Epson 2200 even produces good results on plain paper?

[Greg]

Flycaster - here's a profiling service which does pre-linearization for
their "premium" profiling service:
http://www.profilecity.com/profilekit1p.asp

Greg.

 

[Flycaster]

Flycaster - here's a profiling service which does pre-linearization for
their "premium" profiling service:
http://www.profilecity.com/profilekit1p.asp

Interesting. I've used these guys in the past for a 1280 profile, and used
that "linearizaation" test print they have to determine which media setting
provided the best ink coverage and "neutrality" for the paper I was
profiling (a Red River product, if memory serves me.) At the time, they did
not offer to measure that target and incorporate it into the final profile.

In the end, you will still end up with a RGB profile that attempts to
correct for grossly "un-neutral" hybrid-RGB driver-CMYK device. In that
regard, it is no different from an ordinary ICC profile built from
"unlinearized swatches"...but I'm willing to wait and see what you come up
with.

I love pleasant surprises.

 

[Greg]

In the end, you will still end up with a RGB profile that attempts to
correct for grossly "un-neutral" hybrid-RGB driver-CMYK device. In that
regard, it is no different from an ordinary ICC profile built from
"unlinearized swatches"...but I'm willing to wait and see what you come up
with.

This is *only* true if the profile test chart has enough patches to
characterise the
non-linear response of the printer.

For example, just say the chart has code steps of 128, which is extremely
coarse.
Despite being very coarse, it will produce a better profile if the response
of the printer
is linear to start with, than a profile of a highly non-linear printer. This
is the reasoning behind pre-linearization -
it fills in the gaps (to an extent) that the profiler simply can't model
with a given test chart. Thus,
by measuring stepped wedges with a much finer step than 128, and creating a
pre-linearization curve,
it should allow the profiler to do a better job.

If I had a profiler which could directly make use of, say, a, 4000 patch
target, I wouldn't
dream of doing pre-linearization for my 2200.

I've created my own pre-linearization curves now, and will create another
profile soon.

Note that the first profile I made with pre-linearization was done by
converting a greyscale
wedge into the Epson supplied Premium Semigloss ICC profile, and noting down
the new pixel
values. This is a quick/easy way to get a pretty good pre-linearization
curve, I think.

Greg.

 

[Flycaster]

This is *only* true if the profile test chart has enough patches to
characterise the
non-linear response of the printer.

For example, just say the chart has code steps of 128, which is extremely
coarse.
Despite being very coarse, it will produce a better profile if the response
of the printer
is linear to start with, than a profile of a highly non-linear printer. This
is the reasoning behind pre-linearization -
it fills in the gaps (to an extent) that the profiler simply can't model
with a given test chart. Thus,
by measuring stepped wedges with a much finer step than 128, and creating a
pre-linearization curve,
it should allow the profiler to do a better job.

If I had a profiler which could directly make use of, say, a, 4000 patch
target, I wouldn't
dream of doing pre-linearization for my 2200.

I've created my own pre-linearization curves now, and will create another
profile soon.

Note that the first profile I made with pre-linearization was done by
converting a greyscale
wedge into the Epson supplied Premium Semigloss ICC profile, and noting down
the new pixel
values. This is a quick/easy way to get a pretty good pre-linearization
curve, I think.

What profile building software and test charts are you using with your
Colorsavvy CM2C colorimeter?

 

[Greg]

Colorsavvy CM2C colorimeter?

Pantone Colorvision Profiler Pro V2.3. The largest chart it has is 729
patches, and that's what I'm using.
The patch image data changes by steps of about 32, which is not fine enough,
especially in the shadow tones.

Greg.

 

[Flycaster]

Pantone Colorvision Profiler Pro V2.3. The largest chart it has is 729
patches, and that's what I'm using.
The patch image data changes by steps of about 32, which is not fine enough,
especially in the shadow tones.

IMO, that depends on the paper and inks. Gretag uses either a 918, or a 244
swatch target and, from my experience, the true "photo" papers (ie, Luster,
PPG, and SG) do just fine using only the 244 swatch target. It's the other
papers that really seem to benefit from the addtional LUT adjustments.

It'll be interesting to see if your profile has smoother color and greyscale
ramps (not to mention cast free greyscales.) Getting neutral B&W prints
from any of these printers is a real bitch without RIP, or dedicated inks.

 

[Greg]

It'll be interesting to see if your profile has smoother color and greyscale
ramps (not to mention cast free greyscales.) Getting neutral B&W prints
from any of these printers is a real bitch without RIP, or dedicated inks.

I'll post some results. I expect only a subtle improvement, and mainly in
the darker
shadow tones. If you like I can send you the pre-linearization curves that
I'm using right now,
so you can start doing your own testing. Let me know. (I can be emailed on
(e-mail address removed) )

I've just done a preliminary measurement of a few of the pure green
patches - the curve
deviates again, however it's not as bad as that first deviation. I have a
feeling my brand new
2200 printer might need to be serviced.

Greg.

 

[Flycaster]

E-mail sent.

 

[Greg]

I wrote, regarding pre-linearisation before profiling:

I'll post some results. I expect only a subtle improvement, and mainly in
the darker
shadow tones.

I'm not sure whether anyone else is interested in this topic of
pre-linearisation, but I said
I'd post some results, so here they are.

So far, my attempts at improving the profile by pre-linearising the printer
response before
profiling have failed. Rather than improving the shadow tone linearity, I
feel that it has
made it worse. Not only that, it has introduced a cyan cast, mainly in the
shadows.
The profile is still pretty good (real photos look fine, to me), but I don't
think pre-linearisation
has gained me anything - it has produced a slightly worse result, not
better.

So at the moment, I feel I owe Flycaster an apology - I have been quite
adamant that pre-linearisation
would improve the result, but was wrong.

If I ever learn how to do pre-linearisation properly, and actually manage to
improve the result, I'll
report back. But don't hold your breath. ;^)

Greg.

 

[Flycaster]

[snip]

So at the moment, I feel I owe Flycaster an apology - I have been quite
adamant that pre-linearisation
would improve the result, but was wrong.

If I ever learn how to do pre-linearisation properly, and actually manage to
improve the result, I'll
report back. But don't hold your breath. ;^)

Greg, you owe me no apology whatsoever. I didn't think it would work, but
as I pointed out I was prepared to be pleasantly surprised. For anyone else
who might bother to look at this thread, Greg and I have exchanged a number
of e-mails and (I hope) this exercise had been as fun for him as it was
informative for me. Never up, never in. Nonetheless, I still believe that
linearization that would *bypass* the Epson driver is the ticket - which, to
be sure, is similar to what good RIP software would provide.

Color management and profiling are *tremendous* improvements, but they are
dependant on the accuracy and calibration of the devices being profiled.
The 2200 is a great printer, but as Greg and I (and others) have discovered,
the Epson RGB driver appears to be one of, if not the principal impediment
to "dead-on" profiling. That said, this printer can be profiled more than
adequately for just about anybody's purposes, even for critical observers.