Hardware Profiling a Laptop LCD - Impossible?

Discussion in 'Digital Darkroom' started by rishij, Mar 11, 2007.

  1. This is regarding a consistent blue-->violet hue shift I get in all color-managed (CM) applications after hardware profiling my MacBook Pro (MBP) LCD.
    I use an i1 colorimeter + Eye-One Match 3.6 software.
    I've successfully calibrated, using the above package, my Sony CRT & a 20" Apple Cinema Display. Here are there profiles mapped against the sRGB color space & against each other:
    [​IMG]
    I then tried calibrating my MBP LCD, as did a fellow named Andrew on another forum, and we both keep getting these squarish (indicating a limited gamut) 3D plots of our calibrated profiles, which you can see here ('i1-calibrated LCD' profiles are the profiles generated from my own calibration of my MBP LCD using the i1 colorimeter; 'Color LCD' is the original canned profile for the MBP LCD provided by Apple):
    [​IMG]
    It's not the limited gamut that bothers me. It's the blue portion of the monitor gamuts that, as measured by the i1 colorimeter/software, lies *outside* the sRGB gamut.
    When using these i1-generated monitor profiles, blues turn lighter, and somewhat violet, in CM applications for both Andrew & I.
    Let me see if I can replicate this for you, below. Please use a color-managed browser (Safari); or, if you're a Windows user, please download the files via the links I've provided, and then view them in a color-managed application. If you just don't have access to a CM app, then, for you I have included the third image below, where I have converted my monitor capture of the blue square to the sRGB profile.
    Here's pure blue in the sRGB color profile (0,0,255)
    [a.k.a. What I SHOULD see as blue]:
    [​IMG]
    Linked Here
    Here's a screen capture of this same exact file as viewed on my MBP LCD (i1-generated ICC profile embedded)
    [a.k.a. What *I* DO see as blue, IF *YOU* are viewing in a CM app]:
    [​IMG]
    Linked Here
    Here's the above file (screen capture) converted to sRGB
    [a.k.a. What *I* DO see as blue, IF *YOU* are vieweing in a nonCM app such as IE or Firefox]
    [meaning for you Safari/CM viewers out there, this & the above square should look the same]:
    [​IMG]
    Linked Here
    As you can (hopefully) see, blues get lighter when viewed in CM applications on my MBP LCD. A slight violet hue is also introduced into blues, as can be seen here (exaggerated below, intentionally, by angling the laptop screen at a non-standard viewing angle):
    [​IMG]
    The square on the left (that appears purple/violet) is actually (0,0,255) in a sRGB document in Photoshop (color-managed, obviously). The square on the right is (0,0,255) in a ProPhoto RGB document in Photoshop. For some reason, if you create a new ProPhoto RGB or ROMM-RGB document and dump (0,0,255) paint in it, the color remains blue even when using my i1-generated monitor profile. Create the document in any other color space, dump blue paint, and you get the hue shift I've been mentioning.
    To further analyze where this hue shift is arising from, I took both the blue/sRGB & blue/ProPhoto RGB documents in PS, and 'Converted to' my i1-generated monitor profile (since, essentially, this is what PS is always doing on the fly as a CM application). Here's what I got:
    sRGB to monitor profile: (0,0,255) --> (102,0,255)
    ProPhoto RGB to monitor profile: (0,0,255) --> (0,0,255)
    For the blue in the sRGB document, the red channel was increased from 0 to 102! Hence the hue shift to violet. Not so for the blue in the ProPhoto RGB document.
    I also performed this experiment with my other i1-generated monitor profiles; that is, converting the (0,0,255) blue in my sRGB document to the following monitor profiles:
    Conversion to MBP LCD profile: (0,0,255) --> (102,0,255)
    Conversion to ACD LCD profile: (0,0,255) --> (55,0,249)
    Conversion to Sony CRT profile: (0,0,255) --> (16,0,255)
    As you can see, the violet hues being introduced to blues for Andrew and I are coming from the introduction of reds as Photoshop converts to our monitor profiles on-the-fly. This violet hue is much less noticeable on the ACD, and almost completely unnoticeable on the Sony CRT. Makes sense, given the values above, right? On the MBP, a 0% red in a pure blue image is being changed to 40% red before output to monitor, resulting in the hue shift to violet. On the ACD, a 0% red in a pure blue image is being changed 21% red before output to monitor, resulting in a slight hue shift to violet. On my Sony CRT, 0% red in a pure blue image is being changed to 6% red before output to monitor, essentially allowing pure blues to remain pure blue on the monitor.
    So, why the seemingly bad monitor profiles being generated by i1 for our MBP LCDs? Another user on the Luminous Landscape forum had this same sort of squarish monitor profile with an overextended gamut in the blues that lay outside the gamut of sRGB when he profiled his Pismo Powerbook years ago using the i1 colorimeter... see profiles here:
    [​IMG]
    Note, the 'canned' monitor profiles provided by Apple do not have this overextended gamut in the blues, and also do not display this hue shift with blues. However, the canned Apple profiles are rather inaccurate in other regards (e.g. there's no way the gamut of our displays is almost as wide as sRGB or as that of the iMac display)... I had crazy over-saturated reds/oranges on my MBP using the canned Apple profile... hence, I am mostly happy with the i1 calibration... except for the entirely inaccurate blues in CM apps. Yes, blues remain blue in nonCM applications, since nonCM apps just throw the (0,0,255) values at my screen rather than first converting them to (102,0,255)... which is just insane!
    Any help would be greatly appreciated, as none of us have solved this problem thus far...
    Thanks in advance,
    Rishi
    P.S. We calibrate in total darkness, with native white point and gamma set to 2.2. LCD panels are positioned as normal to the table surface as possible, so that colorimeter does not exert any significant force on the LCD due to its weight. Luminance is usually set between 90-120 cd/m2, but it doesn't really matter: I've tried every combination of 90, 100, 120 cd/m2 with every possible combination of native white point/D65 & gamma 1.8, 2.2... you name it, I've tried it, and still I get this ridiculous behavior with blues...
     
  2. "{It's not the limited gamut that bothers me. It's the blue portion of the monitor gamuts that,
    as measured by the i1 colorimeter/software, lies *outside* the sRGB gamut."

    So? Use a bigger colorspace than sRGB.
     
  3. That's not the point.

    I'm simply stating that I think it's *strange* that my laptop LCD's monitor gamut falls outside of sRGB, and I'm *wondering* if that's part of the explanation for why blues shift to violet.

    I solely work in ProPhoto RGB. But that doesn't change the fact that most people still post images in sRGB and my digital camera probably also uses some variant of the sRGB color space or Adobe RGB and my film scanner... well that's another long story.

    The *only* cases where blue stays blue is when I create a document in PS in the ProPhoto RGB or ROMM-RGB color spaces (yes, that's right, I tried creating a blue document in *every other profile* that exists in Photoshop). Pure blue in *every other* color space shifts hues. And I imagine that the only reason pure blues (RGB = 0,0,255) stay blue for documents in the ProPhoto RGB or ROMM-RGB color spaces is because the blue primaries for these color spaces are 'imaginary'. Hence, for some reason I don't pretend to understand, application of the monitor profile converts this 'imaginary' blue to a pure 0,0,255 blue, rather than a violet 102,0,255.

    The main point here is that blues in pretty much every image become lighter and shift hues in CM apps.
     
  4. Hi Roger, so we meet again... good to hear from you :)

    I'm not surprised that you don't see this hue shift... neither do I on my hardware calibrated CRT.

    Re: my monitor not being able to accurately display blues... I highly doubt that. In fact, blues are the one thing this monitor seems to have a very high response in. Add to that the fact that in a nonCM application, the blue sRGB file looks (to my eye) perfect. And by perfect I mean it looks pretty much exactly like the blue sRGB file in BOTH CM & nonCM applications on my hardware calibrated Sony CRT.

    So my MBP LCD is quite capable of producing that true blue that I see on my CRT... only when my i1-generated monitor profile isn't mucking up the blues by adding all that red...

    Re: Advanced settings in Eye-One for 'doing a look-up table'... Is that under 'Preferences' in the Eye-Match software? Because I didn't see any such option when I went through the 'Advanced' calibration...
     
  5. " I tried creating a blue document in *every other profile* that exists in Photoshop). Pure blue in *every other* color space shifts hues. "

    Hmm, that's really interesting. I wonder if Andrew Rodney or someone can comment on this.

    I created a blue square in Photoshop (0, 0, 255 with fill bucket) and then went to "assign profile" and went through the list. SRGB, EktaSpace, ProPhoto and AdobeRGB all looked quite "blue." My monitor profile also looked blue- pretty much identical to adobeRGB.
    I then assigned blue to SRGB and then looked and "conversion" to Adobe, ProPhoto, etc, and all retained their blueness. I'm using a hardware calibrated CRT.

    Your film scanner ought to be able to give you files in ProPhoto.

    It sounds to me like your screen may not be currently able to accurately display blues. In the Eye-One software advanced settings there is the option of doing a look-up table and I think that's recommended for LCDs rather than the default RGB curves. Did you try the different profiles and see if it made a difference?
     
  6. L G,

    Yes, the Mac does capture the profile altered data. It's the same thing as loading a pure blue (0,0,255) sRGB image in Photoshop, and then 'Converting to' the i1-generated monitor profile.

    Sounds like the profile for your MBP LCD matches up with mine. Please check your i1-generated profile in ColorSync Utility, and confirm.

    Given your post, I'm willing to bet you a nickel, maybe even a dime, that the three blue squares in this post all look about the same (the last, sRGB, one might be a little lighter) color in Safari, but that none of them look AS blue (as in neon-like pure blue) as the first square *as displayed by Firefox* on your MBP LCD. (In Firefox: I predict that for you the second square will be a little more violet, and the third one will be a lighter blue than the first square).

    Am I correct?
     
  7. How does a profile look like for a monitor at it's native state (Native/Native)?

    I would have thought it's an i-1 issue if Andrew over at LL wasn't using Spyder2 Pro and Basiccolor. There could be some videocard LUT weirdness happening that isn't measured correctly.
     
  8. lwg

    lwg

    Using color sync my profile looks very close to yours.

    Using safari the last two blues look the same. The first one looks like the firefox first one. It is darker blue than the other two, but not as intense as when displayed on the Dell LCD.

    You are right about how they look in Firefox.
     
  9. lwg

    lwg

    The profile for my Macbook pro has a much smaller gamut than the Dell 3007WFP that I use as an external monitor with it. The only area that the Macbook LCD has a wider gamut is in a very narrow range of blue. Both were calibrated with an i1 Display.

    I see no problems displaying blues on the Macbook, but the color is noticeably poor compared to other monitors.

    I don't understand how you got the screen captures. Does the mac capture the profile altered data?
     
  10. Aha! Now it's going to drive you crazy isn't it?! :)

    As for the colors in Safari, sorry, I meant to say exactly what you observed. The first one *should* look more 'blue' because its raw RGB values are (0,0,255) while the raw RGB values of the third square (which is the sRGB converted version of what MY MBP LCD is displaying) is (55,125,250).

    Yup, the shipping 'canned' profile reproduces blues just fine. The i1 fixes all other colors but messes up the blues. And the only common denominator I find in all of this is that 'extra strip' of blue in the monitor profile's gamut that lies outside of sRGB. I still suspect that has something to do with it and am willing to bet at least, hmmm, a quarter on it.

    Awaiting patiently for more ideas...

    Thanks for the confirmation L G. Now I know at least 5 people (including myself) that have noticed this problem.
     
  11. Wow, thanks for the dissertation, Rishi :eek:
    I'm *wondering* if that's [monitor gamut falling out of sRGB] part of the explanation for why blues shift to violet.
    Yes. Note that color shift is introduced when you move in a straight line from the sRGB blue (south-east tip of the "iceberg") to the origin in a straight line, to convert to display profile. This is due to the profile connection space not being perceptually uniform, and this has been described by Bruce Lindbloom in his The Blue Turns Purple Problem note.
    The more interesting question is whether profiles with such poor blue performance are unique to EyeOne Match. I would suggest you to download a 14-day trial version of BasICColor Display 4, profiling with it, and then verifying the accuracy of both your EyeOne Match and BasICColor profiles in the blue region (this software produces a nice chart with Delta-E 2000 errors).
     
  12. Great, Eugene, that's the kind of information I've been looking for. Much much appreciated. I will go ahead and give that page a read, and the BasICColor Display a try.

    As for the 'dissertation', let me tell you this is all the perfect intermittent distraction I need to try and learn advanced organic chemistry synthesis, in which I have my final exam tomorrow morning :)

    Cheers,
    Rishi
     
  13. advanced organic chemistry synthesis
    good luck
     
  14. Serge,

    Andrew also tried the DTP94, and got similar results.

    I tried clearing the LUTs using the 'ProfileMenu' utility that was suggested to you over at LL. I also tried Tim's method of first, before calibrating, setting a standard profile (such as sRGB) as your monitor profile instead of a 'canned' profile made from information downloaded from an EDID chip on the monitor. And then calibrating.

    No luck.

    Native/Native calibration put a very offensive blue cast over my entire screen in general (hence my shying away from it), yet STILL somehow caused CM apps to exhibit this blue-->violet hue shift.

    Thanks to everyone for trying to attack this strange issue... Agree with Roger -- a comment from a seasoned expert might be very helpful at this point... i.e. on a very basic level, what would prompt a color-management module (CMM) such as the Adobe CMM or Apple CMM to throw in 40% (102 out of 255) red to a pure blue tone before throwing out to the monitor?

    In other words, what exactly in a device color profile (specifically, monitor profile) would instruct a CMM to add 40% red to a pure blue tone to try and approximate 'pure blue'?? Sounds like insanity to me, but please correct me if I'm wrong!
     
  15. lwg

    lwg

    Trying the blue box in Photoshop I get the same results you are with the i1 profile. The shipping profile looks much better in the blues. Try dragging the small window across to the Mac and the profile isn't applied to the data until you drop the window. You can watch it turn purple.
     
  16. By the way, for anyone that's interested, here's a link to my i1-generated profile for my MBP LCD:
    Rishi's Profile
    -Rishi
     
  17. try attaching again
     
  18. Oops, trying to attach my monitor profile again:
    Rishi's Profile
    -Rishi
     
  19. Eugene,

    This particular description of the "Blue Turns Purple Problem" refers to when the output device is not capable of displaying the blues it is being fed.

    Not so in my case, right? Since the monitor profile shows a gamut *outside* the blues of sRGB, this means that the monitor is more than capable of displaying that blue. So, then, why would the color-management module be doing *any correction* whatsoever to a (0,0,255) blue primary?
     
  20. Roger,

    Whoa... I tried using ICC v4 & LUT (Large Monitor Profile) instead of Matrix (Small Monitor Profile) in the EyeMatch software.

    Using this LUT/ICCv4 option gave me a completely messed up profile... pretty much wiped out most of my screen, rendering my computer useless until I was able to fiddle my way back to applying a new profile in my OS.

    Loading up this monitor profile (LUT/ICCv4) in the ColorSync Utility, I see absolutely nothing in the 3D plot...

    Ever seen this?!
     
  21. lwg

    lwg

    I tried the BasICColor Display 4 package with the i1 Display and got very similar results. The error reported from the blues was no different than the errors for the other colors. So I would guess it is a problem with the i1 unit at this time.

    Has anyone here had a problem with another calibrator and the Mac Book Pro?
     
  22. Oops, I meant to say your monitor is capable of displaying *some* blue, but *not* the (0,0,255) sRGB primary.
     
  23. Finally, you have to realize that (0,0,255) ProPhoto RGB primary is
    a *different* color from (0,0,255) sRGB primary (and is found closer
    to south than south-east), which is why you don't gain red RGB
    values on conversion from ProPhoto RGB blue primary to display color
    space.
     
  24. Since the monitor profile shows a gamut *outside* the blues of sRGB, this means that the monitor is more than capable of displaying that blue.
    If you look at gamut comparisons you posted, you will notice that your monitor is capable of displaying *some* blue, but *not* the (0,0,255) RGB primary.
    So, then, why would the color-management module be doing *any correction* whatsoever to a (0,0,255) blue primary?
    Because your monitor blue primary was deemed by EyeOne Match to be a completely different color from sRGB blue primary. See picture attached.
    00KJKP-35454984.jpg
     
  25. Hahaha -- check out the timing of my last four posts -- it looks
    like Photo.Net switched to Daylight Savings a day and a half too
    late!
     
  26. "Using this LUT/ICCv4 option gave me a completely messed up profile... pretty much wiped out most of my screen, rendering my computer useless until I was able to fiddle my way back to applying a new profile in my OS.

    Loading up this monitor profile (LUT/ICCv4) in the ColorSync Utility, I see absolutely nothing in the 3D plot...

    Ever seen this?!"

    Nope, never seen this. I tried the LUT with my CRT on a WinXP computer and it looked fine. I would suggest contacting the manufacturer as this is very odd.
     
  27. Wow I was just schooled.

    Thanks, Eugene, for pulling my head out of my... I don't know why that didn't solidify in my head earlier... the endpoints being the primaries, that is.

    So, color-managing essentially consists of a *linear* re-mapping in this 3D space (essentially, a representation of the profile connection space)? Can this model be improved on?

    I just wish that color-management wouldn't touch my blues, since, by my eye, the color-managed blue on my Sony CRT (whose blue primary comes close to matching the sRGB blue primary, as per the i1-generated profile) looks like the *unmanaged* blue on my MacBook Pro. I guess there's nothing that can be done about this...

    Yeah, I can see that the blue primary for ProPhoto RGB lies outside the spectrum, and a linear remapping draws a line closer to the blues, not purples, of my MBP LCD profile.

    Thanks a ton, Eugene.
     
  28. Good question; I'm also very curious as to whether or not a LUT-based approach would help here... but I can't seem to get a non-corrupt LUT-based monitor profile...

    Any thoughts on why I'm having issues with a LUT-based approach with my MBP LCD?
     
  29. "So, color-managing essentially consists of a *linear* re-mapping in this 3D space (essentially, a representation of the profile connection space)? Can this model be improved on? "

    Wouldn't a LUT-based approach move beyond just remapping the RGB curves?
     
  30. "Most profiling packages do not allow LUT display profiles to perform the function of VCGT curves because that is mostly unnecessary."
    I didn't realize that.

    In my own experience with a Dell laptop and looking at the results of the profile above suggests even profiled laptops are unsuitable for critical photo work. Have you seen similar limitations with normal LCDs?
     
  31. Agreed, Roger, that even profiled laptops are unsuitable for *any* real photo work (as much as Apple would like to claim otherwise).

    I just find the blues-->purple hue shift rather irksome given that I didn't expect monitor profiling to get rid of one problem (oversaturation) to substitute it for another (hue shift).
     
  32. >>Wouldn't a LUT-based approach move beyond just remapping the RGB curves?</ p>
    What do you mean by "remapping RGB curves"? Matrix gamut conversions do not involve curves. Curves you see in the output of EyeOne Match (VCGT curves) only affect gamma of the monitor to bring it to a better behaved state. If you use a matrix display profile, the only conversion in software is made via a regular matrix multiplication, followed by a pass through videocard VCGT table. If you have a LUT display profile, colors are translated through a 3D table in software, followed by the same pass through VCGT. Most profiling packages do not allow LUT display profiles to perform the function of VCGT curves because that is mostly unnecessary.
    In my experience with several LCD and CRT monitors (admittedly no laptops), I haven't met a single case where a LUT profile would significantly improve upon a matrix profile. In all cases, the gamut of the LUT profile would look exactly like the diamond of a regular matrix profile. This is because the color model of LCD displays is still very similar to that of CRTs (additive color). Inkjet printers, on the other hand, absolutely require LUT profiles (subtractive color).
     
  33. Eugene,

    In the 3D color profile plots, for any particular line with a given hue angle, the further you go out from the center of the plot, the greater the saturation, yes?

    Indicating the extremely limited possible saturation of colors on my MBP LCD monitor (given the very small gamut).

    I understand that, given your example of linear re-mapping of the blue primary of sRGB to the blue primary of my MBP LCD, hue shifts are possible and, often, likely (when going from a larger gamut to a smaller gamut). But, what happens as you are moving along this line of given hue angle? Are any corrections to saturations made when converting from one profile to another?

    I would assume that color-management does affect saturation of images, or is this simply an effect of hue shifts? I say I 'assume' because with one (inaccurate) profile of my ACD, I have to super-saturate colors just to make them look decently saturated. These images end up looking oversaturated with blown-out colors on any other reasonable monitor. With the more accurate monitor profile (calibrated in darkness), however, I no longer need to adjust saturation in this ludicrous manner... images looked well-saturated to begin with. All this, of course, in color-managed applications.

    Thanks,
    Rishi
     
  34. That's what I've read also... that LUT profiles are for the print process.

    Since it IS printers that suffer this 'blue turns to purple' problem, I guess the question was whether or not a LUT profile might help solve this problem by not just selecting a color of similar hue angle (which, if I understand correctly, results from a simple linear re-mapping).
     
  35. Doesn't Photoshop have a feature to desaturate monitor colors by a given percentage? Would that be helpful?
     
  36. I'm just wondering whether or not changing the saturation of colors is a behavior that color-management modules engage in... at this point, for purely academic interest :)
     
  37. With the more accurate monitor profile (calibrated in darkness), however, I no longer need to adjust saturation in this ludicrous manner... images looked well-saturated to begin with.
    That's perfectly normal given your "daylight" profile looks larger than your "darkness" profile. I do not understand why the daylight profile is larger than the other one (would expect it to be the other way around -> with no reflected light we normally see more intense colors). But ignoring that, your "need" to oversaturate with daylight profile seems normal because color management tries to preserve "absolute" positions of colors (in CIE Lab space, which on the ColorSync 3D plots would look like a large gridded cube inside of which the RGB diamonds are housed). So what happens is that, with larger (incorrect) daylight profile, the individual colors become clustered closer to the middle of the RGB diamond, while in the smaller (correct) darkness profile, the colors become spread out in the RGB diamond. Note that in both cases the colors would be found in the exact same place in respect to the grid (CIE Lab color space). Now, assuming your monitor output is the same in both light and dark, you find yourself in the need to "spread out" the cluster in the middle of larger "daylight" RGB space, to achieve the approximately same distribution relative to the RGB diamond as you have in the "darkness" profile. This is equivalent to increasing saturation.
     
  38. In my own experience with a Dell laptop and looking at the results of the profile above suggests even profiled laptops are unsuitable for critical photo work. Have you seen similar limitations with normal LCDs?
    I wouldn't expect much from laptop screens. However, I did recently have a very unpleasant experience with a Dell 2007WFP (LG/Philips S-IPS panel), on which reds and yellows are practically nonexistant apparently because of very high native color temperature. The display looks especially bad compared with my NEC LCD2090UXi (another LG/Philips S- IPS panel). So it appears backlight is as important as the quality of the panel.
     
  39. Eugene, or anyone else who can comment,
    <p>
    What exactly do the grid lines within a color space viewed in the ColorSync Utility represent?
    <p>
    I created a couple of ProPhoto RGB working spaces with different, atypical, gammas, namely, 1.0 & 2.2 (ProPhoto RGB gamma is set to 1.8).
    <p>
    If you look at the white & black points for a given color space in the ColorSync Utility, I notice the following:
    <p>
    In reference to ProPhoto RGB Gamma=1.8:
    <p>
    Gamma=1.0:
    <br> -grid lines become concentrated near the white point
    <br> -grid lines become spaced out near the black point
    <p>
    Gamma=2.2:
    <br> -grid lines become more spaced out near the white point
    <br> -grid lines become concentrated near the black point
    <p>
    So are these grid lines representing how many bits get allocated to darks vs. lights? As I understand it, for working color spaces with higher gamma values, more bits are allocated to shadows and less to highlights. A working space with gamma 2.2 then is ideal since it distributes the bits over the tonal range in a manner that is optimized for human perception (i.e. more bits get distributed to the shadows, where our eyes are pretty sensitive).
    <p>
    I guess another way to look at it would be in terms of where the numerical R,G,B values map to within the color space:
    <p>
    Since there are exactly 8 distinct points along any line connecting black (or white) to a R,G, or B primary, then, in an 8-bit model, with 256 possible shades of any primary, this would mean that any two adjacent points along one of these lines are separated by 32 shades (of whichever primary we're looking at). Therefore, in the gamma=1.0 color space, there's a pretty large tonal range of greens covered in between values 0-32; whereas in the gamma=2.2 color space, the black to green transition is much more subtle (lower range of greens covered) in between values 0-32.
    <p>
    For clarification, see image below:
    <p>
    <img src="http://web.mac.com/rishisanyal/iWeb/Profiling/WorkingSpace_Gamma.jpg">
    <p>
    Many thanks,
    <br>Rishi
     
  40. Hey, forget everything I said about gridlines in ColorSync, because it's wrong. I was thinking of another utility (Linux-only) called iccexamin. The lines in ColorSync look like RGB contours, but do not quote me on that.
    Below is the plot I was thinking about when I wrote the previous post. As you can see, the horizontal axes are a* and b* parameters of CIE Lab, and the vertical axis is L*. So the entire CIE Lab space is represented as a cube.
    00KKwg-35487284.jpg
     
  41. I've replicated the purple blues on my CRT in AdobeRGB, by
    loading two SuperCaL profiles in my system. Each profile
    contains a different XYZ transform formula custom calculated by
    Brock Brandenberg, creator of SuperCaL, one using Vonkrie, the
    second, Bradford.

    Several years ago I was emailing Brock concerning similar
    issues discussed here and he suggested I supply my CRT's
    ROM chip XYZ data extracted with a small app called
    SimpleDDC. He said Apple requires Vonkrie calculations as part
    of their API, but as you can see with the two versions shown in
    the link at the bottom, Bradford is much more accurate and
    doesn't mess with 0,0,255 as evidenced in the Apple Digital
    Colormeter readouts.

    What I think where things got messed up with the Vonkrie
    version is that the colorant numbers generated off the ROM chip
    were originally calculated by the manufacturer of my CRT using
    Bradford and Brock not realizing this flipped the resultant values
    using Vonkrie.

    I emailed him about this, so he supplied the Bradford calculation
    which I had to embed within the app by using Brock's
    instructions hacking SuperCaL using ResEdit. As you can see
    the Bradford version is more accurate and leaves the 255 blue at
    full throttle.

    The image below is a screenshot with my assigned EyeOne
    Display profile of my CRT and should be viewed in a CM app.
     
  42. Eugene,

    Actually, I think you were right in the sense that a particular color in CIE LAB, which would be any given 'absolute' xyz coordinate in the the 3D space of the ColorSync Utility, is what is preserved. Naturally, as this color, in the Profile Connection Space (PCS), is what is meant to observed by the human eye.

    Monitors with larger gamuts have larger 'diamonds' within this 3D PCS, whereas my MBP LCD with a smaller gamut has a smaller diamond. If the grid lines represent RGB values, with (0,0,0) at the black center and (255,0,0) at the red extreme, (0,255,0) at the green extreme, and so forth...

    then, for any given xyz coordinate in the PCS, the RGB value in a smaller diamond would be higher, for any given primary, than the RGB value in a larger diamond.

    XYZ coordinates in the PCS that represent out-of-gamut values for any given profile (diamond), however, are dealt with as you demonstrated above... by drawing a line of constant hue angle, etc.

    I believe this is how viewing profiles in a PCS is meant to be done.

    This also explains why conversion to larger gamut display profiles generally leads to a decrease in apparent saturation (in an un-color-managed viewer); and why conversion to a smaller gamut display profile leads to an increase in apparent saturation (again, in a nonCM app).
     
  43. To add a quick note: I do experience essentially the same color shift using a Macbook (non-pro), calibrated with a Spyder2 colorimeter and its default software.
     
  44. Desist these purple hue shifts immediately, or you WILL force me to draw my weapon! You WILL do as I say. There have been no terrorist attacks in the U.S. soil since Jack Bauer has appeared on television. So perhaps we have a fair shot at obedience.
     
  45. I think it's time to call in Jack Bauer.

    He could figure this out in his spare moments disarming photonet WMDs. And, he has a
    nuclear-capable colorimeter in his manbag that could no doubt help a ton.
     
  46. I think it's time to call in Jack Bauer.
    Will he fix the PN daylight saving time issue?
    -----------------------------------------------
    Tim, it is well known that Bradford transformation is more accurate than von Kries. In Rishi's case von Kries is not to blame because Rishi observed the purple shift in Photoshop, and Photoshop has been using Bradford transformation for gamut conversions for a while now.
    On the same note, I do not think EyeOne Match uses von Kries either. The usual practice with monitor calibration software is to use either CIECAM 1997/2000 or Bradford, all of which have very similar coefficients (with minor changes).
     
  47. Aright, so, is there like a good reference I could read so I could communicate with you guys on your level?

    I feel I'm getting more and more lost as this forum progresses. I'm hoping that my last couple of (relevant) posts weren't way off target... but hard to know for sure.

    I mean, where did you guys learn all this stuff?

    Andrew Rodney's book was great, for beginners... doesn't help at all in having discussions at *this* level.

    Thanks,
    Rishi
     
  48. Yeah, I have a few lectures by Mark Fairchild I could email you..
     
  49. Eugene,

    You've been incredibly helpful, thank you.

    Please, when you get a chance, do e-mail me those lectures. My email address is: rishi.j.sanyal@gmail.com

    Thanks!
    Rishi
     
  50. Eugene,

    I didn't mean to indicate Von Kries is Rishi's problem, but I do
    suspect that the MBP panel type and quality may induce the
    colormeter and Match software to come up with funky transforms
    that may cause similar effects.

    I think it would be interesting to see how RGBCMY purity 0-255
    gradients appear in nonCM apps as compared to the CRT's. If
    they look the same then how can a hardware device come up
    with two so drastically different gamut plots for both.

    Rishi,

    I don't really understand the mathematics involved. I tried reading
    Bruce Lindbloom's articles on the subject but I can't get my head
    around how formulas can cause such issues with displays.
    Printers, I can understand.

    My input is strictly based on dealing with these kinds of issues
    since 1998 with several used CRT's of varying quality and only
    one Powerbook display and iMac LCD.
     
  51. Tim,

    My Sony CRT & MBP LCD were both calibrated with i1.

    Blue purity (0,0,255) color-managed looks the same as non-color-managed on my Sony CRT.

    Blue purity on Sony CRT looks darker and a bit more 'blue' (for lack of better words!) than blue purity in non-CM app on my MBP LCD.

    So i1 profile on MBP LCD is trying to compensate... by adding reds! Looking at Eugene's description on the 3D PCS plot, this behavior makes sense in terms of the manner in which the color-management module is working.

    But is it optimal for what I want to see? In the blues, certainly not. You may be right; there may be some strange way in which the i1 colorimeter/software is working with my MBP LCD... but I don't even know how to begin to tackle this problem.

    However, you're right, it would be interesting to see how RGBCMY purity gradients look in nonCM apps on my MBP LCD vs color-managed on my CRT (since I think that the i1 did a good job with my Sony CRT)... did I get you right in the experiment you were proposing?

    Perhaps this weekend after finals I will give that a shot and post my results.

    Cheers,
    Rishi
     
  52. Just one slight difference in the experiment. The purity gradients need to be examined in a nonCM app on both display's.
    What you want to get the two displays to show is what response/appearance the i1 colormeter is measuring from. To go even further it would be even better to examine these purities at the point the i1 Match software clears the vLUTs which usually has the appearance of the screen becoming slightly darker.
    The issue with doing that though is it requires both displays to be loaded in i1 Match at the section where the LUTs clear. There is a workaround using an app called Profile Gamma Tagger which allows the embedding of a ramped shaped=(without a gamma curve) response to any display profile, custom or canned.
    You can download it from here:Chromix ColorSmarts
    Pick "Linear" when it prompts you with the dropdown menu selection.
     
  53. I have a gradient chart made in Apple's sRGB that would clearly
    show the differences within the two displays because it's
    posterized to show 8 steps, but you can make your own if you
    wish.
     
  54. Sorry,

    Something else comes to mind that might have an influence on
    calibration I started noticing with cheap CRT's picked up at thrift
    stores. During a calibration session when the colormeter is
    measuring the color patches I noticed how fast some colors go
    to black and other colors, especially greens and blues leave a
    latent image or at least a fade effect compared to other color
    patches.

    I was told by a color expert years ago when assessing the quality
    of any CRT that some may have varying rise and fall refresh
    speeds along with what they call short and medium persistance
    green phosphors that can affect colormeter measurements as
    well as judging raster patterns using eyeball calibrators. This
    was to warn me in whether I bought the expensive Barco or my
    cheap $600 Princeton EO90.

    Since this is an LCD using a completely different technology to
    emulate CRT's P22 phosphor response and may be doing it
    badly, it might be the issue here.
     
  55. I have same problem (almost the exact same color shift) using a Monaco colorimeter and a Flexview (IPS) laptop screen (Lenovo T60). I had this problem with my old laptop (HP w/ WUXGA screen) and was hoping the change to an IPS screen would fix it.
     
  56. I've read on the web IPS screens are only made in 19" and above
    sizes. I've never heard of a laptop with an IPS screen.

    My 2000 Pismo which must have a TN panel has the same
    purplish blue using the EyeOne Display profile. Switching to the
    canned G3 Powerbook series profile corrects it but now all the
    other colors especially fleshtones are off.

    I believe the EyeOne software is trying to bend the scrawny
    gamut of laptops to allow accurate representation of a majority of
    colors at the sacrifice of a few.
     
  57. Yes, I believe Tim's right.

    I've tried all different software packages with my i1 colorimeter on my MBP LCD screen.

    Fleshtones and warm colors have been corrected due to profiling, as well as detail in shadows. However, blues turn purple. I guess that's a sacrifice I'm willing to make for the accuracy of other colors.

    I even tried 16-bit LUT display profiles, which actually helped render smooth tonal gradients with minimal posterization on my laptop screen. Still, blues turned purple.

    This is probably an issue that the software companies for the colorimeters would have to take up, if they give a crap about their products properly profiling laptop LCDs. They could probably add a tweak in the software to stop this behavior with blues.

    But who knows if it's worth it to them.

    Thanks for everyone's help. This was quite a long-winded and enlightening discussion thread.
    Cheers,
    Rishi
     
  58. Just a thought... not seen it mentioned in this thread, so i though i might share my findings.
    I too had the blue showing as purple issue, on my recently purchased 20" Cinema Display attached to my 17" Intel iMac.
    The default save location for ColorSync Profiles in the OS X is:
    /Library/ColorSync/Profiles/Displays/
    I used SuperCal to make a new profile this default location was
    ~/Library/ColorSync/Profiles/
    This was when i started noticing the blues being purple.
    It was most odd as it was appearing in some places and not others, name the OS X colour picker app:
    [​IMG]
    I ve had a bit of experience with Tiger being funny about permissions, so i thought it might be a permissions thing? some CM apps needed the right permissions to access the profile, if it was in the users library then they might not be able to access it?
    Anyway so I moved my calibrated profile to:
    /System/Library/ColorSync/Profiles/
    and hey presto blue, lovely blue.
    This might sound to simple but it worked for me! I hope it can help someone else..
    Thanks, Henry
     
  59. I don't understand how this would help.

    I *know* that Photoshop is using my i1-generated calibrated profile, yet it gives me purples instead of blues.

    Actually, re-reading your post, I'm sorry, I just can't make any sense of it. If you'd like, please clarify.

    Thanks,
    Rishi
     
  60. I just sampled your 255 blue in a nonCM browser using Apple
    Digital Color Meter and your blue looks correct but its RGB mix is
    76,34,255 which is basically a luminance difference between
    your SuperCal adjusted vLUT captured in the screenshot and
    how it reads straight off my frame buffer in OS 9.2.2.

    I started experimenting to see how much adjustment it would
    take to see a difference in 255 blue created in an AdobeRGB doc
    in Photoshop using PS's color picker and it seems you can add
    quite a bit of the red but not as much as the green until a
    difference can be seen.
     
  61. Hello, sorry, im not very good at explaining myself!

    Im not sure whether we were having the same problem.

    Bascially I have put the calibrated profiles in:

    /System/Library/ColorSync/Profiles/

    and deleted them from other locations, then relaunched System Prefs and reselected them.

    this made the system & CM apps show the correct blue for me.

    If i put them back, they go purple again.

    Ive have now used a i1 myself to calibrate 2 profiles for each monitor, which has helped no
    end to get similar colours on both screens, which is the idea i suppose. :O)

    does this help you at at?
     
  62. lwg

    lwg

    Henry, are you sure that your profiles are getting loaded from the new location? It is hard to believe the software cares where it is loaded from.
     
  63. Just recently getting to know OS 10.3.9. Everything I've read
    agrees with what Henry states, but only in regards to how the OS
    accesses ICC profiles. I can put profiles in a number of
    Library/Colorsync/Profiles locations-System or Administrator.
    Apple's Colorysync Utility will show them all in the list.

    It doesn't say anything about Photoshop being able to though.
     
  64. But how does this have *anything* to do with blues turning purple? We've already confirmed (I think) here that that it's the *calibrated* profiles that turn blues to purple...

    Plus, I *know* that Photoshop is accessing my profile because under 'Color Settings', in the drop down menu for RGB Working Space, Photoshop shows me: Monitor Profile - LaptopLCD_i1_D65_2.2-LUT, which is my i1 calibrated profile...

    Tim, I have yet to look at some RGBCMY purity charts. I was swamped with finals and had left this discussion aside, understandably, as it required too much brain bandwidth (well, at least, along with everything else). Also, your post about CRT phosphors and problems/interactions with colorimeters is interesting. I would have to check next time I calibrate whether or not some colors go to black quickly during the profiling process. At any rate, I don't claim to understand the technicalities of it nor why it would be affecting just the blues.

    Finally, interesting that you can increase brightness of blue by adding green and getting less subjective 'hue shift' compared to increasing brightness by adding red. But, according to how the compensation for out-of-gamut colors occurs in LAB (i.e. maintaining the hue angle), as Eugene Scherba has graciously shown in a previous post, blues necessarily take on a purple hue given the particular shapes of the sRGB vs. monitor color profiles in question here.

    Now I wonder if fooling with the LCD brightness when calibrating will change the behavior of blues or not. For example, what if I calibrate with my laptop set to full brightness?

    Guess I'll have to try :) Will let you know the results...

    -Rishi
     
  65. L G, Mar 31, 2007; 09:59 a.m.
    "Henry, are you sure that your profiles are getting loaded from the new location? It is hard
    to believe the software cares where it is loaded from."

    The profiles only exist in one place, ive deleted them from all other locations. OS X is
    heavily reliant on permissions, applications cannot access files unless they have the
    correct permission. This must apply to the CM system apps.

    Rishi Sanyal, Mar 31, 2007; 06:56 p.m.
    "But how does this have *anything* to do with blues turning purple?"

    My blues turn purple unless the profiles are in the right place?

    "Now I wonder if fooling with the LCD brightness when calibrating will change the behavior
    of blues or not. For example, what if I calibrate with my laptop set to full brightness?"

    Have you had any luck with this?
     
  66. Thanks for reminding me Henry... I completely forgot to try!

    Please come back and remind me again if I haven't tried this over this weekend. I'm losing my mind these days over school work... :)

    Cheers,
    Rishi
     
  67. -- Thanks for reminding me Henry... I completely forgot to try!

    Any luck still?
     
  68. Thanks for reminding me again... still haven't had a chance to try... however, perhaps part of the problem lies in what is outlined in this lawsuit against Apple...?

    http://www.appleinsider.com/articles/07/05/18/apple_hit_with_class_action_suit_over_macbook_macbook_pro_displays.html

    -Rishi
     
  69. Rishi,

    Now look what you've done and you dragged me in on this to
    boot. Soon we'll both be called as witnesses. Dang! There goes
    my summer vacation.

    Kidding aside, I can't understand what the big deal is about
    having a laptop display or any display for that matter be perfect. A
    commercial press for which the majority of pro color work is
    targeted vary's so much color accuracy can only be used as a
    guide. I'ld like to see how they're going to prove this.

    Interesting link. I wonder if it's just a rumor, though.
     
  70. Hey! For the record, I have nothing to do with that lawsuit :)

    I've been sensible enough to get a CRT & a 2005FPW to do my photo editing work...!

    Cheers,
    Rishi
     
  71. Just found and read through this thread—I'm seeing the same thing with a ColorVision Spyder 2 that I used to calibrate my
    display. Did anyone find a solution, or is it just a hardware limitation? I only calibrated my display to make it look better, not
    so much for color accuracy; is it possible to get CM apps like Safari to ignore the hue shift?
     
  72. Nope, never found a solution. I chalked it up to some strange manner in which the profiling devices interact with the laptop LCD & whatever jacked up algorithm they use for expanding the apparent color gamut. Perhaps this leads to the profiling package 'seeing' a blue primary for the laptop LCD that is very different from any 'standard' blue primary (pretty far off from sRGB's blue primary, as is evidenced above). Then, in correcting it, it mucks everything up.

    You know, what'd be interesting would be if someone tried this on a laptop that is NOT a Mac. Kevin, what's your laptop?

    I just received an email from another guy who has a Mac that says he has the same problem. I wonder if it's a Mac specific thing.

    Someone please try on a non-Mac laptop. I may also when I have a free moment :)

    Rishi
     
  73. I have a MacBook Pro, so I'm afraid I can't help there. I have seen reports of it happening to an HP laptop by searching Google, though. So
    maybe it is just a laptop thing. Seems like they'd fix that, though.

    Out of curiosity, do you end up using the canned profile or your calibration on the laptop display?
     
  74. I end up using the profile generated by i1 because for most other colors/saturation, etc., it works better. I.E. I get colors looking more like my Dell 2005FPW and pics uploaded to the internet look reasonable on other, uncalibrated monitors. I don't mess with blues when I edit my images on my MBP LCD, though, b/c I know my edits would/may be totally off.
     
  75. I have the same problem (blue turning to purple) with my HP 6910p (Vista 64) and the i1 display 2 (tried with 2
    calibrating apps: i1 Match and a demo version of ColorEyes Display Pro).

    And I also have the same problem with my "external" LCD (20'' MVA panel).
     
  76. Did you calibrate the external panel separately, or are you using the profile that you created for the laptop display? I'm by
    no means a color expert, but as far as I can tell, this issue is isolated to laptop displays—it'd be weird if you noticed it on
    your external display as well.

    Side note: New MacBooks in a couple of days. Here's hoping some sort of miracle happens to let us calibrate them.
     
  77. Both screens were calibrated separately, they have their own profile... this problem can also affect standard/external LCD monitors, not only laptop panels.

    If you look at some reviews here (most, but not all, show the gamut after calibration): http://www.behardware.com/html/cat/22/ you'll see that many have exactly the same problem (gamut not wide enough in the blues).

    Btw, I have now tried with a third app (basICColor): still the same problem.
     
  78. Dear all,

    I have the same trouble on three laptops (HP, HP and Acer, all under XP). I use Spyder 3 to calibrate the screens. It looks like "it is not a bug, it is a feature" ;) If it is a bug, then we should stick to the comment in this thread about color change rate and TFT elements response I guess. Some slow reaction if it takes place for laptop LCDs in blues will definitely drive any colorimeter crazy. All the rest is hidden in the profile connection transformations. I'm not common with in mathematical details but I assume there are general STANDARD ways which should give good mappings. Most likely LCD screens ARE NOT STANDARD in one or other way. Is it the matrix response or not: I guess not so many people know. What seems true is that changing the profile creation software will not cure the problem. If the hardware does not drive the colorimeter crazy, then, probably, one has to accept the LCD purple tint and try to find an LCD-specific profile mapping.

    Technically: the blue angle of my monitor(s) profile(s) triangle is just above (the same x, bigger y) the blue angle of the sRGB on the xy-chromaticity chart. NO ONE known to me rendering intent would raise blue point straight up. Thus ... What the laptop users may expect?

    Can people around recommend (or not maybe) the Dell 2408WFP? Is it a good piece for photos? I have an amazing 25% off deal right coming weeks?

    With best regards,
    Alex
     
  79. Yeh, Of course, when I wrote "LCD" I meant "laptop's LCD", sorry.
     
  80. I think this is a Snow Leopard related issue (at least on Apple laptops). I have been using Tiger, Leopard and now Snow Leopard on my old MacBook. In Tiger and Leopard the calibrated profile looked awesome. On Snow Leopard the profile looks really bad - especially the blues (=magenta). I'm using an i1 Display 2 with iMatch 3.6.3. I've also tried a trial version of BasicColor - no change. It's a known problem Snow Leopard won't handle ICC v4-profiles (yet), so we're stuck with v2. Anyway, I can think of several causes (unverified):
    1. Apple has changed the way color management works in SL = old software uses old API:s which no longer work as they should.
    2. Permission problems = maybe the LUT can't be written?
    3. Apple has color managed the GUI and made some mistakes?
    4. The display drivers are messed up in SL? Maybe they've got wrong native white point-values or other wrong parameters?
    5. Some other random bug?
    Once again - I've had my calibrated profile working really good on THE SAME computer, but on Tiger and Leopard - now it's not working anymore in Snow Leopard.
     

Share This Page