Nikolay Pokhilchenko wrote:
Graeme Gill wrote: There is no black generation for the targets. They explore the CMYK space - ie. it's a 4 dimensional characterization of the device. Black generation is done in the profile creation of the B2A tables. But how does the targen do the patch distribution, for example with both -c and -I keys?
There are a variety of different distributions, that's really the point of a lot of those options. There is a brief description in the documentaton: <http://www.argyllcms.com/doc/targen.html#f>. The -c parameter is used for some of the distributions, and not for others. The way -I is handled depends on the type of distribution, some being able to work within the device gamut naturally, and some needing to used cruder methods such a rejection testing.
It seems to me, targen computes grid for device space in Lab(?), then lookups theCMYK value in B2A table from given pre-profile. It is right?
No, it depends on the distribution chosen. For the -m points, the grid is in device space (ie. a 4 dimensional hyper-cube grid for CMYK).
When I want to generate ADDITIONAL qualifying chart, I'd like get patches closer to desired black generation method. I don't interesting in whole 4-dimensional supplementary characterization, because many of full-4D patches almost useless, since I already take shape for blackgeneration.
I don't think there's an easy way to do that, although if you went to enough trouble you might manage it. For instance, one way might be to create an sRGB -> output device link using -G with saturation intent, generate test points in RGB space, then map them to the output CMYK using icclu. Those points could then be edited into the .ti1 file.
Of course, it's reasonable when black generation chosen will not be changed aftertime, including device link. May be it worth to add one more targen feature: "Use pre-conditioning profile black generation" with standard for black generation parameter set, i.e. -k, -K, -L? That feature will allow to get more precision profiles by the less overall patch count.
Maybe, although there are some other features in a similar vein I'd add first, such as being able to target test points at a particular L*a*b*, say the grey axis. Additionally, I think it would be better to add full two pass N channel separation support, which could be used for CMYK, and would achieve the same result with a lot broader application. Graeme Gill.