Why do prints from different profiles look alike?

digidog · Sep 14, 2016

No Lab version. Never was. Here's what Bill wrote which is fascinating:

Hi Andrew,

My 28 Balls test image was created in 16 bit RGB using a program that I wrote. It is meant to be printed without applying a profile, in order to view the underlying raw mapping of a given printer and driver settings. If the underlying driver settings create huge discontinuities when fed smoothly varying RGB values, then any profile built on top of that driver setting will have a difficult time smoothing out these driver anomalies.

The fact that it is tagged with ProPhoto RGB is not supposed to matter at all. It could just as well be be tagged with sRGB, because it is really an untagged RGB test file. It should be printed without a profile, just sent directly to the printer with No Color Adjust or some such equivalent settings, in order to examine the underlying raw RGB interface to the printer. Try printing this test image with different driver settings, and you can quickly eliminate those settings that result in big sudden discontinuities, or those that have gross anomalies near the R=G=B plane.

The reason I use balls instead of linear gradients, is because I discovered that our eyes are more sensitive to anomalies in this presentation.

I hope this clears up my intention behind the 28 Balls test image. Please post or pass this info on to anyone who might benefit from it.

I intended this image to be used without a profile, and the answer about imaginary colors would completely depend on what RGB source profile you tagged it with. All RGB numeric values do fall in the range of 0.0 to full-scale, so if you tagged the image with sRGB, then all colors on all balls would of course be within the gamut of sRGB. You can open the file in Photoshop and assign any RGB source profile to it, then look at the resulting soft-proof for a given printer profile.

Bill

technoid · Sep 14, 2016

digidog said:
No Lab version. Never was. Here's what Bill wrote which is fascinating:

Wow! That does explain a lot of things.

Member said:
Hi Andrew,

My 28 Balls test image was created in 16 bit RGB using a program that I wrote. It is meant to be printed without applying a profile, in order to view the underlying raw mapping of a given printer and driver settings. If the underlying driver settings create huge discontinuities when fed smoothly varying RGB values, then any profile built on top of that driver setting will have a difficult time smoothing out these driver anomalies.

The fact that it is tagged with ProPhoto RGB is not supposed to matter at all. It could just as well be be tagged with sRGB, because it is really an untagged RGB test file. It should be printed without a profile, just sent directly to the printer with No Color Adjust or some such equivalent settings, in order to examine the underlying raw RGB interface to the printer. Try printing this test image with different driver settings, and you can quickly eliminate those settings that result in big sudden discontinuities, or those that have gross anomalies near the R=G=B plane.

The reason I use balls instead of linear gradients, is because I discovered that our eyes are more sensitive to anomalies in this presentation.

That's a great observation of Bill's. Makes a great deal of sense.

Member said:
I hope this clears up my intention behind the 28 Balls test image. Please post or pass this info on to anyone who might benefit from it.

I intended this image to be used without a profile, and the answer about imaginary colors would completely depend on what RGB source profile you tagged it with. All RGB numeric values do fall in the range of 0.0 to full-scale, so if you tagged the image with sRGB, then all colors on all balls would of course be within the gamut of sRGB. You can open the file in Photoshop and assign any RGB source profile to it, then look at the resulting soft-proof for a given printer profile.

Bill

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

Thank you Andrew for your patience and persistence. And please thank Bill for his clarifying comments. This was, indeed, a valuable conversation.

digidog · Sep 14, 2016

technoid said:
digidog said:

No Lab version. Never was. Here's what Bill wrote which is fascinating:

Click to expand...

Wow! That does explain a lot of things.

technoid said:

Hi Andrew,

My 28 Balls test image was created in 16 bit RGB using a program that I wrote. It is meant to be printed without applying a profile, in order to view the underlying raw mapping of a given printer and driver settings. If the underlying driver settings create huge discontinuities when fed smoothly varying RGB values, then any profile built on top of that driver setting will have a difficult time smoothing out these driver anomalies.

The fact that it is tagged with ProPhoto RGB is not supposed to matter at all. It could just as well be be tagged with sRGB, because it is really an untagged RGB test file. It should be printed without a profile, just sent directly to the printer with No Color Adjust or some such equivalent settings, in order to examine the underlying raw RGB interface to the printer. Try printing this test image with different driver settings, and you can quickly eliminate those settings that result in big sudden discontinuities, or those that have gross anomalies near the R=G=B plane.

The reason I use balls instead of linear gradients, is because I discovered that our eyes are more sensitive to anomalies in this presentation.

Click to expand...

That's a great observation of Bill's. Makes a great deal of sense.

technoid said:

I hope this clears up my intention behind the 28 Balls test image. Please post or pass this info on to anyone who might benefit from it.

I intended this image to be used without a profile, and the answer about imaginary colors would completely depend on what RGB source profile you tagged it with. All RGB numeric values do fall in the range of 0.0 to full-scale, so if you tagged the image with sRGB, then all colors on all balls would of course be within the gamut of sRGB. You can open the file in Photoshop and assign any RGB source profile to it, then look at the resulting soft-proof for a given printer profile.

Bill

Click to expand...

digidog · Sep 14, 2016

Member said:
digidog wrote: So considering Bill's intent, considering my (other's?) assumption he built it to be used in ProPhoto RGB, and we can tag it with anything, is there any merit in using it as it is? Or assign another working space? Or just not use it for testing output with an output profile?

As I mentioned to Bill, the one blue ball was interesting when passed through two profiles. One printed nearly toally black, the other looked visually decent. So that suggests this is somewhat useful in seeing what the color engine is doing but then why just one ball that undergoes that condition.

Chris's reply to the question:

I think it depends on the workflow you're testing. If the workflow
calls for mainly ProPhoto RGB images as sources, then use 28 Balls
with ProPhoto. If the workflow calls for something else, use something
else.

While the print it as-is test is what it's designed for, that is still
a conversion test. It's just a conversion within the printer's black
box, one you have non-ICC levers for control: paper type, density
slider (oh god is that f*&*ed up) and so on. That's just as
interesting to punish test as is the ICC conversion. The better the
isolation the better idea you have what's causing discontinuities; and
it's likely that each additional conversion causes the discontinuities
to get worse. Anyway I think it's a good test as long as you don't
overestimate it. Radial blends are hard to render on printers, let
alone through multiple conversions. So it's easy to go wow that looks
like s&*t and wonder if it's time to retreat, when in fact it'd only
rarely be a problem with real images.

All my data ends up in ProPhoto RGB from raw. Of course, nothing I shoot would be anything like that ball among other synthetics (Granger Rainbow comes to mind).

And the ball (among other elements) did show the qualities, or lack thereof, from the two supplied Epson profiles. Going back days to the original question about each; which is correct. The one that prints black isn't as correct, it's more incorrect than the X-rite profile. So I guess, the balls are useful....

technoid · Sep 14, 2016

digidog said:
Member said:

digidog wrote: So considering Bill's intent, considering my (other's?) assumption he built it to be used in ProPhoto RGB, and we can tag it with anything, is there any merit in using it as it is? Or assign another working space? Or just not use it for testing output with an output profile?

As I mentioned to Bill, the one blue ball was interesting when passed through two profiles. One printed nearly toally black, the other looked visually decent. So that suggests this is somewhat useful in seeing what the color engine is doing but then why just one ball that undergoes that condition.

Click to expand...

Chris's reply to the question:

I think it depends on the workflow you're testing. If the workflow
calls for mainly ProPhoto RGB images as sources, then use 28 Balls
with ProPhoto. If the workflow calls for something else, use something
else.

While the print it as-is test is what it's designed for, that is still
a conversion test. It's just a conversion within the printer's black
box, one you have non-ICC levers for control: paper type, density
slider (oh god is that f*&*ed up) and so on. That's just as
interesting to punish test as is the ICC conversion. The better the
isolation the better idea you have what's causing discontinuities; and
it's likely that each additional conversion causes the discontinuities
to get worse. Anyway I think it's a good test as long as you don't
overestimate it. Radial blends are hard to render on printers, let
alone through multiple conversions. So it's easy to go wow that looks
like s&*t and wonder if it's time to retreat, when in fact it'd only
rarely be a problem with real images.

All my data ends up in ProPhoto RGB from raw. Of course, nothing I shoot would be anything like that ball among other synthetics (Granger Rainbow comes to mind).

And the ball (among other elements) did show the qualities, or lack thereof, from the two supplied Epson profiles. Going back days to the original question about each; which is correct. The one that prints black isn't as correct, it's more incorrect than the X-rite profile. So I guess, the balls are useful....

I think they are somewhat useful for people editing in ProPhoto that aren't being very careful with their colors. It's useful if the profiling rendering isn't totally at odds with what people see on their presumably wide gamut screens and the canned ones were pretty bad. The blue balls may have imaginary colors in ProPhoto but the way they are rendered to a display is well defined and those colors are NOT imaginary. One would want the colors from the profile to bear some resemblance to what is seen on the monitor.

Also, now that I understand Bill's intent, I may try using them in device space. I understand his reasoning and it makes a lot of sense. I also know what to look for in that mode and what differences aren't relevant. His observation that round gradients would show transition anomalies better than the standard linear ones makes tons of sense but it had never occurred to me. Very clever of Bill.

My concern with the Epson canned 9800 profiles is that their rendering in Absolute Colorimetric is seriously flawed. It's a side effect of implementing BPC in their RI intent as the tables are also used for AI. So, for critical repro work I just do not use them. But I do not have to as I have I1Profiler and PM5 which generated profiles properly and I augment it with extra neutral patches as people are more sensitive to small errors in those ranges.

For people that do not have profiling capability in house there are custom profiles made by people that know what they are doing (yes you, amongst others). I would particularly recommend that for people with these problematic Epson profiles.

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

digidog · Sep 14, 2016

technoid said:
digidog said:

technoid said:

digidog wrote: So considering Bill's intent, considering my (other's?) assumption he built it to be used in ProPhoto RGB, and we can tag it with anything, is there any merit in using it as it is? Or assign another working space? Or just not use it for testing output with an output profile?

As I mentioned to Bill, the one blue ball was interesting when passed through two profiles. One printed nearly toally black, the other looked visually decent. So that suggests this is somewhat useful in seeing what the color engine is doing but then why just one ball that undergoes that condition.

Click to expand...

Chris's reply to the question:

I think it depends on the workflow you're testing. If the workflow
calls for mainly ProPhoto RGB images as sources, then use 28 Balls
with ProPhoto. If the workflow calls for something else, use something
else.

While the print it as-is test is what it's designed for, that is still
a conversion test. It's just a conversion within the printer's black
box, one you have non-ICC levers for control: paper type, density
slider (oh god is that f*&*ed up) and so on. That's just as
interesting to punish test as is the ICC conversion. The better the
isolation the better idea you have what's causing discontinuities; and
it's likely that each additional conversion causes the discontinuities
to get worse. Anyway I think it's a good test as long as you don't
overestimate it. Radial blends are hard to render on printers, let
alone through multiple conversions. So it's easy to go wow that looks
like s&*t and wonder if it's time to retreat, when in fact it'd only
rarely be a problem with real images.

All my data ends up in ProPhoto RGB from raw. Of course, nothing I shoot would be anything like that ball among other synthetics (Granger Rainbow comes to mind).

And the ball (among other elements) did show the qualities, or lack thereof, from the two supplied Epson profiles. Going back days to the original question about each; which is correct. The one that prints black isn't as correct, it's more incorrect than the X-rite profile. So I guess, the balls are useful....

Click to expand...

I think they are somewhat useful for people editing in ProPhoto that aren't being very careful with their colors. It's useful if the profiling rendering isn't totally at odds with what people see on their presumably wide gamut screens and the canned ones were pretty bad.

Of course. But we have to make a decision: edit colors we can capture and can output that fall outside display gamut (the weak link), soft proof and attempt to adjust such that the final print isn't a shocking surprise.

In my mind, this is no different from my days shooting film professionally, where I shot Polaroids. That damn 35MM back cost as much as the back for my Hasselblad! The Polaroid didn't match the chrome by a long shot. But I learned to interpret what it showed me such that, viewing a transparency on a viewing box wasn't a surprise.

I see this being similar with wide gamut spaces. I've got a wide gamut display. Problem is, my output profiles are far wider in lots of areas than the display. I don't want to clip those colors, when present in the image, that I hope to print.

technoid said:
The blue balls may have imaginary colors in ProPhoto but the way they are rendered to a display is well defined and those colors are NOT imaginary. One would want the colors from the profile to bear some resemblance to what is seen on the monitor.

The idea behind my Wide Gamut Test file, and I'd suggest Bills images, synthetic or otherwise, is as a color reference image. How does it print? What do the differing rendering intents bring to the table? Will saturated blue balls print black or something a lot closer than I expect; blue! We could use them as an aid for soft proofing but we end up with the gamut disconnect again. Or we can tag it in Adobe RGB (1998) for that use.

technoid said:
Also, now that I understand Bill's intent, I may try using them in device space. I understand his reasoning and it makes a lot of sense.

Absolutely. I feel stupid that I didn't know, all these years later, how and why they were designed and that the ProPhoto RGB tag wasn't necessarily intentional. As Bill said, we could tag it with any working space profile for it's designed use. Now the question becomes, can we us it within a color reference print to tell us something useful about the output profiles. I still believe so but I'm open to peer review on that concept.

technoid said:
I also know what to look for in that mode and what differences aren't relevant. His observation that round gradients would show transition anomalies better than the standard linear ones makes tons of sense but it had never occurred to me. Very clever of Bill.

Bill is (and I don't use this word often or loosely) pretty much a genius, #13 employee at Apple!

I recall last century, having lunch with both Bill and Thomas Knoll. There was no question who was the dumbest person at that table! And who got lost in translation in about 2 minutes.

technoid said:
My concern with the Epson canned 9800 profiles is that their rendering in Absolute Colorimetric is seriously flawed.

Sure. That's good to know.

technoid said:
It's a side effect of implementing BPC in their RI intent as the tables are also used for AI. So, for critical repro work I just do not use them.

And you should not and that's a very challenging and specific workflow.

technoid said:
But I do not have to as I have I1Profiler and PM5 which generated profiles properly and I augment it with extra neutral patches as people are more sensitive to small errors in those ranges.

Agreed.

technoid said:
For people that do not have profiling capability in house there are custom profiles made by people that know what they are doing (yes you, amongst others). I would particularly recommend that for people with these problematic Epson profiles.

Well it wasn't that long ago, I had no idea Epson supplied two different profiles from different engines. Only testing Premium Luster and then Exhibition Fiber, using a color reference file with Bill's Balls did it hit me over the head. So all I can say is, it's all good (expect the Seiko profiles, they are pretty bad). ;-)

technoid · Sep 14, 2016

digidog said:
technoid said:

digidog said:

Member said:

digidog wrote: So considering Bill's intent, considering my (other's?) assumption he built it to be used in ProPhoto RGB, and we can tag it with anything, is there any merit in using it as it is? Or assign another working space? Or just not use it for testing output with an output profile?

As I mentioned to Bill, the one blue ball was interesting when passed through two profiles. One printed nearly toally black, the other looked visually decent. So that suggests this is somewhat useful in seeing what the color engine is doing but then why just one ball that undergoes that condition.

Click to expand...

Chris's reply to the question:

I think it depends on the workflow you're testing. If the workflow
calls for mainly ProPhoto RGB images as sources, then use 28 Balls
with ProPhoto. If the workflow calls for something else, use something
else.

While the print it as-is test is what it's designed for, that is still
a conversion test. It's just a conversion within the printer's black
box, one you have non-ICC levers for control: paper type, density
slider (oh god is that f*&*ed up) and so on. That's just as
interesting to punish test as is the ICC conversion. The better the
isolation the better idea you have what's causing discontinuities; and
it's likely that each additional conversion causes the discontinuities
to get worse. Anyway I think it's a good test as long as you don't
overestimate it. Radial blends are hard to render on printers, let
alone through multiple conversions. So it's easy to go wow that looks
like s&*t and wonder if it's time to retreat, when in fact it'd only
rarely be a problem with real images.

All my data ends up in ProPhoto RGB from raw. Of course, nothing I shoot would be anything like that ball among other synthetics (Granger Rainbow comes to mind).

And the ball (among other elements) did show the qualities, or lack thereof, from the two supplied Epson profiles. Going back days to the original question about each; which is correct. The one that prints black isn't as correct, it's more incorrect than the X-rite profile. So I guess, the balls are useful....

Click to expand...

I think they are somewhat useful for people editing in ProPhoto that aren't being very careful with their colors. It's useful if the profiling rendering isn't totally at odds with what people see on their presumably wide gamut screens and the canned ones were pretty bad.

Click to expand...

Of course. But we have to make a decision: edit colors we can capture and can output that fall outside display gamut (the weak link), soft proof and attempt to adjust such that the final print isn't a shocking surprise.

In my mind, this is no different from my days shooting film professionally, where I shot Polaroids. That damn 35MM back cost as much as the back for my Hasselblad! The Polaroid didn't match the chrome by a long shot. But I learned to interpret what it showed me such that, viewing a transparency on a viewing box wasn't a surprise.

I see this being similar with wide gamut spaces. I've got a wide gamut display. Problem is, my output profiles are far wider in lots of areas than the display. I don't want to clip those colors, when present in the image, that I hope to print.

Yep. It's annoying that even a good wide gamut monitor can't always render soft proofs accurately and it comes up from time to time mostly in the more saturated cyans.

Member said:
Member said:

The blue balls may have imaginary colors in ProPhoto but the way they are rendered to a display is well defined and those colors are NOT imaginary. One would want the colors from the profile to bear some resemblance to what is seen on the monitor.

Click to expand...

The idea behind my Wide Gamut Test file, and I'd suggest Bills images, synthetic or otherwise, is as a color reference image. How does it print? What do the differing rendering intents bring to the table? Will saturated blue balls print black or something a lot closer than I expect; blue! We could use them as an aid for soft proofing but we end up with the gamut disconnect again. Or we can tag it in Adobe RGB (1998) for that use.

Member said:

Also, now that I understand Bill's intent, I may try using them in device space. I understand his reasoning and it makes a lot of sense.

Click to expand...

Absolutely. I feel stupid that I didn't know, all these years later, how and why they were designed and that the ProPhoto RGB tag wasn't necessarily intentional. As Bill said, we could tag it with any working space profile for it's designed use. Now the question becomes, can we us it within a color reference print to tell us something useful about the output profiles. I still believe so but I'm open to peer review on that concept.

Member said:

I also know what to look for in that mode and what differences aren't relevant. His observation that round gradients would show transition anomalies better than the standard linear ones makes tons of sense but it had never occurred to me. Very clever of Bill.

Click to expand...

Bill is (and I don't use this word often or loosely) pretty much a genius, #13 employee at Apple!

I didn't know that about Bill. Very cool.

Member said:
I recall last century, having lunch with both Bill and Thomas Knoll. There was no question who was the dumbest person at that table! And who got lost in translation in about 2 minutes.

Member said:

My concern with the Epson canned 9800 profiles is that their rendering in Absolute Colorimetric is seriously flawed.

Click to expand...

Sure. That's good to know.

Member said:

It's a side effect of implementing BPC in their RI intent as the tables are also used for AI. So, for critical repro work I just do not use them.

Click to expand...

And you should not and that's a very challenging and specific workflow.

Member said:

But I do not have to as I have I1Profiler and PM5 which generated profiles properly and I augment it with extra neutral patches as people are more sensitive to small errors in those ranges.

Click to expand...

Agreed.

Member said:

For people that do not have profiling capability in house there are custom profiles made by people that know what they are doing (yes you, amongst others). I would particularly recommend that for people with these problematic Epson profiles.

Click to expand...

Well it wasn't that long ago, I had no idea Epson supplied two different profiles from different engines. Only testing Premium Luster and then Exhibition Fiber, using a color reference file with Bill's Balls did it hit me over the head. So all I can say is, it's all good (expect the Seiko profiles, they are pretty bad). ;-)

I can only imagine what a shock that black, blue ball print was! Funny thing is that the part of the profile that maps the blue ball black is a very different section of the 3D Luts than the part that puts BPC into the AtoB1/BtoA1 tables. That was the killer for me. Didn't know about the strange blue ball mapping until I saw your video on it much later.

Anyway, thanks for what turned out to be a delightful and educational discussion. Again, convey my appreciation for Bill's contribution too.

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

digidog · Sep 15, 2016

technoid said:
I didn't know that about Bill. Very cool.

Like I said, Genius. And a super nice fellow too!

From Wikipedia, the free encyclopedia

Bill Atkinson (born 1951) is an American computer engineer and photographer. Atkinson worked at Apple Computer from 1978 to 1990.

Atkinson was the principal designer and developer of the GUI of the Apple Lisa and, later, one of the first thirty members of the original Apple Macintosh development team, and was the creator of the ground-breaking MacPaint application, which fulfilled the vision of using the computer as a creative tool. He also designed and implemented QuickDraw, the fundamental toolbox that the Lisa and Macintosh used for graphics. QuickDraw's performance was essential for the success of the Macintosh's graphical user interface. He also was one of the main designers of the Lisa and Macintosh user interfaces. Atkinson also conceived, designed and implemented HyperCard, the first popular hypermedia system. HyperCard put the power of computer programming and database design into the hands of non-programmers. In 1994, Atkinson received the EFF Pioneer Award for his contributions.

He received his undergraduate degree from the University of California, San Diego, where Apple Macintosh developer Jef Raskin was one of his professors. Atkinson continued his studies as a graduate student in neurochemistry at the University of Washington, nearly getting a PhD.

Around 1990, General Magic's founding, with Bill Atkinson as one of the three co-founders, met the following press in Byte magazine:

The obstacles to General Magic's success may appear daunting, but General Magic is not your typical start-up company. Its partners include some of the biggest players in the worlds of computing, communications, and consumer electronics, and it's loaded with top-notch engineers who have been given a clean slate to reinvent traditional approaches to ubiquitous worldwide communications.

In 2007 Atkinson began working as an outside developer with Numenta, a startup working on computer intelligence. On his work there Atkinson said, "what Numenta is doing is more fundamentally important to society than the personal computer and the rise of the Internet."

Currently, Atkinson has combined his passion for computer programming with his love of nature photography to create art images. He takes close-up photographs of stones that have been cut and polished. His works are highly regarded for their resemblance of miniature landscapes which are hidden within the stones. Atkinson’s book “Within the Stone” features a collection of his close-up photographs and was published in 2004. The highly intricate and detailed images he creates are made possible by the accuracy and creative control of the digital printing process that he helped create.

Some of Atkinson's noteworthy contributions to the field of computing include:

Macintosh QuickDraw and Lisa LisaGraf
- Atkinson independently discovered the Midpoint circle algorithm for fast drawing of circles by using the sum of consecutive odd numbers. "
Marching ants
The double-click
Menu bar
The Selection lasso
FatBits
MacPaint
HyperCard
Atkinson dithering
Bill Atkinson PhotoCard

Atkinson now works as a nature photographer. Actor Nelson Franklin portrayed him in the 2013 film, Jobs.

To this day, I'm still a fan of HyperCard!

I've got a signed copy of his book, Within the Stone and it's amazing in terms of the color reproduction. He covers the ground breaking color management work he did on his site.

Loading…

www.billatkinson.com

Rich42 · Sep 15, 2016

OK guys,

Can someone now comment on what the Printer Gamut Test file shows (just the colored balls clip) about my Epson canned profile and custom profile?

digidog · Sep 15, 2016

Rich42 said:
OK guys,

Can someone now comment on what the Printer Gamut Test file shows (just the colored balls clip) about my Epson canned profile and custom profile?

IMHO, it still tells you a great deal! In the case of the one blue ball, whether a gamut mapper like CTP shows some of it in printer gamut or not, if you get a black ball (which so far, I've only seen with the Seiko profiles), not good. The X-rite engine and Copra which I also use, map these colors far, far better. The other balls are still useful to see the gradients smoothness, color shifting etc. Do not use JUST one ball, or all the balls alone to make a judgment of your profile! There are real life, actual photography, from raw, encoded into ProPhoto RGB from a raw converter who's processing color space gamut is ProPhoto RGB.

Talk the test file and convert it to Adobe RGB or, dog forbid, sRGB and make a print with the same profile. That tells you a lot as well! Mostly, sRGB is suboptimal for output to a printer.

Rich42 · Sep 15, 2016

I posted this clip from your file way back in this thread.

And these comments:

"The TDI file and all the photographic images in Andrew's file look practically identical with both profiles. The Epson canned profile gives the slight magenta cast as I've described above, but it's only noticeable by comparing the two closely. Either print alone appears to have pleasing and very accurate colors. But the custom profile (X-Rite) is more accurate.

The colored balls in the two are a different matter. Here's a scan of the two. The custom profile at the top. The scan doesn't really show the colors I see here, but it's the best I can do with the scanner I have.

The most noticeable differences are in the blue, green, magenta and cyan balls. Banding in much more apparent in the custom scan. Especially the green, less so in the cyan balls. The green ball has more yellow in the custom, the cyan has more yellow in the Epson. The magenta balls are a dusky rose in the Epson, truer magenta in the custom. All the balls look cleaner and slightly brighter in the custom. The same is true in the grayscale ramps, and the color display to the left of the man in the green boat."

The Custom profile is on top, Epson canned on the bottom.

Any opinion as to what this shows about these two profiles?

digidog · Sep 15, 2016

Rich42 said:
I posted this clip from your file way back in this thread.

And these comments:

"The TDI file and all the photographic images in Andrew's file look practically identical with both profiles. The Epson canned profile gives the slight magenta cast as I've described above, but it's only noticeable by comparing the two closely. Either print alone appears to have pleasing and very accurate colors. But the custom profile (X-Rite) is more accurate.

The colored balls in the two are a different matter. Here's a scan of the two. The custom profile at the top. The scan doesn't really show the colors I see here, but it's the best I can do with the scanner I have.

The most noticeable differences are in the blue, green, magenta and cyan balls. Banding in much more apparent in the custom scan. Especially the green, less so in the cyan balls. The green ball has more yellow in the custom, the cyan has more yellow in the Epson. The magenta balls are a dusky rose in the Epson, truer magenta in the custom. All the balls look cleaner and slightly brighter in the custom. The same is true in the grayscale ramps, and the color display to the left of the man in the green boat."

The Custom profile is on top, Epson canned on the bottom.

Any opinion as to what this shows about these two profiles?

FWIW, I suspect shooting it raw would be a tad better than scanning but let's work with what we have, with the caveat, what we're talking about may be a mile from the actual print.

Keep in mind, there's always a compromise between what a profile does facing saturation/hue and sometimes brightness along with smoothness. The people writing the software have to weigh each and decide what to do. The output profile doesn't know anything about the source color space/gamut. By the time it gets the data, it's in Lab, from the PCS. That's what the PRMG in v4 profiles was supposed to address yet I don't know any product today that supports it. No reason to go there.

IF I'm to view the above two samples and consider them the actual print, we can see some advantages and disadvantages of each. First and most critical to me would be neutrality! The top wins by a mile. The Cyan ball at top isn't as smooth but the color appears 'cleaner' to me. The same seems true for Magenta. In fact, color wise, the top wins in color and loses in smoothness in a general sense. Examine the Granger Rainbow in the same test file, I suspect you'll see similar results.

So is having more saturation but a hue shift 'better'? Is having a purer hue with less smoothness better? Again, a compromise. Blue that maps black, not so good. Grays with a color cast (which is what I happen to see above but viewing this on the web is not at all optimal) isn't something I'd want. The top 'wins' IMHO.

The balls are a torture test!

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

Home

Charles2 · Sep 15, 2016

Rich42 said:
Can someone now comment on what the Printer Gamut Test file shows (just the colored balls clip) about my Epson canned profile and custom profile?

It tells you how the profiles print artificially constructed device values ( which were confused with a color space, two of them actually, until Bill A. explained what he did and why).

Some real photos depend especially on a smooth transition of color for their aesthetic impact. Some real photos depend on accurate representation of the relationship of two or more colors for their aesthetic impact. Printing a sample real photo of one and the other type tells you more about the usefulness of the two profiles than digressing through blue balls territory. At least, no evidence has come forth to conclude otherwise.

As you reported in your original post: "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."

Rich42 · Sep 15, 2016

Andrew and Charles2,

All very interesting. I appreciate the feedback. I'll continue to use the custom profile in preference to Epson's, and of course get more as my machine continues to age.

Other factors must also be my current ink set and paper batch(es). Although they are all Epson, and I imagine they have good quality control, a custom profile catches every factor, not just my machine's "drifting" from factory specs. I'm now wondering how much of the current "magenta tint" is the machine and how much the inks?

I've certainly learned a lot about the canned Epson profiles and I'm glad I ordered a custom file.

Still not yet understanding most of the back and forth between Andrew and Technoid in this thread. But that's ok.

technoid · Sep 16, 2016

Andrew,

Did you ever find out why the blue ball colors are showing as in-gamut with CTP?

I have CTP but haven't used it in ages preferring Matlab for day to day work But if you describe your process with the blue ball colors to show it as in-gamut I can probably replicate it. Hopefully figure out what is going on.

digidog · Sep 17, 2016

technoid said:
Andrew,

Did you ever find out why the blue ball colors are showing as in-gamut with CTP?

I have CTP but haven't used it in ages preferring Matlab for day to day work But if you describe your process with the blue ball colors to show it as in-gamut I can probably replicate it. Hopefully figure out what is going on.

He's what I did.

Take my Gamut Test File. Crop the one blue ball as close as you can to the ball.

Save a TIFF.

Open, or drag and drop over CTP. It will open in the product.

Create a worksheet by selecting Extract Unique Device Values from the dropdown. That can take some time.

You'll end up with a Worksheet of RGB (ProPhoto) and Lab values there. You can save this to disk as a txt file if you wish.

Select Graph List from the dropdown menu of this color list. Then select an ICC output profile to plot along with it. I used a custom Epson 3880/Luster profile I built in i1Profiler.

Set the output profile to be a wireframe which makes it easier to see colors plotted inside the gamut of the printer:

Some of the color list is the gray around the ball and that's easy to see. But some of the blue plots inside the gamut of the printer. Yes, a large amount falls outside printer gamut. But CTP shows blues that are inside it's gamut.

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

technoid · Sep 17, 2016

Thanks for the thorough process description. It will take a while to chunk through the image data, analyze it, and determine if or why it is differing from the results with Matlab.

Iron Mike · Sep 17, 2016

(Sorry for injecting myself into your conversation)

Perhaps there is a faulty assumption that each point on the plot represents 1 pixel in the blue-ball image, however if you look at the message you quoted from Steve Upton, he makes the following points:

- unique color values in ColorThink refer to the number of distinct colors (as per human vision) that exist in a color gamut or a color image or color list or whatever.

- in color images/lists we roughly round color values when counting unique colors to make two very similar colors (less than 1 dE) become the *same* color before counting the number of unique colors.

- the idea behind these calculations was to provide rough approximations to allow for the comparison of color gamuts, image conversions, etc as well as to reduce the number of duplicate (or near-duplicate) color points when graphing image colors in 3D.

(emphasis mine)

It's possible, even likely given all the evidence together, that the parts of the image that lie outside the gamut of the printer on the CTP plot are comprised of 99.5% of the blue ball image, but it's not readily apparent due to the data reduction. The parts that are in gamut are the grey surround, and the edge pixels where grey transitions to blue. We can't assume that the 10-20% (or whatever) of the blue line that's in gamut corresponds to 10-20% of the image pixels.

Thoughts?

Mike

technoid · Sep 17, 2016

Iron Mike said:
(Sorry for injecting myself into your conversation)

Perhaps there is a faulty assumption that each point on the plot represents 1 pixel in the blue-ball image, however if you look at the message you quoted from Steve Upton, he makes the following points:

- unique color values in ColorThink refer to the number of distinct colors (as per human vision) that exist in a color gamut or a color image or color list or whatever.

- in color images/lists we roughly round color values when counting unique colors to make two very similar colors (less than 1 dE) become the *same* color before counting the number of unique colors.

I'll bet that's it. The colors going from the center along any line to the circle boundary would all match, aside from small rounding errors so they would be aggregated. The pixels along the circumference are desaturated to varying degrees towards a neutral gray. There are probably more unique, desaturated pixels than the unique, blue ones.

Iron Mike said:
- the idea behind these calculations was to provide rough approximations to allow for the comparison of color gamuts, image conversions, etc as well as to reduce the number of duplicate (or near-duplicate) color points when graphing image colors in 3D.

(emphasis mine)

It's possible, even likely given all the evidence together, that the parts of the image that lie outside the gamut of the printer on the CTP plot are comprised of 99.5% of the blue ball image, but it's not readily apparent due to the data reduction.

The 99.5% relates to the percentage outside the CIE xy gamut, printer gamuts are all much smaller.

Iron Mike said:
The parts that are in gamut are the grey surround, and the edge pixels where grey transitions to blue. We can't assume that the 10-20% (or whatever) of the blue line that's in gamut corresponds to 10-20% of the image pixels.

Thoughts?

Mike

digidog · Sep 17, 2016

Iron Mike said:
(Sorry for injecting myself into your conversation)

Perhaps there is a faulty assumption that each point on the plot represents 1 pixel in the blue-ball image, however if you look at the message you quoted from Steve Upton, he makes the following points:

- unique color values in ColorThink refer to the number of distinct colors (as per human vision) that exist in a color gamut or a color image or color list or whatever.

- in color images/lists we roughly round color values when counting unique colors to make two very similar colors (less than 1 dE) become the *same* color before counting the number of unique colors.

- the idea behind these calculations was to provide rough approximations to allow for the comparison of color gamuts, image conversions, etc as well as to reduce the number of duplicate (or near-duplicate) color points when graphing image colors in 3D.

(emphasis mine)

I've tried all three options and the differences are not such that no blue colors plot inside the printer gamut.

Also, in terms of calculated lab values compared to the printer gamut, this data might be important (I provided this URL earlier):

ColorWiki - Color Management Myths 26-28

www.colorwiki.com

Why do prints from different profiles look alike?

digidog

Guest

Senior Member

digidog

Guest

digidog

Guest

Senior Member

digidog

Guest

Senior Member

digidog

Guest

Senior Member

digidog

Guest

Senior Member

digidog

Guest

Veteran Member

Senior Member

Senior Member

digidog

Guest

Senior Member

Leading Member

Senior Member

digidog

Guest

Similar threads

Keyboard shortcuts