I did what I believe Martin Orpen suggested, but the results are not what I expected. Actually, I do not think I fully understand them. I downloaded “ISOcoated_v2_eci.icc” from http://www.eci.org/en/downloads and then ran: $ targen testchart $ fakeread ISOcoated_v2_eci.icc testchart $ colprof -kz -O myFogra39-kz testchart which generated “myFogra39-kz.icc”. As Martin said, and according to colprof’s docs too, -kz provides the resulting profile with a zero-black-generation rule. For the purpose of comparing against other black-generation rules, I created additional profiles: $ colprof -kh -O myFogra39-kh testchart # Black Ink generation: 0.5 K $ colprof -kx -O myFogra39-kx testchart # Black Ink generation: max K $ colprof -kr -O myFogra39-kr testchart # Black Ink generation: ramp K Colprof’s docs also state that:

the black level curve is applied throughout the gamut, resulting in GCR (Grey Component Replacement). There is no facility to restrict black to just neutral colors, hence UCR is not currently supported.

which is a pity since, like I said in my original email, I am also looking for “a CMYK profile using FOGRA39 that uses a strong UCR-like separation so that black is ideally used only on neutral values” or a way to create one. I could not get xicclu to help me visualize how colprof’s different black-generation rules affect the resulting profiles. Running xicclu against all four profiles returned the same chart (http://cl.ly/2P1b2S2b0Q12): $ xicclu -g -fb myFogra39-kz.icc $ xicclu -g -fb myFogra39-kh.icc $ xicclu -g -fb myFogra39-kx.icc $ xicclu -g -fb myFogra39-kr.icc even though they do produce different separations (more on this below). Xicclu’s docs describe an optional argument to indicate, yet again, the black-generation rule, the syntax of which seems identical to colprof’s. I do not quite understand what it means to run xicclu with an explicit black-generation rule while its input profile having been created with yet another explicit black-generation rule, identical or different. At any rate, xicclu still provides identical charts for all four profiles for the same black-generation rule. Regarding how the resulting profiles behave, I expected “zero black generation” to produce no black ink whatsoever, but apparently it only means “as moderate with black as possible”. In Photoshop, I applied “Convert to Profile” (Colorimetric Relative with BPC) to a sample image with out-of-gamut colors (http://cl.ly/3S1n0Q2Y1y17), using “myFogra39-kz.icc” as destination. The resulting K channel has values different than zero in the darkest areas, even if those values are admittedly lower than when converting to the other generated profiles (in retrospect, xicclu’s chart above was already saying black ink would be used in the darkest areas, I guess). Colprof accepts an additional optional argument for “overriding black ink limit”, so I tried generating a new profile specifying a limit of 0% black (-L 0). I do not quite understand the meaning of simultaneously providing arguments for both “black generation” and “black ink limit”, and therefore how they interact, but still: $ colprof -kz -L 0 -O myFogra39-kzL0 testchart Xicclu, this time, does plot a chart with no black whatsoever (http://cl.ly/1E0H2V300K0q). Converting the image to this profile “myFogra39-kzL0.icc” does produce a CMY-only separation with all K values equal to zero no matter how dark original colors were, which is what I wanted. Still, colors in the converted image do not look right. The darkest areas, which were 0L in the source Lab image and were converted to 300% CMY, still look pitch black in the converted image―they are reported as 0L with a Lab color picker. Considering the lack of black ink, I expected them to become washed out even at full total ink limit. My intention, again, was obtaining CMY separations that produce colors as similar as possible as when separating to CMYK in the same color space. I was therefore expecting that, by using “ISOcoated_v2_eci.icc” as a source data set, but forbidding the use of black ink, the gamut of “myFogra39-kzL0.icc” would be a subset of “ISOcoated_v2_eci.icc”, with both sharing the same white and black points (even if in “myFogra39-kzL0.icc” no ink mix could actually attain that black point due to zero-K) and the same tone response curve. Using ColorSync Utility’s gamut comparison feature I see that part of the gamut of “myFogra39-kzL0.icc” actually falls outside the one of “ISOcoated_v2_eci.icc”. It also reveals that “ISOcoated_v2_eci.icc” has a table for kTRC (tonal response curve) that “myFogra39-kzL0.icc” lacks, whereas the latter has one for bkpt (Media black-point tristimulus) that the former lacks. It is, if I afterwards assign “ISOcoated_v2_eci.icc” to the converted image, after having converted it to “myFogra39-kzL0.icc”, that all dark colors are washed out. Like I said, I did expect that those would be washed out when converting to “myFogra39-kzL0.icc” (which does not happen), but remain constant when afterwards assigning “ISOcoated_v2_eci.icc” to the converted image (which does not happen either). Furthermore, in cases where “ISOcoated_v2_eci.icc” already produced no black ink “myFogra39-kzL0.icc” produces rather different separations which was not at all what I intended. Again, my intention was preventing the use of black ink and increasing the use of CMY inks to compensate, only when the “standard” profile would indeed produce black ink, but not altering separations that already produced no black ink. Maybe I am making this too convoluted and I should not be trying to do this by creating a custom profile? Is there any other way to convert to a specific CMYK color space, preferably with Photoshop, but altering the default separation algorithm so that K is not used and CMY are increased in order to compensate that lack so that colors remain as close as possible to the original, taking also into account a custom TAC? I do understand that the deviation from the original colors will inevitably be higher on black-ish tones if no K is used.

...

On Thu, Dec 31, 2015 at 1:11 PM, Martin Orpen <martin@idea-digital.com> wrote:

On 31 Dec 2015, at 10:37, Jorge . <chocolate.camera@gmail.com> wrote:

Or, better yet, I would like to know how to create one myself with a software that is free as in beer, so that I can have several with different maximum total ink. Ideally, that software would run on Mac OS X. I have never created a profile and I am not one bit familiar in what it involves and what I should be sure to doublecheck.

You can use Argyll to do all of this.

Steps:

1. Use targen to generate a test chart

2. Use fakeread to fake results from your chart and for example the ISO Coated v2 profile

3. Use colprof to build the ICC profile using -kz to ensure that no K is generated until the CMY hit their maximum

You can also use xicclu on any profile to visualise and test different black generation settings before building your final profile.

HTH

-- Martin Orpen Idea Digital Imaging Ltd

On 5 Jan 2016, at 11:50, Jorge . <chocolate.camera@gmail.com> wrote:

Maybe I am making this too convoluted and I should not be trying to do this by creating a custom profile? Is there any other way to convert to a specific CMYK color space, preferably with Photoshop, but altering the default separation algorithm so that K is not used and CMY are increased in order to compensate that lack so that colors remain as close as possible to the original, taking also into account a custom TAC? I do understand that the deviation from the original colors will inevitably be higher on black-ish tones if no K is used.

I’m not sure what else you’d expect to happen here? If you build the ICC profile in Argyll using -kz -l300 -L0 you are going to create CMYK with no K and the CMY channels compensating as best they can for the missing K. Keep in mind that you'll need a pro CMYK proofing system to appreciate and validate the results of these conversions — Photoshop isn’t going to be very useful… And, if you want to keep things really simple, why not just invert the RGB channels and avoid black generation altogether ;-) -- Martin Orpen Idea Digital Imaging Ltd

On Jan 8, 2016, at 11:44 AM, Jorge . <chocolate.camera@gmail.com> wrote:

that the resulting separations (by converting with Photoshop using Colorimetric Relative w/BPC) are rendered (in Photoshop, again) so differently than when these same separations are rendered as ISOcoated_v2_eci.icc

Quick question: How much do the separations differ when both are done without BPC? Don’t forget, BPC scales between the dynamic range of the input and output profiles. A CMYK profile built without K will have less dynamic range and will separate differently…. I’m not saying that w/o BPC is the desired result, but I’m curious nonetheless. regards, Steve

On 8 Jan 2016, at 19:44, Jorge . <chocolate.camera@gmail.com> wrote:

So I feel I am back at square one: how do I create separations for a color space defined by a profile created from FOGRA39 but using no K and compensating CMY to account for that lack (as far as possible)? That was the original purpose when asking how to create a custom zero-K profile using FOGRA39.

Jorge Perhaps it would be better if you describe exactly what it is that you are trying to achieve? You are already able to do exactly what you describe above. Whether the results look good or bad when rendered back through the profile Photoshop is of no importance whatsoever in CMYK reproduction. The important thing is that the image is optimised for the CMYK device that you are printing on. What device are you printing on? If this CMYK device is *imaginary* and you don’t want K you don’t need any conversion at all as R=C, G=M & B=Y. -- Martin Orpen Idea Digital Imaging Ltd