2017-05-08 at 16:10 #6923
Florian Höch (@fhoech)Administrator
In this post I’ll try to aggregate questions that come up with a certain frequency, adding to it over time as new topics come up.
Frequently Asked Questions
- Can I use software means to adjust RGB gains if my monitor doesn’t have hardware controls (e.g. on a laptop/iMac)?
- I have two or more displays and have calibrated all of them to the same parameters. While they match pretty well, there are differences in apparent contrast in the darker tones. Is there a way to make them match better?
- I have two or more displays and have calibrated all of them to the same chromaticity whitepoint target using appropriate measurement modes and/or corrections for the display technology types or models, yet the visual appearance still seems different after calibration. How come? Isn’t this a problem that calibration aims to solve?
- I’m comparing another calibration software package to DisplayCAL, and the readings are different despite using the same measurement instrument. What’s up with that?
- I’m comparing the results of another calibration software package to DisplayCAL, and certain aspects of the apparent color response do not match, even though I would have expected them to. Is this normal?
Q: Can I use software means to adjust RGB gains if my monitor doesn’t have hardware controls (e.g. laptop/iMac)?
A: Generally, no. These software solutions usually adjust RGB gains via the video card gamma tables (videoLUT), which conflicts with calibration. They cannot be used together.
Q: I have two or more displays and have calibrated all of them. While they match pretty well, there are differences in apparent contrast in the darker tones. Is there a way to make them match better?
A: Differences in contrast can be alleviated so that all displays have the same contrast ratio by adjusting them to have the same black- and white level. The default settings of DisplayCAL do not set a white- or black level target, and thus provide the maximum possible contrast with respect to the chosen whitepoint (moving away from the native whitepoint of a display will lower contrast). If all displays have similar or same native contrast characteristics, this will achieve a good match (provided you match the white level during interactive adjustment), but if some of the displays are capable of a lower black level than others, you may want to sacrifice their higher contrast for a better visual match, by calibrating all displays to the same black level (i.e. the lowest black level that all displays can achieve). You can do so by enabling advanced options in the options menu and setting a black level target on the calibration tab.
Q: I have two or more displays and have calibrated both of them to the same chromaticity whitepoint target using appropriate measurement modes and/or corrections for the display technology types or models, yet the visual appearance still seems different after calibration. How come? Isn’t this a problem that calibration aims to solve?
A: Before we come to a solution, it is helpful to generally understand what the causes of such a mismatch can be, what some of the underlying problems are, and what can be done to overcome them (and limitations thereof).
It is useful to distinguish between an apparent color gamut mismatch, whitepoint mismatch, and a combination thereof.
It is also helpful to understand the difference between calibration (which can be seen as part of device adjustment and like all other device adjustments needs to be done prior to profiling) and profiling (and how only the latter enables full color management when the resulting profile is used by color managed applications). See Calibration vs. Characterization.
If the issue is limited to a gamut mismatch, i.e. the whitepoints match visually, but color does not, due to the individual display devices different gamut sizes and shapes, then employing color management and limiting the material being viewed to the smallest possible gamut that all involved displays can cover, will overcome the issue. E.g. if you have two displays, and one can cover AdobeRGB (a wide-gamut color space), and the other only covers sRGB (a standard gamut color space), then limiting yourself to material in the sRGB color space will allow you to faithfully match color between the two displays (if they are accurately profiled).
It is important to note here that on Linux and Windows systems, the desktop and most applications are not color managed (even on Mac OS X, which employs desktop-wide color management, there are applications that do not manage color). Only a select few applications like some image viewers and photo editors may offer to color manage their output (and not all of them do it correctly or reliably!).
If the issue is a whitepoint mismatch, then some more explanations are in order.
Due to the way the human visual system works, we perceive color in relation to a white reference our eyes are adapted to (I won’t go into detail about mixed adaptation states here).
Generally, calibration and profiling are at their core based on tristimulus colorimetry, which in the case of display measurements means sending combinations of red, green and blue to the display and obtaining the corresponding, individual response in the CIE XYZ colorspace by measuring the light emitted from the display. The idea behind CIE XYZ is that the same XYZ values will produce the same color appearance on different devices despite different underlying light spectra (under the same viewing conditions), which is called a metameric color match – note here that the device RGB needed to produce said same XYZ may (and in most cases will) of course be different from display to display. CIE XYZ is derived by converting the measured light spectra (in the case of spectrometers) or the light filtered by three or more color filters (in the case of colorimeters) using the CIE 1931 2* standard observer color matching functions. The latter is a mathematical model of how the human visual system perceives color – each individual human’s color perception differs to a degree, also based on age and other physiological factors, so the model can be described as an “average” of human vision. It is important to keep this in mind to understand the inherent limitations.
As color instruments “see” color through these color matching functions, even if a match “by the numbers” is obtained, the visual appearance can still be different due to the variations inherent to each individual person’s perception. This will be more likely if the display technologies involved are different, and the emitted light spectra are more peaky due to wider gamuts.
So, how to overcome this problem?
It is relatively straight-forward (although may require a little practice) to match two or more displays that allow sufficiently fine-grained white point adjustment via their hardware controls (note that adjusting the videoLUT channels via potentially available graphics card driver options will not work as it interferes with calibration that uses the videoLUT as well). Pick one display as the reference, display a white (or light gray) patch on both of them, making sure not much else is on screen that could be distracting, and then adjust the other displays to match the reference display. You can use the visual whitepoint editor of DisplayCAL to have a nice, distraction-free white patch on each display. Afterwards, calibrate and/or profile the adjusted display normally (it is important to set the whitepoint target for the adjusted display in DisplayCAL to “As measured”, and if you’re creating a 3D LUT, setting the 3D LUT rendering intent to “Relative colorimetric” so that calibration or the 3D LUT do not change the whitepoint).
If one or more of the displays do not offer sufficient whitepoint adjustment controls, you can use the visual whitepoint editor to do the adjustment in software instead. The process is almost the same, except that after you have achieved a visual match, you hit the “measure” button to measure the white and set it as calibration target instead of using “As measured”.
You may be confused at this point and think, “but if we match the whitepoint by eye, doesn’t this make calibration and profiling redundant?”. The answer is, no it doesn’t, due to the way our visual system works, as described above, which perceives color relative to a white reference. Calibration and profiling will make sure that colors are reproduced correctly in relation to the whitepoint match we have just obtained.
Q: I’m comparing another calibration software package to DisplayCAL, and the readings are different despite using the same measurement instrument. What’s up with that?
A: Make sure to use the exact same measurement mode and/or correction (in case of colorimeters) in both software packages, and that it matches your actual display technology type. If you’re using a Spyder4/5, make sure to select a measurement mode matching the display type selection in the vendor software (e.g. “White LED”), and set correction to “None”. If you’re using an i1 Display Pro or ColorMunki Display, select a correction matching the display type selection in the vendor software, from the list of default X-Rite spectral corrections (e.g. “Spectral: LCD White LED IPS (WLED AC LG Samsung)” is identical to “White LED” in i1 Profiler).
Also make sure that what you’re measuring is exactly the same – DisplayCAL’s interactive display adjustment will always disable any active calibration and display a 100% white patch, and if you want to measure the currently calibrated whitepoint, you need to use use “Tools” -> “Report” -> “Report on calibrated display device”, or create a measurement report using the “Verification” tab. Other software, especially 3rd party products that are aimed at video and TV calibration, may operate differently.
Q: I’m comparing the results of another calibration software package to DisplayCAL, and certain aspects of the apparent color response do not match in my color managed application, even though I would have expected them to. Is this normal?
A: When comparing the results visually, it is important that the application you use to do the comparison (i.e. an image viewer/editor or a web browser) is employing color management and uses the respective display profile. It is hard to do side-by side comparisons when calibration and display profiles are involved, as only one calibration and display profile can be active (per display) at any given time, and applications usually need a restart to pick up on the changed display profile. It is best to use applications with a track record of a correctly working color management implementation (e.g. the Gimp, which uses littleCMS as its color management module, or Photoshop, which uses Adobe ACE).
As it is used so frequently, here are suitable
about:configsettings for Firefox color management:
gfx.color_management.enablev4 true(enables ICCv2 cLUT profiles as well, which is the type of profile DisplayCAL creates by default)
gfx.color_management.mode 1(color-manage everything and assume sRGB for images with no embedded profile)
gfx.color_management.rendering_intent 0(use the perceptual table of cLUT profiles, if available)
Note that Firefox is not multi display color management capable (a limitation that many color mamaged software packages have), so it’ll always use a single display profile (which is typically that of the primary display).
When it comes to test images, it is also important to understand how they are intended to be used, and what they can and cannot show. The often cited lagom.nl test images are not particularly useful in a color managed environment: They cannot show whether or not the display profile is accurate, and in case of an accurate profile, the displayed result will always look as intended.
The lagom.nl black level test for example, is meant to be used as a help to adjust a display, not to test ICC profiles, and should normally be viewed without any form of calibration or color management in use. It is also not particularly useful for digitally driven computer displays, where black clipping due to incorrect adjustment is not typically an issue (unless you deliberately or accidentally set it up wrongly).
A more powerful way to judge profiling results is to measure the actual accuracy of the profile. In DisplayCAL, this can be done on the “Verification” tab. Note that if you want to measure the accuracy of a 3rd party profile using DisplayCAL, you need to assign it to your display (using the means provided by the operating system) and select “<Current>” under “Settings” in DisplayCAL. Note that the profile needs to be an ICCv2 profile, ICCv4 is currently not supported.
Most of the differences seen between profiles created by different calibration and profiling packages come down to the accuracy (or lack thereof) of the created profiles. Profiles of similar accuracy are likely to provide a very similar visual result.