What's in the AdobeRGB profile? (Q related to Samsung XL20/XL30 displays)

<p>Here's some photos of user's displays having uniformity issues with the new iMacs:</p>

<p>http://forums.macrumors.com/showthread.php?t=824940</p>

<p>Looks like they're being pretty picky, but I'm glad my 2004 iMac isn't as bad.</p>

<p>Hope your XL20's white balance is more neutral. </p>

<p>Just a note about the AdobeRGB space Emulation profile built using NCE, that software is having to build additional color tables separate from the simple LUT curves that control neutrality similar to what's shown in the X-rite png demo.</p>

<p>AdobeRGB emulation requires the hue and saturation color levels be reduced by comparison to the XL20's native gamut which is much larger and more saturated. If this software isn't doing this then it would be impossible to get identical previews of the PDI color target in both non-color managed and color managed apps. With this in mind then the NCE software has the ability to adjust the hue and saturation of the display internally to fit the AdobeRGB color response. Just want that clarified.</p>

<p>Sending you the file via yousendit. I usually just use my FTP, but don't want to do that for a public forum. Well I could just send it directly to your mail, but... Ah well :) The important thing is you'll get the file and can check how similar is the measured gamut to AdobeRGB. It should be quite close.</p>

<blockquote>

<p>With this in mind then the NCE software has the ability to adjust the hue and saturation of the display internally to fit the AdobeRGB color response. Just want that clarified.</p>

</blockquote>

<p>Yes. Actually now when I think about it, it's possible that NCE is doing just that (limiting the gamut), but only very little in terms of calibration. Hm. That would make kind of sense. I have to calibrate the display using the Calibrate tab and then measure the result, perhaps that will give a better result.</p>

<p>From the whole NCE package, I get the feeling the engineers had something much bigger in mind when designing it, but didn't get the whole job done. For example, after calibration, NCE saves an .icc profile as a file. Why is that even necessary if it always contains the same information? There is a missing link somewhere.</p>

<p>And OK, I'll give it one more try with that photo picture of the display, but you know that's not very accurate, right?</p>

<blockquote>

<p>Not sure if NCE is utilizing this function at startup or doing it live within its software and maybe X-rite as well but for the video card without a restart.</p>

</blockquote>

<p>I don't think that explanation you got is completely correct. Windows doesn't need to be restarted for the LUT to take place. The LUT (usually referred as a gamma curve btw, which is not precise) can be set or reset at any time. The problem is, most tools that apply the LUT only run on startup, so it seems like it can only be done at startup. Not really; when you run a videogame, it applies its own LUT and resets it on exit. If the game crashes, Windows gets stuck in any LUT the game has set. One of the problem is, applications don't read the current LUT when they change it, and don't bother changing it back; they just reset it. So if your display was calibrated using a custom LUT, that is now lost. A usual resolution is restarting Windows, but it's not technically the only way to do that.</p>

<p>The point is, X-Rite is using the standard Windows API for setting the LUT (just as games etc.). Whatever is Samsung (or nVidia) doing, they are using a completely different method, so Windows, X-Rite or any other application can't see or affect it in any way. So basically when I click before/after in X-Rite E-OM, it's applying its own LUT on top of Samsung's LUT. Is that what you meant?</p>

<blockquote>

<p>LUT for the video card at startup or the LUT for the display for calibration packages that manipulate the displays internal LUT through DDC (Display Data Channel) communicated through the display cable connection.</p>

</blockquote>

<p>Oh, wait. Now I got it. No, I don't think that's the case here either. When I reset the display to factory state, it gets a very strong magenta tint (as I saw yesterday). When I calibrate it using NCE, it becomes neutral. When I exit Windows and boot Linux on the same machine, it stays neutral. So most likely the changes are stored directly in the display and not drivers.</p>

<p>Samsung is using DDC as well, but apparently just for permanent storage in the display.</p>

<p>OK I figured out how to make a good photo of my display, so here it is. The two screens are XnView and Firefox. I've confirmed FF is color managed, because at first it showed a slightly different image - since I forgot to switch color profiles yesterday (the X-Rite one was still active). Here the images are matched, the slight difference you may see is because of the different resizing algorithm used by Xn and FF. They look identical at 100% size. On this photo I tried to match the right one (FF) to the original.</p>

<p>So when I see this photo and the original PDI target from at least 2 m distance, they look fairly the same except:<br>

- the white space is clearly pink (this disappears when I watch it for a minute, as the eyes adapt)<br>

- the Asian lady and the rightmost baby still have a slight orange tint to them. The two middle babies also bu much less so, hardly noticeable (picking hairs here).<br>

- I use neutral gray as my Windows skin, but here it's blueish/purplish<br>

- the yellow PhotoDisc logo is cooler and less saturated<br>

The color checkers look fairly normal, the blue tones are a tiny bit darker. Black is also a little washed out. I didn't try to match the rest of the image.</p>

<p>You should see the same I guess?</p>

<p>Photo.net gave me an error, here's the photo: http://img3.imageshack.us/img3/5195/cam8103.jpg</p>

<p>BTW it's made in AdobeRGB colorspace saved and tagged as AdobeRGB... Hopefully (I'm using an alpha version of RaWTherapee, so I hope it works).</p>

<p>They look identical but they look incorrect. There's way too much brownish yellow in the skintones and the red patch is way too orange. The baby on the far right is suppose to have a pinkish peachy tan with a bit of noticeable yellow.</p>

<p>I take it you didn't edit the image. However, this is the typical digital camera rendering when shooting a display without editing especially if it has an LED backlight like the Samsung's which is often quite blue and spikey in its spectrum that throws off the camera's AutoWB.</p>

<p>Thanks for the effort.</p>

<p>Below is a shot of my iMac of the two babies on the right. The file is in AdobeRGB so it needs to viewed in a color managed app. It shows how peachy pink with a tinge of yellow should be in the tan baby. To test accuracy I get identical CMYK numbers as in the original.</p><div></div>

<p>I've edited the image, but the orange 'tan' is something I couldn't eliminate (see my previous post). But in the middle two people I see it as negligible. When you mention the red patch looking orange... Um, i don't see a difference, but the RGB color values agree with you, it should show a little orange. I just don't see it, so it's either my eyes or the display not showing the difference. The latter might be a limitation of the display when it's trying to emulate a smaller gamut - which is happening mostly in reds. Some precision might be lost in there.</p>

<p>Your photo actually seems to be a closer match to what i see, but the other way around - a little cooler. I'll try again in calibration mode.</p>

<p>No, I see it now. I've swapped the AdobeRGB1998.icc profile for X-Rite's profile from yesterday as my display profile and compared the image in colormanaged FF. Your image is a much better fit and mine shows the symptoms you describe, including the orangeish patch.</p>

<p>So I guess when I calibrate my display with X-Rite SW, it's a good match to your calibrated display. Unless there's a difference between our calibrators at place, I guess Samsung calibration is less accurate. It doesn't show much with the X-Rite's test image, but the difference is more visible on this PDI image.</p>

<p>It's a bummer Samsung didn't go all the way, considering they use technologies for multiple profiles in lower displays. Thanks for your help with the troubleshooting Tim!</p>

<p>This is not an easy image to edit for accurate reproduction as well as getting a match on a commercial presses. That's why it was created. I've seen that orangey, peachy tan baby image printed on retail packaging for baby accessories and it was a dark, dull brown color, way off.</p>

<p>So are you saying the NCE emulation mode and supplied system profile aren't as accurate as the X-rite?</p>

<p>Below is a screenshot converted to AdobeRGB off my system of all three versions-the original, mine I posted above and yours. </p><div></div>

<p>This seem to be the case. Check this attachment - this is how Calibration mode measures. The difference in the darks is very obvious by naked eye (also in Emulation mode, which fired up my original question).</p>

<p>It's obvious that AdobeRGB mode is controlled by different circuitry than Calibration/Emulation mode. That's nice, but if Samsung would want them to look the same (apart from gamut), their calibration process should catch this difference.</p>

<p>But, perhaps the AdobeRGB mode is somehow specially tweaked to conform to Adobe's specifications or something. Not sure. I can imagine it helps when visualizing what the print will look like. Why do you think they would be doing it?</p>

<p>Also, to be fair, I've repeated the X-Rite measurement 3 times and the error of measurement seems to be quite high, almost as high as the measured difference. Also, when I click the before/after button, the difference in the PDI image is mostly in the magenta/green tint (just as indicated). But if you ask me, neither looks especially right. The green added by X-Rite makes the skin look a tad unhealthy, while the Samsung red tint looks like from a Canon digicam. The ideal is somewhere in between, but I'd pick the Samsung if I had to. I often have lunch at Vietnamese place and the Asian lady is closer to those guys.</p>

<p>But without having this picture printed as reference, or having you or someone with similar expertize look at it, I can't really know. It can be just the way my eyes adapt to the LED backlight. It reminds me about the story with the Hahnemuhle Baryta FB paper - http://www.luminous-landscape.com/reviews/printers/tale.shtml</p>

<p>That attachment.</p>

<p>Color temp/cast perception whether it's in the paper or light is not going to have that noticeable difference compared to wrong looking hue/saturation levels which requires selective color table manipulation by mathematical calculation to micro adjust the electronics of a video card and A/D converter affecting voltage going to a display. Every pixel on the screen has a voltage rating.</p>

<p>Believe me this is more complicated than rocket science as confirmed by an actual rocket scientist. Albert Einstein would be drooling and scratching his head if he ever saw the math going on under the hood of a color managed system and how it affects electrons and colorants.</p>

<p>A lot of folks I've noticed who think they have a color cast in their image OR display in reality have an over saturation and/or contrast issue. This happens to me quite often when I edit skin tone and pastel colors to match or look correct. For instance the first edit I did to the digital photo of my display posted above was to back off the saturation and contrast where most of the hue/saturation corrected itself. HSL and custom DNG profile took care of the rest. Color temp correction only affected slight increases of perceived yellow. </p>

<p>These threads below show photos of an Epson 2200 print of the PDI target . The print is from a $10,000 Sony Artisan calibrated professional digital imaging workstation owned and operated by Peter Figen, a professional photographer and fellow color management enthusiast I've had many conversations with in the past 8 years I've been posting on the web.</p>

<p>http://www.photo.net/digital-darkroom-forum/00UbBI</p>

<p>http://www.photo.net/digital-darkroom-forum/00Rr6Y</p>

<p>http://www.photo.net/digital-darkroom-forum/00ULME</p>

<p>http://www.photo.net/digital-darkroom-forum/00UT1m</p>

<p>You'll see the skin tone of that baby slightly change but never go way off due to adaptation affecting saturation adjustment. The images look neutral but are really just over saturated caused by adaptation that's making them look different.</p>

<p>Another issue to keep in mind as Andrew mentioned at the start of this thread is that X-rite's colormeter may not have the right filters that can accurately measure the XL20's LED backlight and its affect on pixel colorants. That Colorsync List discussion I linked to earlier states this by comparing these consumer grade calibration packages like the one's used here with a very expensive Lab grade spectrophotometer.</p>

<p>Also the software/driver like NCE may or may not be tuned/written to formulate accurate calculations with any colormeter whether X-rite's or Samsung branded due to quality control issues. It's all speculation at this point.</p>

<p>Those LUTs you posted don't tell us anything about how Samsung's software or hardware is manipulating circuitry within the display's internal electronics to control hue and saturation. Those curve LUTs only control neutrality within all 255 gray levels and slight tweaks to color temp shown in the curve offsets in the upper right corner of the graph.</p>

<p>The only way to test if AdobeRGB Emulation Mode is actually forcing the XL20 to reproduce the exact color response of AdobeRGB is to view an actual AdobeRGB written image according to ICC transform function calculation after a conversion and have it look correct. If it's off in color then the NCE software is not accurately controlling the internal electronics of the display no matter what LUT curve you derive from the X-rite or NCE.</p>

<p>The pinkish white point is an indication that the filters on the colormeter may not be tuned to the color response tolerances it's designed for as I said previously.</p>

<p>As for color accuracy, tomorrow I'll borrow a laptop and calibrate the display to see how they match. Or at least I'll be able to describe what the differences are. The laptop display is crappy, but at least it can give me some idea.</p>

<p>About that pink tint you see in the photo. When I look at the photo for long enough, that tint completely disappears and matches a white background of the display. Do you experience the same? If not, it might have something to do with the LED backlight and how the eyes react to it. Sometimes I think there's a pink shade to the white, but I only notice I watch a table of gray shades (like the bottom of the PDI image). It's very subtle, only in the white (not grays) and it disappears after a while.</p>

<p>One thing is certain - my D200 sees the white of my display as neutral when the WB is set to around 4600 K. You describe the photo as warm (after my edits to match it to reality), so that fits. Yet, when the ambient light is 6500 (as measured by both the calibrator and the camera, and in actual daylight around noon), the display is completely neutral, even when I compare it to actual white and grey things I have here, including a 'grey card' cleaning cloth.</p>

<p>This can have something to do with the LED backlight. Some people who replaced their incandescent light bulbs with LED lights sometimes say that other people look like corpses under that light - even when those LEDs are actually 6500 K.</p>

<p>So I speculate this is something which can be controlled by our brain thanks to the specifics of the LEDs. It can be something opposite to the effect of ultrasound: we can't really hear 25 kHz sound, but we definitely notice when such sound is turned off.</p>

<p>I understand calibrators need to be tweaked to specifics of LED (perhaps this is the reason?), but actually Samsung claims this calibrator IS tweaked (I haven't really believed it), so I don't know. I've read some of that discussion about calibrators, but the actual effect of LED backlight isn't really very well described AFAIK. I'll do some searching.</p>

<p>I'll also try to get my hands on some actual printed color target. I know a distant photo lab that actually makes them for their clients; perhaps they can send me one.</p>

<blockquote>

<p>About that pink tint you see in the photo. When I look at the photo for long enough, that tint completely disappears and matches a white background of the display. Do you experience the same?</p>

 

</blockquote>

<p>No, I don't. I see pink as I should because the RGB numbers say it's pink. I get a reading of 240,220,240RGB in the white part in question. True white on any display should look white even when you photograph it and give RGB numbers that are close to being identical. I get RGB numbers in my shot of my 6500K display of around 221,226,229 a bit on the blue side as it should since camera's AutoWB reference 5000K internally which is warmer compared to 6500K, slightly cooler but certainly not pink.</p>

<p>The fact you said your shot of your display looks pink but blends in to the white of your display is a red flag that your calibration is way off because 255,255,255RGB white will at least appear white even to a digital camera due to its AutoWB as it does with my DSLR taking shots of my iMac.</p>

<p>Kelvin number readouts measured with colormeters and digital cameras can't be trusted to represent a specific look because there can be several hues that can be considered a coolish 6500K and warmish 5000K and so on. If your eyes first sees pink then its pink.</p>

<p>This is why I think your colorimeter's filters aren't matched to the wide gamut response and LED backlight. Discussions on the web including the Colorsync Mailing List discussion linked above points to this as the cause. Or something is getting real corrupt in the building of the profile and how it integrates with the OS or within the internal electronics of the display.</p>

<p>I guess. That's probably the reason why I subjectively see the people in the PDI target as slightly reddish (but still look quite natural), while X-Rite software overcompensates and makes them look unhealthy green.</p>

<p>And the reason why only 100% white sometimes feels pink, may be that it's only when the LCD pixels are completely transparent, some red pollution from the RGB LEDs can come through to become actually identifiable. The calibrator probably sees too much of it and while NCE isn't bothered much, X-Rite cuts down on the red too much. All in all I agre with your conclusions. It seems to be a limitation of the display, but should be handled better.</p>

<p>Good. This all gives me enough data so I can formulate the questions for Samsung. I'll try to reach someone in the local factory who can perhaps contact me with some engineers far away... Maybe.</p>

<blockquote>

<p>That's probably the reason why I subjectively see the people in the PDI target as slightly reddish</p>

</blockquote>

<p>I can make the PDI target reddish in Photoshop on my system by assigning the XL_20 calib profile to the image. The pink of the white of the display is going to cause your eyes to see the skin as green due to adaptation.</p><div></div>

<p>That's like me switching to Calibration mode in a non-color managed app. No, there's not that much tint in AdobeRGB mode, but it's the same hue. Most likely the gamut in AdobeRGB mode differs from AdobeRGB98 specs enough so it's a little visible.</p>

<p>You're probably right about the pink white = green people effect. Also, this inspired me to test how other modes of my display compare when calibrated with NCE and E-OM (in colormanaged app, with the appropriate display profile set). I've created these profiles a few hours ago, so...</p>

<p>Calibration - NCE looks completely neutral with the people looking good. X-Rite profile gives a slight purple tint to the darker areas, visible in the grey table and the people. Also some obvious clilpping in dark areas thanks to the applied LUT.<br /> Custom - NCE doesn't apply, X-Rite profile adds pink tint to some shades of grey and green to the other. It's all over the place.</p>

<p>Both these profiles are utilizing the full monitor gamut and it seems X-Rite is doing shitty job calibrating them. AdobeRGB mode obviously behaves completely different to other modes. It's probably buggy with those flaws (mostly pinkish white, maybe imperfect gamut) and X-Rite is trying to compensate? I'm checking this now in complete darkness, if that changes anything.</p>

<p>By the way. Tim, again I thank you for all your help and I don't want to bother you anymore, unless you're curious about the outcome of this :)</p>

<p>I think you already explained the outcome of this. I'm all turned around as to what actually works but I think what you're saying is the NCE seems to give better results than the X-rite software probably on account that NCE is programmed to handle the wide gamut of the XL20. </p>

<p>Still not clear if you are using two different colormeters, one from Samsung and the other that came with the X-rite package.</p>

<p>But if you want to keep us posted on the final results or to summarize, we're all ears and eyes here.</p>

<p>I'm using just one calibrator - the Samsung branded one. I've just downloaded the X-Rite software.</p>

<p>I've used the calibrator (my 2nd unit actually) to profile other displays and it seems to be doing the job fine.</p>