Problem 1. When using Sub-Space, the displayed Targets are inappropriate. The Targets are calculated using the selected Patch Scale/Resolution relative to the Sub-Space. But in general, the resulting xyY values are not realizable using the selected Patch Scale/Resolution in the actual Target Gamut/EOTF, and hence do not match the actual patch triplet sent to the pattern generator, when evaluated against the Target Gamut/EOTF.

To provide a simple example:
Target Gamut/EOTF: ST2084 Rec2020
Sub-Space: Rec709
Patch Scale/Resolution: Legal 10 bit
RGB selected in ColourSpace: 940, 64, 64
Displayed Target xyY: .6400, .3300, 2126.39
RGB sent to pattern generator: 897, 687, 553
Correct Target xyY for that pattern: .6403, .3298, 2126.57

To be clear, the RGB sent to the pattern generator is correct; it simply isn't possible to represent the displayed Target xyY exactly using integer RGB values in ST2084 Rec2020. As such, it isn't possible to obtain an accurate characterisation of the accuracy of a display, because the displayed Targets are not accurate for the actual patches used. You can see this trivially by running a Characterisation using Virtual Probe.

Problem 2. When using Sub-Space, using Limit Range destroys the EOTF calculation.

To provide a simple example:

Target Gamut/EOTF: ST2084 Rec2020
Sub-Space: ST2084 Rec709
Limit Range: 722
Patch Scale/Resolution: Legal 10 bit
RGB selected in ColourSpace: 722, 64, 64
Displayed Target xyY: .6400, .3300, 211.306
RGB sent to pattern generator: 468, 301, 218
xyY for that pattern: .6402, .3300, 21.063

There is a 10x difference in Y. The problem is that the RGB sent to the pattern generator is based on a different calculation. Instead of using ST2084 Rec709, an abberant PQ-like EOTF is used, which has 993.731 (the ST2084 Y value for code 722) as max value, rather than 10000. As such, Sub-Space with Limit Range is useless. You can see this trivially by running a Characterisation using Virtual Probe.

Problem 3. Volumetric graphs of Sub-Space characterisations appear to have wildly exaggerated errors.

You can see this trivially by running a Characterisation using Virtual Probe:

Target Gamut/EOTF: ST2084 Rec2020
Sub-Space: ST2084 Rec709
Patch Scale/Resolution: Legal 10 bit
Virtual Probe: OETF: ST2084 Primaries: REC2020
Characterisation: Grey, Primary & Secondary Ramp+

Resulting Volumetric graph:


As such, Sub-Space Volumetric graphs appear to be unusable.

The next ColourSpace Update release has major changes in the area of Sub-Space, as well as throughout large chunks of the underlying ColourSpace code base.
(It is a rather major refactoring.)
Consequently, all the above will need to be re-checked when the Open Beta is released.

We will review the feedback then, and assess/fix as required.
Feedback is appreciated.

Steve