Re: tools for building device links (was "Proper Gamut Mapping")
Re: tools for building device links (was "Proper Gamut Mapping")
- Subject: Re: tools for building device links (was "Proper Gamut Mapping")
- From: Ben Starr <email@hidden>
- Date: Mon, 17 Dec 2007 09:22:20 -0800 (PST)
Terry,
The Epson 4000 was linearized and calibrated using ORIS Color Tuner, as you said full gamut but well behaved.
The same source and destination profile was used for all device link profiles. On GCR I'm sure that the results could have been different had some more aggressive GCR been used in the source profiles. Considering that they all used the same two profiles with fixed GCR settings the results show that the different device link packages handle GCR differently.
As far as the purity tests go, this was done to test the capabilities of the device link profiling packages, profiles built with color preservation did result in large colorimetric discrepancies. In real production preserving pure color on an inkjet would not be done but for the purpose of testing the capabilities of the products this is a valid test as it represents what could be achieved in a press to press or press to proofer situation.
I'm sure there are things that could have been improved in this report such as noting the profiling package and GCR settings used to make the Epson 4000 profiles and I would have used different reporting for delta e going with a best 95% worst 5% instead of average and max. I may have also tried different methods of applying the device links, and investigated using a publicly available software for doing the Delta E calculations, although I can assure us that the software used was thoroughly tested and very sound proven in several other applications.
Let me know if you have any further questions.
Best regards,
Ben Starr
----- Original Message ----
Message: 12
Date: Mon, 17 Dec 2007 09:03:07 -0500
From: Terence Wyse <email@hidden>
Subject: Re: tools for building device links (was "Proper Gamut
Mapping")
To: "'colorsync-users?lists.apple.com' List"
<email@hidden>
Message-ID: <email@hidden>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed;
delsp=yes
Thanks Elie,
I know the author of the report and it would make sense that ORIS
Color Tuner was the RIP platform.
So when the author states that the samples were converted "to the
color space of the Epson 4000", does he mean its "natural" or raw
color space (full gamut), an ink-limited and linearized color space
(full gamut but "well-behaved") OR had the Epson 4000 been profiled
and iterated to match SWOPv2 (TR-001?) within ORIS and that's what the
author was referring to as the "color space of the Epson 4000"?
Not to take anything away from your products (I think they are
excellent and one of only two such products that I recommend) but I
think there's enough holes in that report to bring it under question.
Regards,
Terry Wyse
On Dec 17, 2007, at 3:16 AM, ELIE KHOURY wrote:
> Terry, I was not involved in the tests but I am nevertheless sure
> that a CMYK RIP was used to drive the EPSON. I believe it was CGS
RIP.
> Elie
>
> *****************************************
>
>
> On 17 déc. 2007, at 00:16, Terence Wyse wrote:
>
>>
>> On Dec 16, 2007, at 3:58 PM, Roger Breton wrote:
>>
>>> You need the link to the article in question. I have a hard copy
>>> which Elie
>>> distributed at the GATF CMS Conference, last week. Very
>>> convincing. So many
>>> ways to skin a cat (sorry Marco).
>>
>>
>> I believe there's more than a few flaws with this report although
>> this is the first time I've heard of it or read it. My comments
>> below are after a quick skimming of the report.
>>
>> The first thing I noticed was there didn't seem to be any mention
>> of what profiling package was used to create the destination
>> profiles used to build the links in the first place. The only
>> mention was of the source profile used, USWebCoated(SWOP)v2. A few
>> reasons why I believe this is important:
>>
>> 1) The "quality" or colorimetric accuracy of a device link profile
>> is somewhat subject to the raw profiles used to build the link in
>> the first place. I think some initial testing should have been
>> performed to ascertain what combination of source/destination
>> profiles introduced the least colorimetric error into the profile
>> chain. I suppose as long as all the device links were built using
>> the same pair of source/destination profiles, maybe this point is
>> moot but I can imagine that some device link applications would
>> perform better using one pair of source/destination profiles vs.
>> another, even though the profiles could've been created from the
>> same data set. As noted in the report, differences of .50 delta E
>> or less are probably meaningless. In the colorimetric comparison, I
>> think a single max dE is meaningless as well (a single poor max dE
>> value could be printer variation as much as the device link).
>> Average dE is OK, but I'd rather see a "worst 10%" value. How a
>> profile conversion handles the 10-20% most difficult colors/patches
>> would be very telling I think, even more so than average dE.
>>
>> 2) Ink consumption report. I think it's hardly fair to compare
>> various device link applications that rely on the destination
>> profile for "ink savings" (GCR) as opposed to those link apps that
>> can override the destination profile separation settings. Granted,
>> having the ability to alter the ink savings/GCR directly within the
>> device link application is a nice feature but that doesn't
>> necessarily mean that other device link apps could not have
>> improved upon their ink consumption figures if the destination
>> profile had used more "aggressive" GCR settings. I've done a bit of
>> evaluation of ink savings myself and can tell you that the K
>> generation used in the destination profile has quite an impact on
>> the resulting ink savings. But to be fair, the ink savings that
>> result from those device link apps that offer GCR controls appears
>> to be substantially greater than those link apps that relay soley
>> on the destination profile's separation settings, at least in my
>> testing.
>>
>> 3) "Purity" tests. Apparently this was tested using an Epson 4000
>> as the output device but using USWebCoated(SWOP)v2 as source. As we
>> all know, the pure inks of almost any inkjet device that comes to
>> mind can be anywhere from slightly different to wildly different
>> than an offset press. Testing the preservation of pure ink colors
>> from an offset press profile to an inkjet printer would result in
>> fairly extreme colorimetric errors I would imagine. I fail to see
>> where a test of preservation of pure ink solids or tints using the
>> parameters outlined in the report would have any meaning at all.
>> Preservation of pure inks is really only relevant when repurposing
>> from press-to-press. There's precious few instances where
>> preserving pure inks from a press to an inkjet device is desirable.
>>
>> Other questions:
>>
>> The author of the report only stated that the conversion was
>> performed to the "color space of an Epson 4000" via the Apple CMM.
>> Does this mean the entire test was NOT peformed via a CMYK RIP but
>> was instead printed via the Epson OS driver? If so, then how was
>> this testing CMYK-to-CMYK conversions since an Epson 4000 profiled
>> via the driver must be done in RGB? A better test perhaps would've
>> been to perform this via a CMYK RIP that had some basic
>> linearization and ink limiting features but where the conversion
>> itself was performed in Photoshop instead of the RIP. I can imagine
>> performing such a test with, say, a ColorBurst RIP or even a GMG
>> RIP which offers excellent calibration features but can have
>> profile conversions disabled. This would be a good platform from
>> which to generate very stable/consistent output from the inkjet
>> printer.
>>
>> The colorimetric comparisons and "round-tripping" of profiles was
>> performed using an application not publicly available as far as I
>> know. Knowing the source of this application, I'd be reasonably
>> sure that it's a high quality evaluation software but it prevents
>> any sort of independent verification of the tests results. I'd be
>> much more comfortable if the author had used "recognized"
>> applications such as ColorThink Pro, LOGO ColorLab, X-Rite Measure
>> Tool, etc. I would've preferred that the author had used a more
>> "mainstream" application rather than a custom application that
>> could've quite possibly introduced errors of it's own.
>>
>> Regards,
>> Terry Wyse
>>
>>
>> _____________________________
>> WyseConsul
>> Color Management Consulting
>> G7 Certified Expert
>> email@hidden
>> 704.843.0858
>> http://www.wyseconsul.com
>> http://www.colormanagementgroup.com
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden