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

[nertog]

I'm running a test with great support of @pharmacist

Nice work, both of you! It would be *very* interesting to print some patches and measure the absolute color accuracy of both profiles.

 

[Ink stained Fingers]

It would be *very* interesting to print some patches and measure the absolute color accuracy of both profiles.

Yes, I have something similar in my mind, but that needs a little bit of preparation.

I'm currently printing @pharmacist 's image file - and printing again - I was overlooking a missing nozzle showing some small banding here and there - it happens.............

 

[Ink stained Fingers]

The 3 prints are done - it's with an i1 profile with 96 patches , one with 283 patches and an ArgyllCMS profile by @pharmacist for my printer , I'm comparing them and have an opinion, I'll put the prints into the mail to him to comment as well - without prejudice. I'm curious if we come to the same conclusion.

 

[The Hat]

This great and very old thread was started by @Emulator, and would benefit from more of his inputs, but like many, he may have fallen in the Covid pandemic..

So long to our dear friend..

 

[Ink stained Fingers]

@pharmacist is proposing above in posting # 440 a test image which I'm using as well, particularly for the strawberry colors. This image consists of 2 parts - a collection of photos and a set of B/W-color ramps - a kind of technical section.

I separated these two and have a look to the color ramps in a special view - a pixel coud view where the pixels of the color patches are arranged by their color value in the Lab color space, you see several rows and chains across, and I have overlaid a profile of an L1800 printer showing the gamut of this printer/paper combination

A significant part of the color patches are out of gamut - specifically in the lighter upper color ranges which makes visual judgement of these colors quite difficult, color inconsistencies with these colors in print are in most cases not a profile issue but the out of gamut location; the rendering intent selectable in the printing software relocates such colors elsewhere depending of its intent by cahnging saturation and lightness.

And where are the colors located of the image files of this test image - here the same pixel cloud view

The image pixesl are represented by the pink dots, the blue shade is the gamut of the Netbit paper on a L1800, quite a large number of pixels fit into the printer gamut, but there is as well a significant part in the+a+b yellow to red segment, that's the strawberry, the Utah rocks and the sunset. That's an area you can use to check printouts on different papers, which different inks, these out of gamut reds let you loose image details from different saturation levels.

 

[Ink stained Fingers]

It would be *very* interesting to print some patches and measure the absolute color accuracy of both profiles.

Yes, I have something similar in my mind, but that needs a little bit of preparation.

I created a profile for the Netbit paper on a ET-8550 - this with 2880 patches, and I did a rescan of the 4 patch sheets 3 times to average small printing/scanning errors.

I'll use this profile with its assumed accuracy from the 2880 patches to print various other patch sheets which I can pcocess with the i1 profiler software and do some comparisons with them , I can let the i1profiler software doing the Delta calculations this way; but the coming weekend will be the earlist time for that. And this would allow me as well to test ArgyllCMS created profiles.

 

[Ink stained Fingers]

I'm doing some test with the support of @pharmacist to compare profiles generated either with the ARgyllCMS software or with XRite i1Profiler. Step one was done by creating a profile for my L1800 with Epson 106 inks with both packages, I#m using a i1Pro2 spectro, and @pharmacist a ColorMunki. The gamuts posted above show a very good congruence. I must admit that I didn't expect it that close. But the gamut borders don't tell it at all whether the color corrections for all colors inside the gamut are equally close.
So I created for the next step a profile based on 2880 patches for the Netbit/L1800 combination - that's 4 A4 patch sheets which I scanned 3 times and let the software average the data. I'm assuming at this point that such a profile delivers enough accuracy to use it for comparisons for prints with another i1 or Argyll profile. I'm using in this case a sheet with 283 patches and print it this time with either with this i1 profile or with the Argyll profile, in absolute colorimetric rendering mode - the mode which is supposed to print such that original colors are printed as they are supposed to be - without adjustments caused by the other rendering modes - no black point compensation - no saturation compression etc - the CocaCola red should print as it is specified by a Pantone color.

I'm getting 3 prints on the Netbit paper on the L1800

- one printed with the i1 profile based on 2880 patches
- one printed with a i1 profile based on 283 patches
- one printed with an ArgyllCMS profile

I'm comparing here 2 profiles based either based on 283 or 2880 patches - do more patches deliver more accuracy for the profile generation or not ?



The program displays quite a lot of information - deltaE's for every individual patch and some statistical numbers - averages of all the DeltaE's - in this case a value of 2.6. The patches with the highest 10% of DeltaE values are marked with a yellow border around the patches. I'm not going into details for individual colors at this time, the data can be exported for further processing in a spreadsheet or similar. And I don't know at this time if such a delta of 2.6 is a good or a bad value - I take it as it is, and

I'm now comparing the ArgyllCMS profile against the i1 2280 profile with this result




The summary delta average is now 6.42 for all the 283 patches - quite an increase to the above version. This now raises several questions for the origin of this higher value - is it the ArgyllCMS software or i1 software - are there other parameters which influence this difference. I don't know at this time, and I hope we can find the causes for some of these differences.

 

[nertog]

Hi ISF,

Wouldn't it be a better check to print some known LAB values (so basically a profile sheet) with absolute colorimetric rendering, measure the patches and compare them to the input values? Repeat for each of your 3 samples and you've got a clear answer of which one is most accurate, right?

I any case, keep up the good work

W.

 

[Ink stained Fingers]

Step 1 was a test and comparison of gamuts with a very good match, step 2 is a comparison of printed colors with the abs. col. rendering intent showing quite some variances which need an explanation and further evaluation. The numbers above are averages across all patch fields and do not focus on specific colors. Differences can be as small as deltaE of 0.5 or as large as 10, one check could be if this specifically affects colors close to the gamut borders.
I don't have analysis tools available like those from Babelcolor but will have a look to the gray fields, the L1800 does not use a light gray ink so grays are mixed from the available inks like any other color. The patch sheet contains specific gray steps I'll look to.

 

[nertog]

What I meant was that your analysis compares relative color differences between prints, but not the absolute color accuracy of the profile-printer combination. Never did this either, so I am quite curious as well. I´ll print out a few color patches spread over my R3000´s gamut later this evening and measure them. Let's see...