Display Calibration Basics
White point, luminance and tone response: the three targets a calibration sets, and the things it cannot fix no matter how good the instrument is.
Short answer: calibrating a display means driving it to three agreed targets. A white point, so that white is a specific white rather than whatever the panel felt like. A luminance, so that the brightest white is a known amount of light. And a tone response, so that the values between black and white land where the encoding expects them. Once the display is at those targets, a separate step measures what it actually does and writes that down as a profile. The first step changes the device; the second describes it. Almost every confused conversation about calibration is these two being mixed up.
The three targets
| Target | Typical setting | What goes wrong without it |
|---|---|---|
| White point | D65, roughly 6500 K, for screen work; D50, roughly 5000 K, where prints are compared alongside | Neutrals carry a tint, and the tint follows every color you pick |
| Luminance | Matched to the room, not to the panel's maximum | Images made on a very bright screen look muddy elsewhere, and the reverse |
| Tone response | The sRGB transfer function, or a plain 2.2 exponent | Midtones sit too light or too dark, and every gradient bunches somewhere |
The tone response is the one people get wrong most often, partly because the language is loose. The sRGB specification defines a piecewise curve with a short linear segment near black, and it approximates an exponent of about 2.2 overall without being exactly that. Treating them as identical is fine for most work and not fine for shadow detail. The reason any of this exists at all, and why the encoding is nonlinear in the first place, is on our gamma and perceptual uniformity page.
Software calibration and hardware calibration
There are two places the correction can live, and the difference is real.
Software calibration loads a correction into the graphics card's lookup table. The panel is untouched; the signal reaching it is bent. This works, costs nothing extra, and spends some of your available levels: pushing values around in an 8 bit pipeline can leave visible steps in gradients where there were none.
Hardware calibration writes the correction into the display's own internal lookup table, which is usually higher precision. The graphics card sends untouched values and the panel does the work. This needs a display built for it, and it avoids the banding problem, which is the same banding described on Mach bands and banding.
What a calibration session actually does
A colorimeter or spectrophotometer sits against the screen. Software displays a series of patches, the instrument measures each one, and the software compares what it asked for with what it got. First it adjusts until the three targets are met, using the display's own controls where it can and the lookup table where it cannot. Then it measures a larger patch set and writes a profile describing the calibrated state.
The measurement models involved are not trivial. Characterising a CRT was comparatively easy because the relationship between signal and light was well behaved and could be captured by a simple model with a gain, an offset and an exponent. LCD panels do not behave that way, and dedicated research was needed to model them accurately, which is why software written for one panel technology can quietly underperform on another.
Why your eyes are the worst judge here
Set a display too warm and look at it for ten minutes. It will look normal. Your visual system adapts to the dominant illumination and rebalances what counts as white, which is the same machinery described on our color constancy page. This is excellent for surviving a walk from daylight into a kitchen, and terrible for judging a screen.
Two practical consequences. First, an uncalibrated display that looks fine to you may be far off, and the mismatch only becomes visible when you put a second screen next to it. Second, immediately after a calibration a correctly set display often looks wrong, usually too yellow if you had been living with a cool panel. Give it a day. The adaptation runs in the direction of whatever you look at most.
What calibration cannot fix
Uniformity. Most panels are not the same brightness or the same white across their whole surface. A calibration measures one patch, usually in the middle, and the correction it derives is applied everywhere. If the corners are dimmer, they stay dimmer.
Viewing angle. On many panels the color shifts as you move your head. There is no correction for a value that depends on where you are sitting.
Gamut. Calibration cannot make a display show a color its primaries cannot reach. That is a hardware limit, discussed on sRGB, Display P3 and wide gamut.
The room. Ambient light reflecting off the screen raises the black level and lowers effective contrast. International standards for viewing conditions exist precisely because this is not a detail, and a well calibrated screen in a badly lit room is still lying to you.
Age. Backlights and emitters drift. A profile is a snapshot, and how long it stays true depends on the technology and on how hard the display is driven.
Observers differ, and it is measurable
Standard color matching functions describe an average observer. Real people vary around that average, and the variation becomes visible when a display uses narrow band primaries, as many modern panels do. Work published in 2014 modelled exactly this: observer variability and metamerism failure on electronic displays, where two people with entirely normal color vision can look at the same pair of screens and disagree about whether they match.
This is worth knowing because it sets a floor on what calibration can deliver. Matching two displays to instruments does not guarantee that a given human sees them as matching, and no amount of extra precision in the instrument changes that.
Do you actually need it
If you are producing work that other people will reproduce, print or match, yes, and the instrument pays for itself the first time it prevents a reprint. If you are comparing two of your own screens and want them to agree, yes. If you are a developer picking interface colors that will be viewed on thousands of uncalibrated devices, the honest answer is that calibration helps you trust your own eyes and does nothing for your users, and your effort is better spent on contrast that survives bad conditions. Whether the whole chain is managed end to end is a separate question, taken up on color management and ICC profiles.
The connection to the games
Judging a white point is a warm-versus-cool judgement, made harder by the fact that your visual system is actively working against you by adapting. That is precisely the task in Huember: decide whether a color is warmer or cooler than a reference, with no numbers to fall back on. Play it after switching your screen's color temperature and the difficulty is instructive, because the reference your eye is using has moved.
The theory behind why warm and cool feel like a single axis at all is on our color temperature page. And the usual caveat holds: the game measures you on this screen in this room today, not your visual system in the abstract.
Warmer or cooler? Huember asks for the judgement a white point setting is really made of, without a number to lean on.
Frequently asked questions
What is the difference between calibrating and profiling?
Calibration changes the display to reach chosen targets: a white point, a luminance level and a tone response. Profiling then measures the display in that state and records what it does. Calibration first, profiling second, and a profile made before a calibration change describes a state the display has left.
Should I use D65 or D50?
D65, roughly 6500 K, for screen work, because it is the white point the common encodings assume. D50, roughly 5000 K, when you are comparing the screen against a print under standard viewing conditions, because that is the white the print side is built around. Pick one and be consistent.
What gamma should I set?
The sRGB transfer function if your tools offer it, otherwise a plain 2.2 exponent. They are close but not identical: the sRGB curve has a short linear segment near black, which changes how the deepest shadows land. For most work the difference is invisible; for shadow detail it is not.
Why does my screen look wrong right after calibrating?
Because you had adapted to the old setting. Your visual system rebalances what counts as white towards whatever you look at most, so a correctly set display often looks too yellow to someone who had been living with a cool panel. Give it a day before judging.
Is software calibration good enough?
For most people, yes. It loads the correction into the graphics card rather than the display, which is free and effective, but it spends some of the available levels and can introduce visible steps in gradients. Hardware calibration writes into the display's own higher-precision table and avoids that.
Can calibration fix a screen that is dim in the corners?
No. A calibration measures one patch, usually centre screen, and applies the resulting correction everywhere. Panel uniformity, viewing angle shifts and the gamut limits of the primaries are hardware properties, and no amount of instrument precision changes them.
How often should I recalibrate?
It depends on the technology and how hard the display is driven, since backlights and emitters drift with age. Monthly is a common working rhythm for people whose output has to match something. If nothing you produce is reproduced elsewhere, the honest answer is that you may not need this at all.
Two calibrated screens still do not match. Why?
Possibly because you are the variable. Standard colour matching functions describe an average observer, and individuals vary around it, particularly with the narrow-band primaries used in many modern panels. Published work has modelled exactly this case: instruments agree, observers do not.
Keep reading
- Color management and ICC profiles: what happens after the calibration.
- Gamma and perceptual uniformity: why the tone response curve is shaped the way it is.
- Print versus screen color: where the white point choice starts to matter most.
Sources: International Organization for Standardization, ISO 3664:2009, Graphic technology and photography: viewing conditions; Poynton, C., Gamma FAQ; Poynton, C., Color FAQ; W3C, A Standard Default Color Space for the Internet: sRGB; International Commission on Illumination, CIE 015:2018 Colorimetry, 4th edition; Berns, R. S. (1996), Methods for characterizing CRT displays, Displays 16(4), 173; Day, E. A., Taplin, L. and Berns, R. S. (2004), Colorimetric characterization of a computer-controlled liquid crystal display, Color Research & Application 29(5), 365; Long, D. L. and Fairchild, M. D. (2014), Modeling observer variability and metamerism failure in electronic color displays, Color and Imaging Conference 22, 14.
Understand it, now train it. The daily games stay free forever. A Personal Pass adds progress tracking (trends + export); Teacher / School tools let you assign it and follow a whole class. For teachers & self-learners →