Need for visual verification : adventures in color/colour correction land: Eizo, x-rite i1 +intel
I thought I would briefly write up a problem I had calibrating my screen to save future persons from this trouble, but also to highlight the need to visually verify calibration with onscreen visual guide pics. Googling I found no info on the specific problem I had (which I think is an intel video driver issue), and I suspect that because people are not using onscreen visual guides to verify, they don't realise that they have a problem.
In the past I've simply calibrated using my Spyder 2 device and trusted the result.
But in this case I had bumped up against this visual verification page:
and after using my x-write device found that things were not right if I followed the usual instructions.
The main calibration instruction that made a mess of my results was to "set the graphics card driver software to its defaults". I should mention that this was on the Intel mobile 4 Series express chipset on an Eizo FS2333 IPS screen (Review here). Those defaults are likely fine for the laptop's screen, but not good for external screens.
I reset to the defaults, and then followed the x-write instruction adjusting my screen via its OSD menus to adjust RGB gains, brightness, contrast etc as needed.
There were four problems that the resulting calibration profile produced. Using that web site I could see :
1)the gamma was way out (target 2.2)
2)more than half the contrast checkerboards were not visible (as was obvious just looking at pics which visually had blown highlights galore)
3)the x-write software reported a resulting temperature of 6750K instead of 6500 (selected D65 as target), and general accuracy was reported as somewhat out also (in contradiction of that prad.de review which shows this screen to be accurate enough for semi-pro photography).
4)almost half the blackpoints where not visible.
I tried endless permutations, including eventually adjusting the graphics card away from its defaults. Different driver version were to no avail.
Finally, I hit upon the solution.
I had to find a balance between the graphics card settings (in which it was necessary to adjust both gamma and contrast, to .9 and 40 respectively, likely intel specific) and also the screen's own setting, to a gamma of 1.8. The combo of the two conflicting gammas, neither at their defaults, is what finally got me the closest to a good result.
As it happens, the contrast is still two high on the graphics card, but I can't put it lower than 40. I'm missing just one contrast checkerboard and one blackpoint (here I am assuming that 255 and 0 are meant to be non-visible anyway, so it was 254 and 2 that should have been visible but were not). I think that this is acceptable.
The temperature is now 6476, which is close enough I think, and the gamma is just about right, very slightly off.
I'm still not so pleased with the result since it's evident that the contrast is still something of an issue. Both darker blackpoints and lighter checkboards are very hard to see. You might think that that is to be expected. But when I'm watching DVDs, the darks are blocked up, which is not the case on a TV.
Anyway, the moral of this tale of woe for me is to try out a Radeon or ATI based laptop next time.
correction, those blackpoint and checkerboard numbers are slightly different on the visual verification guide; the main point being that I seem to be missing one of each where the prad.de review shows them present (it even looks like they may be using the same guide).
hmmm, maybe the mods could correct the subject to "Need for visual verification : adventures in calibration land: Eizo, x-rite i1 +intel"
This problem illustrates an issue with monitor profiling. There are two techniques commonly used. One is to set the video controllers gamma conversion table to comp for the requested gamma (usually 2.2) and use a matrix conversion with RGB coordinates for color. The other is to use an unmodified table (ie, a one to one) and have the profile adjust the color and gamma based on a 3D lookup similar to the way printer profiles operate.
The former has the advantage of looking pretty good with non color aware apps like most all browsers. The latter produces better results with color managed apps since a one to one gain mapping produces less banding but much worse results with non color managed apps.
A Third approach is to use a monitor that does all the corrections internally to the monitor. Eizo photo monitors do this. This produces optimal results with both unmanaged and managed apps and always uses a one to one gains on the video controllers output luts. That's what I use. I have both wide gamut and sRGB target profiles set up on the Eizo but use the sRGB profile for all apps except things like photoshop on the infrequent occasion where sRGB is not sufficient.
An update for wandering google users: my problems were fixed by resetting to defaults the intel graphics driver in the 'advanced' section. Resetting to defaults in the normal intel settings had no effect; you have to find the advanced section which I notice on someone else's laptop doesn't exist.
Once I had done this the intel default contrast setting of 50 was perfect and my gamma issues went away. I can see all the black and also white squares. I also didn't need to constantly change the intel contrast setting, which would reset to 50 without my permission after suspends.
It's weird that this was necessary but I have a theory. On someone else's computer using the ADC feature really screwed things up. I reckon it was this that caused the problem on mine. And for whatever reason it couldn't be corrected from the monitor's OSD menus (though I have a new screen that has a 'factory reset' feature that might have worked). I now avoid using ADC in the iprofiler calibration. On the friend's computer there was no way to reset and the only thing that worked was to unintall the graphics drivers and reinstall.
I also discovered that my profile was being frequently unloaded. Because I had adjusted the RGB settings of the screen, as instructed by the x-rite software, the profile was sufficiently close to the unprofiled screen that I didn't realise that this was happening. To make sure I was seeing the profile I produced a test profile without adjusting the screen's RGB settings to D50 (warm paper). With that it was obvious when the profile was loaded or unloaded as the screen would revert to a cold sRGB 6500K. I'm not sure that the profile was even being loaded at all, independent of the iprofile software which I would use to force a loading. (For real D50 viewing (for print puposes) I use a fresh D50 profile with adjusted screen RGB settings since the test d50 profile wouldn't be optimal.)
Fixing this required two changes. Firstly removing the intel gamma loader utility and any others. Adobe Gamma, for instance. the intel one ('persistence') had to be removed via a registrty HKLM->Software->Microsoft->Windows-CurrentVersion->Run key, and adobe via the start menu "Startup" section where you right-click and delete. the intel module comes back with graphics driver updates. Watch out for that with Windows Update (see the 'System' section of Control Panel) if it is set to automatically install 'optional' updates.
Secondly adjusting an advanced option of windows color management as detailed here. Roughly-> Color management->Advanced->Change system defaults->Advanced (again)->Use Windows Display Calibration->tick. (Worth noting is that to change the current profile in the windows color managment window it only happens when you ake the profile the default; not intuitive. To see the unprofiled monitor untick 'use my settings for this device' which presumably doesn't work if you've set one of your profiles to be global in the advanced section.
All in all, I would defintely advise people to produce a test profile that is very different to the screen's unprofiled look, ie, choosing D50 instead of the default D65, and NOT adjusting the screen's RGB settings.
Now I have another issue. My set up is just fine, perfect gamma etc. But on a friend's two intel graphics laptops, it all looks good apart from the gamma which is way out. None of the above has made a difference. They have TN panels, but even at perfect perpendicular viewing the gamma is very wrong. I'll be back to report on that.
I should clarify that in my first post I say that intel's default graphics driver settings (in my case for contrast) were bad but while that was true with the messed up ADC, as documented in my follow up post, once the 'advanced' settings had been 'reset to defaults' this appeared to correct the ADC problem at which point the intel default settings are best for calibration/profiling, just as x-rite claims. But note that I haven't risked using the ADC feature again as maybe it's incompatible with my Eizo FS2333.
So, despite calibration, which is meant to also adjust gamma, two intel based laptops and Dell's 29" ultra-wide screen the u2913wm, appeared to have serious gamma issues.
On both http://www.lagom.nl/lcd-test/all_tests.php and other gamma comparison sites, these screens show significant gamma issues after calibration. However, I'm convinced that in fact it's the gamma visual tests which is the real problem. In other words, the screens don't actually have gamma issues at all.
What I've discovered on the dell is that the sharpness OSD control has a major effect on the visual gamma checks. Other settings can have some effect as well, such as pixel clock and phase, but to a minor extent (and auto-adjust seems to pick the right settings in any case).
It also seems that many online gamma checks are flawed irrespective of sharpness settings. Quite a number of pages make this claim of flaws (and of course that their's is better).
If after calibration the lagom.nl page/site is showing gamma anomalies then I would visit the two links below and use them to adjust OSD settings, like sharpness, to achieve a best fit. If the best fit is way way off then perhaps you have a genuine problem. In any case I wasn't able to remove a slight colour cast in the 2nd link, below, (by using the microsoft gamma utility) so perhaps my gamma is a bit out, which is also reported in the tftcentral review of the calibrated dell.
The 2nd may link may be the best of the gamma checks I've uncovered, although it doesn't tell you how far out you are.
Also worth noting with that 2nd link, it's actually for an srgb gamma which is not exactly 2.2 but a bit different. I calibrated with my x-rite both for gammas of srgb and 2.2 but didn't see an appreciable difference in the gamma check; the slight colour cast didn't go away.
Human vision is reliable enough to do the checks as given.
Calibrate with Argyll+dispcalGUI and after calibration is done, go to
menu Tools>Measurement report.
You've missed the point. The calibration device cannot verify itself. It may be faulty.
Also, as much as it may 'pass' in QA tests (ie, the iprofiler verison of what you mention) nevertheless there can be banding or other undesirable issues. These things can bee seen. In iprofiler, as far as I can tell it only checks colour accuracy in any case.
Trusting to one authority without another form of confirmation means that you can never be sure.
|Louvre Museum pyramid by Didier Quan|
|Oka Frozen Leaf 2002 DP by MarioSS|
from The Dead Leaves of Winter