Printer Manages Color: Down the Rabbit Hole

[unesco]

I think this software is only available on the Epson European website.  I got mine from here:  http://esupport.epson-europe.com/SupportHome.aspx?lng=en-GB&data=0ccCGROWIYM58zV3pqc84W2LhUcmWnHiUk31CNK52sUU003D   I punched in your printer and it was not listed so I think that means the software is not supported for the P7000.

ColorBase appears and dissapears randomly on all possible Epson site. I have seen it on Japanese, US, AU and EU webs but sometimes it becomes invisible, I don't know why.
For P7000 there is another application easy to find via support site.

[Doug Gray]

Here's the tests on my Canon 9500 II

Visually, a quick print from a cell phone snap in sRGB produces a much brighter and more saturated print using Printer Manages with the default driver settings than printing with the driver in ICM mode or using Photoshop Manages. Also, the default drive mode clips colorspaces to sRGB while the ICM driver mode does not. Here is what happens behind the scenes.

There are 8 charts showing the Lab response to Perceptual Intent as well as various Printer manages color modes.

Chart 1 is with the chart in Lab using the standard Photoshop Manages Color using the canned profile.
Chart 2 is with the chart in Lab using the standard Photoshop Manages Color using my custom I1Profiler profile.
Chart 7 is with the chart in Adobe RGB using the standard Photoshop Manages Color using the canned profile.
Chart 8 is with the chart in sRGB using the standard Photoshop Manages Color using my custom I1Profiler profile.
The canned profile shows a hue shift of about 15 degrees and an increase in luminance near the neutrals. The result is a brighter, and less blueish green at high saturation. The custom profile shows no hue shift and a fairly flat luminance that is about 8 Ls lower than the canned profile in neutrals.


These demonstrate conversion to sRGB prior to printing as well as large shifts towards brighter and more luminous color.

Charts 3, 4, 5 and 6 were printed using Printer Manages Color with driver mode (not driver ICM) selected. The image data was in, Lab, ProPhoto, Adobe RGB, and sRGB respectively. All 4 images with differing colorspaces produced very similar charts that indicate the image was first converted to sRGB then the print was processed. It is certain that conversion to sRGB occurs because they all have an a* curve bend at approx. a*= -36 which is the sRGB gamut edge. However, note the actual printed values of *a which vary from about -42 to -46 which is from 6 to 10 units of *a more saturated.  Further, the Luminance is increased even further with all the L values starting off at 60 in the neutral patches, 12 higher than that of the custom profile.  That creates a very large visual difference and is the default for Printer Manages Color.

 
These demonstrate that, when the driver is operated in ICM mode, the image colors are not clipped to sRGB but are converted per the supplied, canned profile.

Chart 9, 10, 11, and 12 have source colorspaces in Lab, ProPhoto RGB, Adobe RGB, and sRGB. The colors are much more properly rendered and the source image data is not clipped to sRGB. They are reasonably close to the results obtained using Photoshop Manages Color with the canned profiles.

[Doug Gray]

Here are the results of a single, specific test of "Printer Manages Color" on Windows 10, Version 1703 (current), x64 using a Canon 9500 II printer. The printer driver selection for color was "ICM" in the manual sub-dialogue. The paper used was Canon Pro Platinum photo paper using the Canon supplied profile: "Canon Pro9500 Mark II <PT> 1/2/3 Photo Paper Pro Platinum"  This profile is selected automatically and grayed out in Photoshop when using "Printer Manages." Colorimetric Intent selected was Absolute which automatically selects "absolute" in the driver dialog. This was chosen as opposed to Relative Intent since the later scales to the media white point.

Green gradients are from Lab(50,0,0), Lab(50,-2,0), Lab(50,-4,0) to Lab(50,-78,0).  This gradient is along the a* axis and has the property that the colors in both Adobe RGB and sRGB are within the printer's gamut. These gradients were first converted to Adobe RGB then sRGB. Then each were printed and measured with an I1 Pro. The gradients were also converted using the Canon profile to determine what the Canon profile expects the printer colors to be.

CORRECTION: Jpgs were "quick exported as jpgs" and the underlying settings automatically converted both to sRGB and didn't tag them with a color space. They now are tagged with the proper colorspace. My images are in 16 bit tifs but I don't know a way to post those on LuLa in other than 8 bit jpegs.

The results clearly demonstrate that, with ICM selected in the printer driver, From Photoshop, Windows does not convert image colors to sRGB using Printer Manages Color.

Here are the sRGB and Adobe RGB strips in 8 bit jpeg. On a wide gamut monitor in a color managed app, the difference between the clipped sRGB and Adobe RGB is very clear but otherwise you may not see any difference in the green gradient strips:

Also included are plots from Matlab showing the tracking of *a as it goes increasingly negative. The Adobe RGB clipping point inflection is around a*=-63, and the sRGB clipping point is around a*=-36. Plots are show for both I1Pro measured values of a* as well calculated a* from the Canon profile.

Attached green images were inadvertently untagged and worse, the Adobe RGB image had been converted to sRGB. Fixed:

[Ferp]

The results clearly demonstrate that, with ICM selected in the printer driver, From Photoshop, Windows does not convert image colors to sRGB using Printer Manages Color.

I'm trying to reconcile your findings (which I don't completely understand) with the statements of Adobe software engineer Dave Polaschek towards the end of the comments section of this TOP thread:

http://theonlinephotographer.typepad.com/the_online_photographer/2015/10/photoshop-vs-printer-managed-color-printing.html

in which he says, among other things (in relation to Windows):

"Printer Manages Colors guarantees a conversion to sRGB IF you are printing with a color profile attached to your image data.  ... So everything gets funneled through sRGB when you print with "Printer Manages Colors" (because ICM is enabled in that case)."

He also says that if you print without an attached profile then sRGB is simply assumed, in which case you can / should specify the ICC in the printer driver.  Your description of your workflow doesn't read like you did, so hence why I'm confused.

My main interest in printer manages colors under Windows has been for ABW printing, and as reporting in the other ABW-related thread, I measured an impact on linearity from the silent conversion to sRGB if I didn't use the null conversion hack.  I'd have expected some impact on color printing as well, unless you were doing as Dave P described and were printing an untagged image and specified the ICC in the driver. 

[Doug Gray]

I'm trying to reconcile your findings (which I don't completely understand) with the statements of Adobe software engineer Dave Polaschek towards the end of the comments section of this TOP thread:

http://theonlinephotographer.typepad.com/the_online_photographer/2015/10/photoshop-vs-printer-managed-color-printing.html

in which he says, among other things (in relation to Windows):

"Printer Manages Colors guarantees a conversion to sRGB IF you are printing with a color profile attached to your image data.  ... So everything gets funneled through sRGB when you print with "Printer Manages Colors" (because ICM is enabled in that case)."

He also says that if you print without an attached profile then sRGB is simply assumed, in which case you can / should specify the ICC in the printer driver.  Your description of your workflow doesn't read like you did, so hence why I'm confused.

My main interest in printer manages colors under Windows has been for ABW printing, and as reporting in the other ABW-related thread, I measured an impact on linearity from the silent conversion to sRGB if I didn't use the null conversion hack.  I'd have expected some impact on color printing as well, unless you were doing as Dave P described and were printing an untagged image and specified the ICC in the driver.

Well, I can't reconcile Dave comments with my results either and the test prints were done with images tagged with a colorspace. One Adobe RGB and the other sRGB and they printed appropriately. Had the Adobe RGB image been automatically converted to sRGB it would have produced the same print and graph as the sRGB one.

It may be that things have changed over the years and additional capabilities have been added that allow this. In particular, the default driver settings (not ICM) do silently convert to sRGB as shown earlier. But in ICM mode (9500), and ICM Host mode (9800) this doesn't happen and color is managed properly.

But there are a lot of options in the driver dialog that seem to be there for forcing different modes. Might be interesting to test them exhaustively but I'm not sufficiently curious enough to do it

I've included screen shots of the various dialogs which will set up the printer driver to manage color correctly.

*As an aside, I've been a bit distracted by testing out my new ISIS 2 XL. It turns out to be a great machine with exceptional repeatability. For instance, subsequent reads of a full, 256 patch gray tone yields dE variation of only .04 on average. This consistency, while really not important for photographic prints, comes in very handy for creating lab charts with more precise reflectance. I've used these in the past to measure the flatness error between the image plane and sensor and want to fine tune some of that prior work. It also turns out that their is some variation based on things like the paper position and even what prior ink patches were printed. It looks like somewhat better results making profiles can be had by taking a patch set and making multiple random patch location distributions then averaging them rather than just re-reading a target and averaging those readings.