A basic guide (see post #1) to setting up ARGYLL CMS profiling on your computer

[Ink stained Fingers]

I found a way to 'edit' a data file which contains scanned color data, I have edited this in a text file with the theoretical color values of the patch sheet which let me compare this data with another data set from another scan , I had to convert the 96 colors of the patch sheet from RGB to Lab data, There are several RGB to Lab converters available on the internet like this one which can deal with 11 different color spaces - scroll down

http://colormine.org/color-converter

I was testing again against these numbers gray step ranges with pretty disappointing results - grays are not gray but pretty colorful, that's easy to see in the scan results when the Lab a and b color values differ very much from 0 . I need to check for the influence of the rendering intents, and there is more to investigate

 

[The Hat]

grays are not gray but pretty colorful

Could you scan and reduce your test sheet to make the dpi grey patches smaller..

 

[Ink stained Fingers]

Could you scan and reduce your test sheet to make the dpi grey patches smaller..

The minimum patch size for the i1io scanner is 7.5 x 8 mm, I can shrink the patch sheet and rescale it during print, I'll keep your recommendation in mind.

 

[nertog]

I had to convert the 96 colors of the patch sheet from RGB to Lab data

I would expect the patches in the I1Profiler file to be encoded as XYZ data, not RGB. Could this be the issue or do your grey patches actually look non-neutral?

 

[Ink stained Fingers]

I would expect the patches in the I1Profiler file to be encoded as XYZ data, not RGB. Could this be the issue or do your grey patches actually look non-neutral?

May be , but that's not visible at the outside , I can store the patch sheet into a .tif file which is RGB code, or i1profiler can directly print to the printer which is RGB data as well. Since the patch sheet is printed without any color correction/adjustment the gray patches don't look exactly gray - depending on the inks and papers , and that's what the profile is supposed to correct at the end when I print with the correct profile. I'm currently on another poster/banner print job - large like 40 cm x 6 m for a city panorama , and banners for a birthday and such.

 

[Ink stained Fingers]

I did a little bit more investigation why I'm not getting good gray patches printed; I was looking to the various rendering intents and recognize that I was under the wrong assumption that the abs. colorimetric rendering intent prints the most correct colors but it does not.
I was reading more than 20 postings what the rendering intents are doing - some are useless - the editor does not know what he is talking about - some are confusing and irritating and some are pointing into the right direction; technical/scientific documents of the ICC consortium as they established the rendering intents are about 30 years old and only accessible to the actual members.
I found this entry here

https://www.johnpaulcaponigro.com/blog/6088/

with a specific statement

Relative Colorimetric

A relative colorimetric rendering intent maps the white of the source space to the white of the output. It reproduces in-gamut colors exactly and clips out-of-gamut colors to the nearest reproducible color.

I verified this with a few patch sheets and got quite good and matching results along the gray axis. Why along the gray axis - the readout is easy - any Lab value with the a or b value not 0 indicates a slight color cast. I'll do some more tests along this assumption about the rel. col. intent.

 

[Ink stained Fingers]

Let me show you a few numbers and the actions around it - I got a patch sheet with 480 colors by @pharmacist which he created with ArgyllCMS, I printed that sheet on my L1800 on the Aldi/Netbit paper with Epson 106 inks and sent him the print for scanning and profile generation.

I'm using a patch sheet with 96 colors with i1Profiler to profile small 4" x 6" photo papers, and I'm using this as well for this test, it contains 12 gray patches - from RGB=0 to 255 - black to paper white. I'm printing this sheet with the Argyll profile with the rel. col rendering intent and scan it for comparison against a perfect patch sheet with 96 patches, this is data which I edited manually and reimported back into a measurement file in i1profiler. All this is done with the M2 UV blocking mode of the spectrometer. These are the numerical results of the gray patches - the gray axis when you connect them.

These are the numbers:

The black point with L=7,03 is not super dark, but that's not the point of discussion at this time, the white point with b=-6,53 is on the blueish side, even w/o UV impact which would increase this number further, the Netbit paper contains a significant amount of optical brighteners. This point already shows what can happen with the gray axis - for all color values of R=G=B , the output of neutral gray values create a color cast caused by the paper, the gray axis needs to bend over to the non-neutral white point. You have a similar effect at the bottom, but to a lesser extent , a b value of -1,86 is still neutral - visibly at least , and this impression can change with varying viewing light conditions, all these measurements and calculations are done with the assumption of a D50 - 5000K - viewing light, strong daylight or lamp light can change all this.
When you look to the middle part of the gray axis range the numbers very well stay neutral from L= 8% up to 65% with small variations , and the gray axis starts bending over slowly to the -b value of -5,53 starting at L=71 with a b=-3,25 stepping down to -6,53 in the following gray spots. The numbers appear to be consistent, there are no outliers or sudden steps in the wrong direction, I think that this is a pretty good performance. And I would assume that this is valid as well for other color spots and ranges - as long as the colors are located inside the gamut. The numbers for the gray values are no special or different than numbers for other color spots so this profile appears to be quite consistent and good to me.

I'm not comparing this profile yet with other profiles - with less patches, or more patches, and/or created with other software like ColorMunki or i1Profiler .

I'll work slowly along these lines when winter time will let me do this.

 

[nertog]

Relative Colorimetric

A relative colorimetric rendering intent maps the white of the source space to the white of the output. It reproduces in-gamut colors exactly and clips out-of-gamut colors to the nearest reproducible color.

I verified this with a few patch sheets and got quite good and matching results along the gray axis. Why along the gray axis - the readout is easy - any Lab value with the a or b value not 0 indicates a slight color cast. I'll do some more tests along this assumption about the rel. col. intent.

The absolute colorimetric intent forces the source and destination to have a common white point, usually D50. This means that white parts of your images will get a yellow hue. It is also the easiest way to do correct greyscale comparisons as you don´t have to worry about differences in white points between your input file and measurements on paper, although you can always calculate them out...

If you print images using the relative colorimetric intent, the white point of your source image (usually D50) is mapped to the white point of your destination (= the paper). In practice, this typically means a shift towards blue to compensate for the optical brighteners in the paper. This shift happens over the full greyscale, from the black point towards paper white.

Both intents clip out of gamut colors to the gamut border of the destination device.

Your profile seems to behave absolute colorimetric with a slow change to the paper white point when approaching near-white....weird.

 

[Ink stained Fingers]

yes, you are describing even more complex situations, I'm using the same D50 settings, the same paper etc, and I'm just looking along the gray axis at this time being in the middle of the gamut and far away from the gamut borders, and the starting point of my question was how well different profiles - created with ArgyllCMS or i1Profiler - with a low or high number of patches - are able to print neutral grays along the gray axis - or deviate one way or another. Another test would be a visual test against known targets e.g. RAL patches, but I don't have a D50 light box available, the L1800 used in this test does not have a gray ink but mixes the grays together which is more susceptible to changes in the ambient light. The test above with the ArgyllCMS profile by @pharmacist was the starting point, I'll do the same with an i1 profile based on 2880 patches or on 96 or 283 patches.

Your profile seems to behave absolute colorimetric with a slow change to the paper white point when approaching near-white....weird.

I checked about 5 times the print setting - I printed with the rel. col. intent - via 2 different programs which create essentially the same output. Prints mentioned above earlier with the abs. col. intent by far don't deliver clean grays.
I'm not looking to small fluctuations - there is a print process on a paper - there is a scan process to create the profile - there is a 2nd print with this profile on the same paper - and I'm scanning this patch sheet 3 x to average it - there is plenty of room for statistical fluctuations.

 

[nertog]

yes, you are describing even more complex situations, I'm using the same D50 settings

That´s not what I meant. The Lab color space has a relative white point that has to be agreed upon before comparing data. When printing with relative colorimetric rendering, that white point = paper white. This means that blueish-white of your paper has now a=b=0 values. The greyscale follows this.
Absolute colorimetric sets the white point reference of the color space to D50, which is slightly yellowish-white. Again, the greyscale follows. Your prints will look more yellow and the paper white is corrected (by printing colored ink) to the chromaticity of a D50 spectrum (= the color coordinates of the spectrum only, not after interacting with your paper or anything else)
When you are now scanning patches, either from your profiler package or another source, you once again have to specify a white point if you analyze that data in Lab format. A wrong white point = wrong measurements. This has nothing to do with viewing conditions or the hardware you are using. In the printing industry, the measurement white point is D50, but Argyll, for example, lets you change this.
To summarize: printing with the absolute intent and then remeasuring those printed patches should be correct. Relative rendering should show a shift towards the paper white.
Confusing stuff, right? I hope my train of thoughts is correct here...