com/ampas/aces-dev/blob/master/transforms/ctl/odt/p3/ODT.Academy.P3DCI_D65sim_48nits.ctl says: “Rolloff white settings for P3DCI (D65 simulation) … const float NEW_WHT = 0.908; const float ROLL_WIDTH = 0.5; const float SCALE = 0.9575;”.
Do the test values in the Technical Bulletin, regarding the P3DCI_D65sim_48nits ODT, take the impact of those parameters into account?

Stockholm, November 19, 2020

(Please, perhaps is Scott DYER the author of the draft Technical Bulletin regarding Output Transforms and able to answer my three questions above? Kindly, Lars)

@lars.haglund, this is an interesting point and quite frankly, I’m not exactly certain. I never really was a fan of supplying the DCDM output transforms as I knew they could create confusion with images mastered using one of the P3 variants. If I wanted to guarantee the exact appearance of my master using the P3 output transform, I would use a third-party conversion tool to create the X’Y’Z’ encoded DCDM files from the P3 files, rather than switching the ODT. In theory it should work, but I’m not certain if it does as there are a number of inconsistencies that I am hoping we clean up with ACES 2.0. (The scaling and roll-off step is one of them - it exists for a specific reason but is not easily extensible to other transforms nor documented well.)

Note this statement from the TB that you mention:

It is not recommended to use the DCDM ODT during the DI process. Rather, the DCDM ODT is provided as a convenience to aid in the conversion of the final RGB graded master into a DCDM version when more appropriate tools are not available. In general, it is recommended that RGB master files be produced with the appropriate ODT for the mastering display device and that those RGB files be converted to a DCDM using a tool designed for DCDM transcodings.

For your questions:

This draft document was never officially published, but you can find it here:
TB_2018_002_v0.2.pdf (256.7 KB) (file last updated 08/06/2018)
(from this thread Request for review and comments on v0.1 Draft of "ACES Output Transform Details")

These are just theoretical values that were run through the transforms and the expected colorimetry on screen, assuming proper calibration and . No EXRs were created (although they could be) - it was expected that they could be made trivially in an RGB generator in a color corrector or other software driving the projector image. I can make these for you if it would be helpful.

The test values listed for each transform were generated by running the ACES values through the transforms and calculating what the screen colorimetry would be. So yes, the scale factors are taken into account for the xyY values listed in that table.

Thank you so much Scott - appreciated.

I now understand why the Colourist should first render an R’G’B’-based master using the P3DCI_D65sim_48nits ODT, not simply replacing that viewing/rendering-ODT with the DCDM_P3D65limited ODT when rendering X’Y’Z’. But … when avoiding the direct use of the DCDM ODTs, does it hurt any of the goals in IMF application #4 (SMPTE ST 2067-40) and/or IMF application #5 (SMPTE ST 2067-50)? I.e. for archival use, with less amount of storage, it would be so convenient to simply swap those two ODTs?

Regarding the nine test-patches: Yes please, thank you! It would be great if you could create those for download by all users here (to get rid of the bias that we all would do something wrong if creating them ourselves). Please, give a thought about the Frame Aspect Ratio, as well as of the use of a 10% rectangle centred in black - as the use should cover all available ODTs for both projectors and monitors. Please, insert a download link in this thread as well eventually.

Kindly, Lars