Home › Forums › Help and Support › 3D LUT creation for madvr – posterization near black / greyscale
- This topic has 13 replies, 4 voices, and was last updated 4 years, 2 months ago by S Simeonov.
-
AuthorPosts
-
2020-01-24 at 11:02 #22410
Hello,
I have googled a lot and unfortunately I am not getting any closer to a solution:
I’m trying to create a 3D LUT for madvr.
Technically this works.
But the greyscales towards black cause problems and I can’t find a setting that is acceptable.
Grayscale near black are clipping into black. (The white Clipping Test is ok)Display device is a JVC x7900 beamer, pre-calibrated.
As measuring instrument I have an EODIS3.
Beamer is set to its largest color profile (BT.2020)
However, a test with a normal computer monitor gave the same result.The problem:
As soon as I use the 3DLUT created by DisplayCAL, the dark grayscales sink/clip towards black.
And there are corresponding posterization effects around black in the movie.
In the Black-Clipping Test (AVS Forum Test-Disk) you can see, that at least gray scale 17 is clipping black.
With other gamma settings also grayscales up to about 20.
I can’t find any setting where the grayscales are cleanly graded.If I switch off the 3DLUT processing and enter “the display is already calibrated; Gamma 2.2” (in madvr),
the gray scale is clean.
The effect is reproducible on the beamer as well as (for testing) on a monitor.
(The monitor was of course also profiled before.)The best result (with 3DLUT) is achieved by the setting
Gamma 2.2; relative; 0% black offset
Here only grayscale 17 is also black.
If I change the gamma to e.g. 2.4 or increase the black offset, more gray levels will shift to black.I have attached examples and my settings (The green color comes from the mobile phone camera)
Where could my mistake be?
Thank you very much!
Attachments:
You must be logged in to view attached files.2020-01-24 at 12:34 #22417Very interesting topic, I also have some odd black clipping if I use 2.4 relative gamma with 60-80% black output offset bar 17 is almost impossible to see. I have no idea if this black clipping test apply only to projectors and tvs, I’m not so sure if it is about pc monitors.
- This reply was modified 4 years, 2 months ago by S Simeonov.
2020-01-24 at 14:16 #22419the black clipping test can of course also be done on pc monitors.
Important is the correct mapping of the video levels (16-235 to the pc levels 0-255)If you decrease the black output offset more, do you get bar 17 visible?
My problem is, I’m already at 0% …so something else must be wrong at my setup…In the AVS Forum I found these post:
SDR Tone curve / gamma for beginnersAs I understand there, I should have a 100% Black offset and go a little lower from there.
But in my case even 0% is not enough…
So I think, there must be another setting, that I have to change….2020-01-24 at 16:04 #22420the black clipping test can of course also be done on pc monitors.
Important is the correct mapping of the video levels (16-235 to the pc levels 0-255)If you decrease the black output offset more, do you get bar 17 visible?
My problem is, I’m already at 0% …so something else must be wrong at my setup…In the AVS Forum I found these post:
SDR Tone curve / gamma for beginnersAs I understand there, I should have a 100% Black offset and go a little lower from there.
But in my case even 0% is not enough…
So I think, there must be another setting, that I have to change….Using 100% BOO will result in crushed blacks, I’m now at 40% and I still 17 barely, I can maybe go to 50% BOO, but anything above that will crush the blacks…
2020-01-27 at 8:20 #22444Could you please give me your settings?
What did you set up for the measurement?Today evening I will have another try.
At the weekend, I tried it at my Desktop Computer Monitor…
There seems all ok with the measurement and the generated 3dlut files.So I think I have to change some settings to correct measure my JVC Projector. I also have now an i1 Pro2 Spectrometer I will try to calibrate the Colorimeter with it before…
For the whole measurement the Spectrometer is to slow, or not?
2020-01-27 at 11:08 #22446I wish I could help you, but I don’t know anything about projectors…I hope Florian will have some solution for your situation.
2020-01-27 at 19:19 #22449In the Black-Clipping Test (AVS Forum Test-Disk) you can see, that at least gray scale 17 is clipping black.
With other gamma settings also grayscales up to about 20.This is working as intended. With a “pure power” gamma curve, you shouldn’t be able to differentiate the lower levels.
2020-01-27 at 19:28 #22453In the Black-Clipping Test (AVS Forum Test-Disk) you can see, that at least gray scale 17 is clipping black.
With other gamma settings also grayscales up to about 20.This is working as intended. With a “pure power” gamma curve, you shouldn’t be able to differentiate the lower levels.
So with 2.4 relative with BOO, we shouldn’t see from 17 and above? What gamma settings should we use?
2020-01-27 at 21:20 #22454In the Black-Clipping Test (AVS Forum Test-Disk) you can see, that at least gray scale 17 is clipping black.
With other gamma settings also grayscales up to about 20.This is working as intended. With a “pure power” gamma curve, you shouldn’t be able to differentiate the lower levels.
Understood, then it is not corrct if I can easily different the lower levels if I choose “the display is already calibrated”?
And what settings should choose in displaycal to get a “smooth” image without posteriztion effects in the low levels/black?
I recently found an advanced option in the profiling tab to setup the gammut mapping. I can here set up a destination viewing condition. Could this setting help?
Thank you in advance.
2020-01-28 at 9:05 #22455Hello Florian,
I have no idea what I did wrong.
But yesterday I made a new measurement and now the result is similar to S Simeonov.
In the range of about 0%-40%/50% black offset, I can distinguish all gray levels close to black.
I will now select a gamma curve with REC.1886 2.4 absolute and 0% black output offset.
Seems to me to be the best of the tried settings, dark grayscales are dark without clipping into black.Concerning the measurement I only changed one parameter to the old measurements:
I have profiled the EODIS3 with DisplayCAL against the projector and an i1Pro2 before and used this correction matrix afterwards.
I left all other settings at default.
But this should have little or no influence on the measurement of grayscales, shouldn’t it?
In other words I don’t understand what I did wrong before.
But I would like to, to avoid the mistake in the future. 🙁2020-01-28 at 15:25 #22460But this should have little or no influence on the measurement of grayscales, shouldn’t it?
Depends on what whitepoint target you’re using.
In other words I don’t understand what I did wrong before.
Nothing probably. Projectors are usually not very stable over time. My recommendation is to use white level drift compensation and limit the number of measured patches to counteract some of the drift.
2020-01-28 at 18:51 #22466What gamma curve should we use in the 3dlut to see bar 17 flashing in the black clipping test?
2020-01-28 at 21:49 #22468If you are using DisplayCAL to make a 3DLUT for madVR HDR tone-mapping, the only acceptable target gamma you should use is 2.2.
This is because madVR tone-mapping is hard-coded to tone-map to 2.2 gamma.
If you want another gamma after this then alter it on the display or in the gamma control in madVR.
2020-01-28 at 21:55 #22469If you are using DisplayCAL to make a 3DLUT for madVR HDR tone-mapping, the only acceptable target gamma you should use is 2.2.
This is because madVR tone-mapping is hard-coded to tone-map to 2.2 gamma.
If you want another gamma after this then alter it on the display or in the gamma control in madVR.
I see, I’m making a SDR 3dlut file for madvr, not HDR. So the gamma is again 2.2 with 100% BOO?
-
AuthorPosts