Display Calibration

[Siso]

Originally Posted by chros

In addition to what huhn said, this is the official guide:
http://sourceforge.net/p/dispcalgui/...0or%20eeColor/
Try it out at first (I don't know the answer to your other questions.)
What kind of instrument(s) do you have?

And I have posted new result in here, if anybody is interested: http://www.avsforum.com/forum/139-di...l#post39630546
I'm getting closer but still not enough

I'm trying to create a 3dlut for madvr, your results look OK, can you share more info on your attempt to create a good 3dlut for madvr?

Here are some settings from my latest windows calibration:

[chros]

Originally Posted by Asmodian

If you need the black point hack you almost always get crushed shadows.

Originally Posted by markanini

I wonder what implications -IB:x.x:2.4 switch or output offset slider has for stuff level adjustments etc.

Originally Posted by Asmodian

The -IBx.x:2.4 value has a fairly subtle overall effect on ~0.15 cd/m² black level (~1000:1) or lower displays. It is noticeable but only in shadows.

Thanks for the detailed explanation, Asmodian!
Well, the -IB:x.x:2.4 switch depends on which rendering intent do you use! (-i switch) Asmodian used -ims , dispalGUI offers -iaw by default. The difference is the -ims is relative, while -iaw is absolute. They state that -iaw supposed to be more accurate (I didn't had the time to test it on my setup though). But, -IB:x.x:2.4 works the opposite way for these 2 groups: to eliminate the fog effect of -iaw I had to set the slider to 100 (that means IB:1.0:2.4).
About the black crush: I noticed that the cause of it is the "-G2.4 -f0 -k0" switches in the dispcal command. If I leave it out and '-b' switch is also present I won't get black crush.

Originally Posted by Siso

I'm trying to create a 3dlut for madvr, your results look OK, can you share more info on your attempt to create a good 3dlut for madvr?

In dipalGUI, select your correction matrix then follow the tutorial on link link you've posted. https://imouto.my/tutorials/madvr/#3dlut
So far this gives me the best result (with the custom testchart) with 2 modifications on my old CCFL-IPS display:
- on calibration tab: leave Tone curve "as measured"
- on 3dlut tab: set 100% for the black offset (under rendering intent) (you have to enable the advanced settings in the menu, I think)
With this I don't get the black crush (17-19) and there isn't any fog-effect, and thanks to the custom testchart I got the most accurate colors (so far, at least ).

[Siso]

Originally Posted by chros

Thanks for the detailed explanation, Asmodian!
Well, the -IB:x.x:2.4 switch depends on which rendering intent do you use! (-i switch) Asmodian used -ims , dispalGUI offers -iaw by default. The difference is the -ims is relative, while -iaw is absolute. They state that -iaw supposed to be more accurate (I didn't had the time to test it on my setup though). But, -IB:x.x:2.4 works the opposite way for these 2 groups: to eliminate the fog effect of -iaw I had to set the slider to 100 (that means IB:1.0:2.4).
About the black crush: I noticed that the cause of it is the "-G2.4 -f0 -k0" switches in the dispcal command. If I leave it out and '-b' switch is also present I won't get black crush.

In dipalGUI, select your correction matrix then follow the tutorial on link link you've posted. https://imouto.my/tutorials/madvr/#3dlut
So far this gives me the best result (with the custom testchart) with 2 modifications on my old CCFL-IPS display:
- on calibration tab: leave Tone curve "as measured"
- on 3dlut tab: set 100% for the black offset (under rendering intent) (you have to enable the advanced settings in the menu, I think)
With this I don't get the black crush (17-19) and there isn't any fog-effect, and thanks to the custom testchart I got the most accurate colors (so far, at least ).

Will try that these days, and I'll share my results , thank you.

A side note about the windows calibration, is it OK (the pics I had uploaded)?

[chros]

Originally Posted by Siso

A side note about the windows calibration, is it OK (the pics I had uploaded)?

Which pic exactly? I was talking about them.

[Siso]

Originally Posted by chros

Which pic exactly? I was talking about them.

Oh, sorry then Will make the 3dlut tonight to see how the things are

[Siso]

Here are the results from tonight:





black clipping test:

[Asmodian]

Screen shots don't show us anything, they look different on my screen than they do on your's.

[Siso]

Originally Posted by Asmodian

Screen shots don't show us anything, they look different on my screen than they do on your's.

I guess it depends on the calibration, I thought one of the purposes to calibrate your screen is to close match other people's calibrated screens anyway, I don't have visible bars from 17-19 with the 3dlut, without it, they are visible.

[huhn]

a 3d lut creates an image that looks right on your screen and only on your screen.

[raffriff42]

Originally Posted by Asmodian

Monitor calibration methodology has been designed for matching different displays to each other and to print work, not to give the best viewing experience. What makes a good calibration for matching prints is not the same as what provides an optimal experience watching video.

I, and others who's screens I have calibrated, much prefer the calibration when not targeting a white point or (shudder) a brightness. Targeting a brightness with the calibration for viewing video is simply a bad idea. There is no way you can notice a small change in brightness but you reduce saturation on all colors in order to target a brightness, which reduces contrast too. The only way to reduce brightness with a calibration is to have the video card output 254 or less when the source pixel's value is 255, this requires lowering all values to avoid banding, reducing the dynamic range for no visible benefit.
[...]
This may give the impression that I think calibrations are somewhat useless but the calibrated gamma curve is generally much better, as long as you do not get banding, and tuning the white point with the meter and the monitor's controls often helps a lot.

Good points, Asmodian. It seems to me any unnecessary altering of an 8-bit video signal should be avoided. I am surprised calibration is not always done at the monitor, and in the analog domain - ie DAC gain & offset. This is preferable I think - except, as you say, for print work and such:

https://en.wikipedia.org/wiki/Color_calibration#Display
Depending on the calibration software and type of monitor used, the software either creates a correction matrix (i.e. an ICC profile) for color values before being sent to the display, or gives instructions for altering the display's brightness/contrast and RGB values through the OSD.

Originally Posted by Asmodian

Screen shots don't show us anything, they look different on my screen than they do on your's.

Yep! Another thing, calibration at the GPU messes up the colors if you are capturing the screen.
https://www.google.com/search?q=fraps+yellow+tint

[chros]

Originally Posted by Siso

Here are the results from tonight

But with which settings?

Originally Posted by huhn

a 3d lut creates an image that looks right on your screen and only on your screen.

Originally Posted by Asmodian

Screen shots don't show us anything, they look different on my screen than they do on your's.

Originally Posted by raffriff42

Yep! Another thing, calibration at the GPU messes up the colors if you are capturing the screen.
https://www.google.com/search?q=fraps+yellow+tint

It's true but it can show us mayor flaws like elevated shadows, etc.

[seiyafan]

Is 3d LUT same as .icc profile? My question is if I create a 3d LUT for MadVR, can I also use that profile on the same monitor for photo editing? Will all the colors look accurate for both movies and photos?

[Arm3nian]

I ended up returning the VP2780-4k. I wasn't happy with it. Cost $100 to ship it all the way to New York from Las Vegas but it's better than paying $700+ for something that doesn't impress you.

I ended up getting the PG279q. A lot of units have QC issues, but they seem to be getting better. Mine doesn't have anything serious, but it does have some uniformity issues. They're all over the monitor when you look for it on lets say a red background. The most noticeable is on a white background when on web pages, where the top 30% looks significantly more dim than the rest.

I don't think the PG279q comes factory calibrated, but after my i1 pro and dispcal calibration, the difference was obvious. A single windows calibration looked better than all the profiles and 3d luts I made with the previous display. The contrast is also really good, 1200+ after calibration, and blacks look nice for an IPS.

The gaming features are a little gimmicky imho. The display goes up to 165Hz, but it's hard to notice much difference after 120Hz, maybe even before it. ULMB pwm flicker just makes my head hurt after 5 min of use, and has ridiculous ghosting, making it unusable. Gsync is cool, but it's badly implemented on the software side. Blizzard games for examples don't work properly with it. For diablo 3, you need to enable vsync in game and turn off fps cap for it to work, which causes micro stuttering .
In WoW, the fps limiter in game is completely broken, and 3rd party doesn't work correctly with the engine. My 980ti can easily maintain 165fps in a specific area (and the game in general), but it drops to 128-153 randomly in the same exact area with the gpu at 65% load and the cpu at 23%. In more demanding games, you don't get a huge stutter, but I can still feel choppiness when it drops to 60's. That's the biggest problem with high refresh rate. After 165Hz, 60Hz is disgusting until your brain adjusts. 1440p is nice. I sit about about a foot away, and often can notice the pixels, where it was impossible to on 4k. But at least I don't have to deal with scaling issues.

I have until January 30th to return it. I should really focus more on my engineering coursework rather than games, so it'll probably be going back if I get free shipping.

Back on topic, I didn't use any of the fancy features in dispcal. I targeted a specific white level and white point to help me set it as desired, then went back and changed to as measured. The windows calibration looks great. From that, I made the 3D lut, and it looks good in movies. I'm guessing it is better to create a new profile for madVR and target Rec. 1866 instead of 2.2 gamma? It feels weird for the tone curve to be gamma 2.2, and then the 3D lut to be set to Rec. 709? Probably worth it to make the new profile.

[e-t172]

Originally Posted by seiyafan

Is 3d LUT same as .icc profile? My question is if I create a 3d LUT for MadVR, can I also use that profile on the same monitor for photo editing? Will all the colors look accurate for both movies and photos?

A 3DLUT is not an ICC profile.

What ICC-aware applications (e.g. Photoshop) do is, they use the ICC profile that's set in the Windows color control panel applet to generate a 3DLUT on the fly and use that to display accurate colors. Non-ICC aware applications don't know how to do that and therefore don't use the ICC profile. However they are still subject to the 3x1DLUT calibration that's stored (vcgt) in the ICC profile which is applied to the entire output of the GPU, making things especially confusing.

Unfortunately, madVR is not ICC aware (I've tried to convince madshi to implement ICC, but to no avail). However, it can load "raw" pre-generated 3DLUTs, so if you generate a 3DLUT from your ICC profile, you can achieve the same end result as an ICC-aware application. The only difference is that it has to be done manually - madVR is not smart enough to deduce the 3DLUT from the ICC profile set in the Windows options.

[huhn]

windowed overlay reads the ICC and ignores the gamma ramp by default. i don't know what it reads from it.

it's nice if you can read an ICC and use this but there is a good reason collink needs minutes or even up to hour to create a 3D LUT.

[seiyafan]

I see, so it looks like I need to calibrate my monitor twice, one for photo (2.2) and the other for madVR (Rec. 1866), because the .icc generated for Rec. 1866 would not work well for photos, I just need to get the 3D LUT file generated to be imported into madVR, and the .icc calibrated for 2.2 gamma would be loaded into Windows. Is this logic correct?

Another question, since my monitor is not wide gamut, should Rec. 709 be used instead of 1866? Excuse my n00bness, I am not good with the technical jargon.

[huhn]

for a 3D LUT only the readings matter.

so you can create a bt 709 profile from the start and use this to create a 3D LUT.

BT 1886 has the same gamut as bt 709 it's just a "different" gamma depending on the "black" your screen has.

[seiyafan]

I see, I have an EIZO IPS with 1000:1 contrast ratio, but nowhere near as black as any OLED. Is it still okay to go with BT 1886?

[e-t172]

Originally Posted by huhn

it's nice if you can read an ICC and use this but there is a good reason collink needs minutes or even up to hour to create a 3D LUT.

3DLUTs can be cached. Also, how do you manage to make collink run for an hour? That's crazy. On my machine it usually runs in seconds. Are you using "high quality" (-q h) options despite the Argyll documentation explicitly telling you not to do that?

Originally Posted by seiyafan

I see, so it looks like I need to calibrate my monitor twice, one for photo (2.2) and the other for madVR (Rec. 1866), because the .icc generated for Rec. 1866 would not work well for photos, I just need to get the 3D LUT file generated to be imported into madVR, and the .icc calibrated for 2.2 gamma would be loaded into Windows. Is this logic correct?

Technically, you can calibrate your monitor to anything you want; anything you do will be compensated back when you generate a 3DLUT, or when your photo software reads the monitor's ICC profile. Calibration is just a way to get the monitor to behave nicely with applications that are not ICC-aware; what matters for madVR or ICC-aware software is correct monitor characterization (i.e. profiling).

Technically, calibrating as close as possible to the actual colorspace of your content might increase accuracy and reduce banding, but by how much is up for debate.

Originally Posted by seiyafan

Another question, since my monitor is not wide gamut, should Rec. 709 be used instead of 1866? Excuse my n00bness, I am not good with the technical jargon.

Rec. 1866 is not a gamut, it's a gamma curve. Rec.709 is a gamut. Using Rec. 1886 "instead of" Rec. 709 makes no sense, because they are complementary. You use Rec. 1886 with Rec. 709.

Originally Posted by seiyafan

I see, I have an EIZO IPS with 1000:1 contrast ratio, but nowhere near as black as any OLED. Is it still okay to go with BT 1886?

Yes. In fact, that's one of the reasons why BT.1886 was introduced in the first place (i.e. achieve correct results in displays with imperfect black).

[huhn]

Originally Posted by e-t172

3DLUTs can be cached. Also, how do you manage to make collink run for an hour? That's crazy. On my machine it usually runs in seconds. Are you using "high quality" (-q h) options despite the Argyll documentation explicitly telling you not to do that?

default dispcal gui settings take about 4-5 min.

-r 256 can take a lot longer time easily an hour. if this is like -qu than don't use it but there is no word about not using high.