Thanks Martin—
I’ll give this method a shot as soon as I can— seems totally logical to me—
On Aug 7, 2020 at 10:37 PM, <Martin Gray (mailto:mdgray@xxxxxxx)> wrote:
Hi Cody,
I had something similar come up a few years back. Had made a print for
someone on my Canon 9500 II. It was a quick print from their sRGB image
letting the printer manage colors. They then wanted a larger version but
with as close as possible to an exact match of the colors that the 9500 II
had printed.
This turned out to be difficult because my larger printer is an Epson 9800
and the 9800 printer produced notable less saturated prints. Now normally I
print with color managed workflow, not the quick and dirty way I had
originally made the guy’s print.
So I investigated what was going on with the 9500 and it turned out that the
perceptual mode letting the printer manage colors greatly increased
saturation in many areas. Also, many were substantially increased in
brightness. Over 10 dE in many areas and even well outside sRGB’s gamut.
Apparently this is done by the 9500’d default to create more pleasing,
colorful images from typical snapshots.
So, having printer profiling hardware, I found a solution which is to
characterize the 9500 II’s default gamut mapping by making a special purpose
profile. It worked perfectly. Here’s what I did.
I created a set of patches and tif files as normal to profile a printer but
instead of printing w/o color management, I assigned the patches to sRGB.
Then printed the targets on the Canon 9500 II using the exact settings
(printer manages colors) used when I had initially printed the guy’s image.
And then from this printed target I made a profile I called
“simulation9500.icm”
Then, with a copy of the original image the image was assigned to
“simulation9500.icm” Note that this profile is not and should not be used
for normal printing. After assignment then the image should be converted to
the desired colorspace using Relative Colorimetric without BPC. I used
ProPhoto RGB but any standard RGB colorspace can be used if the colors don’t
exceed it. Adobe RGB (1998) would probably have been just fine.
Now, the image is printed using using standard color managed workflow with
Relative Colorimetric AND a custom Epson 9800 profile made by the same
spectrophotometer/software. The result was a match. At least for all colors
that are within the 9800 gamut
Give it a shot.
From: argyllcms-bounce@xxxxxxxxxxxxx <argyllcms-bounce@xxxxxxxxxxxxx> On
Behalf Of Cody Ranaldo
Sent: Wednesday, August 5, 2020 10:12 AM
To: Argyllcms <argyllcms@xxxxxxxxxxxxx>
Subject: [argyllcms] Re: Visual Match of Two Printers
Thanks everyone for your responses-- I've experienced essentially the same
thing as Graeme when it comes to having multiple printers of the same type
being run through the same RIP-- Especially when you are defining ink
limits individually. One thing I didn't go over in my original post was
that I was able to calibrate an Epson 9880 using the same everything as the
11880 I was trying to match (except for the spectro) and I was able to get a
very close match. I don't doubt that it would have been even better had I
been using the exact same measuring device to profile both printers.
I guess what I'm ultimately looking for is something slightly different, and
it relates back to the first time I posted about this: some kind of way of
fingerprinting a printer's current output "style" to simulate it on another
printer-- regardless of RIP, CMM, ink limiting setup, age of profile, or
printing color mode (RGB, CMYK, N-color..) etc.
A situation I find myself in often is receiving a small test print from a
client, who loves the color of what they were able to obtain on their home
setup, but they know nothing about color management. Most likely they are
using a canned profile, and maybe not even the correct one for the correct
paper. Then they want me to be able to exactly match that print in order
to make a larger version of it. This generally leads to a large amount of
back and forth (depending on how anal the client is) and a lot of manual
tweaking of colors in the file and many iterative prints to get to a closer
match. The situation I'm in right now is similar-- only in that the
printer being used to make my match print is an older model than the one I
have currently, and the client is exacting to the point of near-insanity.
I've had clients before reject prints because they were printed on the left
side of the printer as opposed to the right, and I've had clients complain of
differences in two prints that came out immediately after one another in the
same exact spot seconds apart (THAT particular conversation is one of
psychiatry).
But I've been doing a lot of thinking about this, and it seems like something
like this should be possible essentially using the tools we use for
calibrating input devices (cameras and scanners). Essentially, a target
having known values, like a colorchecker only with far more patches would be
printed using whatever color management setup is present, which could then
generate a 3D LUT or abstract ICC profile which could be applied to the file
manually, even as a layer in photoshop, which would then alter the colors in
the image before going through a tightly calibrated system. As long as the
gamut of the reference system is sufficiently larger than the target, or the
images are not containing colors that push the boundaries of the gamuts, it
should be possible to simulate the conditions of any machine in any specific
state...
The main issue I think I've been dealing with concerning the p20000 and the
11880 is the black depth of the matte black ink-- the images I'm printing
have very deep blacks which are shown to be out of gamut, and the two
printers have different min L* capabilities on matte paper.
When I get the chance I'm going to experiment using the input calibration
tools in argyll to see if I can make this into something workable.
-------
Profile Printer A, and profile Printer B using an instrument you trust.>Create a proofing transform (i.e. device link ?) from printer A to
>printer B. Print (or color transform) your images as if you were printing
>to printer A, but save the output to a raster, and then transform
>it to Printer B space using the proofing transform. Print on
>printer B using no color management.
>An alternative to creating the proofing transform is simply to take
>your Printer A raster from above and print it to printer B using the
Printer A profile
>as the source profile (and printer B's profile as the destination profile).
>Use relative colorimetric intent.
For whatever reason, when I've tried these methods I've never gotten anything
to shift in an appropriate manner. Most likely because I'm not using the
same measuring equipment. For example, one image has a slight magenta cast
I'm trying to emulate, but in no scenario do the colors shift over to
magenta. This is the reason I was trying to do the printing "through" the
profile-- I had the client assign their profile to the color chart before
outputting it, in the hope that when I measured that chart on my end I would
get some insight into what the profile was doing, but I must admit I was kind
of lost in the significance of what I was doing... I think theoretically the
colors should be the same as they would if printed with no color management?
Also having no reference for those values, I don't see how I could use them
to any reasonable effect...
On Sun, Aug 2, 2020 at 12:47 AM Graeme Gill <graeme@xxxxxxxxxxxxx
(mailto:graeme@xxxxxxxxxxxxx)> wrote:
graxx@xxxxxxxxxxxx (mailto:graxx@xxxxxxxxxxxx) wrote:
> The other times I tried to use the same technique, the results were
not very good.
> Break down of the theory? I'll never know. But it was pure hell.
Depends a bit on how the printers are set-up, and how consistent the inks
are.
If the inks change (different batches, or pigment settling), the it's
hard to keep them consistent, short of re-profiling all the time.
We ran Epson and similar printers in proofing situations all the time
back in Colorbus days, and they matched very well, even using the same
"profiles" (actually the usual proofing setup used device links).
The secret was to have a reliable per channel calibration system, one
that had specific maximum density targets for each channel. The customers
would do a quick 4 channel density calibration each day, and the results
consistently met expectations. Even worked well for our newspaper/magazine
customers who were proofing on colour copiers, using "off the glass"
(i.e. copier scanner) measurement of the calibration charts that I
developed.
Cheers,
Graeme Gill.
