Hello Nikolay On 7-Jan-2011, Nikolay Pokhilchenko wrote: > Hello! > > I've often struck with similar problems while CMYK profiling. There is one > way with Argyll - to play with ink limits and -Kp or -kp parameters. > I've played with Elena's ti3 data and have built rather smooth profile with > heavy black generation. My command lines was: Nikolay, again thanks for taking care of my problem! The profile you sent to me privately is performing very finely, or at least, much better than I could obtain to date. Before receiving your message, I just finished reading a 3000+ patches target, made the profile again, but it shows very little improvement versus the 1300 patches one: the problem in the darks is always there. You seem to instead have the key for success! So let's try to understand HOW !? > spec2cie -f -i D50_1.0.sp -o 1931_2 -n 2xOFPS.ti3 2xOFPScie.ti3 I see you dropped all spectral data. Is there a particular reason why you did that ? > colprof -v -al -al should be the default, right ? > -qh well, I tried once -qh (at least 1 hour on my machine) but it didn't seem a so big key for my black problems > -Kp 1 0 1 0.6 0.55 Ok, you have to explain me well how you came with these parameters. Why -K and not -k ? Also, why start level = 1 ???? > -L90 -l285 ...and how you came to chose these limits. Are you using some special trick or tool? > -r 0.85 I assume this is some "good habit" of yours to give an arbitrarily higher tolerancy of 0.85 and not justified by some specific reason... > -s sRGB.icm -dpp 2xOFPScie > See attached graphs, relative.png and perceptual.png. Well. However, you will admit there's a bug in xicclu. The K plot for perceptual shows no black ramp down, which is not the actual case: a picture converted from srgb to CMYK with Photoshop, Perceptual intent, by using your profile, shows clearly that the K is indeed ramping down after about 50%. Just measure the CMYK numbers. After all, xicclu was showing the same with my profiles also: an always increasing black in perceptual, everything very very smooth (but the visual result not being so). So I please ask Graeme again, if he's reading, to investigate this bug... Bye! /&