EDIT: I have attached the logs from DUCCS to this post, in case they will help figure out the issue.
If you are out there, I would really appreciate your help on this one. I noticed a bit of banding on this monitor, and decided to do some investigation to figure out what the problem was. It turns out that the banding only appears in the CAL1 and CAL2 preset modes after calibrating with DUCCS 1.5.10 with my i1 Display Pro.
In case it matters, I am running Windows 10 and I have a AMD Radeon R9 390X GPU. When viewing a 8-bit gradient from black to white in Photoshop, I see significant banding at the black and white ends of the gradient. The middle of the gradient is fairly smooth. To document the issue, I opened a file containing grayscale blocks from RGB 0,0,0 to RGB 21,21,21. I photographed the display at different OSD preset modes and different ICC profiles. Photoshop was restarted each time the ICC profile was changed so that it sees the new profile.
As you can see from the photos, the grayscale ramp looks fine in both Standard and sRGB modes. However, there is significant banding and discoloration in both CAL1 and CAL2. I repeated the test by isolating each channel, and again, the ramp looks fine in the sRGB preset mode, but shows banding in CAL2.
Since Standard and sRGB modes looks fine, I think I can rule out my GPU or display driver as the culprit. From these photos, the calibration result almost looks GPU calibration with 8-bit LUT rather than the 14-bit internal LUT that this monitor is supposed to have. Does this looks like an issue with the ICC profile, or the DUCCS calibration? On my previous calibration, my settings were Bradford, ICC v2, Table based, large patch set size (462). Are my settings correct, or should I choose something different?
Finally, when I validated the calibration with dispcalGUI, everything looked great. Maximum dE was under 1. However, after doing these tests, I am wondering if dispcalGUI takes the DUCCS generated profile into account, because there is some clear grayscale discoloration when the custom ICC profile is loaded into Photoshop. There is no discoloration when using the default sRGB or Adobe RGB profiles. Also, dispcalGUI is not going to detect banding in the extended testchart because it only measures 22 levels of gray.
Do you have any advice on how I should proceed? Thank you in advance!
Right after posting, I found another thread complaining about the same exact problem:
Chris, would you be able to show this to your support team and let me know what they say? It would seem that either the internal LUT is detective or (more likely) there is a bug in DUCCS.
I tried performing a factory reset, and did another calibration with DUCCS 1.5.9. The results were exactly the same as 1.5.10 with no improvement on the banding.
Chris - can you please provide an official response to acknowledge that this is a problem, and that Dell will investigate to find and fix the problem? I notice that there was no reply from you or anyone else from Dell on the other thread that I linked above.
Since this monitor is marketed to professionals for whom hardware calibration is an important feature, this kind of calibration result is unacceptable. I really hope that Dell doesn't leave UP3216Q owners out in the cold on this.
I can email this this thread to the team. But all are out on holiday until next week. Private message yumichan to get his opinion of the issue.
Social Media Support#IWork4DellDell Support VideosDrivers and DownloadsDriver order
Thank you for your reply. Yes, please forward this thread to your team and let me know what they say after they return.
I did mention yumichan in my original post, but he hasn't been active in the last couple of weeks. Additionally, since I have subsequently done another calibration with different settings and a different version of DUCCS, and there is another report of the exact same issue, I don't think there is much yumichan would be able to add at this point. It is either a hardware problem that affects at least two of these monitors, or it is a software issue with DUCCS, so the fix needs to come from Dell.
You could have Dell setup a monitor exchange. Test the replacement before sending any monitor back to Dell. If the replacement does not have the issue, keep it and send us the original. If the replacement has the same issue, it is very doubtful that the monitors hardware was at fault.
I have already contacted Dell to setup a monitor exchange, but their response is that they have no replacement UP3216Q available to exchange, and only offered me a UP3214Q. Therefore, an exchange is not an option for me at this time.
Maybe I am using the wrong terminology, but if both monitors show the same faulty calibration result, then it is either a design flaw (hardware issue), or a software flaw with DUCCS. What else can be the problem?
I edited my original post to include the calibration logs from DUCCS. Hopefully they will be helpful.
I hope it is a DUCCS flaw.
I played with my system some more, and things are getting stranger.
In my AMD Catalyst Control Center, there is an option to set the color depth to 10, 8, or 6 bits per channel. By default, with the UP3216Q, it is set to 10bpc. I tried changing to 8bpc, but this made no difference. When I changed it to 6bpc, the banding was greatly reduced. This is really odd, because 6bpc means only 64 levels of gray and I am seeing a lot more than that.
I experimented with different ICC profiles in CAL1/CAL2 and found the following:
8bpc or 10bpc, all ICC profiles: same result as shown in the photos in the first post6bpc, with sRGB profile created by DUCCS 1.5.10 (V2 profile, table based): minor banding (around ~10 levels of gray missing out of 256). There is no clipping at black or white.6bpc, with sRGB profile created by DUCCS 1.5.9 (V4 profile, matrix based): very minor banding (around 2-3 levels of gray missing out of 256). However, everything below RGB 7,7,7 is clipped to black.6bpc, with default sRGB profile: no banding at all. However, everything below RGB 6,6,6 and above 250, 250, 250 is clipped to black and white, respectively.
In comparison, in Standard OSD mode with the ICC profile included with the Windows 10 driver, or sRGB mode with the default sRGB profile, there is no banding on any bit depth (10, 8 or 6), and I can distinguish between all 256 levels of gray.
So far, 6bpc with the default sRGB shows the best results with regards to banding. The custom profiles created by 1.5.9 and 1.5.10 both have problems. The V2, table based profile does not clip, but bands about 10 levels of gray. The V4, matrix based profile (which is the default DUCCS setting) bands only about 2-3 levels of gray, but clips 7 levels to black. The total color loss between the two profiles is approximately the same.
Therefore, the results with the display driver set to 6bpc are better, but are still not normal. Normal results would be what happens in Standard OSD mode with the driver ICC profile (no banding, all 256 levels of gray distinguishable, in other words, no color loss at all). I'm going to make another post on AMD's forums to see if I can get some more insight on how the "Color Depth" setting actually works. I hope that this extra information is useful on your end.
EDIT: Some more data:
I tried uninstalling the AMD drivers to see if that made a difference, and the results were the same as the 8bpc and 10bpc setting in the AMD drivers (identical to the photos in the first post). I also tried removing the R9 390X from my system and connected my monitor to the integrated graphics. Again, the results were the same. Unfortunately, I don't have a Nvidia GPU available to test.
The only thing that makes a difference to help the banding is setting the AMD drivers to 6bpc. I am still getting full color in this mode, so the output signal still seems to be 8-bit, but that setting is doing something to reduce banding in CAL1/CAL2.
I am a graphics professional considering purchasing the UP3216Q monitor. Can you please update this thread regarding the calibration problem so that I may make a confident purchase decision?
Here is an update with some additional photos.
After repeating the calibration process again, I made some measurements with my i1 Display Pro using the "Verification" function in dispcalGUI, using the following steps:
- I created a custom test chart that consisted of 256 grayscale patches that ranged from RGB 0,0,0 (black) to RGB 255,255,255 (white).- Ran a measurement report under the following conditions:
- Here are the complete measurement reports:
- I took the luminance data into Excel and made the following graphs (click to see animated GIF):
- I only graphed up to RGB 55,55,55 because the banding stops above that, according to the measurement reports.
- In the default 10bpc setting, CAL2 shows significant banding in the dark areas. Wherever the curve is flat, there is banding. The slope of the curve in the sRGB preset is smooth, but is not calibrated correctly (too dark).- The sRGB mode is not affected at all by the "Color Depth" setting. In CAL2, the 6bpc setting seems to have a dithering or smoothening effect on the output luminance curve. However, even after the smoothening, there is still some unevenness to the curve, which is visible when viewing a black to white gradient, as in the following photos.
- Even though the banding is technically gone in CAL2 when "Color Depth" is set to 6bpc (because each RGB gray level has distinct luminance values), there are still visible bands in the gradient, due to the "kinks" in the output luminance curve.
- Furthermore, setting "Color Depth" to 6bpc is not a long term solution because only AMD has this setting. If I decide to move to an Nvidia GPU, or connect any other device (Blu-Ray player, cable box), the banding will be back to what it shows in the photo with the 10bpc setting.
- I didn't test every OSD preset mode, but I did test both Standard and sRGB. Both of these modes were capable of outputting a smooth luminance curve regardless of the "Color Depth" setting. Therefore, I would think that the monitor itself is fine. The problem seems like it is DUCCS, which is not calibrating the hardware LUT correctly. These results might be fine for a consumer level product, but since this is Dell's flagship monitor marketed towards professionals, I hope that this issue receives the attention that it deserves.
@tendrchkn, plug your monitor to a laptop (to discard broken GPU in your current computer). Open DispcalGUI and clear GPU LUTs.
-If banding is still present... monitor is broken: send it to dell or return for refund or similar options avaliable in the store where you bought it.
-If banding is not present then you have some issue with your computer. Try a Linux Live DVD to inspect gradient again in order to discacrd hardware issue.
Remember that gradients (in banding related issues) SHOULD BE inspected in NON color managed applications, like Internet Explorer.
Color managed applications may introduce banding because of innacurate ICC profiles (innacurate monitor behaviour stored in them) or due to 8bit/channel computations (gammut & gama remapping between monitor ICC profile and image ICC profile)
@yumichan: thank you for your reply. I have tried plugging in a laptop, along with other devices and the banding is always present. The last set of photos were from a non color managed application, so it is not a Photoshop problem either.
I am not quite sure that my monitor is broken because there is another report of the same exact issue (linked on the second post), and it is very unlikely for two monitors to have the same problem independently. Either all UP3216Qs have this defect, or it is a DUCCS issue.
Furthermore, I have already contacted Dell several weeks ago, but they only offered me a UP3214Q as an exchange because they had no UP3216Q available. A new UP3216Q probably wouldn't help me either at this point until Dell fixes this with a firmware update, or a new DUCCS version.
Chris, have you heard anything back from your team?
I see the same on my up3216q, calibrated with both 1.5.9 and 1.5.10 to AdobeRGB with an i1 display pro.
I hope it's the software...
I should probably have said that what I see is banding in a gradient when calibrating with DUCS into cal1/2, but smooth gradients with the AdobeRGB preset.