Hello everyone,
what I am reporting here is a behavior of the conversion algorithms (essentially almost always "Fit space" or "Map space") that I have encountered in recent years in the various attempts to calibrate many 4K projectors (especially JVC, Sony and Epson).
Conversion of projector's native color spaces targeting DCI P3 D65 and gamma 2.2.
Here a last example:



What I always notice when I calibrate using FIT SPACE and MAP SPACE is that the hue of the magenta tones is always shifted towards the blue while and the red tones are always desaturated....as in the image above.
This tendency is always there... sometimes in a very accentuated way, sometimes less so.... but the trend is always the same.
I also include an example regarding a JVC Nx9 (with active DCI filter) where the problem is present but in a mild form.



It almost seems that the FIT SPACE and MAP SPACE algorithms are poorly optimized to adequately correct even this area of ​​the solid color. I would like to raise this question with Steve for those like me who calibrate many projectors...maybe we can find a more optimized solution.

attached also the initial profile of the JVC NZ800 and the post calibration verification .

Thank you all

Attached file: JVC nz800 calib test.zip

From the information you have provided, the results look to be exactly as expected for Fit Space and Map Space.
The algorithms are designed to operate very differently to Peak Chroma, as the User Guides do explain.

However, we would also need to see all your LUT Generation settings, to define the expected result.

Also, due to the 'simplified' nature of Fit Space, Map Space would likely be preferable, although I would probably use Peak Chroma, or Hybrid for that display,

Steve

I have also looked at the profile, mapping it to itself.
There are very few volumetric issues, so I would indeed use Peak Chroma for LUT Generation.

Steve

Self Mapping

Hi Steve,
thanks a lot 4 your feedback.
I also tried Peak Chroma and Hybrid but ... red and green colors at the max saturation level collapsed de-saturated near the centre on the color gamut.
Some day after I've calibrated another projector ... a new Epson EH-QB1000.
The RGB separation of the initial profile was not as precise as the JVC one but ... the gamut native coverage was very similar .... so not so big respect REC709.
I've tried to create the calibration LUTs but the problems are very similar ...
with FIT SPACE and MAP SPACE I've obtained red desaturated and magenta tones shifted to blu....
with PEAK CHROME and HYBRID .... good mid tones mapping but RED and GREEN at the max saturation level .... dramatically desaturated and collapse near the centre of the color space.
So ...two different PJs ...with very similar native gamut .... with similar results.

Please read the User Guides.
The 100% patches will show the max value the display can do of the CORRECT colour!
But, the actual gamut will be far larger, and you need to verify with a larger volumetric patch set to see that.

For example, see: https://lightillusion.com/calibration_issues.html#gamut_coverage

All really is explained in the user guides, and this is also basic calibration knowledge.

Steve

Basically I understood the difference between an algorithm that maps the gamut and one that does not. And, thanks to your example, I was also able to verify that the actual gamut is greater than the one that draws the reading of the 100% RGB patches with the Peak Chroma solution.

What I ask is whether it is possible to have a solution similar to MAP SPACE ....something between Map Pace and Peak Chroma ...... somthing that however tries, with gamuts similar to that of JVC and EPSON (which I attach in this post), to map well the first step of saturations of the primary and secondary colors (at least up to where the gamut coverage is).
In the attached zip you will see also the verification profiles of the various solutions tried.

Attached file: QB1000.zip

Sorry, but I really think you re missing understanding on the realities of calibration.
You can either have accurate calibration, or a compromise.
Nothing else.
What the 'compromise' is will depend on the settings you choose to use when generating the LUT, as the User Guides explain.

There are no other options.
The final choice is yours.

But the new initial profile - 13p_EPSON EH-QB1000 TEST Dynamic normalized.bcs - now shows major volumetric issues, especially towards Red.
These can bee seen by mapping the profile to its own extracted colour space.
And there is instability in the measured black and white levels.
Seem the projector is unstable. That will not be helping.

Steve

FWIW, this is my NZ900 with filter and Map Space:
https://postimg.cc/jntC50L9

The QB1000 gamuts are a mess. It would be a miracle, if somebody is able to create a valid LUT for them.

Hi Karl,
I have here also a JVC NX9. And with the DCI filter ON and Map space I obtain similar results of yours.
What I'm trying to report is that many often , with different projectors, with a less extended native gamut (like a JVC without the filter ON), the MAP space is not so precise with RED and MAGENTA saturation sweeps.
Look at the first image that I posted.
When the gamut of the projector is not so big ... the red saturations are a little bit desaturated and the magenta shifts to Blu.
Sony, Epson and JVC without the filter has similar native gamut coverage.
In all of this cases the Map Space has similar inaccuracies on RED and Magenta.

It seems like Map space is not optimised with this type of native gamut .... and converting to DCIP3 gamma 2.2/2.4

This is

Hi Steve,
thank you.
Yes, what I'm trying to find is a "compromise".... not a perfect calibration.
A compromise that, after the LUT conversion (maybe with MapSpace or something similar) , gives me a good tracking also on red and magenta if I verify it with the standard saturation sweeps in CS.

Like told to Karl .... it seems that map space works very well with much extended color spaces (like a JVC with the filter ON) but has some problems with less extended native spaces ...like a JVC without the filter of a SONY or an Epson.

In your opinion there is some setting in the conversion stage that I can try to obtain a good result with red and magenta saturation sweeps test?

infinite thanks.

As said before, the results are as expected.
The two pre-calibration profiles you have now posted show very different underlying capabilities.
Which is correct, or is the projector just unstable, as the second set of profiles shows?
It seems the projector has some serious underlying issues.

And you have also not provided the requested additional information.
What other settings are you using?

But, there is no 'other' option.
You have to choose which of the available options is working for your specific display.

Steve

Hi Steve,
the two files are different because are referred to two different projectors. A JVC NZ800 and an Epson QB1000.
I'm with you, the best profile is the JVC's.
But ...with Map Space ... both calibrating luts are similar if you measure and look the RED and MANGENTA saturation.

as told to Karl ...it seems that Map Space works in this way if the starting gamut is similar to the one of JVCs without the filter and the one of the Epson QB1000.

I attach the settings used to create the calibrating LUT of all the projectors taken as example in those posts.

As you have Gamut Mapping enabled, you are seeing Gamut Mapping being applied.
Wether you want that enabled or disabled is a choice you have to make.

It is part of the process of dealing with display with low gamut compared to the target colour space.
(And has different effects depending on the LUT Generation option you have chosen to use.)

Steve

@FidelioX
How do you suggest mapping should be done? There isn't enough red and green to work with at 100%.

This is a gamut sweep for 75% and 100% with a NZ9 and Map Space.
I'll take the colour shift of Map Space at 100% over Peak Chroma.

Karl, you should compare the result of Peak Chroma vs. Map Space with a large volumetric profile set.

You will see that Peak Chroma is more accurate, but also covers the same total gamut.

Steve