- This topic has 13 replies, 9 voices, and was last updated 3 months ago by
.
-
Posts
-
Hi all, I just got the Dell AW3423DW and wonder about how to properly calibrate this with the i1 display pro plus.
Previously I have Acer XR341CK and it was very straightforward, just calibrate it and be done since the monitor is only SDR capable.
First I tried to calibrate the AW3423DW without Windows 11 HDR turn on, the result was great. But then when I turned on HDR in Windows and tried to measure delta E again but many colors were skewed off.
So I tried to calibrate the monitor with HDR turned on but then the “measured vs assumed target white point” are never good. It always shown as not ok result and the gamma is also skewed (2.9-3.0) tried many times and after restart displaycal finally got 2.3
But then I think what if the content is only SDR capable? Will the HDR profile skew the SDR color like when the monitor is calibrated with SDR and turn on HDR later?
Or do I need to calibrate the monitor for both SDR and HDR? but then I guess the RGB settings on OSD must be different and it might not be practical to adjust RGB everytime when I switch the profile.
Calibrite Display Plus HL on Amazon Calibrite Display Pro HL on Amazon
Disclosure: As an Amazon Associate I earn from qualifying purchases.I would turn off HDR unless I were to watch a movie or play a HDR game. Multiple conversions are never good in my experience. And even then, I wouldn’t even try to calibrate to HDR response of the display. Most incorporate automatic brightness limiters that make any measurements inaccurate.
This is basically impossible in Windows.
MS is working on something (ACM) to “unify” the SDR and HDR color conversion pipelines for a year. But the problem is they never get it right. And with so many changes in API, almost all previously well-working SDR pipelines will be deprecated, including the source-XYZ-device conversion based on ICC profile.
What current conversion in Windows 11 is doing, as I see it, is as follows (even this has changed significantly in the last 2 years, without the feature mentioned above)
In SDR, everything works robustly as before from Windows XP. There is a global gamma correction ‘vcgt’, that applies to everything you see all the time. After the correction, the color-aware apps will convert the source RGB of content to the display RGB, based on how everything is defined in the ICC, and sent to the video card.
With HDR there is a huge problem: the vcgt correction is basically not true anymore because that curve is measured only in the SDR luminance range, say up to 180nits. It never knows how the display behaves from 180nits to 1000nits. Let alone the display under HDR mode is optically different, which is not captured by the SDR vcgt anyhow.
That’s why if you activate the same ICC, via displayCal, under HDR, usually you will see very weired greyscale, because of the wrong gamma correction. It is almost always better off leave the correction empty. This is just the greyscale part, the color part is even more messed up.
In Windows under HDR, it is ruled that everything be encoded in BT2020 coordinates. 98% of the apps never know this rule (they’re not even color-managing under SDR probably). So their RGBs are treated as sRGB, regardlessly. That’s why if you have a wide gamut display, say P3, under SDR, the desktop and UIs are oversaturated, because they’re stretched to P3 gamut from their intended gamut. Upon HDR on, they’re treated as sRGB, which is so smaller than P3, so things look immediately duller.
Here comes the reason why displayCal cannot calibrate and profile properly with HDR turned on. When profiling, displayCal needs to ask the display to show its raw, uncorrected colors of all kinds, the full red, green, etc. But as a traditional app, Windows thinks it should only output within sRGB, so that full red is compressed and shown in sRGB red, and then displayCal will get the reading and concludes that your display can only do sRGB red, while it can really do something much redder.
If traditional color managing apps like Photoshop, still trying to do the same in HDR, they will first figure out the RGB values under the display gamut of P3. However, these values are wrongfully treated as sRGB, and then converted to BT2020 again by Windows. Just imagine how wrong this will end up.
And even if modern apps get to circumvent this by following MS guidelines, it is still a mystery how Windows defines the display RGB primaries, which are crucial for color conversion. They can simply use BT2020, which is always wrong, or they can use the EDID of the display, which is better only if the manufacturer is serious about their product. But remember the most accurate info is stored in the SDR ICC file from your measurements, but that file has an improper vcgt!! So will Windows extract only the correct RGB primaries part from ICC and ignore the vcgt? Unlikely.
So you see why the calibration and profiling with an ICC is both erroneous and futile. Things are guaranteed to be wrong.
So what is the best way to use HDR monitor? Under SDR, activate the color management using the accurate SDR ICC. Under HDR, switch all ICC profile and calibration off. You have to do this every time you switch.
Why is this so hard for them to get right? Is it just not a priority for microsoft?
We never know. Color fidelity and all aesthetic and artistic work is never the top priority of Microsoft. Ironically, it is even less now.
MS brought in the WCS which let the app handle ICC conversion since Windows 2000. They pretty much knew how it should be done in the first place. But no improvement is made for 20+ years and nowadays even a mobile app treats color conversion better than desktop Windows.
Things with HDR are much more complicated, the 10-12bit handling, HDR tone mapping, HDR/SDR blend and much wider color gamut support, etc, at least makes it 5 times harder than traditional ICC and gamma correction. They knew how to do it with the SDR/ICC thing, and they chose not to perfect it. I feel they don’t even know how to properly handle HDR system to each detail now.
#1:: They shouldn’t do an sRGB –>> Rec2020+PQ conversion but rather an sRGB_primaries+gamma_2.2 –>> Rec2020+PQ conversion. sRGB encoded content was meant to be watched with pure-power gamma 2.2, as analog CRT monitors back then didn’t even have signal conversion options (at least not a fine, parametric control over gamma and definitely not the cheap ones).
#2: Windows 11 nowadays automatically disables the SDR ICC profile when you toggle the OS HDR mode. And there is a new ICC format for HDR mode. I haven’t seen one but some monitor manufacturers are said to ship compatible profiles or you can alternatively run the HDR Calibration utility from the Microsoft Store to create one (it’s helpful and lets you create one that doesn’t alter the saturation but you can set the peak brightness numerically if you know the capabilities of your display and then things like Auto HDR for games will use that information to some extent ).
#2: Windows 11 nowadays automatically disables the SDR ICC profile when you toggle the OS HDR mode. And there is a new ICC format for HDR mode. I haven’t seen one but some monitor manufacturers are said to ship compatible profiles or you can alternatively run the HDR Calibration utility from the Microsoft Store to create one (it’s helpful and lets you create one that doesn’t alter the saturation but you can set the peak brightness numerically if you know the capabilities of your display and then things like Auto HDR for games will use that information to some extent <em class=”bbcode-em”>).
Regarding this, if you hace an Advanced ICC Profile, generated with Windows HDR calibrator, does it affect the default SDR icc profile ?
It does not. However, even the built-in picture viewer of Windows 11 23H2 will show a cian tint if you open a simple sRGB image while OS HDR is active. If you assign this synthetically created profile to an untagged screenshot taken from (or example) an HDR video game (by non-HDR aware tools) using Photoshop and switch to the Microsoft ICM in Absolute mode then it will have a pink tint. So I don’t know what else is supposed to be compatible besides AutoHDR (for select games).
[…] switch to the Microsoft ICM in Absolute mode then it will have a pink tint.
May be PCS white vs actual white issue. When rendering images in A colorspace to screen you should use relative whitepoint.
But IDN have an HDR to test.By the way, how can I calibrate/profile a display in HDR mode using DisplayCAL without using a chained metadata-injector device (like the HDFury Integral) or the service menu of the display to force SDR to be presented as if it was HDR? (Both of these alternatives have some drawbacks, like limited picture format / bandwidth support or an uncertainty if the the synthetic override state accurately represents the normal HDR state.)
I can profile the display with ColourSpace + madVR without issues (well, sometimes it doesn’t seem to trigger correctly but it usually does). CS has options to control HDR but it also works the other way around (if I activate HDR before connecting to madVR then HDR options get locked out as enabled). But it doesn’t seem to work with DisplayCAL + madVR. Even though that preset wouldn’t actually satisfy my needs but I tried the “madVR HDR” preset of DisplayCAL and the output was still limited to ~170nit or something similar (the display can do ~800 nit in uncalibrated state on a 10% window).
In the past, I could use NVAPI HDR and use dispread’s pattern window because NVAPI HDR didn’t convert the legacy Windows desktop elements to HDR, so the windows desktop and the patch window was output at HDR scale. Nowadays the “legacy” parts of Windows desktop are converted with NVAPI HDR as well.
Hmm. It looks like HDR does work as I expect it some of the time. It’s sort of random what I get out of it. I managed to get a ~800nit peak white reading once using Win11 OS HDR. Then I restarted both madTPG and DisplayCAL and now it’s ~225nit (not the same low value as before). I cycled OS HDR on/off and restarted the two software: ~225nit again. 🙁
@Janos there is now a pgenerator client for Displaycal with bit accurate HDR. It should solve your HDR pattern issues.
Thanks but dowgrading madVR from madmeasure_beta to the latest public relese solved the issue. MadTPG is now relieable.
I can use ColourControl to disable dithering (work on both AMD and nVidia GPUs, I am not sure about Intel) and make sure the output format is RGB-Full (the only bit-perfect output format on PC at the moment).
