Re: Profiling Epson 9,500 v 10,000
Re: Profiling Epson 9,500 v 10,000
- Subject: Re: Profiling Epson 9,500 v 10,000
- From: Nick Wheeler <email@hidden>
- Date: Thu, 10 Jan 2002 17:40:07 -0500
Neil:
My first test on the 7000 was going (I hope IS someday soon) to be with
Joe's small gamut inks! The color shifting problem is uniform with the Lyson
neutral black inks on the 3000, I don't believe it is the same problem as
the Epson Pigment Inks. I suspect (but do not know) it is something about
the inks themselves, not how the ink is laid down on the paper. Lyson
alludes to this in their literature, saying they have balanced their inks
for daylight viewing.
I try to avoid that metamerism word as its use seem somewhat controversial.
I also really don't know what it means!
I bought the 3000 as a test bed for all these quad black inks. I originally
thought that I would send it CMYK postcript files (I did prove to myself
that the cheezy RIP they ship with the printer did operate as a true cmyk
setup, I could control ink limits, black generation etc in all four
channels, David Tobie and I had a little back and forth about that).
Initially I tried to build curves in cmyk (you can do it in the duotone mode
as well) that would print only the light gray in the 1/4 tones and then roll
in the other inks for the darker shades. By printing close to a 100% dot in
the 1/4 tones with light ink you could get rid of the visible dithering
there, then roll in the midtone colors and on to an extremely skeletal
black. This is what Jon Cone does and it should work well.
It is not a trivial project and I never really got what I wanted, the
problem is things get muddy in the 3/4 tones. I tried both MIS and Lyson
Inks and finally gave up in frustration. I eventually even bought the Cone
package but found they were really having similar problems in the 3/4 tone
area. There was also the additional problem of not being able to hit the
paper with that driver, and a lot of ink got splattered about the interior
of my poor old 3000. A further problem I had with the Cone driver was their
weaving or screening or whatever you want to call it produced a courduroy
effect I really didn't care for.
Anyway at that point I abandoned the approach completely. Lots of people
love the Cone system so we could be looking at operator error here. But
honestly, the images I was printing here just looked muddy in the 3/4 tones
no matter what I tried. And only about half the image got on the paper, the
rest would be splattered about the interior of the printer.
You know, many years ago I had a Conon with an early Colorbus RIP that I got
fantastic quadtones off of. I just copied Bruce Frasers curves and tweaked
them a little, it was a piece of cake. I don't know why this has proved to
be so difficult.
That's when I moved on to trying the 5500 in b/w. It was much better than
anything I had obtained using the cmyk approach with the quad black inks,
certainly commercially viable but still not artistically satisfying.
By the way - the 5500 problems have been all over the place. The first one
started making clicking noises then one day just groaned and died. Literally
made a horrible groaning sound. Epson has a policy here of sending out
replacement units, all four of which proved to be defective in some other
way. The most recent is a bizarre sort of oversharpening problem. Halos in
areas of high contrast and posterizatiion artifacts everwhere. Also problems
with paper feeding.
This is troubling, I wonder if their quality is slipping. I have a 5000 here
that has been a workhorse for years, no problems, none. It has printed
continously for days at a time with nary a hitch. But a friend who bought a
5000 recently is experiencing the same difficulties I am having with the
5500. Be careful I guess.
I digress yet again. Anyway I revisited the whole quad black question when
work slowed in November. A friend of mine here had told me he was having
great luck just printing with the Epson driver. He told me I was trying too
hard. He was right.
I figured they must have balanced the inks for this driver and sure enough
the first prints were dark but promising (I mean really dark). So I just
lightened the image with curves and tried again. A few go rounds this way
and I had something I really liked. 3/4 tones beautiful. Didn't look much
like the screen though.
The reverse engineering I mentioned in my previous note sounds more
complicated than it should, I'm sorry, I was just being lazy. I'm not too
good at explaining myself, I'll try again.
Once I got a print I liked, the trick was to somehow get the screen image to
match. I tried all manner of using profilemaker to make profiles, proof
setup blah blah get down funky chicken etc and nothing really worked. So I
thought I'm trying too hard again, why not dumb and simple like we used to
do, make the screen match the print!
So in photoshop I opened the final printed file and then in color settings I
just dialed back the gamma in RGB setup to about 1.2 or 1.4 using ROMM RGB
as my color space and bingo, screen matched the print. Save it as your
Quadblack profile. (Oh yeah, that ROMM RGB thing was part of what made the
5500 b/w come out so well come to think of it, conveting from ROMM to the
printer profile resulted in a much better b/w print than any other working
space I tried)
Once you have this profile you have to be careful, open the original scan as
untagged RGB and then do whatever you want to make it into a grayscale image
(channel mixer, desaturate, apply image, every image is different for me,
etc etc) then convert to RGB at which point all will be neutral. Now apply
the profile (apply, don't convert to). The image will shift drastically from
your default working space preview, and it is here that you can now do your
editing to get what you want. Make sure at the end to desaturate as large
corrections in RGB will introduce casts. You need to do this all in 16bit
obviously.
There are many other potential worflows, I have absolutely no idea what
might work best, just make sure you end up with a three channel image where
all channels are equal before you apply the profile, and apply the profile,
don't convert to. Remember you are forcing the screen to match the final
print not vice versa. Once you send the data through the Epson quickdraw
driver, all will be well.
I'm pretty sure you could do this with a grayscale profile as well and
convert to RGB (beware the conversion) before sending the file through the
driver, I just never tried it. Once I got this far I stopped and ordered the
7000 which is where I am stuck today. I would send you the profile I made
but of course I chucked it. But this is really simple to repeat.
Also check out the different media settings. Be careful not to lay down too
much ink and see if the black generation curves change more favorably one
way or another. Too much ink will clog the blacks no matter what.
I would definately recommend the 6 color approach for quadtones, this will
solve the dithering problem that we originally were fighting with. All the
cmyk curves and transfer functions ideas, while they work for those who use
them, simply don't make sense to me if you are using a driver that only
speaks in three channels, so something unseen and uncontrolled is happening
in the driver to take that four channel data and convert it to three. It
feels inelegant to me. I think the inks also are originally designed with a
three channel driver in mind, not to be used with a RIP.
In summation:
1) Start by just printing a 3 channel image where all three channels are
equal. If it is way off try to get it into the ballpark, just ignore the
screen display.
2) Then in edit color settings adjust your working space profile gamma so
the screen display matches the print. Save that edited workingspace profile
with some new name.
3) Now as your normal workflow you would open your image and get it to a
three equal channel state, apply the new profile to your image, edit the
result to taste and print. No conversions!
No doubt there will be some fine tuning and tweakings that have to take
place, but I really think this approach is worth a shot.
Best wishes,
Nick