Display Calibration

[Asmodian]

I got my UP2716d today (ordered it before our discussion on wide gamut).

The first thing that came to mind was wow, vibrant colors, not over saturated and bad looking. It looks a bit like a samsung phone, which imo looks the best out of all phones screens.

It may not be bad looking but it is oversaturated, all the Samsung phone screens are also oversaturated*.

I've found something a bit weird. The monitor on default is in "standard" mode. Everything looks great, and dispcal shows 100%sRGB and 99.7% adobe RGB and 93% dci-p3. When I switch from standard mode to color space, and select adobe RGB, reds turn orange... sRGB emulation mode also looks makes reds look orange. wtf lol

Correcting oversaturated colors always looks worse than having non-oversaturated primaries to start with, in my experiance. The only way to desaturate red is to add green and blue, if you don't add enough blue it will look orange. Also less saturated red will look orangish/pinkish when compared to more saturated reds. A more saturated red is more "red" than a less saturated one. ;)

Trying to figure out why and how colors are off with a wide gamut display taught me a lot about how humans see color and the limits of a calibration. Also, everything is so relative the immediate experiance is sometimes misleading.

edit:
*for sRGB, they are not saturated enough for DCI-P3.

[Arm3nian]

I made profiles to use with the emulation modes, and while they helped a bit, sRGB and AdobeRGB still make reds look orange. Literally. There is no red when looking at a color pallet. sRGB should have red in Windows, but it doesn't. I tried AdobeRGB in photoshop and still no reds. I can't change gain or saturation or anything in the presets, so it has to be an internal defect. No other explanation really.

My panel is going back though. It has 10-20 dead pixels, and 10-20 bright/stuck pixels. I don't understand how this passed any QC control from Dell, especially since it is an ultrasharp model. Seems like it was returned by someone, and Dell or B&H didn't bother to fix or retire it, and it was just repacked and sold as new. Pathetic. Going to stick to Newegg from now on.

The monitor I want is the new OLED from Dell. 400,000:1 CR, 120Hz... too bad it's 5 grand...

Leaning towards the U3415W or the unreleased U2717d. Dell seems to be the industry standard, when you get one without any problems at least :rolleyes:

[chros]

The monitor I want is the new OLED from Dell. 400,000:1 CR, 120Hz... too bad it's 5 grand...
Wha'? :) 5 grand??!!! In which currency? :D That's insane ... :)

[Asmodian]

$4,999 USD. :p

30 inch, OLED, 97.8% of DCI-P3, 3840x2160, 120Hz. :eek:

[Arm3nian]

Wha'? :) 5 grand??!!! In which currency? :D That's insane ... :)
http://www.dell.com/learn/us/en/vn/press-releases/2016-01-06-dell-updates-latitude-and-industry-leading-monitor

Early adopter price. Also no competition from other monitors. I wish they made it somewhat affordable though. Business strategies ruin the fun :devil:

$4,999 USD. :p

30 inch, OLED, 97.8% of DCI-P3, 3840x2160, 120Hz. :eek:
And 0.1ms pixel respone time :D

[chros]

$4,999 USD. :p

30 inch, OLED, 97.8% of DCI-P3, 3840x2160, 120Hz. :eek:
ppffff.. 5000 for 30''. 5000 is the 65'' LG 65EF9500 OLED.

[Siso]

ppffff.. 5000 for 30''. 5000 is the 65'' LG 65EF9500 OLED.

Chros, do you have any recommendation for a 3dlut with gamma 2.4 relative, how many % black output offset should be used, if the black level of the monitor is ~ 0.0871, and the contrast ratio is 1000:1?

[chros]

Chros, do you have any recommendation for a 3dlut with gamma 2.4 relative, how many % black output offset should be used, if the black level of the monitor is ~ 0.0871, and the contrast ratio is 1000:1?
Not really, it depends on your viewing condition and preference. Try out with 100% at first then lower it by 5% all the time (generate a new 3dlut) and check the result with your favorite clip.

[Siso]

Not really, it depends on your viewing condition and preference. Try out with 100% at first then lower it by 5% all the time (generate a new 3dlut) and check the result with your favorite clip.

On 100% I get black clipping in the "black clipping test" video.

I watch mostly in a very dark room at night.

[chros]

On 100% I get black clipping in the "black clipping test" video.

I watch mostly in a very dark room at night.
1. Do you also have clipping without 3dlut?

2. I just checked with HCFR (according to your data and I set 105 cdm for Y white) and compared to my night settings:
- 105 cdm white, 0.14 cdm black, I used 2.4 relative 98 offset
-- this results gamma: 10% white ~2.22 ; 20% white ~2.33 ; 30% white ~2.36
- to achieve similar results you have to use 87% offset (with 2.4 relative) in DisplayCal

But the best approach is to try out different offsets :)

[Siso]

1. Do you also have clipping without 3dlut?

2. I just checked with HCFR (according to your data and I set 105 cdm for Y white) and compared to my night settings:
- 105 cdm white, 0.14 cdm black, I used 2.4 relative 98 offset
-- this results gamma: 10% white ~2.22 ; 20% white ~2.33 ; 30% white ~2.36
- to achieve similar results you have to use 87% offset (with 2.4 relative) in DisplayCal

But the best approach is to try out different offsets :)

With 95% bo offset I can see the 17-25 flashing, 17 flashes and can be barely seen, in a completely dark room I can see it more clearly, I guess I'll stick with 95% bo offset, another strange thing in the 3dlut is some kind of red color clipping...

[chros]

With 95% bo offset I can see the 17-25 flashing, 17 flashes and can be barely seen, in a completely dark room I can see it more clearly, I guess I'll stick with 95% bo offset, another strange thing in the 3dlut is some kind of red color clipping...
Sometimes I get this as well, currently with my monitor. Have you modified the contrast setting? If yes, try out with the default setting.
Otherwise I don't have any idea what can you do about this, apart from rerun the whole measurement and hope the best :)

[Siso]

Sometimes I get this as well, currently with my monitor. Have you modified the contrast setting? If yes, try out with the default setting.
Otherwise I don't have any idea what can you do about this, apart from rerun the whole measurement and hope the best :)

When I calibrate, I never modify the contrast, it is always the native :) I'm now using 87% bo offset, and I'll run few more tests to see how it performs.

[chros]

When I calibrate, I never modify the contrast, it is always the native :)
Good, good :)
I'm now using 87% bo offset, and I'll run few more tests to see how it performs.
What can modify the result:
- light condition in the environment where you take measurement (!) (even with id3pro)
- if you uncheck the "interactive measurement" on calibration tab: first you it to set your device to near to D65 then cancel the process, then start a new madrv process with this setting unchecked.

[Siso]

Good, good :)

What can modify the result:
- light condition in the environment where you take measurement (!) (even with id3pro)
- if you uncheck the "interactive measurement" on calibration tab: first you it to set your device to near to D65 then cancel the process, then start a new madrv process with this setting unchecked.

For the 3dlut I used this tutorial https://imouto.my/tutorials/madvr/#3dlut

My current calibration is set to srgb with linear 2.2 gamma, I got pretty decent result :)

[CarlosCaco]

Hey guys, im trying to achieve a good 3d lut but Having hard times, my display don t get near d65 At all, i have only brightness anda contrast sliders on my tv...
If set white point to as measured should i put relative on rendering intent?
And black level correction and rate? I m getting correct gray scale with blc on auto and rate on 3... Is this correct? Should i use this?
Is i1d2 too old technology to old? Its possible that i1d2 is reading totaly wrong the temperature and the colors?

[Asmodian]

Hey guys, im trying to achieve a good 3d lut but Having hard times, my display don t get near d65 At all, i have only brightness anda contrast sliders on my tv...
If set white point to as measured should i put relative on rendering intent?
And black level correction and rate? I m getting correct gray scale with blc on auto and rate on 3... Is this correct? Should i use this?
Is i1d2 too old technology to old? Its possible that i1d2 is reading totaly wrong the temperature and the colors?

Yes, it is quite possible the i1d2 is too old to measure correct white on your display. Colorimeters use filters to measure color and new backlight technologies require different filters or spectral correction files. The filters also change over time.

Use as measured for white and I suggest the Luminance Axis intent. I also like the less intense Saturation intent, especially for anime or cartoons, it keeps the intensity of colors more correct even if the hue is a bit off. Both Perceptual and Saturation intents are better when making larger changes, relative is only preferred when making small corrections in my experiance.

edit: for black level correction auto and 3 sounds good.

[CarlosCaco]

Yes, it is quite possible the i1d2 is too old to measure correct white on your display. Colorimeters use filters to measure color and new backlight technologies require different filters or spectral correction files. The filters also change over time.

Use as measured for white and I suggest the Luminance Axis intent. I also like the less intense Saturation intent, especially for anime or cartoons, it keeps the intensity of colors more correct even if the hue is a bit off. Both Perceptual and Saturation intents are better when making larger changes, relative is only preferred when making small corrections in my experiance.

edit: for black level correction auto and 3 sounds good.

Luminescente martched apparence results in the Very bad results... I will try others to see

[Asmodian]

Luminescente martched apparence results in the Very bad results... I will try others to see

In what way is it bad?

[Stereodude]

With 95% bo offset I can see the 17-25 flashing, 17 flashes and can be barely seen, in a completely dark room I can see it more clearly, I guess I'll stick with 95% bo offset, another strange thing in the 3dlut is some kind of red color clipping...
I didn't find any shift in black level on my Plasma with the 3D LUTs I made last week. Bar 17 just barely showed up without the 3D LUT and it just barely shows up with the 3D LUT regardless of the offset, whether it's 0% or 100%...

Other bars in the 19-25 range change brightness, but not 17.

For the 3dlut I used this tutorial https://imouto.my/tutorials/madvr/#3dlut
That's pretty complicated. I just used the madVR 3D preset in the drop down in the latest version of DisplayCAL and used the default options on my LCD. For my plasma I enabled white point compensation.

[CarlosCaco]

In what way is it bad?

The colors became washed out, like a white veil in front the image

But i don t know if is the lut or my tv... I do some tests and not so secure about the levels configuration like i say on madvr thread

I forgot that my tv has an hdmi mode option that have movie and graphics modes
I presume graphic modes with madvr and gpu on full its the right mode graphics mode disable all post processing doing by tv and the black don t crush, but the colors are a little under saturated and tv became more brighter like a white/blue wash on image, the white became stronger and more blue
Also in verifications i m getting only not oks but i have to trie others renders intents... A tried set the calibration settings to calibraste white to 65000k but the whites became with a strong green tint...

[Siso]

I didn't find any shift in black level on my Plasma with the 3D LUTs I made last week. Bar 17 just barely showed up without the 3D LUT and it just barely shows up with the 3D LUT regardless of the offset, whether it's 0% or 100%...

Other bars in the 19-25 range change brightness, but not 17.


That's pretty complicated. I just used the madVR 3D preset in the drop down in the latest version of DisplayCAL and used the default options on my LCD. For my plasma I enabled white point compensation.


With 95% bo offset I can see bar 17-25 flashing, with 100% bo offset I only see 19-25 flashing...the tutorial in that site is good I guess. I'm using the 3dlut for my monitor, I don't own a tv.

[chros]

Also in verifications i m getting only not oks but i have to trie others renders intents...
A tried set the calibration settings to calibraste white to 65000k but the whites became with a strong green tint...
You don't have to experiment different rendering intents: the default is the best if the gamut of the device is smaller than the intended one (which is the case most of the time).
And don't calibrate it to D65 if you don't have controls for it on the device: just set temperature "as measured" and you have to live with that your white is not perfect but you'll get use to it.
How much is it now (you can check it quickly)?

[CarlosCaco]

You don't have to experiment different rendering intents: the default is the best if the gamut of the device is smaller than the intended one (which is the case most of the time).
And don't calibrate it to D65 if you don't have controls for it on the device: just set temperature "as measured" and you have to live with that your white is not perfect but you'll get use to it.
How much is it now (you can check it quickly)?

hey thanks for atention...

here my report using madvr ful intel gpu full and tv in graphic mode ( i think is the full mode on my tv)

3:16:00,246 Black level = 0.4396 cd/m^2
13:16:00,246 50% level = 67.27 cd/m^2
13:16:00,246 White level = 220.13 cd/m^2
13:16:00,247 Aprox. gamma = 1.71
13:16:00,247 Contrast ratio = 501:1
13:16:00,247 White chromaticity coordinates 0.3059, 0.3011
13:16:00,249 White Correlated Color Temperature = 7265K, DE 2K to locus = 10.0
13:16:00,249 White Correlated Daylight Temperature = 7287K, DE 2K to locus = 12.8
13:16:00,250 White Visual Color Temperature = 8104K, DE 2K to locus = 9.5
13:16:00,250 White Visual Daylight Temperature = 8548K, DE 2K to locus = 12.2
__________________

[huhn]

what TV is this?

and the i1d2 is maybe to old to be usable.

[CarlosCaco]

what TV is this?

and the i1d2 is maybe to old to be usable.

My tv is a Philco PH24D21D led, is a cheaper tv here in Brasil...


http://philco.com.br/index.php/produtos/info/1419/TV%20PH24D21D%20LED/

[huhn]

the result could be normal for that screen.

graphic mode is unlikely to be a full range mode.

you should start with this test: http://www.lagom.nl/lcd-test/black.php

try a 3D LUT with WTW clipping and set the target to d65.

if you still get a green tint with this the problem should be the meter.

[Asmodian]

I think your brightness might be too high, try turning it down so White level = about 160 cd/m^2 (or even down to ~120 if you are in a pretty dark room). Definitely use as measured for white point and level and set 0 for black level correction. Try the black point hack or compensation, or whatever it is called, too. :o

Your black level is very high, do not use BT.1886 (use gamma, 2.2 for a lighter room or 2.4 for a darker room) and expect some black crush (only ~19+ visible). BT.1886 with a black level that high will look washed out. With a contrast of 500:1 a "correct" calibration will look very washed out. :(

I agree with huhn, that i1d2 might not be useful... but that is a hard TV to calibrate well too.

[CarlosCaco]

I think your brightness might be too high, try turning it down so White level = about 160 cd/m^2 (or even down to ~120 if you are in a pretty dark room). Definitely use as measured for white point and level and set 0 for black level correction. Try the black point hack or compensation, or whatever it is called, too. :o

Your black level is very high, do not use BT.1886 (use gamma, 2.2 for a lighter room or 2.4 for a darker room) and expect some black crush (only ~19+ visible). BT.1886 with a black level that high will look washed out. With a contrast of 500:1 a "correct" calibration will look very washed out. :(

I agree with huhn, that i1d2 might not be useful... but that is a hard TV to calibrate well too.


I did a calibration today before read the answers

I get a good result on verify in dispcall with only withepoint not ok
The resto is ok ok, the best i can get was was with 2.2 relative and black output offset 40 ... By the way only absolute with white point scalling turn my whites green
Only relative intent give me ok ok on reports, all others give-me not ok on reports
No black crushed and 17 bar to above flashs
But on white clipping only 230 and 231 flashs,
I thought graphics mode was full range because in manual was discribed as the mode for computers, but its real too brighter... The tv only has 3 modes graphics and vídeo and auto when set video the image gets overscanned when set auto the image gets correctly just scanned and get the same bright and same level as vídeo mode...
I will post the reports on other moder later... Im really don t know what mode use and want yours opinions But i think i m getting closer to get a decent
Thanks for atention asmodian and hun, here in brasil i don t have anyone to talk about digital vídeo, my friends and my wife don t know nothing about this LOL

[CarlosCaco]

this is the reports on other mode in my tv , very different, i think is more correct

without black bbackground

22:31:45,229 Uncalibrated response:
22:31:45,230 Black level = 0.0859 cd/m^2
22:31:45,230 50% level = 50.89 cd/m^2
22:31:45,230 White level = 198.88 cd/m^2
22:31:45,230 Aprox. gamma = 1.97
22:31:45,230 Contrast ratio = 2316:1
22:31:45,230 White chromaticity coordinates 0.3100, 0.3022
22:31:45,232 White Correlated Color Temperature = 6949K, DE 2K to locus = 11.6
22:31:45,234 White Correlated Daylight Temperature = 6972K, DE 2K to locus = 14.2
22:31:45,234 White Visual Color Temperature = 7891K, DE 2K to locus = 11.0
22:31:45,236 White Visual Daylight Temperature = 8326K, DE 2K to locus = 13.5

with black background

22:33:16,970 Uncalibrated response:
22:33:16,970 Black level = 0.0497 cd/m^2
22:33:16,970 50% level = 40.64 cd/m^2
22:33:16,970 White level = 197.05 cd/m^2
22:33:16,970 Aprox. gamma = 2.28
22:33:16,971 Contrast ratio = 3964:1
22:33:16,971 White chromaticity coordinates 0.3105, 0.3024
22:33:16,973 White Correlated Color Temperature = 6914K, DE 2K to locus = 11.7
22:33:16,973 White Correlated Daylight Temperature = 6937K, DE 2K to locus = 14.3
22:33:16,973 White Visual Color Temperature = 7857K, DE 2K to locus = 11.1
22:33:16,973 White Visual Daylight Temperature = 8289K, DE 2K to locus = 13.6


what do you think guys?

[huhn]

the screen is now dimming or something like that.

it is hard to calibrated if the screen is dimming the backlight.
the measured CR should be higher with an black background not the other way around.

try the madLevelsTweaker.exe you can find it in the madVR folder. and test graphics mode again.

and look at this image and tell us what you see: http://www.lagom.nl/lcd-test/black.php

[CarlosCaco]

the screen is now dimming or something like that.

it is hard to calibrated if the screen is dimming the backlight.
the measured CR should be higher with an black background not the other way around.

try the madLevelsTweaker.exe you can find it in the madVR folder. and test graphics mode again.

and look at this image and tell us what you see: http://www.lagom.nl/lcd-test/black.php

the reports are from video mode that now i think is the correct mode... Its normal that diference between graphics and vídeo mode? Maybe i inverted the reports to with and without blackbackground, i will check and came here to confirm... BTW i was already using madlevelstweaker when did this reports...

I already take a look on the images on the link and with the two modes on full range i can see all squares...

[CarlosCaco]

Hun i checked again and my reports were wrong, i inverted the with without black background, i already edited the post to the right reports...
So now what you guys sugest?

[huhn]

hard to tell i wouldn't try to calibrated a screen with active dimming.

i would disable the dimming feature and calibrated.
after that i would test the dimming.

[CarlosCaco]

hard to tell i wouldn't try to calibrated a screen with active dimming.

i would disable the dimming feature and calibrated.
after that i would test the dimming.

sadly my tv don´t have any dimming feature option to disable :(

its really a bad ideia use the gpu rgb sliders to achieve d65?

[huhn]

just set the white point in dispcalgui to d65 this has the same effect and both lower the CR.

you have an delta e of over 10 there so the white will look more than strange.

your choice.

[CarlosCaco]

just set the white point in dispcalgui to d65 this has the same effect and both lower the CR.

you have an delta e of over 10 there so the white will look more than strange.

your choice.

Not a good choice BTW terrible results... I will have to found someone that have a better colorímeter and wait until i get money do buy a decent tv

[huhn]

the colorimeter isn't a bad one the problem is that it is old.

[markanini]

BT.1886 determines the tone curve of a display device according it's dynamic range. But what if, for any reason, the dynamic range can't be determined and you have to choose predetermined curve?

[huhn]

than the measurements are incorrect too.

[markanini]

than the measurements are incorrect too.
Maybe. I remember reading this article though http://www.chromapure.com/colorscience-gamma.asp and wondering if BT.1886 specified a fallback curve, or on what basis they made the statement "However, if you don't want to bother with a precise BT.1886 calculation, white/black values of 120/0.03 cd/m2 serve as a good rule of thumb.". I think the right context for this question should be devices that aren't color managed but have basic gamma and RGB controls, like some rooted android devices. For HTPCs, of course it's an irrelevant question with MadVr and Dispcal. So what do you think, are chromapure talking out of their ass?

[huhn]

i guess it's about none automatic calibration.
it looks to be pretty old too.

i'm definitely not in a position to say they are wrong.

[e-t172]

Maybe. I remember reading this article though http://www.chromapure.com/colorscience-gamma.asp and wondering if BT.1886 specified a fallback curve, or on what basis they made the statement "However, if you don't want to bother with a precise BT.1886 calculation, white/black values of 120/0.03 cd/m2 serve as a good rule of thumb.".

I can understand the 120 cd/m² white: that's a typical target luminance for calibration. I can't find an original citation for the number, but it appears in most calibration guides (http://www.tftcentral.co.uk/articles/content/calibrating_content.htm#reminder) and even in Wikipedia (https://en.wikipedia.org/wiki/Candela_per_square_metre).

I have no idea why they would suggest 0.03 cd/m² for black, however. That assumes a 4000:1 CR which is quite high end. Some VA panels will reach that number, but it would have been safer to assume an IPS panel (1000:1, so 0.12 cd/m²) which are more widespread. Though I guess that depends on whether we're talking about TVs or PC monitors.

I think the right context for this question should be devices that aren't color managed but have basic gamma and RGB controls, like some rooted android devices.

That wouldn't make sense. What matters is whether you can measure the actual black and white levels. Adjustments come later. If you can't measure the levels, then you're indeed stuck and you'll need to use "best guess" values. If you can measure the levels, then just use that to calculate the target BT.1886 curve; no need to guess. (Whether you can actually make the device reach that target curve using limited calibration controls, however, is a different story.)

To me this is quite an hypothetical situation: if you can't measure something as basic as black and white levels, then you can't do proper calibration anyway, so the whole discussion is moot.

[markanini]

On a different note, on my low contrast display I got about 2.0 effective gamma with BT.1886 which I wasn't satisfied with. Florian Höch suggested selecting 2.2 relative in DisplayCal and I was pleased with the results. You could spend a lot of time tweaking settings, I know I have, but this seems like a solid first recommendation when someone finds standard BT.1886 to bright, or want's something for general use.

[seiyafan]

If the display's contrast is reduced (i.e. black level is higher than native value) after calibration, what's the usual cause of this and how can it be fixed?

[e-t172]

If the display's contrast is reduced (i.e. black level is higher than native value) after calibration, what's the usual cause of this and how can it be fixed?

Some calibration software will try to correct the 0% white point to align it with the desired white point target (e.g. D65). Since the 0% point is already 0% on all channels, the only way to do that is to add some amount of red, green, or blue, which de facto raises the luminance of the black point, and therefore lowers the contrast ratio of the display.

With good calibration software this should not be a noticeable problem, because the human eye is not good at noticing color differences in very dark tones anyway, so it doesn't make a ton of sense to try to correct 0% black. Sometimes this behavior can be customized. For example, with ArgyllCMS, the dispcal -b (http://www.argyllcms.com/doc/dispcal.html#bhack) option can be used to force dispcal to leave 0% black alone.

Note that the same thing can also happen in reverse with 100% white (especially when calibrating toward a specific white point, say D65, as opposed to preserving the natural white point of the display), lowering its luminance and further reducing contrast ratio once again. However, in this case that is intentional and inevitable, since it is a natural consequence of deviating from the native white point of the display. The only way to avoid this is to pick a display that is already close to the desired white point in the first place.

[seiyafan]

Thanks for the explanation. Right now I have a Dell U2415 with a native contrast of about 1000:1, but it goes down to 750:1 after calibrating to a color temperature of 5800K. From what you said the contrast may improve if I calibrate it to say D65?

Another option which I haven't completely understood is the black output offset, I played with different percentages and it did seem to have some effect on the contrast even though not the biggest factor I don't think. Right now I am using 70% output. But how do I know which value is the optimum for a 1000:1 LCD monitor?

Update: You are very smart, I just calibrated to 6500K and contrast went up from 776:1 to 891:1

[Asmodian]

I have a 1000:1 LCD monitor (Acer XB270HU bprz) and my preferred results are 0% output offset and not correcting the black or white points at all, though my display does have an almost perfect D65 white point natively.

The percentage output offset adjusts how fast shadows lighten as you move away from black but does not influence the brightness of pure black. The ideal setting depends on your display and preference; 100% output offset has the darkest shadows. BT.2020 specifies 0% output offset so no detail is lost in the shadows but if your black level is too high this can cause a faded look.

If your white point looks "white" to you normally then don't try to fix it. Human vision adapts to the ambient "white" so small differences in white aren't noticeable after a few minutes. Traditional calibration methods are designed to match monitors to print work, for something like publishing a magazine. Video watchers like us want contrast a lot more than we care about perfectly matching to the appearance of colors in a catalog under natural (or artificial) light.

I know it sounds like you aren't correcting anything if you target the native white and black points but getting everything between them correctly placed relative to them can improve the image a lot without reducing contrast at all. I also suggest at least testing the default perceptual intent. A lot of guides seem to use an absolute intent but there is a reason perceptual is the default. :)

[e-t172]

If your white point looks "white" to you normally then don't try to fix it. Human vision adapts to the ambient "white" so small differences in white aren't noticeable after a few minutes. Traditional calibration methods are designed to match monitors to print work, for something like publishing a magazine. Video watchers like us want contrast a lot more than we care about perfectly matching to the appearance of colors in a catalog under natural (or artificial) light.

That's what you would expect in theory. Unfortunately, while that works great for passive objects, chromatic adaptation doesn't seem to work that well (http://www.color-image.com/2012/02/monitor-calibration-d65-white-point-soft-proofing/) with emissive displays:

The cognitive mechanisms in chromatic adaptation enable an observer to discount the yellow tint cast by D50 illumination on white paper and see the paper as white. This explains why D50 illumination for contract proofs has worked very well for the printing industry for decades. Unfortunately, the cognitive mechanisms in chromatic adaptation do not deliver the same benefits for images viewed on a color CRT computer monitor. [...]

Dr. Fairchild also noted in his research that chromatic adaptation was incomplete for observers who viewed a white patch displayed on a computer monitor with chromaticity coordinates near CIE illuminant A (incandescent light). To the observers, the white patch on the computer monitor retained a yellow appearance. If the observers were able to fully adapt, the patch would have appeared achromatic after complete chromatic adaptation.

Scientific research has shown that D65 is a good white point for color displays, including televisions and color computer monitors. Research has also shown that a color computer monitor calibrated to a D50 white point would retain a yellow appearance—chromatic adaptation would not be complete.

This is also suggested by data in this book (https://books.google.co.uk/books?id=5WkPPU-toHYC) (§3.5.3, "Preferred white point", page 146):

preferred white point ranges between 5800 and 7000K around a mean value of 6500K

I also suggest at least testing the default perceptual intent. A lot of guides seem to use an absolute intent but there is a reason perceptual is the default. :)

I don't think that's wise, especially if the source gamut of the transform is significantly larger than the colors that are actually present in the image (e.g. BT.2020), because perceptual rendering will try to "make room" for large swaths of the gamut that are seldom used, lower color accuracy and lowering effective gamut for the vast majority of colors that actually do get used.

[seiyafan]

Guess what? I bought a second Dell U2415 and calibrated using the same settings as the first and the contrast went up from 760:1 to 960:1. Problem fixed!
Last time I bought a monitor was in 2013 and I had to go through three before finding a decent one.