Experiencing an issue with the Map Space HDR10 LUT Generation Method on a LG C4. Hoping someone can help.

Hello. This is a two-part post, however I believe the two issues may be related.

The first issue is that I have found that when using newer HDR10 profiles (taken with newer CS versions) vs. older ones (taken with older CS versions), the resulting LUT when using Map Space looks very weird. Images are attached and labeled. I do not know if the CS version has anything to do with it, but that's the only thing that changed between the profiles. LUTs are being generated using the newest CS version.

Secondly, if I use the newer profile that gives the weird map space LUT, but instead use the other LUT generation method, where you calibrate it to itself, then map to bt2020 (by adding to existing LUT), it looks normal? (extracted colourspace here should say C4 but I labeled it C5 by mistake, so please ignore that inconsistency)

Does anyone know what may be going on here? Why does the newer profile map space LUTs look so weird?

Even though the new LUT looks so weird, it still seemed to validate well, so I'm just not sure what's driving that weird shape. The profiling methods were identical, so nothing changed. I am also confused as to why the shape of the LUT is completely different (more normal) when using the 2-step method, instead of Map Space.

Old Profile LUT_Looks Normal

New Profile LUT_Looks Weird

Alternative LUT Generation_Looks normal

To stand any chance of understanding this you would need to provide the profiles.

Steve

Hi Steve,

Absolutely, profiles are attached. #4 and #4.1 give me that weird LUT shape. #1,#2,#3 do not. I believe #4 and #2 used the exact same patchlist, so those may be best to compare. The others all used slightly different spacing

I had to change them to .txt files in order for it to let me upload. Let me know if that's an issue.

Thank you so much for taking a look!

Please upload as a ZIP file.

Steve

Zip file is attached.

If it helps, the AWP I am targeting is 0.3100/0.3230

Attached file: Profiles.zip

Ignoring the AWP, as I think that is irrelevant, there are a few issues here.
Profile 2 has measurements that are outside of the CIE triangle.
That is impossible, as means the probe is having so real issues with measuring.

But the difference between the LUT generated with profile 2 vs 4 is as expected, as the measured gamut is different enough, especially as you have used very few measurement points.
(The difference in the LUTs is that one is clipping less that the other - the one that looks 'squarer' is clipping more!)

And yes, there will be a difference between Map Space and the concatenation approach as you have used a Quick Profile with Primary & Secondary patches.
As per the User Guides, that invokes a different colour engine, but that will be different for concatenation, which is Peak Chroma, vs Map Space.

So, the reality is you need a real volumetric profile for this display.
And stop the invalid measurements, which are specifically in the low-lights.
(When you have measurements outside the CIE triangle that also likely means other measurements will be invalid.)

Steve

Thanks Steve. I see that profile 2 does have a good amount of invalid measurments. What typically causes invalid measurements like this? I do see they all appear to be in low light. Should I just not measure down this low at all? I have an Admesy Hyperion, which should be good down to like 0.003 nits. I see that these invalid points have target luminances well below that (i.e. 0.001 to 0.0001 nits.)

Also, is there a method for HDR10 that you recommend that does a full volumetric approach? I thought the quick profile was recommended for HDR10, not a cube?

I very much appreciate you looking at this, thanks!