Using ArgyllCMS + Colormunki to produce excellent printer profiles

[RogerB]

I have used a very common chart the one with portraits of children in the lowest row. The portrait I'm talking about is rather good as it has a great dynamic range.

Roger thank you very much for your analyze! It seems as the variant with bigger patches is as good as the other with more patches. What I think could be interesting is to compare with a profile made with 1152 patches on two A4 papers. I have such a profile and I can download it. It could be interesting to see if the extra work with preprofiling is to any advantage?

I suspect we are splitting hairs now! The 1152-patch profile is similar to the other, but not identical. The colour gamut is a tiny bit bigger, but the accuracy (CC24) is much the same - still essentially perfect to the eye. The B&W response is a bit different, but still shows a gradual fall-off to the Dmax. Not necessarily a bad thing. Here are the screenshots.

Here is the prediction of the errors in the test image - I hope this is the one you refer to.

You will see that the predicted errrors for the faces are very small. The most difficult (for all printers) are the very deep reds and browns, but you printer does very well.

 

[nrdlnd]

I suspect we are splitting hairs now! The 1152-patch profile is similar to the other, but not identical.

Very kind of you analyzing my profiles!

I can see now that I did send the 'wrong' Canson Baryta profile the first time but still interesting how it fared and it fared well! The first one I sent was made with a preprofile on A5 and then I used one A4 chart with 720 patches. The one I intended to send was one with a 576 preprofile on one A4 and then 1152 patches on two A4 (altogether three A4 papers). That was the one I wanted to compare with the profile made on two A4 with 1152 patches and without a preprofile. I send it now!

In this light my earlier post may be a bit confusing! The first two profiles I sent was made with about 7 mm wide patches and the one for the matte paper was modified (made for a larger page and then cropped). The profile made with 1152 patches on two A4 has 8mm wide patches as the one I send now with a preprofile that has used three A4 papers.

 

[RogerB]

Very kind of you analyzing my profiles!

I can see now that I did send the 'wrong' Canson Baryta profile the first time but still interesting how it fared and it fared well! The first one I sent was made with a preprofile on A5 and then I used one A4 chart with 720 patches. The one I intended to send was one with a 576 preprofile on one A4 and then 1152 patches on two A4 (altogether three A4 papers). That was the one I wanted to compare with the profile made on two A4 with 1152 patches and without a preprofile. I send it now!

In this light my earlier post may be a bit confusing! The first two profiles I sent was made with about 7 mm wide patches and the one for the matte paper was modified (made for a larger page and then cropped). The profile made with 1152 patches on two A4 has 8mm wide patches as the one I send now with a preprofile that has used three A4 papers.

Not so different. Very small increase in colour gamut, but otherwise very similar.

I think it would be very difficult to see any difference between the three profiles, but you have to decide that from some real images.

 

[nrdlnd]

Not so different. Very small increase in colour gamut, but otherwise very similar.

This is for the profile made with three A4 papers (1152 patches with preprofile). This profile D-max 2.13 compared to the other 2.12. But it seems to be more linear especially in the green and reds. I do not exactly know what Argyll does with the preprofile but I can visually see especially that there are a lot more green patches.

I think it would be very difficult to see any difference between the three profiles, but you have to decide that from some real images.

I Think both these profiles are usable but the preprofile does something to the better. Thank you again Roger for all help! I may download and try the Gamutvision program as I have more profiles I want to evaluate. Edit: Gamutvision is not possible to install because of a lot of "DLL-hell". I give up! Edit: There may have been some firewall problem. It worked to install Gamutvision when I first downloaded the library Imatest-lib.exe.

My conclusion after all these experiments is that with the i1 Pro I will use the wider -a1.0 patches that gives 576 patches on one A4. These patches are easy to read without faults and when I make the chart it's not necessary to use PS or some other program to make the chart. I think the preprofile is beneficial. I will probably also try to make a profile with a preprofile and then 1728 patches on three A4. I buy A2 papers or 17" rolls and then cut them to smaller formats if necessary. With so much paper I think it's important to have a good profile to start with! I'm soon going to change to the Inktec ink in the big InkRepublic cartridges so I have to make new profiles again.

Per

 

[nrdlnd]

I have now tested one of my 'old' profiles with Gamutvision (trial version). I tried to use the same settings as Roger. The profile I tested was one with 1440 patches about 6.9mm wide on Canson Baryta and two A4 papers. This profile was one of my best profiles visually on the test print. The D-max was only 2.03 (may be because the quality setting was 4 and not 5) and there are some banding. I think the profile with 1152 patches 8mm wide is at least as good as this profile (D-max 2.12 and no banding). Conclusion: There may be important to have enough wide patches. I don't say that it is because that may need more testing and people on this forum have got good results with narrow patches.

Per

 

[RogerB]

I have now tested one of my 'old' profiles with Gamutvision (trial version). I tried to use the same settings as Roger. The profile I tested was one with 1440 patches about 6.9mm wide on Canson Baryta and two A4 papers. This profile was one of my best profiles visually on the test print. The D-max was only 2.03 (may be because the quality setting was 4 and not 5) and there are some banding. I think the profile with 1152 patches 8mm wide is at least as good as this profile (D-max 2.12 and no banding). Conclusion: There may be important to have enough wide patches. I don't say that it is because that may need more testing and people on this forum have got good results with narrow patches.

Per

What exactly do you mean by "some banding"? If you have any banding then there is something seriously wrong somewhere. If the banding is in smooth colour gradients or density gradients then the profile is usually to blame. Poor accuracy in the measurement of the charts is likely to give a profile that is "rough", so patch dimensions that give no mis-reads are essential. With my i1Pro I find that patches 7mm high and 10mm wide give very reliable results.

My approach is very simple. I am happy to use two A4 sheets to produce a profile, and I want the best results from my two A4 sheets. My experience is that 1058 patches with 0% mis-reads is better than 1540 patches with even one or two mis-reads. In other words, I think that accuracy is more important than absolute number of patches. Others may have different views.

 

[nrdlnd]

What exactly do you mean by "some banding"?

What I mean is that in the first view you have showed in Gamutvision in the upper right the rendering of the colors aren't even and smooth with the profile I tested myself. compared it to profiles you tested for me. I called this 'banding'.

My experience is that 1058 patches with 0% mis-reads is better than 1540 patches with even one or two mis-reads. In other words, I think that accuracy is more important than absolute number of patches.

Yes that's what I found out also. I've listened to your advice. I'm now reprofiling the Museo Portfolio Rag with wider patches and also less grey patches (64). I'm using a preprofile and I could read it without any misread (7.8mm wide patches). Tomorrow I will make the profile and publish my results and compare it to the other profile. I have a question: I think I can see in the black and white density response (L*a*b*c*) that the last profile you analyzed for me with the Canson Baryta is more linear (it has a preprofile) compared to the one without. Why I'm asking this is that I wonder if it's worthwhile to use a preprofile that needs an extra sheet of paper.

Per

 

[nrdlnd]

I have now finished the new profile on Museo Portfolio Rag with a 578 preprofile and then embedded in the 1156 profile on LetterR (I did at last find that command that makes the chart horizontal! For A4 A4R). g64 (the old profile g128). The patches are 7.8mm wide. Enlarged the print 99% in the printer driver. Not a single misread! Peak err = 1.378, avg err = 0.3173, RMS = 0.3626. This profile is very similar to the first. D-max 1.63 compared to 1.62. I think the color shift in the red is a little less and it's a darker red. I think I will use this profile for my 'portfolio' and I will try BPC on or off to see what gives the best result. Gamutvision is a nice program and thank you for pointing me to it!

Per

 

[RogerB]

I have now finished the new profile on Museo Portfolio Rag with a 578 preprofile and then embedded in the 1156 profile on LetterR (I did at last find that command that makes the chart horizontal!). g64 (the old profile g128). The patches are 7.8mm wide. Enlarged the print 99% in the printer driver. Not a single misread! Peak err = 1.378, avg err = 0.3173, RMS = 0.3626. This profile is very similar to the first. D-max 1.63 compared to 1.62. I think the color shift in the red is a little less and it's a darker red. I think I will use this profile for my 'portfolio' and I will try BPC on or off to see what gives the best result. Gamutvision is a nice program and thank you for pointing me to it!

Per

Yes, they are very similar. I cannot see any significant difference in Gamutvision. I have tried to compare them using the soft-proof facility in PS, but there is a bit of a problem in that both profiles seem to have the same (internal) description tag. Photoshop uses this tag rather than the filename so the two profiles appear as one in PS. After I changed the tags I was able to compare them - visually I cannot see any difference.

If you have PS you can try this yourself. Open the test image and convert it to profile A using Edit>Convert to profile. You will see a change in the image when you convert. Then soft-proof with profile B using View>Proof setup>Custom, but instead of choosing a normal rendering intent choose Preserve RGB Numbers. Toggling the soft-proof on and off will show the difference in rendering between the two profiles. You will see that for your two profiles they are, for all practical purposes, identical.

This method, although only a simulation, is more sensitive than comparing two prints side by side where differences of a fwe delatE may not be noticed. And of course it is very quick and very low cost!

 

[nrdlnd]

Yes, they are very similar. I cannot see any significant difference in Gamutvision.

The interesting thing is that this last profile was much easier to make and used much less patches! It also had less gray patches.

If you have PS you can try this yourself.

No I don't have PS and I'm not going to get it. I have other programs though where I can soft proof. I can do most of what I need in Capture One and I can use Gimp maybe. Gamutvision is nice though!

Per