Which temperature light is best for evaluating prints?

[digitaldog]

<blockquote>

<p>I always thuoght the first job of the viewing station is to have the print match the monitor...</p>

</blockquote>

<p>Correct. The idea is WYSIWYG. You need a print and a display to produce a match. </p>

<blockquote>

<p>We learned that whatever tempurature your monitor is calculated to, that's what you viewing station should be...</p>

</blockquote>

<p>Nope. The numbers are again somewhat meaningless. One set of values may produce a visual match in which the numbers you view the print and the numbers you calibrate the display to are not the same. In a perfect world, the numbers should match. Its not a prefect world for a number of reasons. A CCT 3400K Solux build and a D55 calibrated display <em>might</em> match. A CCT 5000K Fluorescent box and a D65 calibrated display <em>might</em> match. YMMV.</p>

[richard_pate]

<p>You are correct, Barry. These expensive viewing stations, like from GTI, provide a standardized lighting environment for print viewing that will closely match a calibrated monitor. And under these controlled "laboratory conditions" all is well in the world of color management theory.</p>

<p>But what if you viewed your perfect print under moderate to low intensity 3500K lighting, or 3000K lighting, the most common interior lighting? That print will look rather dull with low contrast and color saturation. Not so good if you would like to impress a viewer enough for them to actually purchase your work. </p>

[joe_c5]

<blockquote>

<p>But what if you viewed your perfect print under moderate to low intensity 3500K lighting, or 3000K lighting, the most common interior lighting? That print will look rather dull with low contrast and color saturation.</p>

</blockquote>

<p>This is probably not true. What I would expect from the different lighting is a different white balance and brightness, with the same contrast and saturation.</p>

<p>If the lighting affects contrast or saturation, then the printer inks or pigments are suffering from metameric failure, and you really may want a printer profile for the 3500K lighting.</p>

<p>Since there is no substitute for viewing the print under the intended lighting, I suggest calibrating the monitor to match that. Start by entering 3500K for the white point into the calibration software, and try adjusting up or down until it matches (probably best to match the color of the light and get the paper color through the “simulate paper color” option when soft proofing, but there may be some argument either way). If does not match at any color temperature, try entering x and y for the white point into the calibration software, which lots of software allows. Some combination of x and y will match, though it may take time to find it.</p>

<p> </p>

[Tim_Lookingbill]

<p>Prints can't mimic realistically with their reflective pigments/dyes the spectral characteristics of differing lights that affect how it renders hue/saturation/luminance on paper no matter how well you can manipulate the color tables digitally on a calibrated system. You pick a standard and you stick with it or else you'll be working your ass off. The standard is a neutral looking editing environment that doesn't add or take away from the source color.</p>

<p>My 2800K incandescent lights make my prints look so reddish yellow there isn't enough blue ink to make a dent in correcting for it. I've never seen what 3500K tungsten/halogen look like except the halogens I've seen at Home Depot. Don't know what their Kelvin rating was. Some are quite bright and somewhat neutral with a warmish tint to their cast. They were way better looking than my incandescent globes which goes to show the Kelvin numbers are meaningless because they're appearance is interpreted and manipulated by the manufacturer throwing any predictable and repeatable standard right out the door.</p>

<p>My advice, Richard, is to judge by your eyes just as your client will.</p>

[Tim_Lookingbill]

 

<p >These other threads I participated in and provided visual samples of the affects of different lights on prints will show how off you can get away with from the standard neutral appearance of your display. </p>

<p > </p>

<p >http://www.photo.net/digital-darkroom-forum/00ULME </p>

<p > </p>

<p > </p>

<p >http://www.photo.net/digital-darkroom-forum/00UbBI</p>

<p > </p>

<p > </p>

<p > </p>

<p >http://www.photo.net/photography-lighting-equipment-techniques-forum/00SN3Y </p>

<p > </p>

<p >http://www.photo.net/digital-darkroom-forum/00Rr6Y </p>

<p> </p>

[Uhooru]

<p>I see what you are saying Andrew. Theoretically they should match, 5000k is 5000k, but if the different lights actually vary then you would have to convert the values some how. I'm thinking that the different lights themselves should have some sort of equivalency table, otherwise, the whole thing doesn't make sense. Then, of course, you do it really scientifically and measure the temp and have a variable light and then match it. All I know, is that at school at least, the prints as viewed in the print stations were extremely close to the monitor given that one is reflective and one is illuminated.</p>

[joe_c5]

<blockquote>

<p>Theoretically they should match, 5000k is 5000k, but if the different lights actually vary then you would have to convert the values some how. I'm thinking that the different lights themselves should have some sort of equivalency table, otherwise, the whole thing doesn't make sense.</p>

</blockquote>

<p>For humans with normal color vision, the white balance of the lights should visually match if they have the same coordinates in LMS (Long Medium Short) color space. The International Commission on Illumination (CIE) does seem to be aware that CIE 1931 XYZ color space has significant limitations, but they have not exactly been speedy about moving to LMS as a conversion space.</p>

<p>Even a match in XYZ color space (which the xyY space from which I mentioned x and y above is a transformation of) really ought to be at least close. For any given “white” light, if a monitor is the correct brightness, then there should be some monitor RGB value that visually matches to within the quantization error. That matching RGB value can be selected by entering the correct x and y into the monitor calibration software.</p>

<p>Each color temperature is a line in xy space, as can be seen in the upper right hand chart at http://en.wikipedia.org/wiki/Color_temperature.</p>

[digitaldog]

<blockquote>

<p>Theoretically they should match, 5000k is 5000k</p>

</blockquote>

<p>Well sort of. Any Kelvin values is a range of colors. The range is pretty large too. </p>

[leighb]

<blockquote>

<p>Any Kelvin values is a range of colors. The range is pretty large too.</p>

</blockquote>

<p>Actually, a given Kelvin equivalent only measures the ratio of the red and blue components, totally ignoring the green. So three lights one of which appears white, one very green, and one very magenta, could all have exactly the same color temperature.</p>

<p>The "color temperature" concept comes from the fact that if you heat a black body (a piece of carbon) it will start to emit light. The higher the temperature, the bluer the light. The energy from this body has a uniform and predictable distribution throughout the visible and infrared spectrum. There are no significant spikes plus or minus from that smooth curve.</p>

<p>So comparing the red and blue components of the radiated energy is sufficient to define the entire spectrum FOR THIS TYPE OF RADIATOR. For this reason measuring the green component was never required.</p>

<p>Here's a graphic from Wikipedia comparing a 2800°K incandescent with a 5000°K fluorescent.</p>

<p>The radiated energy is shown as black against a spectrum, rather than white. The green spike in the fluorescent spectrum is quite obvious. The 2800°K spectrum does not look like a true curve because much of its energy is in the infrared, which is outside the range of the graph.</p>

<p>Sorry, guys...<br /> This brain-dead software won't let me post the image. Here's the link:<br /> http://upload.wikimedia.org/wikipedia/commons/b/b0/Spectral_Power_Distributions.png</p>

<p>- Leigh</p>

[leighb]

<p>Here's the graphic, I hope.</p>

<p>on edit: I give up. Why a small PNG file is "undisplayable" is beyond my understanding.</p>

<p>Click on the link to view the file.</p>

<p>- Leigh</p>

[digitaldog]

<blockquote>

<p>Actually, a given Kelvin equivalent only measures the ratio of the red and blue components, totally ignoring the green. </p>

</blockquote>

<p>Indeed, the lines of correlated color temp do alter roughly perpendicular to the true black body curve (which is based on a theoretical object), from green/yellow to orange/violet. I believe the illustration in figure 1, in this article is far more clear:<br>

http://www.ppmag.com/reviews/200512_rodneycm.pdf</p>

<p>Bottom line is, a CCT kelvin value is a range of colors. <br>

If you want to define an exact color, best to stick with a standard illuminant or an SPD.</p>

[Tim_Lookingbill]

<p>Leigh, this site will display jpegs inline within a thread as long as the widest dimension is 700 pixels and under.</p>

[leighb]

<p>Hi Tim,</p>

<p>I think the PNG file format is the basic problem. </p>

<p>I didn't convert it to a JPEG since it's not my image in the first place.</p>

<p>Thanks.</p>

<p>- Leigh</p>

 

[leighb]

<blockquote>

<p>Bottom line is, a CCT kelvin value is a range of colors.</p>

</blockquote>

<p>Correct.</p>

<p>In fact it defines the relative amplitude of all wavelengths in the visible spectrum using a single number.</p>

<p>Obviously that only works when the emitted spectrum matches that of the standard "black body" curve.</p>

<p>That's exactly the point I've been trying (unsuccessfully) to convey.</p>

<p>Thanks.</p>

<p>- Leigh</p>

[richard_pate]

<p>So gentlemen, in "light" of this in-depth discussion, would it be best for me to try to target the settings of my monitor calibration to emulate a 3500K print viewing environment, or would it be best to try to come up with equivalent printer profiles? Thanks for all the input.</p>

[photomark]

<p>There's a lot in this thread, so I'm a little hesitant to stir it up, but there are a couple things that could be a little clearer. First, there is an

important distinction between color temperature as described above as the temperature of an ideal Plankian radiator, and 'correlated color

temperature' which is based on the <i>perception</i> of color matches between a whitish light source that is not a black body radiator and the

ideal. Standard illuminant F8, for instance, has a 5000K correlated color temperature but because it is describing a fluorescent source, it has

a significantly different power distribution than D50. In this case 5000K is not really 5000K—or as Andrew Rodney said, "Well sort of."</p>

 

<p>From a practical standpoint, I've found that the white point of a display is the least of people's trouble

matching prints. Rather it's the luminance of the display and ambient light levels. I've used ImagePrint with it's alternate tungsten/daylight profiles and have found the differences to be so subtle that unless everything else is dialed in perfectly they would be perceptible to most people only if compared side by side. It's possible the media I'm printing on don't show much color shift between light sources, so the difference may be more dramatic on different paper—I don't know. Also the perceptual differences between the way we discount the illuminant

of a reflective print have been shown to be quite different from the way we discount the white point on a monitor. For those who might advise

changing the white point of the monitor to the color temperature of the ambient light, consider this from Mark Fairchild's <i>Color Appearance

Models</i></p>

 

<blockquote>

<p>

When hard-copy images are being viewed, an image is perceived as an object that is illuminated by the prevailing illumination. Thus both

sensory mechanisms, that respond to spectral energy distribution of the stimulus, and cognitive mechanisms, that discount the 'known' color

of the light source, are active. When a soft-copy display is being viewed, it cannot easily be interpreted as an illuminated object. Therefore

these is no 'known' illuminant color and only sensory mechanisms are active. This can be demonstrated by viewing a white piece of paper

under incandescent illumination and comparing the appearance to that of a CRT (or LCD) display of a uniform field with exactly the same

chromaticity and luminance viewed in a darkened room. The paper will appear white of just slightly yellowish. The display will appear relatively

high-chroma yellow. In fact, a white piece of paper illuminated by the display will appear white while the display itself retains a a yellow

appearance.

</p>

</blockquote>

 

<p>

It's just not as simple as basic colorimetry can at times make it seem—our perception of color is really complicated. Even if you can match

the cone response on the retina (as Joe C suggests above), it is no guarantee that we will perceive the same color—we need to consider

a host a psychological and physiological factors that are not accounted for in the basic color models used in everyday software. For a

dramatic example of this consider <a target="_blank" href="http://www.lottolab.org/illusiondemos/Demo%2014.html">this illusion</a>. Our

basic color models tell us that the center patch on the front and top face should be the same color, but clearly they look different.

</p>

[Uhooru]

<p>@Andrew<br />Thanks for the link and your patient answers. So if I understand the article you linked above (and I probably don't) the D standard is a standard illuminant and therefore if you calibrate your monitor white point to D65 and use a reviewing light also rated at D65, the viewing environment should be identical or nearly so to the monitor? </p>

[joe_c5]

<blockquote>

<p>So gentlemen, in "light" of this in-depth discussion, would it be best for me to try to target the settings of my monitor calibration to emulate a 3500K print viewing environment, or would it be best to try to come up with equivalent printer profiles?</p>

</blockquote>

<p>Myself, I would calibrate the monitor to match the lights first because I think it would be easier. I would only bother with a 3500K printer profile if it still seemed to be necessary. If time and/or money allow, you can do both since they are not mutually exclusive.</p>

<p>Without knowing how you personally value either the costs of these two actions or the benefits of different degrees of color matching, I cannot speak to what course of action is “best”. I think calibrating the monitor to match the lights might take as little as an hour if you match the RGB values first and then measure that x and y with the sensor, that coming up with a lighting specific printer profile would take significantly longer to do for the first time, and that either one by itself should get you matching that is close enough to keep wasted paper and ink to reasonable levels.</p>

<blockquote>

<p>Even if you can match the cone response on the retina (as Joe C suggests above), it is no guarantee that we will perceive the same color—we need to consider a host a psychological and physiological factors that are not accounted for in the basic color models used in everyday software.</p>

</blockquote>

<p>For a perfect match, you would need to match the cone response of the surroundings as well, not just the image, as well as the rod response. If all three cones and the rods have the same response, then the other psychological and physiological factors will also match because the inputs are indistinguishable. For completeness, that should include three dimensional frames matching the ones that will be used in the gallery, including glass, a monitor finish matching the gloss of the print, and fake backdrops around the room to look like the gallery. Having been serious for the entire thread, that last sentence is true yet completely over the top.</p>

<p>I really think that just the cones would be close enough with sanity level matching of the surrounds instead of pitch black behind the monitor and white behind the print.</p>

[digitaldog]

<blockquote>

<p>So gentlemen, in "light" of this in-depth discussion, would it be best for me to try to target the settings of my monitor calibration to emulate a 3500K print viewing environment, or would it be best to try to come up with equivalent printer profiles? </p>

</blockquote>

<p>This is being made more complex than it needs to be. Forget the numbers, they are rather meaningless in describing an absolute color value. We are not working with true backbody radiators. We are working with differing reference media (prints and emissive displays). Target the values such that you get a visual match. Your viewing booth is probably a fixed entity. Say a Solux CCT 3500K bulb. What target white point (defined in a standard illuminant, kelvin or XYZ) and luminance value (defined in cd/m2 or NITS) produces a visual match? Too cool on the display? Lower the white point value to get a match. Too dark a display compared to the print? Up the target calibration luminance (or move the Solux bulb farther from the print). <strong>The right numbers are those that produce a visual match. End of story. </strong></p>

<p>Most printer profiles are built to assume a D50 illuminant. Few products (notability ProfileMaker Pro) allow one to measure the light source, save out a CFX file and build the profile using that data instead of D50 (some prebaked CFX files are also available in the app). It makes a small but visual difference but its a tiny tweak <strong>compared</strong> to having the screen and print match using appropriate target calibration aims (white point and luminance). </p>

[Tim_Lookingbill]

<p>DON'T CALIBRATE YOUR MONITOR TO MATCH 3500K HALOGEN LIGHTS! PERIOD!</p>

<p>You will be reducing the color gamut of your display in doing so.</p>

<p>Geez! People! You're making this more complicated than it is.</p>

<p>The backlight of your display isn't halogen. It's either fluorescent or LED depending on your model. Both have varying spikes in their spectral distribution make up. Both can't replicate the spectral distribution characteristics of a halogen light, black body radiator or daylight exactly. It will never happen.</p>

<p>You might be able to replicate the color cast, but the spectral distribution portion which mainly affects levels of hue/saturation of individual colors within an image can't be emulated without severe contorting of the video LUTs and/or mathematical matrices written into the calibration/profile ESPECIALLY WHEN USING A CHEAP COLORIMETER to build the display profile.</p>

<p>Combine that with the severely warmish color cast and you've reduced the gamut of your display to the point certain colors like greens, blues, aquas desaturate or change hue and reds and oranges start to glow where no editing in the world will fix. You'll be maxing out those Hue/Saturation sliders and get nothing. </p>

<p>Your display is an editing station, not a light emulation station. You work neutral or not at all.</p>

[Tim_Lookingbill]

<p>Maybe this will make it more clear.</p>

<p>Top image is a shot of an inkjet print off my Epson NX400 and how it appears under 4700K/D50 Solux task lamp. Bottom is how the same print appears under 2800K incandescent.</p>

<p>Making a display emulate any other light other than NEUTRAL looking D50 will be difficult because of the various spectral reflectance properties of different lights, inks and papers.</p>

<p>I calibrate to 6000K just to reduce the intensely annoying blue violet cast making my display appear a lot more NEUTRAL. Kelvin number is meaningless because as you can see the 4700K Solux print looks neutral and brings out the spectral properties in the inkjet rendering a level of hue/saturation/luminance that matches the image viewed on my 6000K display.</p><div></div>

[joe_c5]

<blockquote>

<p>DON'T CALIBRATE YOUR MONITOR TO MATCH 3500K HALOGEN LIGHTS! PERIOD!<br>

You will be reducing the color gamut of your display in doing so.</p>

</blockquote>

<p>The first line is a suggestion, and therefore at least arguable. Personally I doubt it merits all caps, exclamation points, or the word “period”, but I’m going to try a calibration to match a 100W tungsten bulb and see what happens.</p>

<p>The second line seems inaccurate. The worst thing that is likely to happen to the display is increased posterization (thresholding or banding). The gamut should not significantly be changed. Well, the brightness of blues and whites will be reduced a lot; I suppose in three dimensions that is a reduced gamut, but monitors tend to be brighter than they need to be to begin with. Note that the brightness of blues and whites will also be lower for the prints viewed under this lighting.</p>

 

<blockquote>

<p>Maybe this will make it more clear.<br>

Top image is a shot of an inkjet print off my Epson NX400 and how it appears under 4700K/D50 Solux task lamp. Bottom is how the same print appears under 2800K incandescent.</p>

</blockquote>

<p>Nearly everything I have said in this thread applies only to human eyes. For cameras, instead of LMS, the camera’s RGB response would need to match, plus more importantly in this case, the camera’s white balance would need to behave the same way as human eyesight.</p>

[Tim_Lookingbill]

<p>Wasn't directing my rant at you in particular, Joe. Sorry, you took it that way.</p>

<p>Reduction of gamut is inevitable when using a calibration package that expects a certain standard or response (AdobeRGB or sRGB gamut) from a display. Along the lines of what I said before, deviating from this standard by calibrating a fluorescent/LED backlit display to match a 3500K halogen light source is pushing the limits and capabilities of colorimeters and their software in accurately compensating for the errors to hue/saturation/luminance in color managed images caused by the completely different physical properties of these two light sources.</p>

<p>Also it is well known from reading web postings on this that colorimeters and calibration software have difficulties even matching exactly two displays in appearance using 2.2 gamma, 6500K, 110-120 cd/m2, imagine trying to get them both match to 3500K. </p>

<p>And about human eyesight. Can you tell me why the white painted walls in the two images I posted look completely different from the white inkjet paper? I can tell you it has nothing to do with human eyesight.</p>

[joe_c5]

<p>I just tried to calibrate a monitor to x=0.443 y=0.393, but for reasons I have not been able to figure out, my video card look up table would not take the resulting values. There might be hardware limits or sanity checks or both.</p>

<p>I would not want to turn down the green and blue on the monitor because that takes more steps than switching between profiles.</p>

<p>So now, after having tried it, one would need to either be luckier about what their hardware would do or far more motivated than I am in order to get that to work.</p>

<blockquote>

<p>And about human eyesight. Can you tell me why the white painted walls in the two images I posted look completely different from the white inkjet paper? I can tell you it has nothing to do with human eyesight.</p>

</blockquote>

<p>That is because of a combination of the camera’s three spectral response curves, the spectrum of the lighting, and the reflective spectrum of the paper and white paint. Neither LMS nor XYZ match the response of the camera. A typical digital camera therefore does not capture enough information to reconstruct what a human eye would see.</p>

<p>Similarly, what they look like to the eye is because of a combination of the rod and three cone responses of the eye, the spectrum of the lighting, and the reflective spectrum of the paper and white paint. The paper and wall colors in LMS space should correlate very well to what a human sees, and XYZ should correlate at least reasonably well.</p>

[Tim_Lookingbill]

<blockquote>

<p>A typical digital camera therefore does not capture enough information to reconstruct what a human eye would see.</p>

 

</blockquote>

<p>The images posted are exactly as they appear in my studio under those lights. They did have to be edited to look as they do. </p>

<p>There's something else I discovered about scene gamut (or other color science theory oddity) as captured by my DSLR's sensor while shooting those images in Raw. When I used Auto WB shooting the incandescent scene, the blue sky on the print was nearly a dull gray blue and the skin tones were way too red orange to the point I couldn't edit them to look as my eyes saw it under those lights. The skin tones were all one monochrome color of red orange losing all color distinction. </p>

<p>Adjusting hue/saturation to select channels using both ACR and Photoshop's tools didn't make a dent. Selecting a dual 6500K/2800K table custom DNG created camera profile didn't work either. It worked when editing a shot of the X-rite CC chart under these lights which is what I built the 2800K table from, but for some reason the inkjet print posed a challenge.</p>

<p>Shooting using the incamera's Tungsten WB preset did allow a match as shown in the final edited image posted here. Getting that blue to look as it does wasn't easy and did introduce posterization in some areas. The thin white wispy clouds turned a bit pink.</p>