Output Transforms Architecture VWG - August 21, 2024

Here is the meeting recording . Notes to be posted when ready.

1 Like

The notes from meeting #166 are now available.

I did some testing following up on @remia’s comment in the meeting that he got small negative values with the cyan and blue patches of the DLAD and the P3-D60 2.6 gamma rendering.

The v60 Blink does not have an option for P3-D60 output. But if you output linear AP0, followed by a Nuke ColorSpace transform from AP0 to P3-D60, you do indeed get -1.0e-7 in the red channel of the cyan patch, -7.5e-9 in the red channel of the blue patch, and -1.5e-8 in the red channel of the green patch.

In theory our code for that transform goes via P3-D60 and clamps off negatives. So I can only assume that the small negative comes from precision issues in the subsequent matrices going to XYZ and back again. Perhaps we should use a spow() in the final inverse EOTF encoding to mitigate that.