What's new with that driver? Z negative values are not capped to 0 from the driver.
This comes from some inaccuracy I have noted in probe matching done starting based on highly saturated Red - You see x+y = 1.00000000000 and in the info point you have a confirmation of Z = 0.000000000000. I have seen that on my C1 and on a new Sony Bravia QDOled.
Reported to Steve, he queried the 3rd party developer and he fixed the issue. For understanding the impact in probe matching have a look to the following:



On your right the outcome of probe matching on my c1 with the driver capping Z to 0. Gamut was set to native. The other cases, from the right has given accurate correction using as correlation WOLEDExample (Z=0 disappeared), using the previous XRITE drive (the CMF correlation also reported negative Z) and at the end, the one with the fixed 3rd party driver.

This, obviously has some impact when you are using probe matched with capped Z to 0 with the new update drivers:
To asses that, I have verified that outcome, extending the investigation to the outcome in case the probe matching is done in SDR.
See here the native case (the case I am interested to is the one on the right)



and the SDR one:



Generally talking I see errors with a delta ITP below 3, so not so visible. Anyhow that is related to my C1.

PS: Last and not least - I have checked with Calman (anyhow an old version). Z is capped to 0 (when negative)

It is worth noting that if the display is not wide gamut, or is wide gamut but is set to a smaller gamut, there will be no clipping of Z values.

It should also be noted that performing Probe Matching using the target colour space, and NOT the display's native gamut, can produce better results when the filters in the colourimeter are not well matched to the CIE observer, especially when NOT using a Multi-Point Volumetric Probe Match.
To use the target colour space you just need to set the display to the target colour space using its own preset colour spaces, such as Rec709, or via a quick LUT.
(The accuracy of the display to the target colour space is not relevant for probe matching.)

And you should really not rely on an old Probe Match.
A new Probe Match should be performed before each and every calibration.

Steve

My only concern about using a reduced gamut is that it is generated through a LUT, and LUT might alter panel linearity.
And, in this case, I have extended my testing patch-set and repeated it also for the SDR case to be comfortable on the fact that negative Z's don't come from filter linearity issues.

No - using the reduced gamut is the opposite.
It will improve Probe Matching as you are performing the match in the target gamut.
It is like the approach for a Focused patch Set.
How you put the display into the smaller gamut is irrelevant, as you are matching the exact same input RGBW value (or more, if using MPVM).

Negative Z value come from poor CIE filters within the probe - nothing at all do do with filter linearity.
It is technically impossible for any RGB monitor to ever actually have negative z values!

Steve

Is this based on empirical data?
It sounds strange to me because we are usually measuring the native gamut (which is usually a wide gamut nowadays) to create the calibration LUT. The reduced gamut (usually emulated through the LUT for, say, Rec709) only comes into play during verification (which is usually no more than a sanity check of the workflow if the probe is the same for both the calibration and the verification steps).
It makes even less sense to me if the target colorscape is roughly the same (for example, DCI-P3 on a Xenon projector) or even wider than the native gamut (for example, Rec2020 on HDR displays).

By the way, what is the fear of negative numbers? I would take a negative minus one over positive two error even if the negative number makes no physical sense. Engineers are utilizing imaginary numbers in their calculations to designs real-world structures. It's nothing more than (immediate) math (real numbers go in, real numbers go out but imaginary numbers are there for intermediate calculations...).

It's based on both mathematical certainty, backed up by empirical testing.

When performing a Probe Match using the target colour space you are focusing the match onto the colours that are important.
Additionally, poor CIE filters basically show a greater error when measuring wide gamut, so you are reducing the likelihood of having inaccurate matching.
This all has nothing to do with LUT generation.
The key is to limit the probe match errors, which will potentially be greater if the probe match is on a gamut that is greatly different to the target, and the probe filters are not well matched to the CIE observer.
If the native colour space is close to the target colour space, the issue is really just to avoid issues with the probe's filters when pure primary colour is displayed.
That is where any optical filter will fail the most, so best avoided.

And there is no fear of negative numbers.
They just show there is an issue with the probe's filters, and that they are not well matched to the CIE observer.

It is all very logical.

Steve