AVS Forum Special Member
Join Date: Jan 2003
Location: Portland, Oregon, USA
Mentioned: 61 Post(s)
Tagged: 0 Thread(s)
Quoted: 402 Post(s)
A couple comments on some posts, and news on the next Dynamic Tone Mapping release:
The Radiance Pro CMS only does what it is told. If you have color issues after a calibration then the calibration is suspect. The only calibration software (at this point) that I would recommend as good enough to do a 3D LUT, especially for Rec 2020, is Lightspace. I know Portrait Displays has hired a Color Scientist to improve their 3D LUT Rec 2020 calibration. So I expect Calman will have a much improved 3D LUT capability at come point in the future. My attempts at 3D LUTs with Chromapure did not work out for me.
For the Lumagen Demo Theater I use Lightspace for 3D LUTs. I have also used Calman, but did not have great results for 3D LUTs using Calman. If you have a projector with good Rec 709, and Rec 2020, response to its Gamut limit (such as the RS4500) a 1D LUT may be sufficient for excellent performance.
In the Lumagen theater Radiance Pro I have two different Calman calibrations (1D LUT only, with one each for SDR and HDR) that look good. I also have a full 3D LUT for SDR and another for HDR using Lightspace. I use the 3D LUT calibration on daily basis, and believe it is better than the 1D only calibration.
The advantage of using Lightspace to do 3D LUTs for our RS4500 is that I can set the Color Mode in the RS4500 to Off. This bypasses its internal CMS and in my measurements gives a slightly larger Gamut for Green and Red, but necessitates doing a 3D LUT calibration. I use this for both SDR Rec 709 and HDR Rec 2020. At least in our RS4500 the Rec 709 mode shows a slightly under-saturated Green, but the Color Mode = Off allows me to get a fully saturated Green for Rec 709 using Lightspace 3D LUT.
For reference I use a CR-100 tri-stimulus probe calibrated against Kris Deering's CR-250 spectro-radiometer for the Lumagen RS4500. For the 3D LUTs I did a 21x21x21 profile run (took about 4 hours each since I increased the minimum measurement time).
When we change Tone Mapping, if you see color changes, the issue is in the projector/TV, and/or potentially in the calibration. The Tone Mapping modifies red, green, and blue by exactly the same percentage and so cannot change the color. Of course intensity changes the perception of the color, but this would be a change in perceived saturation as the intensity is changed rather than a change in color (e.g. red versus green).
Precision issues in the projector can also come into play. For example some projectors that are DCI based truncate to 8-bit internally. If you have a dark color (say 30 in the 16 to 235 range used for 8 bit color) a change might affect only one channel. This is due to how "fractional bits" come into play. In the above case if only red changed by one LSB it changes the percentage of red from (30-16)/(235-16) = 6.39% to (29-16)/(235-16) = 5.94%. Then since Green stays at 6.39% (assuming near the gray vector) in the example the percentage change versus red is (6.39%-5.94%)/6.39% = 7%. This artifact of the limited pipeline precision away form red due to the truncation to 8-bits in the projector can be a noticeable color shift.
In contrast the calibration pipeline of the Radiance Pro is 12-bits (and much higher in critical spots) this type of precision error would be no more than 0.44% (1/16 that of 8-bit) and so is very likely not noticeable. In testing with the RS4500 we see no color shifts due to changes in Tone Mapping. Of course we do see changes in the intensity due to the improvements in tone mapping affecting the transfer function.
We have have been testing the next version of Dynamic Tone Mapping, and it is IMO (and others) yet another improvement in the Radiance Pro's DTM. This upcoming release includes a new blending algorithm between dark and bright transfer functions and well as additional improvements in scene detection.
We should have a release with the new DTM, and a couple other improvements, in the next couple of weeks.