Optimizing teschart

2014-02-22
2014-02-26
  • Florian Hoech
    Florian Hoech
    2014-02-22

    Do you think it's worth tweaking the testchart? If so what parameters would you suggest?

    For a matrix profile, probably not. For a LUT profile, the higher the amount of measured patches the better usually the accuracy, although you may hit a point of diminishing returns at some stage. What generally seems to work well when using a custom chart is to choose a low number of single channel, some more neutral and a lot of optimized farthest point (or alternatively multi-dimensional body centered cubic grid) sampled patches. As a rough guide which combination of patch amounts to choose you can look at the pre-defined testcharts (or use one of them). For an already reasonably accurate LUT-profile the default testchart for LUT profiles should work (271 patches). If you want and don't mind the additional waiting time during measurements you can go higher of course.

     
  • John
    John
    2014-02-22

    I usually select monitors on reviews from this site but they also maintain a data base of brightness, contrast and RGB values that should provide a good starting point.

    http://www.tftcentral.co.uk/articles/icc_profiles.htm#h

    They haven't reviewed your monitor so the figures are from some one else. The reviews normally set contrast to the default value for calibration and give starting points for RGB values and brightness. Usually brightness can be adjusted a fair amount without upsetting the rgb balance much.

    John

     
  • roadtonever
    roadtonever
    2014-02-23

    John, I have noted dispcalGUI and Argyll documentation recomend the factory contrast setting in OSD. The TFT central settings recommend a higher value than factory. I did this earlier as it allowed me to reach my luminance target, But I tried again with the factory value and it seemed to produce a better calibration to my eyes.

    I have one more question. If I create an XYZLUT profile a matrix is included as a fallback. Does this mean applications that don't support the LUT won't fail to use the matrix?

     
    Last edit: roadtonever 2014-02-23
  • John
    John
    2014-02-23

    I don't know enough to answer that really. I have always used the LUT and haven't had any problems using the rev option. I understand the LUT aspect is handled in the graphics card.

    I would be inclined to forget your eyes and concentrate on the measurement report results and aim at the lowest errors you can get and ideally double ticks on all items in the summary.

    I have being having problems due to blue jumping when the initial rgb is balanced aiming for 6500K. This is probably because Dell choose to adjust 0 - 100 where as the hardware probably works 0-255. I found that the brightness can be adjusted making little difference to the balance so started using the as read options and now have a max delta E error of 0.8, average 0.25 but at 6600K where the rgb values do balance precisely. Previously I have found the black point compensation option helped but in this case some errors in the grey scale were always over 1.5 if that was used. Looks like it's best to start without that and add later if the rgb is all over the place at the dark end of the tone curve that is produced.

    I sometimes wish there was a little bit of layman documentation on what for instance the black point adjustments attempt to do. The lowest one shown in DispCal appears to attempt to tidy up serious rgb variations at low luminance levels but going on this monitor can make it worse if not needed.

    John

     
  • roadtonever
    roadtonever
    2014-02-26

    OT: Is the Spyder 3 Express supposed to be a slow device? Ran a default LUT testchart with high calibration speed and it took a bit over an hour. According to one guide this should take half an hour.

     
  • Florian Hoech
    Florian Hoech
    2014-02-26

    If I create an XYZLUT profile a matrix is included as a fallback. Does this mean applications that don't support the LUT won't fail to use the matrix?

    Basically with XYZLUT you have two options:

    • XYZLUT + swapped matrix. The swapped matrix inverts colors, so it will be very visible if an application uses it instead of the actual LUT.
    • XYZLUT + matrix. This will contain an asccurate matrix as fallback for applications that don't deal with LUT profiles.

    OT: Is the Spyder 3 Express supposed to be a slow device?

    The Spyders (well atleast up to the Spyder 3, don't know about Spyder 4) are notoriously slow unfortunately, especially on dark patches.

     


Anonymous


Cancel   Add attachments