Profiling macBook XDR screens

Update: I’ve performed two additional tests to try to better understand the logic. The first test is based upon the fact that Screenshot.app on macos saves the icc profile into the snapshot. So I can confirm that the macbook XDR takes into consideration a mix of both the reference mode (preset) and the icc file. Here’s the full story: with the factory icc set and the screen preset in any of the P3 modes, Screenshot will save the factory icc into the picture. Ok, correct, so what? If I however choose either HDTV Video BT.709  OR Internet and Web (sRGB) as active preset, then the factory profile is overridden by “sRGB IEC61966-2.1”.  This is also correct, because with such presets the snapshot is taken on an image rendered in a sRGB colour space, not in a P3 one. Now, if I perform this very same test after applying a custom ICC (such as the one created by the calibration process) things gets funny:  Screenshot will always save my custom icc profile into the picture, regardless of the colour space of the active preset! Which means that a screenshot taken on a screen with a sRGB or BT.709 preset on a mac with a custom profiled ICC will have wrong colour space information, if the custom ICC is not in the same colour space of the preset. This seems to confirm my very initial doubt: since Apple uses a mix of icc and reference preset to achieve a given result on screen, probably I should profile twice: one for P3 and one for sRGB. And than switch between the two icc files when switching reference mode…Given the hassle this means, I performed a second test to attempt to answer the paramount question if there is a perceivable difference in colour rendering after screen profiling on my three year old macbook XDR. And this both under P3 and under BT.709. The short answer is: no one  that I can  spot with my eyes. To have something to share, I ended up setting up the following, which is for sure not a scientific test but a reasonable practical comparison. I took RAW pictures of a Rainbow chart with my camera, as displayed on the macBook screen both with factory and profiled icc and with both the Apple Display (P3-500 nits) and HDTV Video (BT.709-BT.1886) presets. I than developed the pictures into tiff files and finally created a difference between both image pairs (factory ICC and profiled ICC) using Photoshop. I was interested in perceivable differences, not absolute colorimetric values. The results I think speak for themselves and are attached below. Yes, there are differences (see red arrows), but they are quite negligeables - at least to me.

Source Rainbow chart

Difference file (Apple Display P3-500)

Difference file (HDTV Video BT.709)

Update 2. To my true amazement, no matter what I tried out (see what I’ve described in my previous posts above) in spite of the good results reported by the calibration software I could never nearly match what I was seeing in the color window of DaVinci Resolve on my macBook with what was displayed on my REC709 calibrated EIZO 2740 (connected to the very same macBook as reference with a Blackmagic’s Ultrastudio monitor 3G box). No way. Contrast, saturation and certain hues were off one way or another, with any of the preset and with the various ICCs. I was about to give up when I finally found a way.  All the “good” calibration reports I’ve been generating and studying in the last couple of weeks where of no use for what really mattered to me: the blatant difference in Resolve was there to be seen.

So, here’s the procedure,  if you want to give it a try. The key is using DisplayCal in connection with DaVinci Resolve as pattern generator to generate a 3dLUT which will be loaded into Resolve itself. This workflow is usually used to create 3DLUTs for external displays or LUT Boxes and can't be done with Calibrite Software. I’m simply using it to “calibrate” the macBook display from inside Resolve, the concept and the process are however basically the same.

The caveats: 1) you need to download and configure DisplayCal version 3.9.12 (Python porting) in order to to run it on macOS sonoma  2) you have to run the process as described with two main differences: a) you can’t influence the RGB balance of a macBook screen b) you’ll end up applying the 3DLUT to the “Color viewer lookup table” instead of “Video monitor lookup table” 3) you have to find a correction file for DisplayCal which takes into account the particularity of the macBook XDR screen and your calibrating device.

Before starting the DisplayCal / Resolve calibration session I’ve 1) reset the ICC profile on the mac to its factory default and 2) I’ve cloned the all-purpose "Display Apple P3 500 nits" preset to one whose brightness is limited to 100 nits (but keeps the P3 native space). In other words my macBook is now back to its factory defaults. But, with the 3dLUT applied to the Color viewer, Resolve is now able to show me identical images on both screens. I'm appalled.

[Edited by Moderator]