You don't need much patches to create a good printer profile

Hi @Ink stained Fingers The 10x15 cm target is already up to the maximum I can do (192 patches). The i1iO target can be easily scanned with the manual method using the scanning board. I have done that alrady many times. The normal target patch set using the scanning board is without separation bars, which I do not like, especially when two colours that are very closed to another there is sometimes a problem because the spectro sees it as one patch throwing an error when finishing the row scanning. The separation bars between patches is very usefull to counter this problem.

According to my calculation a target for 13x18 cm photo card (5x7 inch) should hold up to 285 patches (15 rows of 19 patches). The RGB colour cube can be extended up to 6x6x6 (216 patches) + 15 grey steps between RGB 0,0,0 and RGB 255,255,255 and also can hold 9 triplets of near neutral patches of RGB (27) + 9 triplets of near neutral patches of CMY (27) around the neutral axis up to 285 patches in total for an even better sampling of the RGB colour cube around the neutral axis.

For the A5 target to be optimized for the i1iO2 I think just over 400 patches should be possible.

I'm not using 5x7 inch cards, and I'm quite happy with a patch sheet with 283 patches on A5; I don't think I need an upgrade here at this time. That was an import of a standard profile of the old ProfileMaker package. I remember from longer time ago that missing separation bars were impacting the reliability of the scanning process, but that proglem is gone.

Thanks for all the very valuable Input @pharmacist and @Ink_stained_fingers.
I have the suspicion that the Colormunki approach of scanning few patches, creating an profile, printing and check by the second measurement how good the correction worked and then creating the final icc profile is superior.

When controlling machines, the control loop with verification (=Colormunki) also produces better results than a pure control (i1pro). So could the workflow of i1pro be improved, by a second scan and a second profile calculation?

Lothman said:

Thanks for all the very valuable Input @pharmacist and @Ink_stained_fingers.
I have the suspicion that the Colormunki approach of scanning few patches, creating an profile, printing and check by the second measurement how good the correction worked and then creating the final icc profile is superior.

When controlling machines, the control loop with verification (=Colormunki) also produces better results than a pure control (i1pro). So could the workflow of i1pro be improved, by a second scan and a second profile calculation?

Click to expand...

The Colormunki method is based on an estimation of a perfect linear behaviour of how the ink is laid down. I have used a rather not so good LC ink in my previous Epson P800 that was too light in hue, causing terrible banding and clipping from very light blue to dark blue when producing a profile using the Colormunki method (50 + 50 patches), because this method expect a linear behaviour. Because it did not measure the intermediate points (for example in steps of 12.5% from 0 to 100%). I did the profile again using the ArgyllCMS method using the Colormunki and the clipping and banding was gone, because the printer could use the dark cyan to compensate for the malfunctioning LC ink by adding smaller droplets of dark cyan (variable droplet technology) in the area's where the LC ink should be printed and could not reproduce that dark tone. However I can tell the way how i1Profiler distribute the measuring point in the RGB cube is not well optimized. ArgyllCMS does a much better job to evenly distribute the measuring points in the RGB space to maximize avoiding each other within a given patch count. I find the profiles using ArgyllCMS better than the ones produced with i1Profiler using the same amount of patches. I can add as much grey and near neutral points I want using the targen command within a given patch count. The i1Profiler only gives you a given amount of grey and near neutral points at given patch counts. Sometimes 1 extra patch will remove all the grey and near neutral points from the previous patch count (for example between 2005 and 2006 patches).

As you know these grey steps and near neutral points are very important for the internal sampling around the neutral axis (the most important area to create the visually most pleasing profiles). The TC9.18 profile has a lot of greys and especially near neutral patches (stained grey ramps) in the patch design, giving a a much better profile compared to a profile by the autogenerate function of i1Profiler using the same amount of patches. I try to mimic this by making my own patch designs by removing redundant patches and substituting them with much more near neutral patches and extending the amount of grey patches.

There is a optimize function in i1Profiler and I tried it once and to my surprise it generates alot of similar looking near neutral patches that I took for the Spyderprint "Extended Grays" target and also that can be seen of the Black and White profile method in i1Studio/ccCalibrite (Colormunki) software that uses 50 + 2x50 patches to produce a profile. The second patch set using 2x50 instead of the standard 50 patches of the color method; has several stained grey ramps: also to neutralise black and white printing and to neutralise greys in colour printing as well.

I would like to see a function in i1Profiler that not only says: I choose 700 patches of which I want 64 grey steps and 64 near neutral steps and having the rest of it 700-64-64=572 patches to be evenly distributed in the RGB color space. That is not possible unfortunately. The patch generator is very rigid and 1 extra patch and the whole previous amount of gray and near neutral steps are gone.

pharmacist said:

The Colormunki method is based on an estimation of a perfect linear behaviour of how the ink is laid down. I have used a rather not so good LC ink in my previous Epson P800 that was too light in hue, causing terrible banding and clipping from very light blue to dark blue when producing a profile using the Colormunki method (50 + 50 patches), because this method expect a linear behaviour. Because it did not measure the intermediate points (for example in steps of 12.5% from 0 to 100%). I did the profile again using the ArgyllCMS method using the Colormunki and the clipping and banding was gone, because the printer could use the dark cyan to compensate for the malfunctioning LC ink by adding smaller droplets of dark cyan (variable droplet technology) in the area's where the LC ink should be printed and could not reproduce that dark tone. However I can tell the way how i1Profiler distribute the measuring point in the RGB cube is not well optimized. ArgyllCMS does a much better job to evenly distribute the measuring points in the RGB space to maximize avoiding each other within a given patch count. I find the profiles using ArgyllCMS better than the ones produced with i1Profiler using the same amount of patches....

Click to expand...

You are way beyond anything I have done on profiling, but it might be worth pointing out that with the i1Studio, one can generate at least one additional sheet of patches with custom colors. I don't have enough experience with it to know if it is useful, but I like the idea in principle.

Incidentally, inkjet colors are usually thought to consist of purely subtractive colors, but a close examination of the light-colored areas immediately dispels that notion. On close examination, one sees non-overlapping dots with different colors. These colors are additive. But if the dots consist of mixed inks, the colors must be subtractive. Accurate control of color under these circumstances seems extraordinarily difficult, and I am not surprised by some nonlinearity. I suppose the printer software manages most of that difficulty, and that profiling -- fortunately -- does not need to deal with it.

@ThrillaMozilla Yes I know the optimization method. You can either choose a image to have the software to pick out the colours it will have difficulty to reproduce and measure them to improve the existing profile. However in the past I have tried this method by first generating a profile by 50 + 50 patches and improving it with 2 interations of selected images: one with black and with ramps and out of gamut colours and the second one using real life images with skies/vegetations/water/buildings and skin tones of every ethnicity....Turns out the ArgyllCMS 480 patch target and Colormunki produces much better profiles with better tonality and more neutral gray ramps.

I tried the optimization function of the i1Profiler software only once long time ago, I cannot judge if profiles are getting better that way or running just one cycle with the double number of patch colors at one time. I typically let patch sheet prints dry overnight since the colors of such fresh prints drift for a while, but I wouldn't like to wait another day for the 2nd cycle.
@pharmacist above explains the options for the patch sheet layout which appears to be rather flexible and let you focus on a subset on colors - you could use portraits to extract more skin colors or you can add a range of Pantone colors with even more colors than the original profile was made with. and multiple cycles are possible as well. I'm happy enough with just one cycle.
The spectro just reads intensity of wavelenghts in a spectrum , and a recognition whether it is a subtractive or additive color is not possible when the spectral data are converted into RGB color data.

Nelson I didn't mean bad colordistributed patches nor insufficient grey patches, that's another problem. But if I want to correct the ink layout and I do not know how the printer/paper/inks combo reacts to a change, how will the software estimate the proper correction? In the Colormunki approach the software should have the possibility to measure how good/far the corrections worked and then can further improove it on the second run.