Why do prints from different profiles look alike?

[Rich42]

I'm printing on Epson Hot Press Bright on an Epson 7890 24" printer. I've been using the "canned" Epson paper profile as well as a custom ICC profile that I purchased that is supposed to be "very high quality." The custom profile was made from targets consisting of 2600 total patches and I'm told, "read on X-Rite equipment." So I hope the quality is as advertised.

I'm printing from Mac Photoshop CS6, 16 bit files with 16 bit enabled in the print dialog, Photoshop manages colors, "normal printing," Relative Colorimetric Intent, Black Point Compensation,.

I've examined the two profiles in the Mac Colorsync Utility. The canned Epson space is 10-15% larger than the custom profile. They are exactly the same shape. They touch at the white point, but since the Epson profile is larger, it extends significantly "lower" than the custom one at the dark "bottom" end.

My screen is calibrated and profiled (Mac Cinema display, i1 Pro display profiling hardware/software and 3rd party software).

Prints made from the two profiles are visually almost indistinguishable. Deep blacks are the same despite the very noticeable difference in their 3D representation in ColorSync. Hue and saturation everywhere, in a wide variety of subject matter is the same, except for a barely noticeable pink tint in some neutral grays of some images (so slight that you have to go back and forth between prints to see it) in the canned Epson profiled print. The custom profiled prints in those cases are more true to the screen image.

So why is this happening? Why does a custom profile show a smaller gamut than the canned profile? Why does it then perform the same, visually, across a wide variety of images?

I guess it matches the screen image better (by a gnats hair) because it IS more accurate, being a measure of my current machine/inks/paper batch. But why would it look significantly smaller in gamut and still perform identically?

Thanks,

Rich

 

[Howard Moftich]

How do you know that the images you're printing are stressing the gamut of the paper/profiles in question? The gamut 3D graph just shows the surface of the space and nothing about how any image fits into that space.

 

[Rich42]

Howard,

I don't know how far I'm "stressing" the gamut, but the images contain a lot of nearly saturated colors, deep blacks. I'll try to post a few.

 

[technoid]

I'm printing on Epson Hot Press Bright on an Epson 7890 24" printer. I've been using the "canned" Epson paper profile as well as a custom ICC profile that I purchased that is supposed to be "very high quality." The custom profile was made from targets consisting of 2600 total patches and I'm told, "read on X-Rite equipment." So I hope the quality is as advertised.

I'm printing from Mac Photoshop CS6, 16 bit files with 16 bit enabled in the print dialog, Photoshop manages colors, "normal printing," Relative Colorimetric Intent, Black Point Compensation,.

I've examined the two profiles in the Mac Colorsync Utility. The canned Epson space is 10-15% larger than the custom profile. They are exactly the same shape. They touch at the white point, but since the Epson profile is larger, it extends significantly "lower" than the custom one at the dark "bottom" end.

My screen is calibrated and profiled (Mac Cinema display, i1 Pro display profiling hardware/software and 3rd party software).

Prints made from the two profiles are visually almost indistinguishable. Deep blacks are the same despite the very noticeable difference in their 3D representation in ColorSync. Hue and saturation everywhere, in a wide variety of subject matter is the same, except for a barely noticeable pink tint in some neutral grays of some images (so slight that you have to go back and forth between prints to see it) in the canned Epson profiled print. The custom profiled prints in those cases are more true to the screen image.

So why is this happening? Why does a custom profile show a smaller gamut than the canned profile? Why does it then perform the same, visually, across a wide variety of images?

I guess it matches the screen image better (by a gnats hair) because it IS more accurate, being a measure of my current machine/inks/paper batch. But why would it look significantly smaller in gamut and still perform identically?

Thanks,

Rich

The reason the profiles look different is that gamut is determined from the colorimetric tables (BtoA1 and AtoB1 as defined in the spec at color.org) Epson has long used profiles that wrongly implement black point correction in these tables. Back, about 15 years ago, there was some ambiguity as to whether this was a correct way to create profiles but the ICC clarified things about 10 years ago. Epson has apparently decided to continue it's initial practice. Perhaps to prevent confusion from a change.

The reason you don't see much difference is that, aside from this anomaly, typical printing uses the Perceptual Intent and Epson and X-rite are very similar. And they both do BPC in their Perceptual Intent tables .

If you want to see the impact of this print using Relative Colorimetric but leave BPC (Black Point Correction) unchecked. The differences will show up in deep shadows.

OTOH, if you check BPC both profiles will create very similar prints. Correctly made profiles used in Relative Colorimetric w/o BPC can produce blocking in the darkest shadows and even a slight color shift. This is because the profile will try to print the patch as close as possible to the requested color and an Lab(0,0,0) could well be closer to Lab(6, -3, 0) than the neutral Lab(6,0,0).

Incorporating BPC in the profiles themselves has a slight impact on profile accuracy and can show up in softproofing but Epson avoids consumer questions about shadow blocking and color shift by people that use Relative Colorimetric without understanding some of the subtleties.

 

[Rich42]

Technoid,

Thanks, that seems to describe what I'm seeing.

Here are some images. All from a D800e, Nikon 180 f/2.8, wide open. Summer evening in Carlsbad, CA.

The prints are crops of these and others. There's a pretty good range of gamut here.

 

[Gesture]

You have to print with each profile on a standardized image designed to test the limit of an ink set, paper, printer. Keith Cooper and The Digital Dog have plenty of such images on their web sites. I would start by testing the gray scale ramp.

 

[technoid]

Nice pics. Thanks for sharing.

 

[Rich42]

Good point about standardized test images. I'll try that. I've used them many times to see differences in papers. Didn't think of using them to test profiles on the same paper.

l'll post my results. But even if they show significant differences, a wide range of images do not.

 

[technoid]

Good point about standardized test images. I'll try that. I've used them many times to see differences in papers. Didn't think of using them to test profiles on the same paper.

l'll post my results. But even if they show significant differences, a wide range of images do not.

Digitaldog has some test images in ProPhoto that has colors well outside a printable gamut. The Epson profiles made by Seiko do a pretty bad job of mapping these out of gamut colors to printable ones and you will see really large differences between prints made with the X-rite and Seiko (Epson canned) profiles. X-rite maps these to much more pleasing colors.

 

[digidog]

I'm printing on Epson Hot Press Bright on an Epson 7890 24" printer. I've been using the "canned" Epson paper profile as well as a custom ICC profile that I purchased that is supposed to be "very high quality." The custom profile was made from targets consisting of 2600 total patches and I'm told, "read on X-Rite equipment." So I hope the quality is as advertised.

I'm printing from Mac Photoshop CS6, 16 bit files with 16 bit enabled in the print dialog, Photoshop manages colors, "normal printing," Relative Colorimetric Intent, Black Point Compensation,.

I've examined the two profiles in the Mac Colorsync Utility. The canned Epson space is 10-15% larger than the custom profile. They are exactly the same shape. They touch at the white point, but since the Epson profile is larger, it extends significantly "lower" than the custom one at the dark "bottom" end.

My screen is calibrated and profiled (Mac Cinema display, i1 Pro display profiling hardware/software and 3rd party software).

Prints made from the two profiles are visually almost indistinguishable. Deep blacks are the same despite the very noticeable difference in their 3D representation in ColorSync. Hue and saturation everywhere, in a wide variety of subject matter is the same, except for a barely noticeable pink tint in some neutral grays of some images (so slight that you have to go back and forth between prints to see it) in the canned Epson profiled print. The custom profiled prints in those cases are more true to the screen image.

So why is this happening?

Two sets of profiles from Epson, one made using X-rite's products and color engine, some not.

The CS utility's gamut map isn't an ideal way to get info about the profile’s gamut IMHO, here's why:

ColorWiki - Color Management Myths 26-28

www.colorwiki.com

Then watch this:

Not all ICC profiles are created equally

In this 23 minute video, I'll cover:

The basic anatomy of ICC Profiles

Why there are differences in profile quality and color rendering

How to evaluate an ICC output profile

Examples of good and not so good canned profiles and custom profiles on actual printed output.

High resolution: http://digitaldog.net/files/Not_All_Profiles_are_created_equally.mp4

Low resolution (YouTube):

--
Andrew Rodney
Author: Color Management for Photographers
The Digital Dog

Home

 

[digidog]

Good point about standardized test images. I'll try that. I've used them many times to see differences in papers. Didn't think of using them to test profiles on the same paper.

Two to try (the wide gamut for sure):

http://www.digitaldog.net/files/2014PrinterTestFileFlat.tif.zip

http://www.digitaldog.net/files/Gamut_Test_File_Flat.tif

 

[Petruska]

https://www.dropbox.com/sh/ka35g3gtyd10823/iRGJexS5hA/Twenty-Eight Balls.tif.zip

The file is safe, I just downloaded it, and malware scanned it with 3 different apps.

As Rodney stated these types of reference test photos will show you instantly how the 2 profiles differ.

Bob P.

 

[technoid]

https://www.dropbox.com/sh/ka35g3gtyd10823/iRGJexS5hA/Twenty-Eight Balls.tif.zip

The file is safe, I just downloaded it, and malware scanned it with 3 different apps.

As Rodney stated these types of reference test photos will show you instantly how the 2 profiles differ.

Bob P.

It will show differences in how profile map out of gamut colors but will not show how good a profile is at rendering in gamut colors. For instance the lower right hand blue ball is composed entirely of colors that are not only unprintable, but imaginary in that they do not, and cannot, exist. Some Epson canned profiles render these as black. Profiles that X-rite creates render them with a dark blue gradient. Which is correct? Who knows? They are imaginary colors. However, they are not imaginary when rendered on a monitor where they are converted to the monitor's color space making them "real" but, of course, different. That said, the way they look on a monitor is at least somewhat closer to they way they are printed with an X-rite profile.

But it's complicated. Even actual physical colors that can be rendered on a lowly sRGB display may not be possible to print on a theoretically perfect printer. The set of unprintable colors for a given printer is smaller still. All colors that can't be printed will, by definition, be mapped by the profile and color engine into colors that can be printed. How that is done is completely up to the profile vender and is completely independent of how good or bad the profile is at rendering colors that are in gamut.

And this is where soft proofing becomes critical.

I depend on soft proofing to show, on a display, what the printer will print. It's not perfect either. There are limits. Wide gamut monitors can't typically render printable saturated cyans but they still do a good job on the large majority of printable colors.

So when I look at a profile 99% of my concern is how good the printer is at printing in gamut colors.

For that I prefer to use the old Kodak standby, PDIO_Target-DCP image recognizable by the four faces at the bottom. Even that won't show perturbations in the profile and these can be problematic. Abrupt changes in hue, saturation, or luminosity across a gradient is highly annoying. Luckily, they tend to show up only occasionally where you have smooth color gradients that just happen to be in the areas of high error.

And that's what bugs me and it bothers me much more than color accuracy. A printed color can be off a deltaE of 3 without being very annoying but if it is jumps from being off -2 in one direction to +1 in the opposite and does so abruptly when the requested color has only changed 1 deltaE then it stand out like a sore thumb.

So ideally I would love to see a large set of gradual gradient images that are within gamut so I could spot these before printing large numbers of images then having someone notice that sort of thing when the image happens to bring out the problem.

One possibility is a set of 19, 2D Lab slices of constant L from 5 to 95 in steps of 5 covering only the portion of the print gamut printable at each L.

Does anyone know of any images that try to accomplish this? I haven't run across any to date.

For those interested, here is what Atkinson's 2 lower rows of balls print like on a 9800 with prem. glossy using a canned profile and an X-rite profile. Clearly the X-rite profile produces better looking balls. At least it doesn't render the imaginary blue ball as black.

 

[digidog]

https://www.dropbox.com/sh/ka35g3gtyd10823/iRGJexS5hA/Twenty-Eight Balls.tif.zip

The file is safe, I just downloaded it, and malware scanned it with 3 different apps.

As Rodney stated these types of reference test photos will show you instantly how the 2 profiles differ.

Bob P.

It will show differences in how profile map out of gamut colors but will not show how good a profile is at rendering in gamut colors. For instance the lower right hand blue ball is composed entirely of colors that are not only unprintable, but imaginary in that they do not, and cannot, exist.

Sorry no, the text in bold is incorrect.

All of Bill's synthetic's were built in L*a*b*; nothing imagery in color. And if you examine the values in ProPhoto RGB, in LR where Lab isn't truncated as it is in Photoshop, there are plenty of areas within that blue ball that fits within Lab so no, it isn't entirely colors that are unprintable and imaginary.


Many of the Blue Ball IS Printable. Epson 3880 Luster gamut plot compared to the ball.


Another view showing the ball indeed contains printable blues. The lower right hand blue ball is NOT composed entirely of colors that are unprintable.

Some Epson canned profiles render these as black. Profiles that X-rite creates render them with a dark blue gradient. Which is correct?

That's easy to answer: the profile who's print of that ball isn't black! Just LOOK at the prints. And not just at those blue balls which is telling; other portions of the print.

--
Andrew Rodney
Author: Color Management for Photographers
The Digital Dog
http://www.digitaldog.net

 

[technoid]

https://www.dropbox.com/sh/ka35g3gtyd10823/iRGJexS5hA/Twenty-Eight Balls.tif.zip

The file is safe, I just downloaded it, and malware scanned it with 3 different apps.

As Rodney stated these types of reference test photos will show you instantly how the 2 profiles differ.

Bob P.

It will show differences in how profile map out of gamut colors but will not show how good a profile is at rendering in gamut colors. For instance the lower right hand blue ball is composed entirely of colors that are not only unprintable, but imaginary in that they do not, and cannot, exist.

Sorry no, the text in bold is incorrect.

OK, I confess, you're right. Only 99.95% (on the 8k pixel diameter big file) of the pixels are imaginary.

Every color in the lower right hand blue ball is imaginary with the exception of the 4 pure black pixels in the middle which are RGB (0,0,0) and the 2 pixel, jaggy suppression transition region on the outer edge of the ball.

All of Bill's synthetic's were built in L*a*b*; nothing imagery in color.

What makes you believe that just because a Lab value can be stated that the color exists. That just isn't so.

And if you examine the values in ProPhoto RGB, in LR where Lab isn't truncated as it is in Photoshop, there are plenty of areas within that blue ball that fits within Lab so no, it isn't entirely colors that are unprintable and imaginary.

Truncation has nothing to do with whether the colors are imaginary. The blue ball Lab values that don't exceed the a:b limits of +127/-128 are just as imaginary as those clipped at -128.

Some Epson canned profiles render these as black. Profiles that X-rite creates render them with a dark blue gradient. Which is correct?

That's easy to answer: the profile who's print of that ball isn't black! Just LOOK at the prints. And not just at those blue balls which is telling; other portions of the print.

As I said, when one prints imaginary colors one can't really state with any authority what the printed color should be. The same holds true for displays but at least they have a defined methodology for converting imaginary colors into something displayable. The ICC defines the conversion process for ProPhoto into XYZ PCS and from there into a RGB matrix profile based display.

However, I quite agree that X-rite's profiles print both out of gamut and imaginary colors more pleasingly than Seiko's.

--
Andrew Rodney
Author: Color Management for Photographers
The Digital Dog
http://www.digitaldog.net

 

[digidog]

https://www.dropbox.com/sh/ka35g3gtyd10823/iRGJexS5hA/Twenty-Eight Balls.tif.zip

The file is safe, I just downloaded it, and malware scanned it with 3 different apps.

As Rodney stated these types of reference test photos will show you instantly how the 2 profiles differ.

Bob P.

It will show differences in how profile map out of gamut colors but will not show how good a profile is at rendering in gamut colors. For instance the lower right hand blue ball is composed entirely of colors that are not only unprintable, but imaginary in that they do not, and cannot, exist.

Sorry no, the text in bold is incorrect.

OK, I confess, you're right. Only 99.95% (on the 8k pixel diameter big file) of the pixels are imaginary.

I don't know that stat is correct, how did you come to that conclusion?

Every color in the lower right hand blue ball is imaginary with the exception of the 4 pure black pixels in the middle which are RGB (0,0,0) and the 2 pixel, jaggy suppression transition region on the outer edge of the ball.

How did you come to that conclusion?

All of Bill's synthetic's were built in L*a*b*; nothing imagery in color.

What makes you believe that just because a Lab value can be stated that the color exists. That just isn't so.

Well that's what Lab is based upon. The Lab color space includes all perceivable colors.

Check out Bruce Linndbloom's site and the gamut efficiency of Lab at 97%:

Welcome to Bruce Lindbloom's Web Site

www.brucelindbloom.com

The Lab Gamut Efficiency % indicates the percent of the entire Lab Gamut (i.e. all colors visible to the eye) that the working space encompasses.

And if you examine the values in ProPhoto RGB, in LR where Lab isn't truncated as it is in Photoshop, there are plenty of areas within that blue ball that fits within Lab so no, it isn't entirely colors that are unprintable and imaginary.

Truncation has nothing to do with whether the colors are imaginary. The blue ball Lab values that don't exceed the a:b limits of +127/-128 are just as imaginary as those clipped at -128.

What did you use to come up with those values? And did you examine Bill's original Lab created image?

Some Epson canned profiles render these as black. Profiles that X-rite creates render them with a dark blue gradient. Which is correct?

That's easy to answer: the profile who's print of that ball isn't black! Just LOOK at the prints. And not just at those blue balls which is telling; other portions of the print.

As I said, when one prints imaginary colors one can't really state with any authority what the printed color should be.

A blue ball shouldn’t print black; simple as that.

We can go into accuracy and dE metric's but of the two profiles, just looking at what comes off the printer, which would you guess is far, far less accurate? And does it matter? Who wants a saturated blue to map to black? Not I.

 

[Gesture]

"However, I quite agree that X-rite's profiles print both out of gamut and imaginary colors more pleasingly than Seiko's."

Are the "pro" profiles that Epson sometimes offers (and require special download) any better?

 

[technoid]

"However, I quite agree that X-rite's profiles print both out of gamut and imaginary colors more pleasingly than Seiko's."

Are the "pro" profiles that Epson sometimes offers (and require special download) any better?

Good question. Since I found out that Epson's canned profile for my 9800 incorrectly applied black point correction to the Relative Intent I made my own with X-rite. They work quite well and account for variations in my printer. Also, it makes it easy to use other papers.

I believe Digitaldog has indicated Epson has some profiles that work much better and are made with X-rite tools. He might chime in.

 

[digidog]

"However, I quite agree that X-rite's profiles print both out of gamut and imaginary colors more pleasingly than Seiko's."

Are the "pro" profiles that Epson sometimes offers (and require special download) any better?

Watch my video referenced earlier, it gives you the special decoder ring where you can figure out if the profiles are X-rite's color engine or not.

 

[digidog]

https://www.dropbox.com/sh/ka35g3gtyd10823/iRGJexS5hA/Twenty-Eight Balls.tif.zip

The file is safe, I just downloaded it, and malware scanned it with 3 different apps.

As Rodney stated these types of reference test photos will show you instantly how the 2 profiles differ.

Bob P.

It will show differences in how profile map out of gamut colors but will not show how good a profile is at rendering in gamut colors. For instance the lower right hand blue ball is composed entirely of colors that are not only unprintable, but imaginary in that they do not, and cannot, exist.

Sorry no, the text in bold is incorrect.

OK, I confess, you're right. Only 99.95% (on the 8k pixel diameter big file) of the pixels are imaginary.

Let's use some facts based on colorimetry. Here's a plot of Bill's Blue ball that printed black that you state is 99.95% composed entirely of colors that are not only unprintable, but imaginary in that they do not, and cannot, exist. Of course the first bit about unprintable is wrong, the gamut map provided earlier shows while many are, many are not. They fall within the printer's gamut.

Now let's examine the same blue colors plotted compared to the spectrum locus and ProPhoto RGB:


Wire Frame is ProPHoto RGB. Blue dots are of course Bill's image.

It's quite clear that ProPhoto RGB's gamut, as both of us know, define device values that are 'imagery', not colors. You can see in this plot it falls outside Spectrum Locus.

Another angle where again, quite clear that ProPhoto RGB falls outside the Spectrum Locus but the blue of BIll's image? None that I see:


ProPhoto RGB wire frame gamut falls outside the Spectrum Locus. Bill's blue ball?

Now examine the blue. 99.95% of the pixels are imaginary?

Again, where did you come up with such a metric? Can you explain the above?

--
Andrew Rodney
Author: Color Management for Photographers
The Digital Dog
http://www.digitaldog.net