Argyll CMS printer profile workflow update

[pharmacist]

Thanks to member CLS and his tips of creating a printer profile I did some experiments on both a simple printer profile and a high quality profile. Now the whole thing is about that amazing X-rite Colormunki device, capable of creating amazingly good printer profiles.

CLS was able to squeeze 675 patches on a single sheet of A4 paper, but I had significant problems to have these tiny patches to be scanned reliably. It was too cumbersome to have these patches read in just one go, so I decided to lower the amount of pathces to 648 (24 rows of 27 patches). Now why 648 ? I used the following calculation: 8 steps in each RGB-direction will yield 8^3=512 individual steps in each XYZ-direction of the RGB-axes and adding 128 grey steps to neutralize black&white printing + 2 patches for white + 2 patches of black makes: 512 + 128 + 2 + 2 = 648

The results are really amazingly good for out of gamut colours and black&white prints, I find it even better compared to the profiles I make with the standard X-rite Colormunki workflow with 2x50 patches + 2 optimization charts, especially in visually rendering the out-of-gamut colours. Argyll CMS created printer profiles tend to work best with the relative colorimetric intent for out of gamut colours, unlike the Colormunki produced profiles, which prefer perceptual content.

What about more patches ? Well, I also created a profile with twice the amount of patches on 2 sheets of A4 paper (1296 in total, including more than 10^3=1000 steps in each RGB-axes). I softproofed with both profiles on a test file (PrinterEvaluationImage_V002_ProPhoto.tif) and I could hardly see any difference (actually nothing at all) in those 2 "different" profiles. To be sure, I also printed the test file using both profiles to compare visually. I cannot see the difference with a profile made with "just" 648 patches compared to the higher patch count profile using 1296 patches. Skin tones are the same, Out of gamut color ramps are equally vivid without any artifacts and the black and white area's of the picture are equally good and neutral.

My workflow (A4profile is the name of the file I have used):

targen -v -d2 -G -e2 -B2 -g128 -f648 A4profile

printtarg -v -ii1 -a0.90 -A0.90 -b -T360 -m0 -M0 -P -p210x297 A4profile

Now the produced A4profile.tif file must be manipulated and cropped in Photoshop to remove the right text column and to crop and resize the whole image to A4 and 360 dpi resolution.

No more 2 or more step printer profiling with the Colormunki software and just one sheet of A4 is enough to produce a more than satisfactory printer profile.

 

[Emulator]

Interesting information, I have often read comments to the effect that above a few hundred patches shows little improvement. What is the finished profile file size of the best profile and does the file size vary with number of patches?

 

[pharmacist]

Hi Emulator,

The 648 profile was about 1.5 MB and the 1296 was 2 MB, so yes: more patches seem to increase file size. Most printers demand RGB input, which is internally translated into CMYK data and feeded to the internal RIP software. For CMYK profiles at least 2000-3000 patches are needed to obtain good profiles, not so for "ordinary" RGB profiles. CMYK profiles are needed for RIP software/drivers. According to the argyllcms website 500 patches should suffice for a medium quality printer profile (pseudo RGB).

 

[cls]

Tried your setting and works great
this a ready to go 2 page A4 profile
https://docs.google.com/file/d/0B9pRlaqsNzTdckhGQ1hZNzhqQUE/edit?usp=sharing (722kb) (uncompressed size is 143MB!)
i tried it for one page and works just fine had no problem ready the patches at all with a aloy 40cm ruler as aligment help

iRGB test chart
Compostie Grey steps = 128
Full spread patches = 1296
Degree of adaptation: 0.100
Adaptation weights: Device = 0.935, Perceptual = 0.065, Curvature = 0.010
Perceptual cache resolution = 13
Acceleration grid res = 12

There are 128 unique fixed points to add (129 total fixed points)
There are 1167 far spread points to add
Added 1295/1295
Patches = 1296
Test patches per row = 27
Rows per page = 24, patches per page = 648
Rows in last strip = 24, patches in last row = 27
Total pages needed = 2
Worst case delta E = 12.843847
Worst case direction distinction delta E = 53.054618
Optimising layout for strip reader:
100%
After optimisation, worst delta E = 67.551844
Worst case direction distinction delta E = 75.702921

I could even offer scanning this target for you if you send both pages to me ! conact me

 

[Emulator]

Thanks pharmacist, it suggests Argyl CMS builds its profiles strictly to user requirement, whereas ColorMunki starts with a full profile assuming a linear system and adjusts according to additional input.

My comment in another post:

If you take a look at the profile file sizes Canon profile size is typically 322KB, ColorMunki typically about 2400KB, a vast difference in size.
An Optimisation in CM may add about 35KB to the total each time.

suggests that a Canon profile may well only contain adjustments relating to the selected paper and leaves the colour correction to user profiles.

I was surprised that ColorMunki automatically changed to CMYK target prints when used with the PrintFab version of the Canon Pixma Pro9000 ll printer, but your comments explain this, it must bypass the printers in-built RIP conversion.

 

[Grandad35]

If you take a look at the profile file sizes Canon profile size is typically 322KB, ColorMunki typically about 2400KB, a vast difference in size.
An Optimisation in CM may add about 35KB to the total each time.

My i1 uses 918 targets and its profiles are 882 KB. Profiles generated from about 1100 patches by Dry Creek's profiling services for commercial printers are in the 1000-2000KB range (click on a state/country and chose a site to see the profiles available for the printers and papers used at that site). There is obviously more than the number of patches involved in determining the profile size.

 

[mikling]

http://www.nifty-stuff.com/forum/viewtopic.php?pid=66101#p66101

Looks like nobody went to class. Read Chapter 4 and you'll see that there are different methods of processing the data.
The beginning will give an explanation of the field contents and the bits and bytes of how to interpret. The other thing is that some field are padded and some not. It depends on how the data is defined.


For the studious, they'll figure it out.

 

[Emulator]

Or read Chapter 7. But it doesn't really help answer our questions about ColorMunki etc., which are dependent upon how the company has configured their system.

You can use ICC profile inspector and wxProfileDump to look at the internals of the profile, but there is a lot of data.

I haven't got an Argyll CMS profile to look at, perhaps that would help in the comparison,

 

[Emulator]

I find that the ICC Profile Inspector software tool will provide the answers on profile sizes. The front page image of one of my Sihl paper profiles is below, click to enlarge.



The profile size in the Windows directory listing is 2380 KB. The Profile Inspector front page shows Size as 2436776 bytes or 2436KB. The total of the Tag Table "Size" column is 2436456 bytes, all reasonably similar, except Windows!

You can click on the individual tags in Profile Inspector and they open up additional pages.
Now I would like to see the equivalent for an ArgyllCMS profile or better still, be sent a copy of a profile!

Also see http://www.color.org/ICC_Minor_Revision_for_Web.pdf, document page 11. (PDF page 20), for a profile map.
See also Table 20 PDF page 28, document page 19, for Tag details.

Summarising the Tags:- Description, White point, Black point, Chromatic Adaptation, Copyright, A2B and B2A, Gamut, DeviceSettingsTag - Specifies the device settings for which the profile is valid, Characterization target.

The BToAxTags represent the inverse operation of the AToBxTags, where 'x' = 0,1 or2 represents 'perceptual', 'colorimetric' or 'saturation'.

The Look Up Table algorithm for profile connection space to device space transformations, process data sequentially through a
matrix, input tables, a color Look Up Table, and output tables.

Watch this space!

 

[pharmacist]

Hi Emulator: send me a PM with your email, so I can attach one or two to you.