Incorrect calibration and profiling

Home Forums Help and Support Incorrect calibration and profiling

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #719

    anonymous SourceForge
    Member
    • Offline

    I have two monitors, a Samsung 305t 30″ IPS and a BenQ xl2411z 24″ TN. My OS is arch linux and I use dispcalGUI 3.0.0. I have an x-rite i1 Display 2 colorimeter and I calibrated and profiled both monitors on the same day with the same exact settings.

    Here are the verification reports:

    Samsung verification report
    BenQ verification report

    As you can see from those two reports, both report an excellent calibration with an avergae DE under 0.5 and with almost perfect gamma and grayscale. Also the colors are on spot on both monitors.

    But by comparing the displays with the naked eye using the same image, it is clear and obvious that the image on Samsung is (much) reddier and that the image on the BenQ has washed out midtones (brighter). The image is significantly different on both monitors.

    According to the verification reports, the image should be almost identical to the naked eye, but it isn’t. So what is the problem?

    If I used the same software and hardware and came up with different results it makes me highly doubt if my colorimeter is at all capable for a even a remotely correct calibration, as I have no way of verifying which display is more close to a correct image. Or even worse maybe they are both incorrect and a proper calibration is completely infeasible.

    #720

    anonymous SourceForge
    Member
    • Offline

    (I’m the OP, I just forgot to log in first)

    #722

    Florian Höch
    Administrator
    • Offline

    x-rite i1 Display 2

    How old is the instrument? The organic filters in these instruments degrade, and they are not produced anymore afaik so even if you buy one “new” it may be fairly old in reality.

    According to the verification reports, the image should be almost identical to the naked eye, but it isn’t.

    These specifc reports only test profile accuracy (closed loop test of how well or not the profile predicts the measured CIE values when displaying specific RGB combinations), it doesn’t say much per se about the accuracy of color transforms. To check the latter, enable “Simulation profile” and set it e.g. to sRGB on the “Verification” tab. Also, use a larger verification chart to get a more comprehensive idea about color errors.

    But by comparing the displays with the naked eye using the same image, it is clear and obvious that the image on Samsung is (much) reddier and that the image on the BenQ has washed out midtones (brighter). The image is significantly different on both monitors.

    Which program are you using to compare the images? How visually similar (or not) is fullscreen white on both displays?

    #723

    anonymous SourceForge
    Member
    • Offline

    How old is the instrument? The organic filters in these instruments degrade, and they are not produced anymore afaik so even if you buy one “new” it may be fairly old in reality.

    Well it’s pretty old and it is degraded. That’s why I tried to correct it by hand. I used a method of my own device, by measuring a 80% white patch on a THX certified mode of my Panasonic TV plasma I come up with a correction matrix (through trial and error) that was accurate to that. I then used that correction matrix (created a ccmx by hand) with dispalgui for the calibrations. I know this is wrong and that it is off, but I feel that even if it is wrong, I should get 2 identical calibrations even based on those wrong readings. At least that’s what logic says.

    Which program are you using to compare the images? How visually similar (or not) is fullscreen white on both displays?

    I use XnViewMP for linux (with the icc profiles correctly applied on KDE by colord ). 100% White is different, on samsung is warmer.

    #724

    Florian Höch
    Administrator
    • Offline

    Well it’s pretty old and it is degraded. That’s why I tried to correct it by hand. I used a method of my own device, by measuring a 80% white patch on a THX certified mode of my Panasonic TV plasma I come up with a correction matrix (through trial and error) that was accurate to that.

    Even if a matrix obtained in such a way were to give you an accurate result on the plasma, a colorimeter correction 3×3 matrix is always only valid for a specific instrument + display combination, so it’s only going to help when measuring that plasma.

    I know this is wrong and that it is off, but I feel that even if it is wrong, I should get 2 identical calibrations even based on those wrong readings. At least that’s what logic says.

    No, it doesn’t work that way. Ideally a colorimeter needs to be corrected for each display using a separate matrix. Only instruments whose filters are a good fit to the CIE 1931 color matching functions (i.e. the i1D3) will work well with generic corrections, and the i1D2 wasn’t one of them even when it was brand new.

    I use XnViewMP for linux (with the icc profiles correctly applied on KDE by colord )

    Hmm. I don’t see an option in XnViewMP to use the system display profile (atleast under Arch Linux, XnViewMP 0.72 from AUR). I had to set the user defined profile to the exact path of the display profile under “Settings” -> “General” -> “ICC”, and set the default profile to sRGB to enable color managed display.

    100% White is different, on samsung is warmer.

    If it’s a visually significant difference, I’d recommend re-calibrating one of the monitors to match the other visually, and then re-profiling it.

    • This reply was modified on 2015-06-20 23:31:45 by fhoech.
Viewing 5 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS