Untitled Document

Those preparing files for commercial printing are often counseled to "ask the printer" what separation settings to use, on the theory that good printers should know such things. In the real world, printers rarely do, often giving horrible advice that would result in wrecked jobs. In late 2007, the following long thread appeared, complaining about the information that one such printer had provided. In the following three months, four similar complaints appeared about different printers. Although not discussed at the same length as the first case, they're worth a read.

I am preparing some RGB files to be printed on a cmyk press in Hong Kong. These images are going to be used in "coffee table" book. The printer has supplied me with a profile, but I am not sure the profile is any good. It really flattens out the color, alot more than the SWOP2006_Coated3v2 profile. The profile I was given is for an offset press. I know this because the profile has "OffsetMatt" in the name. I have an image that has a red of 253,39,31. When I convert with the printers profile I get 0,72,75,0. When I convert using the SWOP profile I get 0,97,83,0. I have a sample do the paper and it looks close to the Epson proofing semi-matte, which is what I am going to use to proof on my Epson 4800 with the Colorburst X-Proof rip. I have been trying to get in touch with the printer, but I have only been able to speak with the publisher. The publisher says that they have just used the standard swop profile in the past and then supplied them with proofs for them to match. But my concern is that when I convert with the printers profile I get a TIL of 340. I am under a tight deadline and I am not sure if I can wait to get more info. I don't really have to much more time to wait for an answer from the printer. So what is the best way to proceed? Do I just use the swop profile or do I create a custom cmyk setting using 20% dot gain and a TIL 340?

Thanks,

Steve Peters

___________________________________________________________________________

Steve, I presume that the RGB space to assign to your values is A98 (you did not say, not that it really matters in this case though).

What rendering intent did you use that resulted in 72m75y for the red (the RGB values are out of gamut for common press conditions). If perceptual, does relative colorimetric with BPC provide better results? Perhaps try all four intents and see if one is better for saturation, but still acceptable for TAC, black point scaling etc. Are you using the Adobe ACE or another (MS ICM or Apple ColorSync)? Perhaps try a different colour engine to see if results vary with this particular profile. Obviously the SWOP and this printer are using different stock and ink, but I would presume that one could do better for reds than what this profile provides with your current conversion settings.

You mention, 340 TIL, but what is the K limit? If you convert a 0r0g0b patch to CMYK, what shadow build do you get (it should vary between render intents). What is the character of the K plate? What GCR or UCR is going on for neutrals and colours? Where does K start, how 'steep' is the ramp from start to the K limit? Gray gradients and step patch test files are handy for this evaluation.

You ask about simply converting with another profile. Convert to say SWOP v2 and then assign the OffsetMatt to see what the image would be like in this new condition (CMYK values remain the same). Or better yet, perform a softproof with preserve values checked, instead of/in addition to the assign step. Using a neutral step wedge, one can evaluate neutrals via LAB readings to see what happens to the SWOP v2 gray balance when it is used in the OffsetMatt condition.

Repeat and do the same for Custom CMYK too, to see how this compares to the other two profiles.

Another thing to evaluate is dot gain/TVI. If you are going to use SWOP v2 (TR001), does this printer in HK have similar dot gain to the TR001 files that you may supply?

It is best to get the separations right to begin with, but one can also edit the less than satisfactory CMYK file and make the reds and or other hues or tones richer if desired. Unlike inkjet printing, you are working in final print space and can thus directly control the values (one is not forced to accept the profiles values).

Hope this helps,

Stephen Marsh.

___________________________________________________________________________

You've got a potentially troublesome situation on your hands with this book. You can't get hold of the printer, but you do have a profile that supposedly describes their press condition. The profile, however, doesn't make a whole lot of sense.

Your client (I think), the publisher, tells you that in the past they have not used the printer's profile and simply converted to SWOP.

Is there any reason you *have to* supply CMYK to the printer? If you can show your client what your images look like on an industry-standard proof, and they are OK with the printer doing the conversions and matching to your proofs on press, then if it doesn't look good (like the proof) the blame will be on them.

If you take over a fundamental part of the printing process and don't have a solid grasp on the kind of jargon that will get thrown around after the job gets bounced you will be left holding the bag. And it sounds like the printing will be expensive.

If it were me, I'd do everything I could to get in touch with the printer and get a comfort level with their profile, or even better, agree on something similar (like GRACOL or SWOP). No matter what happens with that I'd have a frank conversation with my client, explaining the potential problems, and getting them on the same page in the decision.

If you absolutely have to convert to something I'd say do what the publisher told you, not the printer.

--

J Walton

___________________________________________________________________________

There was one other thing. I just tried converting that same file using custom cmyk settings, swop (coated) 17%, gcr, light,K limit 85%, til 350. The red printed pretty flat with those settings as well, and the proof and monitor are way off. I like the way the colors are reproducing with the SWOP2006 profile and I am able to use soft proofing. If I find out that I need to have a TIL of 340, what is the best way to go about getting there? Could I just use the GRACoL2006_Coated1v2?

___________________________________________________________________________

Thanks guys. The rendering intent was relative with BPC and I am using Adobe ACE. I did try other intents and rendering engines and none where any better. I know that this could be a big problem and I have been in contact with my client and they understand the issues.

So I have a 21 step white to black. Using the supplied profile, 0,0,0 comes out to 88,80,84,91, so that actually 343. Black comes in at the 166,166,166 which is 33,24,25,1.

I know that I really need to get more information from the printer. 99% of the time my file are going to be printed in standard us pubs, which I would usually use the latest SWOP2006_3v2 profile that I got from SWOP.org. It's very rare that I have to actually deal with a printer, but the times that I have they have never been able to give me any more information that "convert to cmyk". Or if I am lucky they would at least know the TIL or dot gain. So I am trying to at least get the TIl and dot gain. The publisher has told me that they have just used the standard swop profile with in photoshop and then supplied them with a proof. But if the TIL is actually 340 and I give them files that are only at 300%, and supply them with a cmyk proof from my Epson, will they be able to match it? The reason we don't just give them the rgb files is because there are alot of very bright saturated colors that will not reproduce so we are going to need to make some artistic adjustments that we are not comfortable letting some one else do. If I can at least get the TIL and dot gain should I just create a custom cmyk setting? I am assuming that soft-proofing with a custom cmyk setting is pretty much pointless? I do know that the Colorburst X-Proof rip comes with Spectravision pro and I do have the X-Rite pulse. Is there any way I could create a profile using the Spectravision Pro software and use the TIL and dot gain of the actual press? I know that this is only going to be profiling my printer and paper, but it seems like there should be some way to make this work. Wouldn't I at least get better soft proofing? Which in this case is kind of important so that we aren't having to pull a bunch of proofs when we are trying to adjust those out of gamut colors?

I know that I am still going to be taking a gamble with all of this, but in lieu of getting all of the pertinent info, I have to figure out the best way to minimize any issues.

___________________________________________________________________________

While this sort of problem is out of my field, I can't help but wonder what would happen if you sent an sRGB file along with a proof print and ask the printer if he can give you a good match. If the printer is so confident of his operation that he feels it unnecessary to work with you, would it be unreasonable to expect him to convert your sRGB file to the CMYK profile that produces a good match on his press under his printing conditions? I would think that an sRGB image file that he converts to his custom profile should be expected to give as good a print on his press as one might expect if you sent it to a printer who uses a SWOP profile for a press that uses SWOP. The RGB file would be the same in either case. If you're not the printer, and you can't get any useable answers, doesn't it become the printer's problem if he has to make the conversion? If he is unable to convert the file himself with the expectation of at least coming very close to your proof print, shouldn't the publisher be aware of that?

Howard Smith

___________________________________________________________________________

Steve,

Have you tried Stephen's suggestion of converting to the supplied proifle and assigning your swop profile to see just how different they are tonally and color wise? That'll tell you a lot. If there is a big shift when you assign your swop profile, then it'll be pretty clear that your printer is not using something close to swop. You can't judge how your colors are looking when you use the wrong profile - that is, if the swop profile which you say makes your colors look good, is not correct for your press, then it's just a waste of time. Better to use the right profile and tweak the image after converting if that's what it takes. Who's to say they're not using some non standard ink that needs different percentages to make the colors you're after.

___________________________________________________________________________

Peter Figen wrote:

Thanks Peter, this is one important aspect of evaluation, although I suggested converting to SWOP v2 then assigning the OffsetMatt profile/softproofing as this will indicate how the SWOP numbers will reproduce in this new setting (if the profile is acceptable for this soft-proofing or proofing task).

I agree on the first sentence Peter, if using the 'false' SWOP conversion for the OffsetMatt condition and proofing is not acceptable.

You could be correct about the inkset, but we are talking of Adobe RGB values of r253 g39 b31, for all intents and purposes a highly saturated red, out of press gamut (even in sRGB this is OoG for typical press conditions). A sensible separation approach would deliver a highly saturated result in CMYK. No matter what ink is used, one would want to increase to full/near full MY with no or little CK values (in a flat area that contains no detail).

Steve also reported that generally speaking, the images are coming out rather flat and disappointing. These are probably the tricky saturated images that he spoke of. It is unclear to me if more subdued images within press gamut convert similar to the SWOP.org profile, or if RGB colour within press gamut is weak too.

Having access to the profile would answer many of these questions.

Sincerely,

Stephen Marsh.

___________________________________________________________________________

Steve, as you may know - there is a lot of variability in profiles that describe the same/similar condition.

Your Adobe RGB red hue values of:

253r 39g 31b

Convert to (RelCol+BPC):

SWOP TR001 Chromix - 80m 84y

SWOP Press - 91m 87y

SWOP v2 Adobe - 93m 85y

SWOP TR001 - 97m 100y

With Perceptual rendering, the Chromix profile really washes out with a result of - 71m 74y.

So it may just be the profile and an update to the profile creation software or another profile package or perhaps different operator choices may produce a better profile for the same condition from the same measurement data.

Reducing saturation before conversion for OoG hues may be required to produce a 'richer' conversion.

You could try doing two conversions, one to the OffsetMatt and one to SWOP...Layer the SWOP file over the OffsetMatt with no conversion (retain values over appearance) - try setting the blending mode to darken to compare to normal. In the advanced blending (layer) blending options for the upper SWOP layer, uncheck the K channel check box and perhaps use blend if sliders to blend in the darker values (so as not to mess with gray balance in the light end of the tonal scale). One may need to layer mask (saturation or density mask) this in key areas that require a saturation boost. This can all be put into an action, although it may just be easier to use selective colour to boost values in the red hues etc. It will depend on the image if simple edits or blends of two conversions or more work will be needed.

Hope this helps,

Stephen Marsh.

___________________________________________________________________________

A SWOP printer would soot himself in the foot by having files reduced to sRGB, a packaging (Hexachrome?) printer would aim higher.

On the Mac platform, open the Color Sync Utility

click on the Profiles icon (top bar)

chose sRGB

Control click on the color windowhold for comparison

chose US Sheetfed Coated v2

Observe that some of the CMYK colors fall outside the sRGB space.

(that's the foot shot : )

Hexachrome is a larger space than CMYK

(that could be the higher target—the crotch? the head? : )

Laurentiu Todie

__________________________________________________________________________

On Oct 31, 2007, at 3:00 AM, Stephen Marsh wrote:

Absolutely. Too many profiles provided are also built with ink limits, black generation or rendering options that suggest whoever built the profile didn't understand the printing process. I will often dump provided profiles back into ProfileMaker and rebuild them to suit, or pull them into ProfileEditor and tweak specific response.

Peter Figen

___________________________________________________________________________

So I posted the profile in the files section of this forum. The name

is Midas_KBA_OffsetMatt_V3.icm.zip

I still have no more printer info today and I need to get started.

___________________________________________________________________________

Steve, no time for a deep reply during my working day...

My advice is to Convert to SWOP v2 or SWOP.org ICC first, then go from SWOP to MattCoated.

Yes, that is right, RGB > CMYK and then CMYK > CMYK (I have never had to do this for good press conditions, only newsprint!).

I'll post more later if required, if the numbers and preview do not make sense with this brief explanation!

Stephen Marsh.

___________________________________________________________________________

Thanks, Laurentiu! All this time I thought (mistakenly) that sRGB was the ideal space for conversion to CMYK when one wants to reduce the color loss as much as possible. I've found that the service bureaus with which I worked in the past all seemed to share this opinion. I'll never forget one highly recommended "expert" looking down his nose at me and sniffing that not many people know the right color space for a scan when the final image is intended for CMYK printing (he was convinced that sRGB was the obvious choice) . I'm sure others have covered your points in print, but somehow it never made an impression. Probably because I've not had any hands-on experience with CMYK printing other than--long ago--working with a commercial printer's press supervisor to make sure the output was close enough to the proofs prepared by others from my 8 x 10 transparencies.

But a question lingers. Dan has repeatedly suggested submitting an RGB image file (or a LAB image file) to a printer when there is some question about which CMYK profile to use for conversion. Wouldn't an Adobe RGB 1998 file go to the opposite extreme and give colors that wouldn't convert well? I'm beginning to see why so many different solutions are being provided in response to the original post. There don't see to be any easy answers.

Howard Smith

___________________________________________________________________________

Truth be told, there aren't legions of CMYK colors falling outside of the sRGB space. You'd have to be kind of greedy to want pure cyan, "better" greens and bright yellow (I am : )

Besides, those colors are not visible on most monitors (they are on paper) and as better (16bit) monitors become available, better CMYK inksets seem to be made so we're chasing our tails.

The "expert" scanner technician may not be aware of Plato's observation from 2500 years ago. which is that we can only be experts in our own… cave : ) My cave is not that of a color expert, and it is definitely smaller than say,… Stephen Hawking's who's range of expertise goes from neutrinos to black holes, but I'd give the offset printer a file in the "capture space" (unless the edit space is larger and some of it's additional colors were obtained through curves or other means) This of course means that the capture was not done by one of NASA's super scanners and that the edit was done by an expert who didn't trade shape for saturation.

At times like this I wish that Bruce Fraser was still around, not because his cave was larger, but more focused and his command of the English language was kicking the ass of my Romanglish (Romanian native—American by choice)

PS

I (personally) would give the offset printer a converted (to his space) file with a black plate who's deep SD was extended to close to 98% if needed, but printers are jealous of their spaces or don't know their cave to save their own lives or think that we are stupid : ) Well,.. some of us are as some of them are too.

Laurentiu Todie

___________________________________________________________________________

Further to my earlier reply -

Steve, now that I have had time, I can make some observations on this Midas profile and then I will comment on the conversion workflow.

It appears that this profile has major problems converting highly saturated red and green hues, such as found in the primaries of common RGB working spaces (and the opposing hues cyan and magenta). Blues (and the opposing hue yellow) are "better" for some unknown reason. The profile does not appear to have issues with in-gamut colours and provides similar results to other common profiles.

Rendering intents - relative colorimetric with black point compensation seems to be the best choice, particularly so for out of gamut reds and greens. If one wishes to depart from the original RGB slightly, one can use the perceptual intent, although this is not recommended for saturated originals.

As my previous brief post indicated, after testing the profile I now strongly recommend that you convert highly saturated images that have problems with a regular conversion workflow using my suggested two-step workflow approach. You may wish to just convert all images with this two-step conversion workflow, just for consistent results and to simplify things (you will obviously be working on dupes and will have access to the RGB files if you need to reconvert or channel blend a tricky file).

1. Convert from RGB to Adobe U.S. Web Coated SWOP v2 (TR001). You will likely use RelCol+BPC in this step, although the conservative/subtle gamut compression offered by the Adobe profiles perceptual intent may help if there is detail in highly saturated areas of the RGB file.

2. Once the file is in CMYK, do a second conversion to the Midas CMYK profile using RelCol+BPC. The previous conversion to a "well behaved" press profile will have tamed the out of gamut colours, thus not leaving you to the mercy of the Midas profile's handling of OoG colours. The Midas profile will have no problems converting from the more similar TR001 press space than say direct from Adobe RGB. You will now have saturation similar to that offered by other profiles.

This would be put into an action. It would then be your choice on how you would use this action. One may simply automate>batch the action, for attended or unattended conversion and saving of all the source files.

One would then inspect all the conversions and perhaps tweak some of the images. Special cases may require converting via perceptual intent with a *non* Adobe CMYK profile (less conservative) in step 1, before converting to the Midas profile in step 2 and then performing post conversion edits (curves, channel blends etc).

As an earlier post of mine indicated, reducing saturation in the OoG RGB hues before converting to the Midas profile would create more pleasing results. I find that converting to a "better" intermediate CMYK profile is a faster process and provides better results than say reducing saturation via the hue/saturation command slider (normal or color blend) or with AB curves in LAB.

I hope the rest of your project runs smoother than the CMYK conversion process!

Regards,

Stephen Marsh.

___________________________________________________________________________

Sorry about this "vanity" post : )

On Nov 1, 2007, at 5:27 AM, Laurentiu Todie wrote:

"it's" is a typo

The forum software seems to have removed my (em?) hyphen should read: "Romanian native, American by choice"

and for Dan (who banned me for sarcasm, in the past) the language barrier can lead to frustration, frustration can lead to sarcasm well,.. that's my explanation/excuse : )

Laurentiu Todie

___________________________________________________________________________

On Oct 31, 2007, at 8:10 AM, Laurentiu Todie wrote:

Huh? sRGB would actually work fine for SWOP. The only areas where SWOP is actually out-of-gamut compared to sRGB is in a small portion of pure cyans. Otherwise, sRGB works fine for conversion to many typical CMYK press color spaces. Not true of course for typical CMYK inkjet gamuts. AdobeRGB is better suited for full-gamut inkjet printing.

Not necessarily. Any decent (good media) CMYK inkjet profile is going to be larger than Hexachrome. Even if you're talking "press" CMYK spaces, it's not a given that Hex is larger since there are several "wide-gamut" CMYK ink sets designed for offset printing.

Regards,

Terry Wyse

_____________________________

WyseConsul

Color Management Consulting

G7 Certified Expert

704.843.0858

http://www.wyseconsul.com

http://www.colormanagementgroup.com

___________________________________________________________________________

Steve, here's what I can tell you about your profile....

* Total ink limit is closer to 350% (348% to be precise) with a K limit of 95%. This is way too high of a TIL in my opinion. Like I said earlier, with a K limit of 100%, you really don't need more than about 280-300% TIL to achieve maximum printable density.

* Profile was made using Agfa ColorTune. I can't say one way or another if this is a good profiling application or not but I can tell you that NONE of my colleagues use this application.

* Precision is probably not very good. This profile only uses 9 grid points in the A2B table and only 17 grid points in the B2A (separation) table. Most profiles will use at least 25-33 grid points for the separation table.

* Probably the most disturbing things about the profile showed up when I viewed the hue/chroma plot. In comparison to SWOP2006_Coated3, I observed the following:

* Cyan is "weak" (lacking in chroma) and has a fairly large shift towards blue/magenta. I would call this a very "dirty" cyan.

* Magenta is slighty weak but the hue is OK.

* Yellow has good chroma/saturation but the hue is slightly shifted towards green.

* Blue (C+M) is a bit too saturated but is very "purple" or magenta casted probably due to magenta contamination of the cyan ink. Bottom line, if you were touse the SWOP2006_Coated3 profile to separate with, you'll end up with purple blues on press, if they print like this profile predicts. The relatively high chroma of the blue has me wondering whether this came from an actual offset press or did it come from a dot proofer such as a Kodak Approval? I've seen both dirty cyans and too-magenta blues with particular "ink sets" on Approvals. Dot proofers with their perfect "wet trap" characteristics will generally show a higher blue chroma compared to an actual press plus a hue shift towards magenta.

* Red (M+Y) is fine as far as hue angle but is a bit weak in chroma.

* Green (C+Y) at 100% is fine for both hue and chroma but has a yellow hue shift between 60-80%.

* The GOOD news about this profile is the gray balance characteristics. At the midtone, it's within 2-3% of standard GRACoL/ SWOP gray balance but it does get a bit worse in the 3/4 tones. If you separated using SWOP2006_Coated3 and printed on this "press", you'd see a decidely warm/reddish cast in the neutrals assuming, again, that the press prints like your profile predicts.

As to your problems when using this profile to separate, I believe that it's an issue of rendering intents. If you're using perceptual, there can be a HUGE difference in how different profiling applications convert using this intent. Monaco PROFILER is typically the "best" in terms of it's perceptual rendering but ProfileMaker is pretty decent also. I have not idea how an Agfa ColorTune profile behaves using different rendering intents.

Steve, I'm uploading a new profile as we speak for you to try. It's build using ProfileMaker. The name is

"Midas_KBA_OffsetMatt_TIL320.icc.zip". Give it a try.

Regards,

Terry Wyse

_____________________________

WyseConsul

Color Management Consulting

G7 Certified Expert

704.843.0858

http://www.wyseconsul.com

http://www.colormanagementgroup.com

___________________________________________________________________________

On Nov 1, 2007, at 10:43 AM, Terry Wyse wrote:

…none of which print Orange.

Look, I'm just giving the gist of color spaces, the way I see it. A loss is a loss, and sRGB won't do for me. You can do what you want.

Laurentiu Todie

___________________________________________________________________________

I just looked at Steve's Midas profile this morning. Overall, it's pretty close to SWOPv2 in its basic look. The saturated red thing is definitely a problem, so I used ColorLab to build a data file (this was an Agfa profile that had no embedded measurement data) and then imported the new measurement data into ProfileMaker and built a new profile that has much better red rendering. I sent the new version directly to Steve. The whole process took about three minutes.

Peter Figen

___________________________________________________________________________

Just to reemphasize Stephen's point, there are three important lessons to be drawn from this instructive thread.

1) When you ask a profile to convert something that's as wildly out of printer gamut as this particular red is, you're playing Russian roulette. There isn't any "right" way to handle this situation because the color being demanded is far, far, too intense for the printer to handle. To get an in-gamut color from this extremely intense light red, you'd have to, if using Hue/Saturation, take saturation down to -50 or so.

In such cases the only question is not whether a horrible mismatch can be avoided, but which horrible mismatch is the least objectionable. That will depend on the image. If there's a fire engine or a Canadian flag in the picture, which would both be in the 100m100y range, then the SWOPTR001 rendition of the lighter red would be a big mistake, because we wouldn't be able to tell these highly dissimilar colors apart. OTOH, the more ink being used, the more intense the color. If there aren't any competing reds, then the SWOPTR001 lie is probably the preferable one: better a relatively bright red that's too dark, than one that's the right weight but too dull.

In evaluating a profile's performance, then, colors like these are best ignored, because there isn't any right answer. Concentrate on ones that the printer has some remote possibility of hitting.

2) I haven't looked at the profile myself, but Terry, Stephen, and Peter all seem to find fault with it. If so, it's no surprise. Profiles handed out by commercial printers are like stock market tips handed out by printers. If you have some evidence that the printer knows WTF he's talking about, then his profiles and stock market tips are both to be given some respect. But an unknown printer--I wouldn't assume that the profile was worth anything. Which is a a very strong argument for those who may find themselves at a printer's mercy to learn enough about the process so that that lamentable situation is avoided in the future.

3) Both Peter and Terry report that they were able to generate corrected profiles in short order. That serious professionals find it necessary to edit CMYK profiles in real-world contexts like this one is a strong argument for the inclusion of profile editing in Photoshop.

My thanks to those who contributed to this good thread.

Dan Margulis

___________________________________________________________________________

On Nov 4, 2007, at 7:52 PM, Dan Margulis wrote:

No, if you carefully read what Terry wrote, he "built" a new profile "using ProfileMaker." He further noted that not all profiling applications give equivalent results (particularly with perceptual rendering intents), and that the profile generation software (Agfa ColorTune) used to generate the printer's profile is generally not used in the industry and that it is in that respect an "unknown." He also noted that the ColorTune profile uses only 9/17 (vs 17/33) LUT grid points. X-Rite's ProfileMaker and Monaco PROFILER are the runaway favorites for profile generation. But in general, most experts agree that it is much better to generate a new profile than it is to try to edit a profile. My preference is ProfileMaker, but I've spent very little time with PROFILER.

Rather than an argument that Adobe should bundle a profiling editing application within Photoshop, this thread makes the point that good tools are already available, and they can be purchased (like Photoshop) by those who need to use them. (That should have included the printer!) Of course, one also needs to know *how* to use these specialized tools (or hire a consultant). And if you don't own a spectrophotometer, you won't be generating many profiles, regardless of what software you use, or whether or not it's bundled with Photoshop. Profiling is all about colorimetry, plain and simple. Good profiles are the answer. A bunch of hacked (and deceptively modified / named) profiles is not. If you want to edit or create profiles, there is already good software available. Asking Adobe to bundle equivalent software is not reasonable. These are full-blown applications. And if you do much profile generation / evaluation, you should also have Steve Upton's ColorThink Pro to examine and evaluate the profiles.

Perhaps Terry could shed some light on how he evaluated the bad profile and how he built the new one.

--Rich Wagner

___________________________________________________________________________

Peter Figen wrote:

Peter, I don't have the hardware or 'proper' software, so my evaluations are usually performed in Photoshop with RGB and CMYK test images. I can compare the gamut in 3D in WinXP and also look at a gamut plot in ColorSync on the Mac. On the PC, I can also use two free apps - "ICC Profile Inspector" and "ICC Inspect" to find out some basic under the hood details on profiles.

There are 17 tags listed (starting from #0), with tag #8 with a signature of DMDD or Device Model Description tag. This is listed as "Euro Standard". Go figure. PCS=LAB, Default Intent=RelCol.

I suspect that if there is any measurement data, it is hidden away in tag #3 with a signature of AGFA. This is a private or unknown tag. It shows that an IT8.7.3 was used with an Xrite DTP41 with Agfa ColorTune Pro 4.2.1 Mac OS Classic. This tag is huge, larger than any of the A2xx or B2xx tags.

Tag #16 has a signature of VUED or Viewing Condition Description Tag, which lists what appear to be some sort of reference to measurement data in the header which may mean nothing (nm380, nm390 etc). It also states a TIL of 350%, K limit of 95% and K start of 30%.

As I only had access to Photoshop, I found great results with converting to Adobe SWOP v2 first, then from SWOP to Midas (RGB>CMYK, CMYK>CMYK). This still leaves one at the mercy of this profile for TIL etc, but it cures the saturation issue for out of gamut reds/greens as it has no problems with in gamut colour (despite it's other problems).

Regards,

Stephen Marsh.

___________________________________________________________________________

Thank you for the in-depth reply Terry.

I agree with you on the TIL Terry, as the profile defaults to 95% K.

For comparison, I looked at the Adobe SWOP v2 profile, which uses a CLUT resolution A=9 and B=33. The Quark Xpress 6.5 SWOP TR001 profile uses A=17 and B=33. Chromix SWOP TR001 uses A=11 B=25.

Can you expand on this point Terry? I only used Photoshop to evaluate the ink:

Adobe SWOP v2 Cyan 100% = 62L -44A -50B

Midas Offset Matt Cyan 100% = 56L -29A -49B

I presume that you are referring to the differences in the L and A values (not that TR001 is an "ideal" reference or the exact condition that Midas may use, but it makes a good general yardstick for comparison).

Side note: I had RelCol set in my Color Settings when performing the above evaluations, a more "accurate" evaluation could perhaps be obtained by temporarily setting the rendering intent to AbsCol before performing info palette evaluations of single solids, tints and overprints of CMYK values.

The Device Model Description tag states "Euro Standard", for what this is worth.

I asked this of Steve earlier and he said that RelCol was being used, although all intents were tried will similar results.

Something deeper is going on here Terry. It appears that this profile simply falls apart when presented a wildly out of gamut colour. The results for in-gamut colour are similar to other profiles.

Steve's A98 red of 253r 39g 31b converts to Midas CMYK:

Perceptual/Image = 75m83y

Relative Colorimetric + BPC = 72m75y

Compared to RelCol+BPC:

SWOP TR001 Chromix - 80m 84y

SWOP Press - 91m 87y

SWOP v2 Adobe - 93m 85y

SWOP TR001 - 97m 100y

Your adjusted/recreated profile with the 320% TIL converts to:

Perceptual rendering intent = 88m73y

Relative Colorimetric + Black Point Compensation = 91m95y

It is not just OoG red hues that are improved, green is better and so is blue (you appear to have been conservative with the ratio of magenta to cyan in OoG blues, which I applaud).

As I only have Photoshop, my hack was to first convert from A98 RGB to Adobe SWOP v2, then from SWOP v2 to the Midas Offset Coated profile. This resulted in the red hue discussed above converting to: 90m88y1k

The low K tones introduced are not ideal, but can easily be removed with selective colour, or a curve without affecting the quarter tones or higher or any true K plate content (with the start point being 30 on the grey ramp). This provides better results than RGB to Midas CMYK if one does not have access to dedicated profile software.

Terry, thank you for the illustration of the difference between software and operators and how this affects a given profile for the same condition.

P.S. If you are up for some limited talk on v4 profiles (and how the ICC now seem to state how profiles should be used [not so simple!]), perceptual rendering and "preferential" conversion aspects of profiles, particularly in regards to CMYK perceptual transforms - then I for one would be interested in such a discussion.

Sincerely,

Stephen Marsh.

___________________________________________________________________________

On Nov 5, 2007, at 2:43 PM, Richard Wagner wrote:

I cheated, that's how! :-)

Actually, I'm a little embarrassed that Richard pointed out something that I simply left slide by; the fact that I *did not* edit the profile in question but I in fact built an entirely new profile from the colorimetric data of the original profile. With the exception of Photoshop's "Custom" CMYK engine, there's no real profile editing applications that I know of that can take a profile and re-generate it's separation settings. Neither Monaco's built-in profile editor nor ProfileMaker's Profile Editor app has any features to change the GCR, total ink limit, etc. To do THAT you need to build a new profile.

As far as what I used to evaluate Steve's profile, I used the following:

ColorThink Pro 3.0 by Steve Upton at Chromix, the "swiss army knife" of profile evaluation apps. I used Colorthink primarily to view all the tags in the profile, looking for a tag that MIGHT have had the original colorimetry (measurement data) that I could extract and use to build the new profile (alas, it did not). I also used ColorThink Pro to evaluate and compare the gamut plots between Steve's profile and SWOP2006_Coated3 (note to all folks on this forum, "SWOPv2" is NOT the official SWOP profile anymore, hasn't been for over a year now, it's the new "G7" SWOP2006 profiles that are the official "blessed" ones these days). Also note that a typical gamut plot display the "volume" of a profile is not all that telling. What I was able to do in ColorThink Pro is look at the hue/chroma plot of the primaries and secondaries which showed that, contrary to some statements, this profile was nothing like SWOPv2, especially in the cyans/blues. The gamut plot might have been similar looking (most "press" CMYK profiles look similar in that regard) but that doesn't always reveal what's going on with the primaries/secondaries.

The other profile evaluation program I turn to is ColorShop X by Chris Halford at X-Rite. I generally use this for a second look at the primaries/secondaries since this app allows you to look at the different rendering intents. For evaluating a profiles "separation" characteristics, looking at the relative and perceptual rendering intents is more telling that using absolute colorimetry. I also use ColorShop X to evaluate the gray balance and K generation settings. With their "Neutral Analysis" plot, you can look at the CMY gray balance, view the K start and K limit, and also look at the "stability" or accuracy of the gray balance in the profile.

Getting to back to how I was able to actually build a new profile, the way I did it was to use the LOGO ColorLab application to "extract" the measurment data from the profile. I did this by opening up a text reference file of an IT8.7/4 chart and converted this to LAB and saved this as a text file of the original measurements. This is far from an ideal way to do this because I'm relying on the inherent accuracy of the original profile's (already suspect in my view) CMYK-to-Lab (A2B) conversion. Once I had the Lab data, it was a simple matter to import this into ProfileMaker and build the new profile (bonus points to anybody that knows of an easier way (there is) I could've done the same thing).

To Richard's point, we don't need no friggin' new profile editor/creator as part of Photoshop; the program is already bloated enough in my view. As a color management consultant, I also don't want to see even more poorly built profiles floating out there in the ether because of a possible half-baked attempt by Adobe. It's taken us this long (10+ years) to get to the point we're at today in terms of having really good measurement instruments and profile creation/editing/ evaluation applications and we're still not QUITE there yet in my opinion. I'd hate to see us take a few steps backwards.

Now, back to the Funny Farm,

Terry Wyse

_____________________________

WyseConsul

Color Management Consulting

G7 Certified Expert

704.843.0858

http://www.wyseconsul.com

http://www.colormanagementgroup.com

___________________________________________________________________________

Peter Figen is correct - there is no measurement data embedded within the Agfa profile, which is why he used ColorLab to build a (synthetic) datafile from the profile, and then used ProfileMaker (as did Terry Wyse) to build a new profile. Note that neither Terry nor Peter "edited" the existing profile, even though ProfileMaker 5 includes a separate profile-editing program that they both had access to, creatively named ProfileEditor.

The private tag named "agfa" of type "data" cannot be read outside of ColorTune, and I have no idea what the data is used for. Agfa notes that "ColorTune Pro uses ICC-based profiles and unique mathematical algorithms to calibrate each of your output devices. Then, it uses the device characterization within each profile to automatically compensate for differences in color ranges between your monitor, printer, proofer and printing press." As the tag contains proprietary data, all bets are off. If you look in ProfileMaker profiles (like the one Terry posted), the data sets are stored in ASCII text. This makes it easy to open and re-build profiles using different parameters, or to examine the original measurement data.

--Rich Wagner

___________________________________________________________________________

Can't speak exactly for Terry, but I'm pretty sure he used the same tools I did. Since the Midas profile had no measurement data included in the profile itself, I used Gretag's free but unsupported ColorLab program to generate a new set of measurement data. It works basically by loading the text data from a standard CMYK profiling target and then converting that information to Lab using the Midas profile and Absolute Colorimetric rendering. ColorLab will then save out a new set of measurement data based on that conversion. While it's not exactly the same as using the original measurements, it seems to be so close as to be inconsequential. From there it's a simple matter to import that data back into ProfileMaker and generate a new profile with whatever ink limits and black generation you desire.

I don't bemoan the fact that these tools are not included in Photoshop. As an imaging professional, I just see owning a good spectro and a good profiling application are part of the basic toolkit. Yeah, there is a certain price to it all, but it also give you a competitive advantage that a lot of others do not have.

Peter Figen

www.peterfigen.com

___________________________________________________________________________

Comments in-line...

On Tuesday, November 06, 2007, at 02:44PM, "Peter Figen" wrote:

That's precisely how I did it (great minds think alike?!).

In terms of accuracy, there can be a bit of comprising of the "measurement" data when you convert using this method. Depending on the originator (author?) of the profile, it can be upwards of .25 dE. Doesn't sound like much but if you're using this method in a proofing environment, that level of compromise can be unacceptable because you have to assume the proofing system will compromise things even further. Before you know it, you've got a real 3 dE mess on your hands! :-)

As far as the Agfa ColorTune profile, I would expect an even higher degree of inaccuracy due to the lack of grid points in the A2B side of the profile.

Agree with that 110% Peter.

Regards,

Terry wyse

___________________________________________________________________________

Comments in-line...

On Tuesday, November 06, 2007, at 02:41PM, "Stephen Marsh" wrote:

Last I checked, both Monaco PROFILER and ProfileMaker use 17 grid points max on the A2B side (CMYK-to-Lab). In the other direction, B2A, Monaco defaults to 33 grid points but it's user selectable. ProfileMaker defaults to 25 ("Large Profile") but I think will drop to 17. Not sure on that because I always use the largest number of grid points for accuracy.

Yes, mostly the a* value. It's lower than the SWOP a*, indicating that the ink is heading in the direction of magenta. The L* is also telling us it's darker than SWOP, hence the "darker, dirtier and too purple" comment.

I'd bet dollars-to-doughnuts that the profile came from a Kodak Approval running Euro cyan material. Seen it before.

I usually default to using absolute since I'm usually evaluating the profile relative to proofing but using relative is a worthwhile excercise if you're primarily interested in how the profile performs for separations.

I'll confess that I didn't even bother looking at the actual numbers it produced going from RGB-to-CMYK. The fact that the perceptual rendering produceds something wildly different than other profiles doesn't surprise me but the relative+BPC numbers kind of surprise. USUALLY you see more consistent behavior with that intent regardless of where the profiles came from.

I'd love to take some credit but since I didn't *edit* the profile in any way, the credit goes to ProfileMaker and their "Colorful" gamut mapping option. Personally, my first choice would've been to use Monaco PROFILER but I would've had to jump through a few more hoops to make that happen.

Happy to. Beats listening to the personal wrangling that's gone on in recent months. I was THIS CLOSE to leaving this contentious list a couple of months ago. It's "fun" having a constructive thread for a change where I almost have something to contribute.

Eek. I can barely spell "v4 ICC profiles" let alone speak with any kind of authority on the subject. I've been mostly using v2 profiles only because many of the front-end RIP systems and proofing systems I have to deal with haven't been able to handle v4 profiles reliably. Ya stick with what works in my line of work. Having said that, I need to educate myself on v4 profiles and see if they offer any advantages in my work. Somebody smarter than I can pick up the "v4 ICC" discussion gauntlet and possibly educate us all!

Regards,

Terry

___________________________________________________________________________

On Nov 6, 2007, at 1:40 PM, Terence Wyse wrote:

OK, you made me go back and check and build profiles with each, because I was sure ProfileMaker used 33 grid points for large.

It goes like this with ProfileMaker:

"Default" = 11/25 grid points (A2B / B2A)

"Large" = 17 / 33 grid points (A2B / B2A)

I have always used Large as well. (The profile you uploaded does in fact use 17/33.) I don't own Monaco PROFILER, so I can't say what it uses.

--Rich Wagner

___________________________________________________________________________

I'll take those bonus points if nobody beat me to it. In Profilemaker you can pick your favorite profiling chart in the Reference Data drop down and then in the Measurement Data drop down, the second selection is "Import ICC Profile..."

I usually work with Monaco Profiler, so I don't know if you can then get the measurement

data out of the Profilemaker profile.

--

Bret Hesler

L.P. Thebault Company

Parsippany, New Jersey

___________________________________________________________________________

On Nov 7, 2007, at 5:06 AM, Bret Hesler wrote:

You can only do that if the profile you are importing has the requisite data embedded in the profile. I tried that with the Agfa generated Midas profile and came up with a missing tag error message. You get the same result with PrintOpen profiles.

Peter

___________________________________________________________________________

Terry writes,

Certainly as a color management consultant it seems undesirable for the general public to be able to edit existing profiles without the aid of a color management consultant. Similarly, portrait photographers used to believe that eliminating negatives was a bad idea because it seemed to them undesirable for the general public to be able to reprint existing work without the aid of a photographer. Prepress houses thought that Macintoshes were a bad idea because it seemed to them undesirable for the general public to be able to modify their own projects without the aid of a prepress house.

Even today, at the Javits Center in New York, somewhat restrictive rules exist because it seems undesirable to certain parties that anything should be plugged into an electric outlet without the aid of an electrician and an electrician's assistant.

Wait a minute. This particular PBP was generated without Adobe's assistance, no? It was causing a major problem in a paying job, right? It was editable in three minutes, right? The new SWOP profiles that you were recently recommending replaced PBPs with certain well-documented peculiarities/defects, right? Given the prevalence of PBPs, and their ability to mess up paid work, and the ease of repairing them in three minutes or so, is it all that unreasonable to want to do it without spending $2,500 or hiring an electrician (er, color management consultant)?

Saying that you don't want users to be able to modify PBPs because you're afraid somebody may make them even worse is like me advocating take curves out of Photoshop because so many people color-correct badly with them. Having these capabilities do NOT make it worse than it is in their absence.

To be sure, many people correct badly using curves. No matter how incompetent the execution, however, they almost always make the picture look better. Because even the incompetent know to do a Command-Z after they apply curves, and if they don't think they made the picture better, they conclude that they've wasted their time and they stick with the original. And if they *do* think they've made the picture better, it really doesn't matter what a professional retoucher thinks.

Similarly, if a person is sophisticated enough to believe that a certain PBP has a problem that he feels competent to attempt to fix, it is highly unlikely that the resulting new profile will be worse than the original. Because if the person making it doesn't think it's better afterward, he will conclude that he has wasted his time and goes back to using the original PBP. And if he *does* think it's better, it really doesn't matter what a color management consultant thinks.

I certainly agree that progress in the field has been glacial, just as I predicted it would be in 1998, unless it became possible for CMYK professionals to alter canned profiles for things like black generation, dot gain, total ink, and minor color changes, without spending $2,500 or hiring a color management consultant. So, yes. There has been significant progress over the last ten years--probably so much so that we have now reached the point we would have been at in Y2K had rudimentary profile editing been a part of Photoshop from 1998 onward.

Because color management consultants have been, shall we say, slow to grasp that they probably benefit more from the wide availability of profile editing than any other group, they have been reluctant to raise their voices to demand that Adobe make the trivial programming effort needed to include a rudimentary CMYK profile editor in Photoshop. If that reluctance continues, it can confidently be predicted that progress will continue--and that by, say, 2050, we'll be where we should have been in 2007.

Dan Margulis

___________________________________________________________________________

On Nov 6, 2007, at 3:12 PM, Terence Wyse wrote:

.25dE?

3dE mess?

I can appreciate the desire for precision, but for me comments like this have forever insinuated a level of precision that is unrealistic to expect on-press, real world.

Henry Davis

___________________________________________________________________________

Hello Dan,

I won't take the bait right now since I have to prep for a BSA father- son(s) camping trip this weekend. Mind if I take a rain check for next week? My two young sons (14 and 7) need me far more than me taking the time in writing a thought-provoking, insightful and witty response to your comments (I flatter myself, I know). Just pray that I don't freeze to death tonight camping in the mountains of North Carolina otherwise the world will be a worse place for not hearing my rebuttal to some of your silliness (I'm not saying "silliness" is a bad thing, by the way).

I just didn't want my silence for a few days to be interpreted as any kind of tacet agreement of your points and an "admission" that I had any kind of agenda in mind when I wrote my editorial comments about the possible inclusion of more advanced profile editing/creation tools in Adobe products.

I've already taken too much of my time writing this note stating that I don't have the time right now to properly "rebutt" you.

Get back to you soon!

:-)

Regards,

Terry Wyse

_____________________________

WyseConsul

Color Management Consulting

G7 Certified Expert

704.843.0858

http://www.wyseconsul.com

http://www.colormanagementgroup.com

___________________________________________________________________________ .

On Nov 9, 2007, at 8:41 AM, DMargulis wrote:

No, when looking at the effect of the general public having the power, but not the knowledge to properly "edit" profiles, it becomes very UNdesirable to add this functionality to Photoshop. We are finally getting to the point where there are good profiles available for common inkjet printers and papers, and good quality monitor calibration hardware, software, and profiles, and high-quality, standardized profiles (and data) for offset printing. Having people hack these profiles would not be good for color communication. And very few "color management consultants" routinely "edit" profiles. Rather, they GENERATE or BUILD new profiles using appropriate software packages. Ideally, this is done from ACCURATE colorimetric measurements of target data. This requires having the printed target to measure and an accurate spectrophotometer, as well as the software. So should Adobe bundle an Eye-One with proposed profile editing capability in Photoshop CS4, too?

No, the "poorly built profile" wasn't "editable in 3 minutes." As was clearly pointed out more than once, the "bad" profile was not edited at all. Rather, the bad profile was used to regenerate approximated measurement data, and that data was supplied to an expensive profile generation software package to build, de novo, a new profile. It would have been preferable to use the original measurement data to re-build the profile, but that was not available. The old profile was NOT edited in any way whatsoever, and profile editing software was never used. Both Terry (a color management consultant) and Peter (a very sharp photographer) used essentially the same tools and techniques to solve the problem. The tools are standard parts of their toolbox. ProfileMaker can be purchased for a fraction of the cost of a pro digital camera body or digital back, or many prime lenses. If you need the tool, you purchase it. Don't complain because Adobe won't bundle the same functionality that costs 3+ times the cost of Photoshop for free. There's a lot in a profiling package (http://www.xrite.com/ product_overview.aspx?ID= 794), and the least important part is "profile editing." Many people on this List have found this software to be a useful expenditure and an essential tool. If you feel that editing or creating profiles is essential, don't wait for Adobe. Just take out your credit card and buy ProfileMaker or PROFILER. And don't forget to pick up a spectrophotometer.