Re: [Gimp-print-devel] Strategy for profiling Epson R280 with Ultrachrome K3 inks

[Marko M. <mar...@gm...>]

Thank you for detailed response.

On Fri, Aug 8, 2008 at 10:52 AM, Hal V. Engel <hv...@as...> wrote:
> On Friday 08 August 2008 03:36:18 am Marko Milisavljevic wrote:
>
>> I tried printing CMYK but can't figure out how. I think the correct
>> procedure, on a Mac at least is to take a CMYK image and print it
>> while also selecting "CMYK" in Gutenprint driver settings. This
>> appears to be printing RGB, since K gradient looks indistinguishable
>> from gray CMY gradient.
>
> You need to make sure that the CMYK file you are printing actually has a K
> channel. Most generic RGB -> CMYK conversions will use only process colors
> (IE. no K channel). But you also need to remember that GutenPrint does use
> the K channel when it does it's RGB to printer seperations when you are
> printing an RGB image so even if there is a K channel being printed with
> your CMYK file this many not be too different from what GutenPrint is
> sending to the printer. In addition if process grays are a close match to a
> K then you may not be able to tell the difference between a process gray
> gradiant and a K gradiant with out the use of a measurement device. Do you
> have one?

I have a CMYK file with side-by-side K and CMY gray gradients. They
looks identical on screen, but would be unlikely to look identical
when printed without some top-notch profiling, even if just
eyeballing. I do have a device to measure with, but I think I just had
incorrect settings when printing.

>> How do you make Gutenprint work in CMYK?
>
> You send it a CMYK file and you set the colormode in GutenPrint to CMYK. You
> need to make sure colorsync is not messing with your stuff.
>
>> At least from Photoshop.
>
> I don't know what Photoshop does. Photoshop was designed for Windows and Mac
> and originally both of these only had RGB print drivers. GutenPrint is
> closer to being a RIP than a printer driver although you have to use third
> party front ends to fully utilize this functionality. You might want to have
> a look at the PhotoPrint front end for GutenPrint. It is specifically for
> printing images to a GutenPrint device and gives you complete control over
> how color management is used and how this interacts with GutenPrint assuming
> that you scan get colorsync to not cause problems.

What I've been doing in Photoshop is tagging images with sRGB and
sending them like that and I think they are not going through another
set of conversions. When printing from Photoshop, reagrdless of color
management option set in print preview window, Colorsync options in
driver settings are grayed out. In other apps there is a choice of
CMYK/RGB profiles. I am not sure if that means Photoshop works as if
defaults were selected (sRGB/generic CMYK) or what. I'll test it some.

> Any image type (even RGB) -> CMYK device profile -> device specfic CMYK.
>
> If you are using a CMYK profile to transform the image the file that the
> print driver gets is CMYK no matter what color space was used for the
> original. Of course the original image file must be tagged with a profile or
> you will need to somehow provide this information during the transform.
>
> .... But I suspect that GutenPrint only accepts 16 bit/channel CMYK
> files. Robert is this correct? If that is correct then you must make sure
> that the CMYK files coming out of the device transform are 16 bit/channel.
> Since you are using 16 bit/channel images this is likely not an issue for
> your normal work flow. But targen creates 8 bit/channel target images and
> these will need to be converted to 16 bit/channel before they are sent to
> GutenPrint. Not a big deal but I though that I would warn you up front.

When I tried to print in CMYK yesterday, the file was 8-bit. I will
retry at 16-bit.

About targen... As far as I can tell, it doesn't think in bits - it
outputs .ti1 file which contains floating point coordinates. Only with
printtarg is the choice about bits made if outputting to TIFF with -t
or -T for 8 or 16 bits respectively. Outputting in 8-bit and
converting to 16-bit would, at least theoretically, be less precise
then creating 16-bit directly from floating point. I suspect the
difference is a small fraction of typical error of my spectro though.

> One other ArgyllCMS thing. ArgyllCMS has the ability to produce profiling
> targets that are optimized for a specific device if there is already an
> existing profile for that device. The way this works is that you first
> create a profile using a relatively small number of patchs (perhaps 300 to
> 500 - IE. one or two pages depending on your spectrophotometer). Using a
> small number of patches makes it fairly quick to to get this first profile
> and it should give OK results even with this limited numbr of patches if you
> should need to use it. Then you use that profile as input into targen which
> will look at the profile to find areas where it shows signs of having issues
> such as non-linearities and in those areas it will generate more patches so
> that the profiling software can get a better handle on what is happening in
> those places. For this second pass you will want to generate a bigger patch
> set of perhaps 2000 to 3000 patches. This will take a lot more effort to
> measure but should result in a very high quality profile.
>
> Hal

I will use ArgyllCMS first for RGB profiling to get a hang of that
software, but I would like to eventually do CMYK profiling. How did
you actually go about figuring out appropriate GCR settings in
ArgyllCMS? Any advice beyond what is written here:
http://www.argyllcms.com/doc/Scenarios.html#PP6

Marko