Any way to make a MadVR 3DLUT Profile that uses 0-255 RGB instead of the TV RGB?

[Darkmatter]

Hi,

I wanted to know if there was any way to make a MadVR 3DLUT Profile that uses 0-255 RGB instead of the TV RGB? I know MPC has a filter that can convert the two but didn’t really want to use the CPU/GPU power to do it that way unless I had no choice. I also don’t know if doing that degrades the quality of the image. I will often watch a show on my monitor but I use MadVR so using a truncated color gamut isn’t needed or desired (as far as I am aware, although Florian has set me straight on more then one occasion. lol)

I was also wondering about MadVR calibration. At the moment I am telling MadVR that my monitor is already calibrated so it doesn’t reset the calibration when a video loads. How is telling MadVR that a 1DLUT icc/icm profile (made with XYZ + Matrix) is already active different from pointing MadVR specifically to a 3DLUT made specially for MadVR?

Last question for MadVR. You mention madTPG in your notes for this release when talking about MadVR, as well as needing a special (linked) version of ArgyllCMS if you use Windows as your OS. Is this something I should be using if I am using MadVR? If yes, unless you have links ready to paste into a reply, or unless you know there isn’t anything easy to understand online, just let me know if it is important or not and I’ll see what I can get from Google before bothering you with having to explain something that is probably explained 100 times already online.

Also, when calibrating a Sony Bravia KDL-50W790B, I wanted to know what affect using 0-255 RGB, instead of TV’s 16-235 RGB? Also, whether I should calibrate it to a specific gamma that best matches the lighting in my room, or should I use sRGB, or should I use Rec709? This is a non-HDR display FYI.

Thanks!

[Darkmatter]

Either I’m blind or there is no edit feature, or it has timed out. Chances of blindness… 68.7%. lol

I just wanted to add that I just read that MadTPG is for making MadVR Profiles so I now know what it does, I just don’t know if there’s a way to make it use 0-255 RGB.

Sorry but I still have a few other areas of your program that I still don’t know anything about.

1. Blackbody Locust vs White Point. I can’t find on the main page’s extensive description of features, what exactly a Blackbody Locust is, and how it is different from a White Point, and which is better.
2. I still don’t know why or when I should or would use a simulation profile when doing a verification check of an installed profile.
3. What exactly are multidimensional patches, how do they work, and how many should I have for an XYZ + Matrix profile?
4. If I use an XYZ + Matrix profile type is there any point in doing a 2000 or 4000 or whatever huge number of patches profile calibration? Is the difference in accuracy worth it?
5. Greyscale is a huge deal from what I’ve read of your site, so I am assuming that making a profile with all 256 points would give the best accuracy. Would I lose much by going to 128?
6. If I want to make an slowest calibration speed XYZ + Matrix profile that more or less hits every important, semi-important, and edge of gamut points, how many patches would I need for a non-HDR, 8 bit monitor? Would I use the Auto-Optimized feature or set it up manually? If manually, I would need a general idea of how many grey, iterative, and multidimensional patches to use. I generally use 2 or 3 Black, White, and R,G,B, patches so I get more then one measurement during the test.
7. I have no idea when or why I would use a preconditioning profile but I fully admit that there is a lot about the profile editor that I don’t know.
8. Should I use “Optimized Farthest Point Sampling” or another setting?
9. Lastly, (I hope, and I’m sure you do too…) Does it matter what sorting order I choose? I understand that “Minimize Display Response Delay” is slightly faster but I’ve been using either “Maximize RGB difference” or “Vary RGB Difference.” I don’t even really know what “Interleave” or “Shift & Interleave” do.

Sorry, that was meant to be 2 questions. I first had to change it from “a couple” to, “a few.” I probably should change it again to “a lot” but I’m afraid that’ll scare you away! lol

Thanks,

DM

[Vincent]

madVR works with 16-235. Once processed you output to whetever your display support limited or full range, so there is no need to do what you ask for.
Anyway if you wish to use it in the wrong way, use command line argyll to get LUT3D file and set manually full range in ful range out.

“Locus”, not “locust”, meaning “place”.
-Blackbody locus is a curve (in a CIE xy plot or equivalent 2D plot) with all posible “colors” of an idealized incandescent object (short explanation). Since it is a physics concept you’ll find extensive information on the web.
When you heat something the peak of radiation starts glowing red (low temperature in kelvin degrees), then orange, then white, then bluish (extremely high temperatures). So you can take it as some model of idealized light sources.
-Dayligt locus is something close blackbody but tunned to match Sun’s star light + Earth’s atmosphere glow. Very close to blackbody but moved a little towards green.
Since you see the plot in DIsplaycal profile info app, try it.
You’ll see both curve segments run almost parallel at ~4dE from each other.
Computer displays and TVs meant for movies, internet & games usually aim for D65, daylight curve (not blackbody), 6500K (6504) correlated temperature. An intersection between daylight curve and isotherm 6500K (curve) segment.

People use simulation profiles when:
-want to test TV/monitor inner calibration (use simulation profile as display profile) without GPU calibration
-want to test TV/monitor calibration while working with color managed apps (do not use simulation profile as display profile) , for example Photoshop.

Think of patch number as nodes in a 3D cube mesh. Cube edges are R,G, B axis.
1000 patches would result a 10 step edge division. That is the resolution of that “cube” = 10. 10x10x10=1000.
Usually a lot of profesional apps  use 17x17x17 cube luts, so… try to figure out the number of patches needed to capture your calibrated/uncalibrated display behavior with 17 step resolution.
You may want to use higher step resolution for cube even you do not have enough patches because you want to calculate inter step values in a explicit way.

1DLUT for GPUs have 256 entries at n bits (usually 16bit). There is no point storing it at 128.
ArgyllCMS (DisplayCAL) could measure and correct up to 96 entries in 256 grey ramp while computing 1DLUT calibration. ~3 RGB value step between measured corrections. AFAIK this is the most accurate 1DLUT correction out there.
256 entries for 1DLUT calibration are computed from that set of corrected measurements (up to 96) in a smooth way between those measurements (interpolated values).

I understand that “Minimize Display Response Delay” is slightly faster but I’ve been using either “Maximize RGB difference” or “Vary RGB Difference.” I don’t even really know what “Interleave” or “Shift & Interleave” do.

They are meant to solve issues with some technologies like WOLED consumer TVs. In those displays you do not want that the same or close color stays too much time on screen in order to prevent automatic brightness level  kicking in (darkening the patch) and such.
For typical consumer displays without those issues 1st one works good.

• This reply was modified 7 years, 7 months ago by Vincent.

[Author]

Posts