Canon PRO1000--colours off as far a dE 26--is it broken?

[Prospero]

Hi All,

I have had a Canon Pro1000 now for a year and was never really happy about the color rendition.  I am not an experienced color manager, but still decided to get to the bottom of it with some test charts now and discovered that some specific tones are completely off.  One very striking aberration occurs in the cyan tones. 

Concretely, when using the x-rite color checker chart (https://xritephoto.com/ph_product_overview.aspx?ID=820&Action=support&SupportID=5159#), the patch number 18, which is supposed to have Lab values 51, -28, -28, comes out exactly as dark and saturated as the blue patch (number 13), and when I measure it, it shows Lab values 31, 15, -37, which is a whopping dE(76) of close to 26 off.  Some other colours are also off (green to dark, red too orange), but not as far.

I get the exact same results, when printing the test chart from the x-rite iProfiler Tool and when printing it from Photoshop.  I also get the exact same results when using my own profile vs. using cannes ones from Hahnemühle.  So it can't be a broken profile.  The differences are also so big that they cannot be due to the settings (black-point comp., etc.).

My understanding is that acceptable color differences should not be more than dE of, say, 3-5.  So does someone have a suggestion what else I could try, or should I assume that my printer was broken from the start?

Many thanks in advance!
Markus

[Doug Gray]

Yep. Something is badly broken or something in your process is way out of whack.

What instrument are you using to measure color? Are you using the Canon profiles, custom, or ones you make yourself?

The proper way to test ColorChecker colors is to print an image with these colors. You can download ColorChecker images made from various vintages from here:

https://www.babelcolor.com/colorchecker.htm

Print the image from Photoshop using Absolute Colorimetric intent. Then measure the colors and compare to the Lab values in the image.

Doesn't really matter which one you pick as you can measure the Lab values directly in Photoshop and they are all in gamut using most papers with the exception of the whitest and darkest neutral patches which are sensitive to the printed paper but should be quite close even if slightly out of gamut.

[Prospero]

Thank you very much, Doug, this is really helpful!  Yes, definitely need to continue digging deeper.

I understand your approach and the link is very helpful!  I assume I need to be very careful that any checker image has the right profile embedded and that PhotoShop uses that.  When printing absolute colorimetric from PS, I am getting a warning that I should rather use the Adobe Color Printer Utility, if I want to to print a calibration target without color management.  I find this a bit confusing, since there is clearly a difference between printing with absolute colorimetric vs. no color management and I am wondering, if the former still works perfectly, or if PS has partially switched off support for this.

I am generally surprised that there isn't a more systematic way for this kind of verification process included in my x-rite i1Profiler application.  Isn't it a very common use case to test the accuracy of a printer profile?  So why can I not just choose a calibration patch set in the app, print that *with* the profile I want to test, scan the print, and get a bunch or error metrics, like average/max/min dE, etc.  After all, the i1Profiler has all the info needed (correct Lab values, etc.) and 90% of that work flow is identical to creating a new target, so it's already there, except for the final comparison step.  Maybe I am missing something.  Also glad for any pointer to a different app that would provide this kind of workflow, which would avoid having to figure out the Lab values for my checker image and then do very manual comparisons.

Lastly, what should I expect good to look like?  I vaguely remember seeing a post that said dE of 1 is about as good as it gets.  But not sure, if I should expect to reach that with my printer/setup and if I even need that.

Thanks again!

[Doug Gray]

Thank you very much, Doug, this is really helpful!  Yes, definitely need to continue digging deeper.

I understand your approach and the link is very helpful!  I assume I need to be very careful that any checker image has the right profile embedded and that PhotoShop uses that.  When printing absolute colorimetric from PS, I am getting a warning that I should rather use the Adobe Color Printer Utility, if I want to to print a calibration target without color management.  I find this a bit confusing, since there is clearly a difference between printing with absolute colorimetric vs. no color management and I am wondering, if the former still works perfectly, or if PS has partially switched off support for this.

I am generally surprised that there isn't a more systematic way for this kind of verification process included in my x-rite i1Profiler application.  Isn't it a very common use case to test the accuracy of a printer profile?  So why can I not just choose a calibration patch set in the app, print that *with* the profile I want to test, scan the print, and get a bunch or error metrics, like average/max/min dE, etc.  After all, the i1Profiler has all the info needed (correct Lab values, etc.) and 90% of that work flow is identical to creating a new target, so it's already there, except for the final comparison step.  Maybe I am missing something.  Also glad for any pointer to a different app that would provide this kind of workflow, which would avoid having to figure out the Lab values for my checker image and then do very manual comparisons.

Lastly, what should I expect good to look like?  I vaguely remember seeing a post that said dE of 1 is about as good as it gets.  But not sure, if I should expect to reach that with my printer/setup and if I even need that.

Thanks again!

First, getting the " Adobe Color Printer Utility" warning has nothing to do with what Intent you select to print. it's a result of having the image colorspace the same as what you select in the printer dialog. That is not normal when printing and is a way of bypassing color management that works in Windows but doesn't reliably in Apple OS. This produces the warning. You should be selecting the printer/paper profile for the Pro1000 in the print dialog.

Appears you have an i1Pro2 or i1Pro3 and i1Profiler software. Is this correct?

[Prospero]

Yes, I am indeed using an i1Pro3Plus with the i1Profiler application.
Thanks again!

[Doug Gray]

Yes, I am indeed using an i1Pro3Plus with the i1Profiler application.
Thanks again!

Do be careful making profiles using M3 (polarized light). These can produce pretty off colors depending on media. The M3 mode is great for profiling fabrics and certain coarse materials and are especially good for fine detail in deep shadows using Perceptual Intent or Relative with BPC. The downside is they tend to read colors as more saturated and darker than they really are. Especially on matte paper. This is because the polarized light does not reflect as much from matte surfaces and so measures colors that are darker and more saturated than they are when observed under normal lighting.

The upside is that with fabrics M3 will give more consistent (less noisy) readings.

I'm curious as to why you chose to buy a i1Pro3plus. They really are something of a specialty item. but they do make great profiles at the cost of a lot more paper charts for the same number of patches.

You should easily be able to make profiles that print colorchecker (and other) colors with less than 1dE errors.

[Prospero]

I solved my problem!  The issue was a buggy/outdated printer driver.  At least reinstalling the latest one made the problem disappear.  Big thank you to henrik olsen!

Thanks for the tip--had not figured out the differences between M1-3 yet.

I actually only borrowed this i1Pro.  In what sense is it "specialty"?  So far, it seems to offer mainly the standard/mainstream use cases (monitor, print, projector calibration) with what seem to be not many more options/functions than, e.g., the datacolor spyder print kit I owned 10 years back.  I do like the built quality and the fact that you can scan entire lines of patches in one quick go.  The only downside I found so far is that the i1Profile application wastes a lot of paper--it should be possible to print about twice as many patches on one sheet, compared to the layout the application uses.  Is that what you mean by "a lot more paper charts"? 

Great to know that 1dE is possible and that should really help keep *tones* consistent vs. what I see on the screen.  With regards to *luminosity/saturation/contrast* though, I would argue you need an incredible amount of experience judging what you see on the screen and predicting that this is really what you want on paper, meaning the human error in misjudging luminosity/saturation/contrast, etc. seems to be easily worth 5-10 dE in precision.  Even the proofing views in PS don't really help that much to that end, I find.  Just a thought.

Is there a rule of thumb as to how many patches I need in my target to reach this 1 dE precision?

Can you (or someone who reads this) recommend any application that would facilitate the "profile verification" workflow I mentioned above?  I.e., selecting/creating a patch set target, printing it with the profile to be verified, scanning the result, creating error measures between the original Lab values and what was actually printed?

One solution could be to get the i1Profiler to save a file that includes both the original target patch Lab values alongside the scans or the printed sample, in a format that I could read easily with python for example.

Many thanks again!

[Doug Gray]

I actually only borrowed this i1Pro.  In what sense is it "specialty"?  So far, it seems to offer mainly the standard/mainstream use cases (monitor, print, projector calibration) with what seem to be not many more options/functions than, e.g., the datacolor spyder print kit I owned 10 years back.  I do like the built quality and the fact that you can scan entire lines of patches in one quick go.  The only downside I found so far is that the i1Profile application wastes a lot of paper--it should be possible to print about twice as many patches on one sheet, compared to the layout the application uses.  Is that what you mean by "a lot more paper charts"?

Yes, on a letter size sheet, the Pro3+ requires 16x16mm patches while the Pro3 uses 8x7mm patches. This is 140 patches for the Pro3+ v 693 patches for the Pro3.

People usually profile with at least 600 patches and usually over 1000. Some will regularly use 3000 or more. This is a lot of paper with the Pro3+.

Is there a rule of thumb as to how many patches I need in my target to reach this 1 dE precision?

I find 957 patches (i1iSis which has mins. of 6mm patches) and fits on one page produces deltaE 2000 of around .5 average with about 1.3 dE in the worst 1% of random, in gamut, patches. You should be fine with 1120 patches on 8 pages or even 560 patches in 4 pages. I wouldn't try for real high patch counts until you are comfortable with the overall process. Too easy to get frustrated printing and scanning 20 pages when something isn't working elsewhere.

Can you (or someone who reads this) recommend any application that would facilitate the "profile verification" workflow I mentioned above?  I.e., selecting/creating a patch set target, printing it with the profile to be verified, scanning the result, creating error measures between the original Lab values and what was actually printed?

One solution could be to get the i1Profiler to save a file that includes both the original target patch Lab values alongside the scans or the printed sample, in a format that I could read easily with python for example.

To verify the accuracy of a profile you need to print a set of colors that are different from those in the patch set used to make a profile. A common way is to print a colorchecker image with Photoshop (Lightroom can't do this) using Abs. Col. Intent then measure the printer patches. Then measure the patches' Lab values and compare them to the Lab values as read out from the image in Photoshop. You can also make your own set of colors directly in Photoshop using Lab colorspace and print/measure those. But make sure they are printable colors (in gamut). As for tools, check out PatchTool by BabelColor. It allows easy creation of sets of colors spread out in a printable gamut. You can use i1Profiler to read in arbitrary color patch grids and save the result as a CGATs file which is standard text and easily read.

[Prospero]

Thank you very much, once more--really helpful advice!  I will take it from here now.
 
(Except that my problem has come back again this morning...which is super frustrating...posted examples here:
https://forum.luminous-landscape.com/index.php?topic=133310.0)

[Doug Gray]

Thank you very much, once more--really helpful advice!  I will take it from here now.
 
(Except that my problem has come back again this morning...which is super frustrating...posted examples here:
https://forum.luminous-landscape.com/index.php?topic=133310.0)

Based on your post in the printer section, it sure looks like a bug in MacOS. Probably something in the process is not applying CM to your image and it's getting bypassed. Unfortunately I've never seen anything like that kind of bug but I've only used Windows and Dos before that. Hopefully Mac people will be able to help.

One possible test is to make a second image by converting, not assigning, a CC image from Adobe to ProPhoto. Print both. If color management is being applied, regardless of all other settings, both prints should look the same. If CM is not being applied due to some OS bug then the ProPhoto one should be darker and less saturated. That won't fix the problem but might narrow it down to where someone with Macs can help.

Best of luck to you.