A lot of people overcomplicate this. It's actually very simple.
1. Photoshop converts on the fly from the document profile to the monitor profile. As long as the monitor profile is an accurate description of the monitor, it displays correctly regardless of the original document profile. This is the whole point of color management, translation from one color space to another maintaining appearance/intent.
There's no particular reason to work in sRGB just because your monitor is (or close to). However, you won't directly see the colors in the file that are outside the monitor gamut.
2. Proofing to Monitor RGB is a quick test to see how the image looks without any color management at all. I suppose that could be useful in some cases when you need consistency over accuracy. Other than that it's completely pointless, it only shows how it looks on your system. And anyway, there's no reason today to not use fully color managed browsers.
3. Proofing is an extra round to the conversion mentioned above. The point is to limit the gamut to the proof profile, so that you can see waht clipping will do to the image. So it goes document profile > proof profile > monitor profile. If the proof profile has a gamut equal to or larger than the monitor profile, you won't see any difference.
Important point to keep in mind: If source and destination profiles are the same, no conversion happens and color management is by definition disabled or turned off.
The first part of the proofing chain, document to proof, can be either a "convert" or an "assign". Convert maintains appearance, while assign maintains numbers but changes appearance. This is the "Preserve Numbers" checkbox in the proof setup dialog.
Hey and thanks 21, that clears some things up!
After another couple of hours of research I did realise where my knowledge was lacking - I was unaware of the differences between color calibration and profiling (also seems to be called 'monitor characterization'). I was under the impression that the whole 'calibrating a monitor' process just creates an ICC profile that simply sits on top of everything running in the OS, affecting non-colormanaged and colormanaged apps alike.
Now from what I gather, there are actually two sides to it - the calibration and the profile. The calibration does indeed sit on top of everything in the signal processing chain but the created monitor profile is used only by colormanaged programs (a lot of my info came from http://www.cambridgeincolour.com/tutorials/monitor-calibration.htm, very nice resource if somebody's curious/trying to understand this thing).
Am I correct in this?
If yes, would I be correct in assuming that both, the profiling information and the calibration information are stored in the SAME icc file, which then is used by OS and colormanaged apps in different ways (for applying calibration information for the former and using profiling information for conversons for the latter)?
If I'm correct, a little broader (and existential ) question would be - why is it being done like that? Why wouldn't you have just one profile+calibration that sits at the end of the image processing chain and processes everything coming out of the OS?
Am I correct in this?
Yes, you are indeed. Understanding the distinction between calibration and profiling is pretty crucial to understanding the whole process - but once there, it all falls nicely into place.
In brief, calibration is a modification of the monitor itself. So it's global, but it is not part of the color management chain as such.
The profile is a description of the monitor in its present, calibrated state. So the software builds the profile after the calibration is finished.
The calibration look-up table (LUT) is just stored inside the profile because it's a convenient place to keep it. But much of the common confusion is rooted in this. The two - profile and calibration LUT - are in reality separate entities performing separate functions.
(Edit: Where they in some sense overlap, is in setting the white point temperature and luminance. This is an important part of the calibration, but it also indirectly defines parameters for the profile, in that it defines what is to be treated as 255/255/255 white).
Just to dig one step deeper - so high-end monitors that have hardware calibration through adjustable internal LUTs (like Eizo CGs) - do these internal monitor LUTs act as a replacement to the OS applied calibration LUTs and they still have an ICC profile for color managed apps to use, or is the profling part also taken care of in the monitor internally somehow?
No, it works the same way. Monitor LUT equals video card LUT, it's just much more precise (higher bit depth) with much less chance of banding and other artifacts.
The profile is still separate, and adds another precision level that the LUT cannot account for (like the exact position of the primaries).
Ok, i think it makes sense to me now.
My last question would be regarding the use of calibration, the end goal of it. I understand that, for example, the goal of monitor's profile is to provide consistency when using different color spaces, but what about the goal of calbration?
Should the calibration target a specific color space? Like, if I use only sRGB for my work, I would target calibration to sRGB, so that my monitor is as close as possible to sRGB space, so then sRGB images would 'pass through' the color managed apps without any/very little adjustment from the monitor profile.
Or should the calibration be sort-of un-targeted in terms of a particular color space, in sense that it's main goal would be to try to maximize the potential of the monitor, without adhering to a standard.
I suppose the essence of my question is whether the monitor should be calibrated to a particular standard color space or simply to show the most colors it can?
1 person found this helpful
Calibration has several aims, and their relative importance depend on viewpoint (but you need to take them all into consideration).
The most important is to define a white point. This relates to output, getting a good match from screen to final output. I work mostly for offset print, and so I set a white point that translates as directly as possible to my perception of paper white. For me, a temperature of 6300K and a luminance of 100 cd/m² works, but since this is largely perceptual it depends very much on the ambient light.
Then you want to standardize the response beyond that (gamma, neutral color balance). This matters mostly in a non color managed environment (which I frankly don't care much about...). In a color managed environment these parameters will be remapped profile to profile anyway, and are "invisible".
And last, but certainly not least: you want the monitor to behave well. Forcing the monitor into non-native behavior will often lead to less than optimal performance, especially if it's done in the video card. This is where hardware calibration has an edge.
What you don't do, is calibrate "to" any color space, that's both futile and unnecessary. The profile describes the monitor as is, period. It has its own native color space.
From a color management perspective, and if the monitor is a good one, calibration isn't even needed beyond setting the white point. I have a hardware calibrated Eizo at work, but at home I have an NEC without hardware calibration (P232 EU edition). So what I do is that I set the white point (in the monitor hardware), but don't calibrate it any further. The video card LUT is a straight line. I just profile it, using ArgyllCMS and an i1 Display Pro sensor. In color managed software, like Photoshop/Lightroom, the profile takes care of the rest and it's as accurate as the Eizo.
1 person found this helpful
In terms of Calibration, it's useful to keep this in mind: Calibration ideally places a device in a known, desried and repeatable state. In terms of a display, the 'ideal' calibration for most is one in which the display and the print produce a visual match. YMMV a great deal! You can calibrate a display such it doesn't produce a match quite easily, getting calibration to produce the match takes work! If any of the targets aim points (White Point, Luminance and contrast ratio) is less than ideal, the calibration doesn't produce a visual match. See: http://www.luminous-landscape.com/tutorials/why_are_my_prints_too_dark.shtml
In theory, if you didn't need to match the display to a print, you don't have to even calibrate the display for an ICC workflow. But you should <g>. You could simply profile it's current behavior which would work in an ICC workflow because you built a profile that reflects the current display conditions. The display will change over time because a display, unlike say an Epson Pro printer, is unstable and it's behavior will change over time. You need to calibrate or get that display back to it's original condition ideally. With the Epson, it's very, very stable (although I'd point out, not necessarily calibrated ideally, if we had more control over the native driver, we could 'adjust' or calibrate a more linear behavior). So with an Epson, you build a profile on top of the less than ideal driver conditions but the driver and the rest of the system is super stable and consistent. You simply need a profile to define that behavior and you're done. The printer and inks are very stable so you could also in theory place a different driver or RIP and calibrate better behavior and profile.
I'd forget the 'calibrate' a display to sRGB on a modern LCD unit short of perhaps an Eizo or NEC SpectraView system. The sRGB color space is based upon a theoretical CRT display circa 1994 or so, with very defined primaries and conditions. There is no real 'standard' to how one can and should define the display, especially if you want to visually match a print next to the display. The illuminant used to view the print plays a huge role! So providing one set of calibration aim points and expecting all displays to produce a match is pie in the sky. Might be close, might be spot on (I wouldn’t depend on that), might be way off. All explained in the URL above. If one setting worked properly, we'd all use that one setting and calibration products wouldn’t vary from only a few presets (kind of uselss) to products like Eizo and NEC which provide a huge number of options for setting white point as just one example.
Thanks a lot guys for clearing these things up, very valuable info.