or Connect
AVS › AVS Forum › Display Devices › Display Calibration › MadVR - ArgyllCMS
New Posts  All Forums:Forum Nav:

# MadVR - ArgyllCMS - Page 25

Quote:

My recommendations are now mostly on the first post of this thread. With only one exception: I recommend "dispcal -qu", at least if you have a fast meter (e.g. i1 Display 3 Pro).

Please note that what I recommend is just what worked best for *me*. That doesn't mean that the same settings will work as well for anybody else. Although those who have tried seem to have had good success with my recommendations.

I see, but unfortunately the first post doesn't include the .ccss examples of correction files. But no problem, will search for them myself.

Done the calibration. I just wished there was a way to use the calibration without raising my Black Levels. My Dell U2410 already has very poor Black Level, with the .3dlut it gets even worse.
Edited by Alec246 - 9/30/13 at 3:15pm

### AVS Top Picks

Quote:
Originally Posted by Kukulcan

Relative or Absolute colorimetric for collink? I agree with madshi, absolute intent! And now I try to explain why Graeme Gill may be "wrong" suggesting relative...

If the white point after the calibration curves are applied is the correct target (D65 in this case), and the source and destination viewing conditions are the same, then there should be little or no difference between absolute and relative intent. If there is, then it would be worth investigating why. I'd also be interested if you think there is any difference between -ial and -irl.

The disadvantage of absolute intent is that you risk gamut clipping if the white point drifts between calibration and profiling. Given that there should be no difference otherwise between the intents, there is no reason to choose absolute.
Just to add another datapoint, I found that differences between -ir and -iaw greatly decreased after averaging several .ti3 files together. So if you're seeing a big difference, that might be caused by noise in the measurements (presumably because the noise negatively affects the fit, more than because the measurements are entirely wrong).

Incidentally, I only really see a difference on my desktop monitor, which has a wider gamut than my laptop's monitor (which seems to make calibration more difficult).
Edited by VerGreeneyes - 10/1/13 at 7:43am
I have (once again) a weird problem:
When using
Code:
dispcal.exe -v -dmadvr -Yp -y7 -qm -m -w 0.3127,0.3290 -gs -o output\TVmtx.icm output\VideoLUT > logs\2.Calibrate.VideoLUT.log

my black levels get all messed up.

madVR is configured to 16-235, the graphics card outputs 0-255 and the TV is expecting 16-235 (not changable).

I wanted to see what dispcal did so far and applied the icm file: Steps below 16 are visible and the white balance isn't good at all. If I change the configuration to madVR 0-255, gfx 16-235 the black levels are all right, and the greyscale looks pretty good with a weird anomaly at 40%.

Does anybody have an idea of what's happening there?

I will do another calibration run with the changed settings, but it takes some time (Colormunki Display).

edit 1: the run with the changed settings didn't change anything, although I didn't try to use it with original settings

edit 2: I'm now running
Code:
dispread.exe -v -dmadvr -Yp -y7 -K output\VideoLUT.cal output\2000_g128_s32 > logs\4.Read.Pattern.Chart.log


The "-y7" switch should activate the plasma spectral sample, and in the log for dispcal it said "Display type is 7". Dispread says no such thing. How do I know if the sample is active?
Edited by monvo - 10/1/13 at 9:21am
@monvo, when using madVR/madTPG as a test pattern generator, ArgyllCMS relies on madVR to take care of TV vs PC levels. Basically dispcal produces a PC levels icm file in that situation, I think. So if your GPU is set to 0-255 and you apply the icm file produced by dispcal/madTPG, you get 0-255 output to your display.
Quote:

@monvo, when using madVR/madTPG as a test pattern generator, ArgyllCMS relies on madVR to take care of TV vs PC levels. Basically dispcal produces a PC levels icm file in that situation, I think. So if your GPU is set to 0-255 and you apply the icm file produced by dispcal/madTPG, you get 0-255 output to your display.

Thanks, so if I understand you correctly, I should be able to use the resulting 3dlut regardless of the other settings?

Now that I have finished the process, I have another problem, but first the command lines for the entire thing:
Code:
dispcal.exe -v -dmadvr -Yp -y7 -qh -m -w 0.3127,0.3290 -gs -o output\TVmtx.icm output\VideoLUT > logs\2.Calibrate.VideoLUT.log
targen.exe -v -d3 -G -f750 -g128 -s32 -e16 -coutput\TVmtx.icm output\2000_g128_s32 > logs\3.Create.Pattern.Chart.log
colprof.exe -v -qh -aX output\2000_g128_s32 > logs\5.Create.ICC.Profile.log
collink.exe -v -3m -qh -et -Et -Ib -G -ir -a output\VideoLUT.cal ref\Rec709.icm output\2000_g128_s32.icm 3DLUT\2000_g128_s32.ir.b22.icm > logs\6.Create.3DLUT.log


Everything looks pretty good with the 3dlut, except for the delta luma for blue, which has a whopping +25,9% at 100%sat/amp. I have the same problem without the 3dlut on my pc as well as on my bluray player and with the different icm files that are created in the process. This brings dE to 4.1 at that spot. What could be the reason for this?

I've tried -ir as well as -iaw, the problem persists while everything else stays basically the same.

nvm, got it :)

Edited by YxPi - 10/2/13 at 11:39am
Can be disregarded, after a second try the problem doesn't present itself anymore. Warning: Spoiler! (Click to show)
Ok, so I've run into an even weirder problem than before. Again, the steps I took were:
Code:
dispcal.exe -v -dmadvr -Yp -y7 -qh -m -w 0.3127,0.3290 -gs -o output\TVmtx.icm output\VideoLUT > logs\2.Calibrate.VideoLUT.log
targen.exe -v -d3 -G -f750 -g128 -s32 -e16 -coutput\TVmtx.icm output\2000_g128_s32 > logs\3.Create.Pattern.Chart.log
colprof.exe -v -qh -aX output\2000_g128_s32 > logs\5.Create.ICC.Profile.log
collink.exe -v -3m -qh -et -Et -Ib -G -ir -a output\VideoLUT.cal ref\Rec709.icm output\2000_g128_s32.icm 3DLUT\2000_g128_s32.ir.b22.icm > logs\6.Create.3DLUT.log

The problem is this (and it is the same with -ir and -iaw):
With 3dlut:

Without 3dlut:

There's obviously something very wrong in the first picture, and I can't really make sense of why that is. Although the blue luminance is off at 100%sat, it shouldn't have this effect. Here are all the measurements I've got. I took the measurements with HCFR using my Colormunki Display and MadTPG as a generator: Warning: Spoiler! (Click to show)
CIE:

Gamma:

Primaries and secondaries data: (Blue luminance is way off)

RGB Levels:

Saturation Luminance:

Saturation Shifts:

Any help would be greatly appreciated!

edit: I've added the calibration logs for anyone interested.

Edited by monvo - 10/2/13 at 8:09am
Now I have a pretty good 3dlut, but the black level is considerably elevated. I can use the brightness controls on my plasma to get it back to normal, but then the <10% of my greyscale suffer, obviously gamma at the low end, as well as all the color luminances at low saturations (0-50%). The elevated black level is already visible in the first icm that dispcal produces and translates into all subsequent icms and the 3dlut. Is there an easy fix? Something that I could edit myself in the 3dlut? These are my steps and they took almost 3 hours to complete, I would hate to have to go through them again:
Code:
dispcal.exe -v -dmadvr -Yp -y7 -qu -m -w 0.3127,0.3290 -gs -o output\TVmtx.icm output\VideoLUT > logs\2.Calibrate.VideoLUT.log
targen.exe -v -d3 -G -f2000 -g128 -s32 -e16 -c output\TVmtx.icm output\2000_g128_s32 > logs\3.Create.Pattern.Chart.log
colprof.exe -v -qh -aX output\2000_g128_s32 > logs\5.Create.ICC.Profile.log
collink.exe -v -3m -qh -et -Et -IB -G -iaw -a output\VideoLUT.cal ref\Rec709.icm output\2000_g128_s32.icm 3DLUT\2000_g128_s32.iaw.IB.b22.icm > logs\6.Create.3DLUTiaw.IB.log


-ir or -iaw as well as -IB or -Ib make no difference in this.

Something that got my attention is that after the initial 6 color patches in dispcal the first 10 or so of the first 128 patch run are black. Are they supposed to be? It doesn't make much sense to me, especially since essentially no meter can measure this far down anyway. So even if there was black crush on my display and dispcal would try to account for that by raising the black level, I don't think virtually anyone herehere posses a meter that would be good enough to measure these low stimuli.

edit: I just realized, that something completely different weird is happening:
The colors go crazy above 90% white. Although my pc is set to output 16-235, steps above 235 appear when i look at contrast patterns. Also they are (light)blue. They disappear as soon as I disable the 3dlut and everything behaves as expected, 235 being the whitest and everything above that cut off. I have no idea, how that can even theoretically happen. When I use the icc profile created with colprof, this problem isn't there, and the blacklevel is better, but not good. Also the colors aren't corrected in the icc profile.
Edited by monvo - 10/2/13 at 9:39am
Which controls have you set to 0-255 and which to 16-235 exactly? And have you left this controls the same way during measurements and video playback or did you change them somewhere on the way? From what you're describing (black level raised + WTW visible) it sounds like there's a double expansion going on somewhere. Please try clearing the VideoLUTs manually by using "dispcal -c". Does that make any difference?

I'm not sure how dispcal handles the situation if the black level is lower than your meter can measure. That would be a question for Graeme.
Quote:

Which controls have you set to 0-255 and which to 16-235 exactly? And have you left this controls the same way during measurements and video playback or did you change them somewhere on the way? From what you're describing (black level raised + WTW visible) it sounds like there's a double expansion going on somewhere. Please try clearing the VideoLUTs manually by using "dispcal -c". Does that make any difference?

I'm not sure how dispcal handles the situation if the black level is lower than your meter can measure. That would be a question for Graeme.

At the moment I have madvr set to 0-255, the gpu set to 16-235 and the tv (unchangably) set to 16-235. It was this way through the whole process. What bugs me about this is, that the gpu shouldn't even send anything above 235, and in my experience doesn't. I can't clear the VideoLUTs at the moment, but I'm confident that they were cleared. I will however try this as soon as I'm finished with the current dispcal run.

In the dispcal documentation it says, that it will try to correct the hue of the black point. Furthermore it says that it can happen, that this elevates black levels. Luckily there is a switch to turn that off and I'm doing another dispcal run at the moment to see if it works.

edit: -k doesn't make a big difference and I've cleared the VideoLUTs with dispwin -c and the problem is still there.

I really don't get it, even if my gpu sends YCbCr (confirmed by my AVS) the steps above 235 appear, and while the backround is perfectly white, the steps are blue. Here's a picture:

Again, Madvr is set to 0-255, gpu 16-235, TV 16-235.

With a 3dlut I created yesterday everything seems normal in that department, but it has other problems, which I managed to solve through simple reruns. It seems that everytime I do another run the results end up vastly different, although I didn't change anything.

@Graeme: Could it be that the -k switch in dispcal doesn't work? After putting in -k 0.0 the log still contains this:
Code:
Adjusted target black XYZ 0.01 0.01 0.02, Lab 0.07 -0.01 -0.08
Target black after min adjust: XYZ 0.009 0.009 0.017, Lab 0.072 -0.009 -0.079


On the other hand the resulting .cal file says
Code:
KEYWORD "BLACK_POINT_CORRECTION"
BLACK_POINT_CORRECTION "0.000000"


But I definitely see dithering (I have a plasma) if I apply the icm profile that wasn't there before.

I would just like the programs to leave my black alone.
Edited by monvo - 10/2/13 at 11:25am
If you've setup your GPU to output 16-235, but still see the bars above 235 on your display then something is wrong either with the GPU behaviour or with the display settings. Are you testing this with madVR in windowed, overlay or fse mode? FWIW, in overlay mode sometimes levels can be unexpectedly different.
Quote:
Originally Posted by monvo

Something that got my attention is that after the initial 6 color patches in dispcal the first 10 or so of the first 128 patch run are black. Are they supposed to be? It doesn't make much sense to me, especially since essentially no meter can measure this far down anyway. So even if there was black crush on my display and dispcal would try to account for that by raising the black level, I don't think virtually anyone herehere posses a meter that would be good enough to measure these low stimuli.
The whole point is to characterize the display. So dispcal can't know how the display behaves without trying out a whole range of device values, some of which may turn out to be quite black. And yes, if the display has a "flat spot" from black the idea is for the calibration to skip that bit, so that the calibrated response is progressive from the minimum. And yes, good meters will be able to measure quite low light levels (ie. a JETI, i1d3, dtp92, dtp94, i1pro, ColorMunki spectro, spectrolino etc.). Yes, cheaper & older meters may struggle, and it is hard to get reliable readings at some levels, but there is no way to simply ignore such issues. If someone has a display with very good blacks, then they expect to be able to use them!
Quote:
Originally Posted by monvo

@Graeme: Could it be that the -k switch in dispcal doesn't work? After putting in -k 0.0 the log still contains this:
Code:
Adjusted target black XYZ 0.01 0.01 0.02, Lab 0.07 -0.01 -0.08
Target black after min adjust: XYZ 0.009 0.009 0.017, Lab 0.072 -0.009 -0.079


I don't see any problem there. The first line shows the target to 2 digit precision, while the second shows it to 3. Round the second line to 2 digits and you get precisely the first line values.
Quote:
I would just like the programs to leave my black alone.

There is a bug in the current release of dispcal that may worsen the black level slightly. You could try the one in http://www.argyllcms.com/dispcal_win.zip and see if that improves the situation or not.
Quote:

If you've setup your GPU to output 16-235, but still see the bars above 235 on your display then something is wrong either with the GPU behaviour or with the display settings. Are you testing this with madVR in windowed, overlay or fse mode? FWIW, in overlay mode sometimes levels can be unexpectedly different.

I've tried it both in fse and windowed mode. I doubt that it is the gpu or the TV, as I've said as soon as I disable the 3dlut or take an older one everything goes back to normal. I'm just gonna try another run today and hope the problem goes away.
Quote:
Originally Posted by gwgill

I don't see any problem there. The first line shows the target to 2 digit precision, while the second shows it to 3. Round the second line to 2 digits and you get precisely the first line values.
There is a bug in the current release of dispcal that may worsen the black level slightly. You could try the one in http://www.argyllcms.com/dispcal_win.zip and see if that improves the situation or not.

Thanks, I'll try that.
Quote:
Originally Posted by gwgill

The whole point is to characterize the display. So dispcal can't know how the display behaves without trying out a whole range of device values, some of which may turn out to be quite black. And yes, if the display has a "flat spot" from black the idea is for the calibration to skip that bit, so that the calibrated response is progressive from the minimum. And yes, good meters will be able to measure quite low light levels (ie. a JETI, i1d3, dtp92, dtp94, i1pro, ColorMunki spectro, spectrolino etc.). Yes, cheaper & older meters may struggle, and it is hard to get reliable readings at some levels, but there is no way to simply ignore such issues. If someone has a display with very good blacks, then they expect to be able to use them!

I know they can measure quite low levels, I have the ColorMunki display and it measures 5% grey somewhat reliably. But I wouldn't really trust it beyond that, which is my point. I own a 2012 panasonic plasma and black is pretty black.
Quote:
Originally Posted by gwgill

If the white point after the calibration curves are applied is the correct target (D65 in this case), and the source and destination viewing conditions are the same, then there should be little or no difference between absolute and relative intent. If there is, then it would be worth investigating why. I'd also be interested if you think there is any difference between -ial and -irl.

The disadvantage of absolute intent is that you risk gamut clipping if the white point drifts between calibration and profiling. Given that there should be no difference otherwise between the intents, there is no reason to choose absolute.

Thanks Graeme! Actually I was curious to let you think about my idea of giving up something at 100IRE white in order to get more neutral greys below. But it was just an intuition and I'm not really an expert in digital color science, so I take for good your suggestions.

I made other attempts but, as VerGreeneyes suggested, I probably have severe problems of inconsistent behaviour of my display (the meter is iD3), it's a cheap lcd Samsung not an Eizo Color Graphics....

I'll try some other tests with my vpr Planar, a DLP that behaves much more better and consistently than most LCDs, but since it's a vpr, it's much more annoying and time consuming making these kind of test (darkening the room, tripod positioning...)
@Graeme
Out of curiosity, if no -IB or -Ib parameter is specified with a collink.exe command, what is applied to the resulting 3DLUT?
@Graeme:
I've tried the new dispcal and it thankfully resolved the issue at this step of the process, but, if I remember correctly, still only after using "-k 0.0". Sadly in either the dispread, colprof or collink step the black level got raised again. But this might have to do with the number of patches I ran (only 250 for testing), and was easily resolved by turning the brightness control down one step. This didn't mess up the bottom part too much. Otherwise the result I got with only dispcal -qm and 250 patches is great!

I think the blue issue was a result of the display clipping blue at the high end and the 3dlut trying to account for that. However I still can't figure out how the steps above 235 got displayed in the first place. Could it be that the madvr with a 3dlut applied can override the output settings of the gpu at that fundamental level? If so, why doesn't it do so with other 3dluts?
Quote:
Originally Posted by N3W813

@Graeme
Out of curiosity, if no -IB or -Ib parameter is specified with a collink.exe command, what is applied to the resulting 3DLUT?
Nothing is applied. The BT.1886 is an extra step applied between the source profile and the destination profile which modifies the apparent source profile. Something is typically needed because video colorspaces are encoding definitions, not output definitions. -IB/b transforms the encoding definition to an output definition. In principle something similar can also be done using the viewing condition settings in combination with CIECAM02 based gamut mapping, but in practice BT.1886 seems to be closer to what's expected in the video world.

Hi.

I am trying to replicate with madVR 3dlut what a typical .ICC profile would do in MPC-HC or any other Color Managed software.

No matter what I tried I can't get the same gamma curve, with just colors being fixed.

The 3dluts I get are always too bright.

Is there any way to implement this 3dlut as just a Gamut fix (like .Icc would) without destroying the gamma curve?

Never mind I have managed.

Rec709_gamma22.icm did the trick (X:\dispcalGUI-1.2.7.0\ref).

Again and again and again....

Rec709 is for Encoding ONLY !!!!

Somehow the Rec709.icm has the encoding Gamma in it which should NOT be used in a Display profile.

Use this for 2.2:

collink.exe  -v -3m -qh -et -Et -G -a "Measured.cal"  -iaw  Rec709_gamma22.icm  "Measured.icm"  "3DLUT.3dlut"

This for BT.1886 (It works only with Rec709.icm):

collink.exe  -v -3m -qh -et -Et -G -IB -a "Measured.cal"  -iaw  Rec709.icm  "Measured.icm"  "3DLUT.3dlut"

This for calibrated Gamma (changes only color/gamut):

collink.exe  -v -3m -qh -et -Et -G -a "Measured.cal"  -iaw  "Measured.icm" "Measured.icm"  "3DLUT.3dlut"

Yes the same .icm file.

Works wonders for wide gamut monitors that cover 100% sRGB/Rec709 Color.

Cheers.

Edited by James Freeman - 10/4/13 at 7:26am
Quote:
Originally Posted by James Freeman

Hi.
The 3dluts I get are always too bright.

Me too ... I have abandoned argyllcms the same reason.
Quote:
Originally Posted by James Freeman

Never mind I have managed.

Rec709_gamma22.icm did the trick.

Also, -iaw should be -ir you will get your blacks back  (no elevated blacks).

Use this:
collink.exe  -v -3m -qh -et -Et -G -a "Measured.cal"  -ir  Rec709_gamma22.icm  "Measured.icm"  "3DLUT.3dlut"

Can you / anyone explain why you get a better result with Rec709_gamma22.icm
Because the effective gamma curve for r709 is 1.95. (Best fit power law)
Quote:
Originally Posted by zoyd

Because the effective gamma curve for r709 is 1.95. (Best fit power law)

It does not matter what effective gamma it is...

You should never calibrate you display device to 709.

In fact, forget about 709 for Gamma completely.

Rec709_Gamma22 means:

Rec709 color space with 2.2 Power curve gamma, which IMO is ideal.

Edited by James Freeman - 10/4/13 at 6:43am
Quote:
Originally Posted by James Freeman

It does not matter what effective gamma it is...

You should never calibrate you display device to 709.
In fact, forget about 709 for Gamma completely.

Rec709_Gamma22 means:
Rec709 color space with 2.2 Power curve gamma, which IMO is ideal.

Everybody recommends using -IB or -Ib(Ib:2.2, Ib:2.3, Ib:2.4) with collink, both of which give BT.1886 gamma as you have said yourself. There's not much sense in ignoring the recommendations and then complaining about gamma.
Edited by monvo - 10/4/13 at 7:20am

I will post it again just in case.

Use this for Gamma 2.2:

collink.exe  -v -3m -qh -et -Et -G -a "Measured.cal"  -iaw  Rec709_gamma22.icm  "Measured.icm"  "3DLUT.3dlut"

This for BT.1886 (It works only with Rec709.icm):

collink.exe  -v -3m -qh -et -Et -G -IB -a "Measured.cal"  -iaw  Rec709.icm  "Measured.icm"  "3DLUT.3dlut"

This for your calibrated Gamma (changes only color/gamut) (Best option IMO):

collink.exe  -v -3m -qh -et -Et -G -a "Measured.cal"  -iaw  "Measured.icm" "Measured.icm"  "3DLUT.3dlut"

Yes, the same .icm file.

Works wonders for wide gamut monitors that cover 100% sRGB/Rec709 Color.

Hope this help.

Edited by James Freeman - 10/4/13 at 7:44am
Quote:
Originally Posted by monvo

Everybody recommends using -IB or -Ib(Ib:2.2, Ib:2.3, Ib:2.4) with collink, both of which give BT.1886 gamma as you have said yourself. There's not much sense in ignoring the recommendations and then complaining about gamma.

With all due respect.

I did not write collink.exe, Graeme Gill did.

I don't know the reason but for BT.1886 it wants only Rec709.icm, anything else results in crashed blacks (I have tested it).

In the post right before your first Graeme says that IB/Ib is needed for BT.1886. I really don't understand where your problem is.
Quote:
Originally Posted by monvo

In the post right before your first Graeme says that IB/Ib is needed for BT.1886. I really don't understand where your problem is.

If you read carefully my posts you'll see I haven't typed a single bad word about IB/Ib.

I was talking about -i which is a completely different animal.

I edited to remove the -i stuff because it had no effect (after I was quoted).

The problem WAS retaining nice Gamma curve, but now its gone, thank you very much.

Quote:
Originally Posted by James Freeman

collink.exe  -v -3m -qh -et -Et -G -a "Measured.cal"  -iaw  "Measured.icm" "Measured.icm"  "3DLUT.3dlut"
I tried this on my profile, but I got the following error:
Code:
collink.exe: Error - rspl.radial: failed to find radial triangle

New Posts  All Forums:Forum Nav:
Return Home
Back to Forum: Display Calibration

### AVS Top Picks

AVS › AVS Forum › Display Devices › Display Calibration › MadVR - ArgyllCMS