Home › Forums › Help and Support › Banding when using profile
- This topic has 25 replies, 3 voices, and was last updated 3 months, 1 week ago by Old Man.
-
AuthorPosts
-
2024-01-26 at 17:49 #140420
I’m on macOS, using an Asus PA279CV with a Macbook Pro.
My problem is that after calibration, there are visible banding issues on lower gray scales. Actually, not just banding, but I think some grayscales look visibly darker than the neighbors. Without using a profile, everything is super smooth.
Can you help me diagnose why is this? I used the default Office 2.2 D6500 profile. Sensor is i1D3. I used one of the 3.3nm uploaded CCSS files (ASUSTek COMPUTER INC ASUS PA279CV by 4KM (i1 Pro 2).ccss).
Is there a way to “smoothen” out a curve profile? For example TVs have 22 point calibration, which I think would be better in my case than the current profile.
So far, I’ve reverted the profile and just use the RGB Gain controls on the monitor to set the correct white level. This way the banding, etc. is fixed, but my gamma is of course not perfect.
- This topic was modified 3 months, 1 week ago by Hyperknot.
- This topic was modified 3 months, 1 week ago by Hyperknot.
- This topic was modified 3 months, 1 week ago by Hyperknot.
Attachments:
You must be logged in to view attached files.Calibrite Display Pro HL on Amazon
Disclosure: As an Amazon Associate I earn from qualifying purchases.2024-01-26 at 18:48 #140428Banding can be a side effect of using a VCGT. Try doing a basic calibration of the monitor in hardware like you’ve started to do (white level/point and EOTF), then go to the calibration tab and set everything to “as measured” and redo the profile. When a dialog pops up with two options, only check the first option.
2024-01-26 at 19:05 #140430This monitor can do:
– Six-axis Hue
– Six-axis Saturation
– Gain
– Offset
– White point (already set at 6500K)But based on what I’ve read here, I should only touch RGB Gain, right?
Then, on calibration, I should change both Whitepoint and Tune curve to “As measured”? If I do this, then the button changes to “Profile only”. Is this what I want?
2024-01-26 at 19:12 #140431White point/gain and *maybe* offset (probably not though – you can test it). Profile only is what you want, yes
2024-01-26 at 19:29 #140432This is what you meant, right?
Attachments:
You must be logged in to view attached files.2024-01-26 at 19:30 #140434Yes
2024-01-26 at 19:35 #140437There are two sources of banding, one is GPU LUT: truncation to 8bit or lack of dithering or both, usually on nvidias and iGPU. AFAIK silicon GPUs should not have these issues, but maybe i’m wrong about this, I’ve not used apple silicon chip.
The other one is color management (8bit undithered) and since macOS desktop is color managed this looks like the culprit: you are seeing rounding errors.
My problem is that after calibration, there are visible banding issues on lower gray scales. Actually, not just banding, but I think some grayscales look visibly darker than the neighbors. Without using a profile, everything is super smooth.
Can you help me diagnose why is this?
Explained above. It’s mostly unavoidable at 100% because color management at desktop level needs to be fast = simplified. For example it lacks of BPC so previews on OS default image viewer may cause black crush, while Photoshop do not show it.
Is there a way to “smoothen” out a curve profile? For example TVs have 22 point calibration, which I think would be better in my case than the current profile.
Aim for a TRC closer to the TRC in default ICC generated by macOS when you plugged the Asus. sRGB like or sRGB/2.2 with a stright line near black.
This way the banding, etc. is fixed, but my gamma is of course not perfect.
On color managed apps display gamma does not matter as long as display profile ICC stores actual display gamma.
But due to limited precision in computations for color management it may cause banding because the lack of dithering. Most color managed apps suffer it.
C1 and LR(develop) do dither, so this should not happen there. ACR do dither. PS do not dither but on >8bit files + 10bit OpenGL output may hide it.
I do not r emember about Safari.2024-01-26 at 20:26 #140444I did a profile only run, based on Old Man’s suggestions. It looks really nice to me, banding is gone and the extended verification chart looks visually nice for me. I mean I cannot see color difference with my eyes.
What is happening here exactly? The LUT is linear / not doing anything. But then what is happening when I’m using this “profile only” profile on system wide macOS?
Attachments:
You must be logged in to view attached files.2024-01-26 at 20:47 #140446Your monitor is handling what the VCGT (think 1D LUT) normally would. VCGT’s often cause banding, so I just don’t use them and do the basic calibration in hardware instead.
The profile contains everything needed for the volumetric color correction (think 3D LUT (plus whatever remains needing correction in the 1D land)), so either macos or color-managed programs handle the rest. This would be true even if you used a VCGT
That report looks excellent, btw
- This reply was modified 3 months, 1 week ago by Old Man.
2024-01-26 at 21:10 #140448What’s more important for me is the system wide color management, not specific apps, like C1, LR or ACR.
I think I prefer the Profile Only behavior for now. It might not be perfect in the lowest levels, but I prefer not to have banding like this.
Thank you both for your help!
2024-01-26 at 21:23 #140449If mac color management is system-wide, then your whole system should look correct now. Again, the report looked excellent. The lowest levels will never be accurate with current hardware limitations
2024-01-26 at 21:44 #140450I did a profile only run, based on Old Man’s suggestions. It looks really nice to me, banding is gone and the extended verification chart looks visually nice for me. I mean I cannot see color difference with my eyes.
What is happening here exactly?
The LUT is linear / not doing anything. But then what is happening when I’m using this “profile only” profile on system wide macOS?
It’s happening that your display grey ramp has no color tint issues out of the box, hence you can try to model it with a single TRC for 3 channels.
If it had color tint you’ll need to apply grey calibration, hence no linear LUTs for GPU.
It works by “chance” (good quality control on that unit), no joke.AFAIK M1 had no banding issues due to GPU LUT. Which model do you own? macintel? silicon? which GPU output?
You can try to test this with an “uncalibrated report” in tools/report menu. It’s a yes/no test, “LUT >8bit” or not.If mac color management is system-wide, then your whole system should look correct now.
Only if display gray is “calibrated” with no calibration applied at all.
2024-01-26 at 21:52 #140451I told him to calibrate grey in hardware. And since mac has system-wide color management, a profile is enough to correct grey system-wide. It’s not chance. Did you look at the report? Grey is perfect. I doubt the monitor is naturally that good.
He said he had banding with the VCGT and none without. I think that speaks for itself, don’t you?
Again, grey is roughly calibrated in hardware. Macos handles the rest with the profile. It’s not like windows
2024-01-26 at 22:02 #140452Here is the uncalibrated report:
21:55:38,156 Setting up the instrument 21:55:38,156 list_iccss: xdg_bds returned 19 paths 21:55:38,156 list_iccss: returning 19 ccss's 21:55:38,156 Product Name: Colormunki Display 21:55:38,156 Serial Number: CM-20.B-02.507496.03 21:55:38,156 Hardware Rev.: B-02 21:55:38,156 Firmware Version: v2.28 21:55:38,156 Firmware Date: 29Jan14 21:55:38,156 dispcal: Warning - new_dispwin: Frame buffer depth 8 doesn't match VideoLUT 10 21:55:38,156 Uncalibrated response: 21:55:38,156 Black level = 0.1209 cd/m^2 21:55:38,156 50% level = 26.94 cd/m^2 21:55:38,156 White level = 125.88 cd/m^2 21:55:38,156 Aprox. gamma = 2.22 21:55:38,156 Contrast ratio = 1041:1 21:55:38,156 White chromaticity coordinates 0.3140, 0.3312 21:55:38,156 White Correlated Color Temperature = 6420K, DE 2K to locus = 5.2 21:55:38,156 White Correlated Daylight Temperature = 6422K, DE 2K to locus = 0.7 21:55:38,156 White Visual Color Temperature = 6239K, DE 2K to locus = 5.0 21:55:38,156 White Visual Daylight Temperature = 6401K, DE 2K to locus = 0.6 21:55:38,156 dispcal: Warning - Unable to determine effective Video LUT entry bit depth
Also, there is something strange with the very near-dark levels, my curves look like this, with red-green drop in the lower 10%. But when using the calibrated profile, I was noticing a brownish tint on the lower colors. Is there any change that it’s actually my instrument’s problem in dark areas? It’s a ColorMunki Display, I mean the cheaper version, possibly binned.
I mean visually, the gray scale tint is much more uniform without LUT than with LUT.
- This reply was modified 3 months, 1 week ago by Hyperknot.
Attachments:
You must be logged in to view attached files.2024-01-26 at 22:04 #140455Although, was the verification done with the profile active? Maybe the monitor is that good. Either way, macos should correct everything with just a profile, right? I don’t use mac, but that was my understanding
-
AuthorPosts