Print Service ICC Profiles

[digitaldog]

This makes no sense to me. If printers need to reprofile periodically because of printer drift, each profile will have somewhat different transforms to preserve colour accuracy. That being the cause, the softproof may also look different to the discerning eye from one profile to the next, depending on how radical the drift.

Yup, makes no sense to me either Mark! Anyway, I'm at a loss to understand how providing a profile solely for soft proofing, "accurate" (providing something close to WYSIWYG) or not, can be considered a color managed workflow when you're forced to send sRGB and unable to use the profile fully. I'm at a loss to fathom how Costco can profile individual machines and substrates and supply them to their customers, allowing a fully color managed workflow but that's not possible with other shops.

[Mark D Segal]

Yup, makes no sense to me either Mark! Anyway, I'm at a loss to understand how providing a profile solely for soft proofing, "accurate" (providing something close to WYSIWYG) or not, can be considered a color managed workflow when you're forced to send sRGB and unable to use the profile fully. I'm at a loss to fathom how Costco can profile individual machines and substrates and supply them to their customers, allowing a fully color managed workflow but that's not possible with other shops.

Well, of course it's all about the difference between what is technically feasible and what different providers consider to be "commercially sufficient".

[Doug Gray]

This makes no sense to me. If printers need to reprofile periodically because of printer drift, each profile will have somewhat different transforms to preserve colour accuracy. That being the cause, the softproof may also look different to the discerning eye from one profile to the next, depending on how radical the drift. Anyone hoping to receive a print with expected colours back from the printer is best advised to prepare the files under soft-proof using the profile that the printer will use for making the prints, otherwise all bets are off. Being consistent in the use of the printing profile for file preparation under soft-proof is the best assurance to avoid unwanted surprises.

Think about it. A profile for a particular printer will accurately render within gamut colors even though the specific tables that convert Lab space to printer space, and printer space to lab space have changed. The only differences will appear beyond the gamut boundaries and only to the extent the actual gamut boundaries have drifted. Since proofing is done by use of both BtoA tables and the inverting AtoB tables, most of the drift in any particular printer will be nulled out in round tripping.

Now I haven't tested this but it's not hard to check. I would even posit that soft proofing via a profile made from any one model of printer will not differ much from other printers of the same model using the same paper. As I said, it's not hard to check given that numerous profiles exist on different Costco printers of the same model.

Sounds like an interesting thing to investigate.

The extant question being how much variation would occur soft proofing using these different profiles from the same printer models? I would expect soft proof variations to be somewhat greater on different physical printers than from the same printer over time but we don't easily have a set of those. Costco printers are re-profiled every 6 months so that data is there presumably. Perhaps someone has archived those profiles.

Seems like something anyone here can test. How about it?

[digitaldog]

Think about it. A profile for a particular printer will accurately render within gamut colors even though the specific tables that convert Lab space to printer space, and printer space to lab space have changed.

There you go again with that accuracy nonsense again.

Now I haven't tested this but it's not hard to check.

Do your testing, then get back to us.

I would even posit that soft proofing via a profile made from any one model of printer will not differ much from other printers of the same model using the same paper.

Yeah, sure. Might as well just skip soft proofing if that's your view.

Sounds like an interesting thing to investigate.

That's your windmill to chase. Interesting you've provided a fair degree of text on the subject prior to the investigation!

[Doug Gray]

Might as well just skip soft proofing if that's your view.

Listen carefully. Soft proofing is valuable because it represents the actual colors printed. At least within the capabilities of current tech. My speculation is that the additional proofing color shifts that occur as printers age or those that just differ between different printers of the same model is minimal and rarely material but that the shifts seen by soft proofing itself relative to the reference image remain significant.

If this is true then it is not necessary to have current profiles for proofing though, or course, it is often necessary for current printers to produce consistent prints. Printing does not round trip the profile tables so printer drift directly impacts print colors.

[digitaldog]

My speculation is that the additional proofing color shifts that occur as printers age or those that just differ between different printers of the same model is minimal and rarely material but that the shifts seen by soft proofing itself relative to the reference image remain significant.

That's easily the most accurate text, written without assumption and speculation you've made today; you're admitted speculation! 

So let me see if I have this is clear: the lab has to re-profile for whatever reason. But the differences that warrant the new profile that presumably affects output doesn't affect the soft proof so the customer can keep using that old profile. Sounds nonessential to me. Anyone else?

[Doug Gray]

That's easily the most accurate text, written without assumption and speculation you've made today; you're admitted speculation! 

So let me see if I have this is clear: the lab has to re-profile for whatever reason. But the differences that warrant the new profile that presumably affects output doesn't affect the soft proof so the customer can keep using that old profile. Sounds nonessential to me. Anyone else?

Yes, you have that right! Differences that would affect an actual print quite a bit would make little difference soft proofing. Specifically, you might have a printed color of Lab(70,20,20) on a current profile that shows up at Lab (70,19,20) on the soft proof. This is pretty typical of how profiles operate. The same image might print that color using a profile several years older at Lab(68, 25, 23) yet it would return a soft proof color quite close to that of the new profile, say Lab(70, 20, 19).

The reason the soft proofs comes back close to the same is because it is round-tripped and that removes most all of the device drift.

I would think this is obvious.

[luxborealis]

Lots of discussion... The whole thing can be solved by following the age old adage...TRY IT!

I would suggest sending the service bureau the same file prepared a few different ways (w/ and w/o ICC; as sRGB, ARGB(98), w/ and w/o soft proofing, etc.) and judge from the results.

They don't need to be full-sized enlargements, provided the same set-up is used.

[Mark D Segal]

I would think this is obvious.

The only thing that's obvious is what I said: the safest procedure is to soft-proof using the same profile the printer will use for the making the print.

The next most obvious point is - for those of us who have actually printed and examined these comparisons - that it usually does take more than one L* or a* or b* step between one sample and the next to perceive the difference in print. The number of steps required before a difference of tone or hue is perceived varies depending on the colour. Therefore it is the case that humans will not generally perceive very small differences between the reference value and the printed value, or very small differences between two values under soft proof in printed output of the two samples. If this is all you are trying to say, fine.

[digitaldog]

What's obvious to me is lots of speculation and zero colorimetric proof or as just suggested, actual output!

[Doug Gray]

From 5 Noritsu 3411s at different Costcos:

Printing RelCol with a requested color of Lab(70 20 20). It's arbitrary but is the one I used as an example earlier.

This is a list of the printer locations and proof colors generated:

Profile: Costco-CA-Carlsbad-Gls   SoftProof Color: 70.1, 20.0, 20.1
Profile: Costco-CA-Danville-Gls   SoftProof Color: 70.2, 20.0, 20.1
Profile: Costco-CA-Santa Clarita-Gls   SoftProof Color: 70.1, 20.0, 20.0
Profile: Costco-CA-San Diego-Carmel Mtn. Rd-Gls   SoftProof Color: 70.1, 20.0, 20.1
Profile: Costco-CA-Sacramento-Expo   SoftProof Color: 70.1, 20.1, 20.1

This is a list of the last 4 printers using the Carlsbad printer to show that the printers do have some variation, though fairly small.

Profile: Costco-CA-Danville-Gls   Printed on Carlsbad printer - Lab: 71.2, 20.5, 21.1
Profile: Costco-CA-Santa Clarita-Gls   Printed on Carlsbad printer - Lab: 72.0, 19.0, 19.3
Profile: Costco-CA-San Diego-Carmel Mtn. Rd-Gls   Printed on Carlsbad printer - Lab: 72.9, 20.9, 20.0
Profile: Costco-CA-Sacramento-Expo   Printed on Carlsbad printer - Lab: 72.2, 20.6, 20.5

As you can see, the variations between printers make very little difference soft proofing. I would expect more differences using different profiles when soft proofing the entire sRGB aRGB, or ProPhoto gamuts but I would still expect those differences to pale in comparison to the differences between the initial image and the proofed (or printed) image.

[Mark D Segal]

I'm curious to know what application you are using to measure soft-proof colours, because as we know this is not possible in Photoshop. If you establish a soft-proof condition in Photoshop that produces an image appearance very different from the non-softproofed file view (say selecting a profile for a matte paper), the colour picker returns exactly the same colour value information in the Info palette whether the soft proof condition is switched on or off. Of course for what matters most - how the prints look - differences between the results of using different profiles, papers and printers are best read directly from the printed output of a target having known reference values, using a high quality spectrophotometer and a supportive application that allows measuring of patches. Grateful if you could clarify your procedure.

[digitaldog]

As you can see, the variations between printers make very little difference soft proofing.

Yes, profiles that define devices properly, devices that are kept in consistent behaviors (which your testing doesn't define) work by doing as their name implies; defines (profile) device behavior. So what?

[Doug Gray]

I'm curious to know what application you are using to measure soft-proof colours, because as we know this is not possible in Photoshop. If you establish a soft-proof condition in Photoshop that produces an image appearance very different from the non-softproofed file view (say selecting a profile for a matte paper), the colour picker returns exactly the same colour value information in the Info palette whether the soft proof condition is switched on or off. Of course for what matters most - how the prints look - differences between the results of using different profiles, papers and printers are best read directly from the printed output of a target having known reference values, using a high quality spectrophotometer and a supportive application that allows measuring of patches. Grateful if you could clarify your procedure.

Yes, Photoshop makes getting actual values of soft proof colors a real PITA. I recently noticed that Photoshop has made fractional Lab values available and I started looking into it again. There still is no direct way to get soft proof Lab values but there is an indirect, somewhat circuitous way. I'll write about the specifics later when I get back to my main computer.

I didn't use Photoshop for the numbers above. I used Matlab's image toolbox with some custom functions to make their color engine stuff more usable. As an example this will plot the perceptual intent tone curve of "myprofile.icc":

x=[]; for i=0:100;  x=[x; ConvertProfile([i,0,0],'myprofile.icc', 'f', 0, 'r', 1)] ;end; plot(0:100, x)

As for actually measuring colors on the monitor I just use a I1Pro. Not perfect but averages around 1 dE and rarely off more than 2.  Monitors are tricky. There's variation depending on where on the surface you measure. I have an Eizo 318 which is quite good but still has some variation.  And, of course, the spectro operating in radiometer mode, has some drift. Sources of error are endless though. Instrument variation, differences between the graphic/industrial arts and photo folks (XRGA).

I really like PatchTool.  Very flexible and makes recording data for input into Matlab easy peasy.

[digitaldog]

Yes, Photoshop makes getting actual values of soft proof colors a real PITA.

Apple's Digital Color Meter anyone?

[Mark D Segal]

Yes, Photoshop makes getting actual values of soft proof colors a real PITA. I recently noticed that Photoshop has made fractional Lab values available and I started looking into it again. There still is no direct way to get soft proof Lab values but there is an indirect, somewhat circuitous way. I'll write about the specifics later when I get back to my main computer.

I didn't use Photoshop for the numbers above. I used Matlab's image toolbox with some custom functions to make their color engine stuff more usable. As an example this will plot the perceptual intent tone curve of "myprofile.icc":

x=[]; for i=0:100;  x=[x; ConvertProfile([i,0,0],'myprofile.icc', 'f', 0, 'r', 1)] ;end; plot(0:100, x)

As for actually measuring colors on the monitor I just use a I1Pro. Not perfect but averages around 1 dE and rarely off more than 2.  Monitors are tricky. There's variation depending on where on the surface you measure. I have an Eizo 318 which is quite good but still has some variation.  And, of course, the spectro operating in radiometer mode, has some drift. Sources of error are endless though. Instrument variation, differences between the graphic/industrial arts and photo folks (XRGA).

I really like PatchTool.  Very flexible and makes recording data for input into Matlab easy peasy.

Thanks Doug. How did you get fractional Lab values in Photoshop. I looked through the Preferences for the Info palette in the latest CC subscription version and couldn't see an option for that. Must be somewhere else?

[digitaldog]

How did you get fractional Lab values in Photoshop.

You can't. You can in Lightroom and you can set Lab in PS like this (assuming this is what you guys mean by fractional):

[Doug Gray]

I'm curious to know what application you are using to measure soft-proof colours, because as we know this is not possible in Photoshop.

Here's a screen shot of the Lab 70,20,20 in Photoshop. Global color settings are Relative Colorimetric w/o BPC. The top snap is with Photoshop view proof set to the Carlsbad Noritsu, also with Rel Col.  The RGB values in the proof column are exactly what I get from Matlab.

The middle snap is after profile conversion to the Carlsbad 3411. Values are shown to 2 decimals but, as you will note they correspond to the Matlab round trip posted earlier.

These proof colors are scaled so the white point is L=100,0,0. To see the actual color printed one can convert using Abs Colorimetric. This is the same as "show paper white" in view proof.

[Doug Gray]

You can't. You can in Lightroom and you can set Lab in PS like this (assuming this is what you guys mean by fractional):

Didn't used to be able to get fractional values in PS. Sometime last year they changed it, YAY! just set the option for 32 bits.

[Mark D Segal]

Didn't used to be able to get fractional values in PS. Sometime last year they changed it, YAY! just set the option for 32 bits.

Thanks, that indeed works, but does exactly the same thing with more precise read-out. Still no softproof values - only picks up the values without softproof.