Severe gamut reduction in profiles Colormunki vs i1Pro2 in ArgyllCMS

[vikcious]

Hi everyone,

I have recently upgraded my spectrophotometer from Colormunki Design to i1Pro2 (actually an ES-2000) and eagerly proceed to create some new paper profiles. In the past 99% of my profiles were created with ArgyllCMS since Colormunki design i1 studio was too lame.
I like a lot the powerful ArgyllCMS and most of the time the results were outstanding and yet the command-line experience didn’t scare me that much probably given my past experience being a Unix geek. Anyway…
Here comes the trouble. In short, given pretty much the same Argyll patches and commands, the difference in gamut size (for the same type of paper) between profiles created with Colormunki Design vs the ones created i1Pro2 is quite severe, almost 15% less gamut in the profile created with i1Pro2.

Basically the Colormunki profile (ArtMakerHighGlossy260_QM) is 776k cubic gamut units while i1Pro2 profile (ArtMakerHighGlossy260) is 672k cubic gamut units.

I started using i1Profiler and, despite my limited proficiency with the tool and its famous lack of proper documentation, I have managed to get quite a good profile but with even less gamut than the profile I would get from ArgyllCMS using pretty much the same amount of patches.
I have a feeling that the profile generation workflow in ArgyllCMS could be different for the two spectrophotometers but so far I haven’t been able to track the proper differences that would result in such a gamut loss.

I am confused and not so sure what is wrong! Is this just plain normal, as a difference between different hardware, is there a catch in the ArgyllCMS workflow or am I plain dumb… ‘cause confused I really I am! 😊

Any guidance of any type would be appreciated.

My basic ArgyllCMS profile workflow creation is like this:
Paper I am trying to profile now is a RC, high OBA/FWA content (for sure!) produced by the same factory that produces many of the Canon papers, here in Romania. Actually I have measured this paper spectral reflectance and it’s incredibly similar to the results of “Glossy Photo Paper 240gsm” Ernst Dinkla has captured in his SpectrumViz treasure box! 😉

Target Preparation
targen.exe -v -d2 -c CN_PRO-2000_520_GlossyPhotoPaperHG255.icc -R -e8 -G -g128 -f1046 ArtMakerHighGlossy260

Target Printing
printtarg.exe -v -iCM -h -R1 -L -m10 -M10 -T300 -p A4 -P ArtMakerHighGlossy260 (for Colormunki)
printtarg.exe -v -ii1 -h -R1 -L -m10 -M10 -T300 -p A4 -P ArtMakerHighGlossy260 (for i1Pro2)

Target Reading
chartread -v -H -T0.4 ArtMakerHighGlossy260 (for Colormunki)
chartread -v -H ArtMakerHighGlossy260 (for i1Pro2)

Profile Generation
colprof -v -qh -r1.0 -S "AdobeRGB.icc" -D "ArtMaker High Glossy 260g" ArtMakerHighGlossy260 --> OR -->
colprof -v -qh -i D50 -o 1931_2 -r1.0 -cmt -dpp -S "AdobeRGB.icc" -D "ArtMaker High Glossy 260g" ArtMakerHighGlossy260

[Doug Gray]

Looks like the larger differences are on the violet side. That's likely due to the impact of M0 (I1Pro 2 default) v M2 (CM).

If you're willing, zip the two profiles and attach the zip file to a post. I have pretty good tools that can detail out the impact on dE at a range of in gamut colors. Alternately, I can PM you my email if you'd rather not post them. But posting the zips here would let others analyze the profile differences as well and thus could provide a larger range of views.

[vikcious]

Looks like the larger differences are on the violet side. That's likely due to the impact of M0 (I1Pro 2 default) v M2 (CM).

If you're willing, zip the two profiles and attach the zip file to a post. I have pretty good tools that can detail out the impact on dE at a range of in gamut colors. Alternately, I can PM you my email if you'd rather not post them. But posting the zips here would let others analyze the profile differences as well and thus could provide a larger range of views.

Hi Doug,
Thanks for jumping in to help. No worries, I can share the profiles here. I'm not that "paternal" about them and if other can learn from my mistakes I am always willing to share. Here they are!

Also, depending on how proficient are you with Argyll, could you please validate my workflow specifically for the i1Pro2 usage? I feel like I'm missing smth when trying to profile high OBA/FWA papers with the my new i1Pro2 and Argyll.

Thanks!

[Doug Gray]

I downloaded the profiles. A quick look sees nothing wrong. The Argyll profile, as is the default, produces identical tables for Perc. Rel. And Sat. The paper shows a b* shift around 6 between the two which is typical for a high OBA paper between M0 and M2.

I'm running some other checks on the Rel Col. Abs isn't useful as it will vary considerably due to the M0/M2 differences. Will let you know much more detail in a hour or so.

Had some other stuff to do but am back on this. Interestingly, the profiles contain spectro data in 3.3nm increments instead of the standard 10nm. I'm going to look at both instrument's C Y and M spectra for clues.

[Doug Gray]

I'm going to create an I1Profiler CGATs files out of the spectral data and make profiles using I1Profiler for comparison. Should be interesting to compare them.

[Doug Gray]

The tone curves on neutral colors differ significantly. L* is 10 to 15 higher on the I1Pro generated profile. It's not the profile. The spectral data shows the same thing.

Somehow the target printed for the I1Pro was not printed with color management turned off while the CM one was. What workflow did you use to print the I1Pro targets?


BTW, Profiles created from the data sets using I1Profiler are quite close to the Argyll ones. Problem was in the printing of the I1Pro patch set.

The graph shows printed, Rel Col,  L* v device space RGB neutral values from 0 to 255. The blue line is the CM profile. The orange line is the I1Pro profile. The latter was not printed with color management off. Your Argyll s/w creates 16 bit tiff files which should be printed in device RGB space with color management off using the same settings when printing with a profile in Photoshop. Are you using Windows, iOS or some other OS?


BTW, my first exposure to Unix and command line stuff, aside from DOS, was running an IBM-XT with 2 extra user terminals on RS-232. We did device driver coding, source control, and documentation on it.  Quite amazing in retrospect as it only had 640KB of RAM. Somewhere around '82 or '83.

[dehnhaide]

I'm going to create an I1Profiler CGATs files out of the spectral data and make profiles using I1Profiler for comparison. Should be interesting to compare them.

Niiiice, I wish I could learn how to do that too. I am very much into learning everything about color management since have a little crush on it since I bought my Canon PRO-2000! Is there information available how to do that? Thanks!


Sent from my iPhone using Tapatalk

[vikcious]

The tone curves on neutral colors differ significantly. L* is 10 to 15 higher on the I1Pro generated profile. It's not the profile. The spectral data shows the same thing.

Somehow the target printed for the I1Pro was not printed with color management turned off while the CM one was. What workflow did you use to print the I1Pro targets?

Hmm... then we might have a problem! Because my workflow for printing the targets is always the same.

BTW, Profiles created from the data sets using I1Profiler are quite close to the Argyll ones. Problem was in the printing of the I1Pro patch set.

The graph shows printed, Rel Col,  L* v device space RGB neutral values from 0 to 255. The blue line is the CM profile. The orange line is the I1Pro profile. The latter was not printed with color management off. Your Argyll s/w creates 16 bit tiff files which should be printed in device RGB space with color management off using the same settings when printing with a profile in Photoshop. Are you using Windows, iOS or some other OS?

So it goes like this. Windows 10 + Canon PRO-2000. You can see from the images I attached that the printing workflow is pretty basic. And using Canon Print Studio Pro to print without color correction was specifically indicated in one of two Canon PRO-2000 major reviews as the way to print without color correction on this printer. Can't remember if it was on northlight-images or on Mark Segal's one.
Is this wrong? And if so why was it OK for the CM targets and any different for the i1pro2? I don't get it...

BTW, my first exposure to Unix and command line stuff, aside from DOS, was running an IBM-XT with 2 extra user terminals on RS-232. We did device driver coding, source control, and documentation on it.  Quite amazing in retrospect as it only had 640KB of RAM. Somewhere around '82 or '83.

I am definitely younger than you and my UNIX odyssey started back in the 1990 when the very fist exposure to UNIX was via Chip magazine that included the mighty Slackware cd-rom. I remember I was impressed at that time by how unfriendly it looked and cumbersome was to actually get that Xorg graphic server up and running on my Tseng 6000 video card. And from that point on ... I build my carrier on it! Anyway... those days were amazing and I keep them dear memories. 

BTW... I have attached another three profiles:
The profile I created with i1Profiler for the same paper we're discussing here (ArtMaker_HighGlossy_260g_1701p_i1pro2)
There other profiles I created for under the same circumstances for a different paper (high OBA, RC, Crystal Luster, 270g):
- one older created with CM and Argyll (ArtMakerCristaLuster_840p_CM_HQ)
- one new created with i1Pro2 and Argyll (ArtMakerCrystalLuster270_i1Pro2)
- one new created with i1Pro2 and i1Profiler (ArtMaker_CrystalLuster_270g_546_i1pro2)

Same observations + 1 new... Reduced gamut CM vs i1Pro2 and the new is that the black levels on the i1pro2 profile got with Argyll are messed up vs the profile crated in i1Profiler. I have attached a screen shot of what is happening when I am soft-proofing!
 

[vikcious]

Had to split the initial archive because it was exceeding the 4096KB limit. He I've attached the last one of the three profiles described in the previous post.
 

[vikcious]

The tone curves on neutral colors differ significantly. L* is 10 to 15 higher on the I1Pro generated profile. It's not the profile. The spectral data shows the same thing.

Somehow the target printed for the I1Pro was not printed with color management turned off while the CM one was. What workflow did you use to print the I1Pro targets?

Found it! It was Mark Seagal's review and he noted:

"This reassured me that the methodology *can be* fine. But clearly not “all else is equal”; so “what gives” between ACPU and the Canon driver with this Pro-2000 printer? Further consultation with Canon revealed that they too had worked on this issue and they recommended I should not use ACPU for generating the profiling targets in these particular printers, but rather their own utility – Photoshop Plugin Print Studio Pro Version 2.1.0 (hereafter PSP), with the Color Mode set to “No Color Correction”. *Why* has never been clarified, this being a matter best left to the engineers in Adobe and Canon, but my basic objective was to see what works best empirically and then use it to print."

[Doug Gray]

Found it! It was Mark Seagal's review and he noted:

"This reassured me that the methodology *can be* fine. But clearly not “all else is equal”; so “what gives” between ACPU and the Canon driver with this Pro-2000 printer? Further consultation with Canon revealed that they too had worked on this issue and they recommended I should not use ACPU for generating the profiling targets in these particular printers, but rather their own utility – Photoshop Plugin Print Studio Pro Version 2.1.0 (hereafter PSP), with the Color Mode set to “No Color Correction”. *Why* has never been clarified, this being a matter best left to the engineers in Adobe and Canon, but my basic objective was to see what works best empirically and then use it to print."

That's quite disturbing. I don't like unexplained phenomina. I've never seen a difference between printing directly from IK1Profiler or the tif file using ACPU. Or, for that matter using the null transform trick to print directly from Photoshop.

My Canon 9500 has only intermittently worked with Canon's Print Studio Pro. To gprint through Photoshop, I just use the null transform approach (Which does not work with iOS but does with all versions of Photoshop and Windows) .

That said, whatever Studio Pro is doing it should be the same for the different targets and that's not the case. No idea why.

My standard test for checking profiles is to print a Colorchecker image from BabelColor using Abs. Col. then visually compare it to an actual Colorchecker.

Here's the 2D gamut at L*=25 for the CM and I1 profiles you made. IK also stripped the spectral data and made profiles using I1Profiler.

The white lines are from I1Profiler profiles, the yellow from Argyll..  The outer group is the ColorMunki data, the inner group is the i1Pro.  The Argyll and IK1Profiler profiles are quite close with small differences since they have different numbers of 3D LUTs.

[Panagiotis]

Canon PSP is not a stand alone app, so targets must be opened first in PS and the plugin is called from there. Maybe the iPro2 targets accidentally got a profile conversion when opened in PS?

[vikcious]

Canon PSP is not a stand alone app, so targets must be opened first in PS and the plugin is called from there. Maybe the iPro2 targets accidentally got a profile conversion when opened in PS?

Absolutely no chance of error on assigning or converting the profile. You can actually see from my screenshot above that I clicked on OK without triggering the profile assignment. I do the all the time, all the same for all targets.

[Panagiotis]

Absolutely no chance of error on assigning or converting the profile. You can actually see from my screenshot above that I clicked on OK without triggering the profile assignment. I do the all the time, all the same for all targets.

(EDIT: Sorry I just noticed the screenshot from PS above!)

This is strange then. Was it the same version of PSP for both targets sets?

[vikcious]

That's quite disturbing. I don't like unexplained phenomina. I've never seen a difference between printing directly from IK1Profiler or the tif file using ACPU. Or, for that matter using the null transform trick to print directly from Photoshop.

I will be traveling for the next 3 days but when back I will try to print either directly from i1Profiles or from ACPU. Never did but willing to try.

My Canon 9500 has only intermittently worked with Canon's Print Studio Pro. To gprint through Photoshop, I just use the null transform approach (Which does not work with iOS but does with all versions of Photoshop and Windows).

Could you please clue me what is this "null transform approach", please?

That said, whatever Studio Pro is doing it should be the same for the different targets and that's not the case. No idea why.

My standard test for checking profiles is to print a Colorchecker image from BabelColor using Abs. Col. then visually compare it to an actual Colorchecker.

Here's the 2D gamut at L*=25 for the CM and I1 profiles you made. IK also stripped the spectral data and made profiles using I1Profiler.

The white lines are from I1Profiler profiles, the yellow from Argyll..  The outer group is the ColorMunki data, the inner group is the i1Pro.  The Argyll and IK1Profiler profiles are quite close with small differences since they have different numbers of 3D LUTs.

I am curious if you are able to detect the same mismatch in the printing condition from the latest i1Profiler profile that I attached. Its absolutely the same workflow and I have real doubts the first one was a mistake.

Thanks a lot for taking the time to tinker with this.

[vikcious]

(EDIT: Sorry I just noticed the screenshot from PS above!)

This is strange then. Was it the same version of PSP for both targets sets?

Absolutely! I actually haven't updated the software (PSP) since I got the printer last October.

[Panagiotis]

I will be traveling for the next 3 days but when back I will try to print either directly from i1Profiles or from ACPU. Never did but willing to

There is also the new standalone and plugin application from Canon called Professional Print & Layout which is also has a "No Color Correction" option presumably for printing profiling targets.

[Doug Gray]

Could you please clue me what is this "null transform approach", please?

In Windows, one can print w/o color management by doing the following:
Open the tiff target image but make sure it isn't automatically converted to a working colorspace. This doesn't normally happen but there are custom settings that can do it.

1. Assign any printer profile to the image. But note the exact name of the profile. I like to pick one close to the list top to make it easier to find.
2. Now open the print dialog. Make sure the driver has color management turned off. Set Photoshop Manages Colors then choose the exact same profile as in step one. You will get a scary warning that you appear to be trying to print w/o color management and it suggests and links to a program (ACPU). Ignore the warning and cancel the warning box. It's there because this doesn't work in iOS.

Ignore the other settings like Colorimetric Intent, BPC, and such. They have no effect when the printer profile of the image is the same as the printer profile selected in the dialog box.

This technique is also used when creating a montage and desiring to print the images with differing Intents such as Perc., Rel. Col. Abs. Col. with/without  BPC. Convert each to the printer profile using desired intent. Insert them into a blank canvas that has been assigned the printer's profile.  Then just print using the same profile.

I am curious if you are able to detect the same mismatch in the printing condition from the latest i1Profiler profile that I attached. Its absolutely the same workflow and I have real doubts the first one was a mistake.

Thanks a lot for taking the time to tinker with this.

The profile appears to be quite a different paper (matte?) with a white point at L*=92 and black point at L*=13. Also a ton of OBAs. No way to compare against the other two.

[vikcious]

In Windows, one can print w/o color management by doing the following:
Open the tiff target image but make sure it isn't automatically converted to a working colorspace. This doesn't normally happen but there are custom settings that can do it.

1. Assign any printer profile to the image. But note the exact name of the profile. I like to pick one close to the list top to make it easier to find.
2. Now open the print dialog. Make sure the driver has color management turned off. Set Photoshop Manages Colors then choose the exact same profile as in step one. You will get a scary warning that you appear to be trying to print w/o color management and it suggests and links to a program (ACPU). Ignore the warning and cancel the warning box. It's there because this doesn't work in iOS.

Ignore the other settings like Colorimetric Intent, BPC, and such. They have no effect when the printer profile of the image is the same as the printer profile selected in the dialog box.

This technique is also used when creating a montage and desiring to print the images with differing Intents such as Perc., Rel. Col. Abs. Col. with/without  BPC. Convert each to the printer profile using desired intent. Insert them into a blank canvas that has been assigned the printer's profile.  Then just print using the same profile.

Wow! That's quite a workflow! Thanks for sharing, Doug! Never seen this before!

The profile appears to be quite a different paper (matte?) with a white point at L*=92 and black point at L*=13. Also a ton of OBAs. No way to compare against the other two.

Well, surprise surprise! All the last three profiles I've submitted are for the same paper, luster semi matte type! At least from a gamut perspective the situation is / should be similar to the initial one!

[Doug Gray]

Here's a summary of the three profiles you made:

Profile: "C:\WINDOWS\System32\Spool\Drivers\Color\ArtMakerHighGlossy260.icm"
Descr: "ArtMaker High Glossy 260g 2nd" Ver: 2.2.0   Copyright: "Copyright, the creator of this profile"
WP:  95.2   1.8  -6.3     BP_PI:   2.7  -0.5  -1.1    BP_RI:   3.0   0.3  -0.5

Profile: "C:\WINDOWS\System32\Spool\Drivers\Color\ArtMakerHighGlossy260_QM.icm"
Descr: "ArtMaker High Glossy 260g" Ver: 2.2.0   Copyright: "Copyright, the creator of this profile"
WP:  94.7   0.2  -0.8     BP_PI:   3.2  -0.2  -0.1    BP_RI:   3.2  -0.2  -0.1

Profile: "C:\WINDOWS\System32\Spool\Drivers\Color\ArtMaker_CrystalLuster_270g_546_i1pro2.icm"
Descr: "ArtMaker_CrystalLuster_270g_546_i1pro2.icm" Ver: 2.1.0   Copyright: "Copyright X-Rite, Inc."
WP:  92.0   2.4  -9.7     BP_PI:  13.1   1.9   0.6    BP_RI:  13.1   1.9   0.6

Notice the white point shift (b*) for the two made with an I1Pro.

Also noticed the huge increase in black point in the last one with L* going form 3 to 13.  That's why the soft proof looks like it does. Something is messing up your results.

I suggest you use a single page, small (200 to 400)  patch set for all future tests until whatever is causing these variations is found and fixed. By using the same patch set you can directly compare patches printed on each. They should, of course, be quite close and visibly the same. This also makes extracting the data and comparing them much easier.

I noticed the last profile has 8 bit RGB values and spectral data at 10nm intervals. The first two had 16 bit RGB values (scaled 0:100) and spectral data at 3.3nm intervals. Argyll has this higher resolution option, XRite doesn't and uses only 10nm intervals.

I checked the RGB(0,0,0) spectral values and they show about 1.4% reflectance which is quite a lot higher than the other two profiles which were about .3%.

These results are extremely strange.  Too many variables. Each of the 3 profiles has different patch sets, which makes troubleshooting unnecessarily complex.