QTR-Quadtone RIP

I am trying some odd things with inks that lay down a great deal of ink.
Don't ask why <G>.

Two questions-

Is linearization iterative? In other words, will additional
linearizations over a previous fine tune it, or will any linearization
always be a correction from the "native state"?

If one enters a gamma of something other than 1 in the gray curve
section to initially bring things more in line, will linearization
correction include that gamma assuming the setting remains the same
from there on out?

Thanks,
Tyler

Hi Tyler.

I'm a little less proficient in QTR than I am with IJC/OPM, but I am 
pretty sure they work the same way in regards to linearization.  RIPs 
like StudioPrint allow you to multiple linearizations, one after 
another, but I know IJC doesn't work that way, and I am reasonably sure 
QTR doesn't either.

When you linearize, you are always doing it based on the raw curves 
built into the profile.  So relinearizing always goes back to raw data, 
or native state.

The gamma is a bit more arcane to me.  In IJC, I know you can set a 
gamma target point using a slider during profile creation to lighten or 
darken the midpoint of your grayscale.  I am assuming the gamma in QTR 
does the same thing.

Lou

--- In QuadtoneRIP@yahoogroups.com, "Tyler Boley" <tyler@t...> wrote:
> I am trying some odd things with inks that lay down a great deal of 
ink.

> Don't ask why <G>.
> 
> Two questions-
> 
> Is linearization iterative? In other words, will additional
> linearizations over a previous fine tune it, or will any linearization
> always be a correction from the "native state"?
> 
> If one enters a gamma of something other than 1 in the gray curve
> section to initially bring things more in line, will linearization
> correction include that gamma assuming the setting remains the same
> from there on out?
> 
> Thanks,
> Tyler

--- In QuadtoneRIP@yahoogroups.com, "Louis Dina" <lbdina@c...> wrote:
> Hi Tyler.
> 
> I'm a little less proficient in QTR than I am with IJC/OPM, but I am 
> pretty sure they work the same way in regards to linearization.  RIPs 
> like StudioPrint allow you to multiple linearizations, one after 
> another,

Right, I can start out with something pretty wacky, and bring it into
line after a coarse initial linearization. But then one would hope so,
for the price.

> but I know IJC doesn't work that way, and I am reasonably sure 
> QTR doesn't either.

I didn't really think so, but it was worth asking.

> When you linearize, you are always doing it based on the raw curves 
> built into the profile.  So relinearizing always goes back to raw data, 
> or native state.

Got it.

> The gamma is a bit more arcane to me.  In IJC, I know you can set a 
> gamma target point using a slider during profile creation to lighten or 
> darken the midpoint of your grayscale.  I am assuming the gamma in QTR 
> does the same thing.

I'll play with that a bit, thanks.
Tyler

Right now you can only do one linearization. Roy might correct me on 
this, but looking at my Quadprofile source code I don't think there 
would be anything to prevent me from building in a few more iterations. 
(Quadprofile is only gluing together various components made by Roy so I 
don't always have a complete overview of how everything works).

Linearization is applied after gamma, shadow and highlight. You're free 
to set those to anything you like before you do the linearization.

--
Daniel Staver
http://daniel.staver.no


Tyler Boley wrote:

> I am trying some odd things with inks that lay down a great deal of ink.
> Don't ask why <G>.
> 
> Two questions-
> 
> Is linearization iterative? In other words, will additional
> linearizations over a previous fine tune it, or will any linearization
> always be a correction from the "native state"?
> 
> If one enters a gamma of something other than 1 in the gray curve
> section to initially bring things more in line, will linearization
> correction include that gamma assuming the setting remains the same
> from there on out?

Hi Tyler,

As things stand Linearization is a one-shot operation.  I.e. you always apply it
to the same "raw" output. However the gamma and (highlight/shadow) values
are applied first so you can get closer to linear before getting to the final
linearization.  These help a whole lot because the 21 steps are more evenly
spaced out.   

I've thought about allowing iterative linearizations or possibly linearizations
at other levels i.e. gray or toner linearizations.  How valuable do you think
this would be versus how more complicated it would be?

Roy


--- In QuadtoneRIP@yahoogroups.com, Daniel Staver <daniel@p...> wrote:

> Right now you can only do one linearization. Roy might correct me on 
> this, but looking at my Quadprofile source code I don't think there 
> would be anything to prevent me from building in a few more iterations. 
> (Quadprofile is only gluing together various components made by Roy so I 
> don't always have a complete overview of how everything works).
> 
> Linearization is applied after gamma, shadow and highlight. You're free 
> to set those to anything you like before you do the linearization.
> 
> --
> Daniel Staver
> http://daniel.staver.no
> 
> 
> Tyler Boley wrote:
> > I am trying some odd things with inks that lay down a great deal of ink.
> > Don't ask why <G>.
> > 
> > Two questions-
> > 
> > Is linearization iterative? In other words, will additional
> > linearizations over a previous fine tune it, or will any linearization
> > always be a correction from the "native state"?
> > 
> > If one enters a gamma of something other than 1 in the gray curve
> > section to initially bring things more in line, will linearization
> > correction include that gamma assuming the setting remains the same
> > from there on out?

--- In QuadtoneRIP@yahoogroups.com, "Roy Harrington" <roy@h...> wrote:
> 
> Hi Tyler,
> 
> As things stand Linearization is a one-shot operation.  I.e. you
always apply it
> to the same "raw" output. However the gamma and (highlight/shadow)
values
> are applied first so you can get closer to linear before getting to
the final
> linearization.  These help a whole lot because the 21 steps are more
evenly
> spaced out.   

Thanks Roy, I'm going to give that a try. Makes sense to have these
controls do some of the work before linearization under some
circumstances.

> I've thought about allowing iterative linearizations or possibly
linearizations
> at other levels i.e. gray or toner linearizations.  How valuable do
you think
> this would be versus how more complicated it would be?

I think it would be user dependant. Obviously many are up and going
with no problems without it using a number of ink configurations, and
it's going a fine job. People using sets with toners have not seen the
need for the toner to be linearized on it's own I assume?
I would call it an advanced feature, and maybe one that could be left
off as a default. It could be a very useful tool for some,
particularly trying unusual things, or unusual inks or media.
On the other hand, you have to consider the level of control you are
offering for such a low price, and the level of support such features
might require.
It did occur to me that something like this would be useful for what I
was just trying. But I have to admit, the more control options I get,
the more trouble I get myself into, and wind up asking questions...
It's also occurred to me on many occassions that this code could be
turned into a great CMYK color driver, with individual color
linearization, etc.. I don't know if it's possible, but the level of
control QTR has lead me to think it might be. Of course many issues
like GCR would have to be handled in profiling, so mostly for people
experienced in CMYK and color management.
I'm actually sort of suprised someone hasn't already brought it out.
Many more color users than B&W users. There was a beta 5 or something
last year that had all kinds of controls as an RGB driver, but the
developers speak a completely different language than the rest of us,
and working through all the options became way too tedious.

Great stuff Roy, still trying some things with it...
Tyler

Hi Roy, Tyler, Daniel,

Just some thoughts about linearization, etc.

I have generally had very good luck with both quad inksets and more
recently UC with the Paul Roarks PKN and LKN. While maintaning the
ability to print color I can get a neutral BW with minimal toning 
by the LC and LM.

It has always amazed me just how well the single linearization does
given that there are many nonlinearities that the process would seem to
ignore.  In general, it is just great.

For some reason the PKN on Luster paper builds up density in a very
nonlinear manner. The density curve is very steep, then flattens 
out. But you don't want to ignore that top part of the curve by setting
an early ink limit because it means the difference between a dmax of 
2.0  vs 2.3.  The LKN has a much more gradual slope throughout. As 
such, when linearization scales down the output values,
which it must do strongly for PKN, the effect on density contributed
by the different inks is not proportional. Thus, there is some error in the
final density.  For this situation it would be nice to linearize 
this single ink before any partitioning is done. (This is my 
understanding, consistent with experience-please correct me if I 
am wrong).

Perhaps more of a challenge is that density gets really wacky at the
dark end for many inks, papers, and printers. I have had a lot of trouble
with getting  good uniform separation above 95%. Linearization as it 
stands is implemented by fitting a curve to the whole density function, 
then scaling back the output values to fit an "ideal"  (perceptually uniform) 
curve.  Unfortunately, the heavy ink loads refuse to behave nicely, making 
it very hard to get a good curve fit.  Wiggles in the curves result. For this 
reason using a 21 step linearization  often produces a smoother curve than
 the 51 step. A second linearization at the end would  help a bit here, but 
would not  take out isolated bumps in density.

I have been  thinking that an approach would be to make and measure a 
test chart with  many dense values from say 95-100% made up of
different output values of the darkest inks. A program could then build 
up the curve that  best  produces the desired densities rather than trying
to deduce the closest single scaling factor. I'll do this if I can find the time.
 
Short of this, experimenting with the shadows and highlights helps a
huge amount. I just find the trial and error of printing many curves with
different values to get a less bumpy starting point more time and effort
than if we just had an (optionally) more involved linearization process.

just my two cents.

Costa Colbert

--- In QuadtoneRIP@yahoogroups.com, "Roy Harrington" <roy@h...> wrote:
> 
> Hi Tyler,
> 
> As things stand Linearization is a one-shot operation.  I.e. you
always apply it
> to the same "raw" output. However the gamma and (highlight/shadow)
values
> are applied first so you can get closer to linear before getting to
the final
> linearization.  These help a whole lot because the 21 steps are
more evenly
> spaced out.   
> 
> I've thought about allowing iterative linearizations or possibly
linearizations
> at other levels i.e. gray or toner linearizations.  How valuable do
you think

> this would be versus how more complicated it would be?
> 
> Roy
>

Hi,

I've posted an experimental cool curve for UC with PK ink on Epson 
Premium Semigloss in the files section. I know there's been some 
requests for this.

\ufffd'm using the MIS archival 7600 inkset with MIS PK (not PKN) in a 2100 
so there may be differences between my setup and one using Epson inks in 
a 2200. Nevertheless, this curve should be much more neutral/cool than 
the original one, which was actually quite brown and not very neutral at 
all.

I cooled the curve down by adding LM in Toner 1 and LC in Toner 2 with a 
50% ink limit for LC and 38 for LM.

Let me know what you think. The curve is not linearized yet, but has 
good separation between all the steps on my printer.

--
Daniel Staver
http://daniel.staver.no

--- In QuadtoneRIP@yahoogroups.com, "ccolbertbw" <ccolbert@u...> wrote:
...
> For some reason the PKN on Luster paper builds up density in a very
> nonlinear manner. The density curve is very steep, then flattens 
> out. But you don't want to ignore that top part of the curve by setting
> an early ink limit because it means the difference between a dmax of 
> 2.0  vs 2.3.

Costa, I haven't had time to finish my testing, but I have learned a
few things so far. You are right about the aparent lack of sense using
ink limits to initially bring things in line, therefore losing dmax.
But, it's the way QTR seems to work best, and Roy designed in K boost,
which allows you to bring all the way back to native unlimited output
in that ink (if needed), so you can "unlimit" there. For straight
quads, there is way too much ink going down to get effective
partitioning unlimited.

Until I learn otherwise, it seems the best route is to bring things
well into line with limits, gamma, highlight, and shadow controls,
then bring dmax in with K boost (and maybe a bit of overlap), and
linearize.

Tyler

--- In QuadtoneRIP@yahoogroups.com, "Tyler Boley" <tyler@t...> wrote:
> --- In QuadtoneRIP@yahoogroups.com, "ccolbertbw" <ccolbert@u...> wrote:
> ...
> > For some reason the PKN on Luster paper builds up density in a very
> > nonlinear manner. The density curve is very steep, then flattens 
> > out. But you don't want to ignore that top part of the curve by setting
> > an early ink limit because it means the difference between a dmax of 
> > 2.0  vs 2.3.
> 
> Costa, I haven't had time to finish my testing, but I have learned a
> few things so far. You are right about the aparent lack of sense using
> ink limits to initially bring things in line, therefore losing dmax.
> But, it's the way QTR seems to work best, and Roy designed in K boost,
> which allows you to bring all the way back to native unlimited output
> in that ink (if needed), so you can "unlimit" there. For straight
> quads, there is way too much ink going down to get effective
> partitioning unlimited.
> 
> Until I learn otherwise, it seems the best route is to bring things
> well into line with limits, gamma, highlight, and shadow controls,
> then bring dmax in with K boost (and maybe a bit of overlap), and
> linearize.
> 
> Tyler

Tyler,
   As the controls exist now, I agree completely with your assessment.
My  comments were mainly looking forward towards ways of making
the last bit of fine tuning less fiddly.  The boost does offer a pretty 
good fix for the dmax. It is easy to get a  99-100%  black with good
dmax. For me, it is more difficult to get  a smooth well-linearized 
transition from 95  or so up to that final dmax.  I have made many
curves that looked great except for weirdness in the deep shadows
that result in  black blobs where the density abruptly jumps to the
final dmax. I haven't found this jump easy to fix with either the
shadow controls or the final linearization. 

I absolutely love QTR. It works in a logical, coherent, direct manner. 
There are few surprises, because you can predict what it will do. It 
also ends up with a table (the profiles) that you can print out yourself 
and see what is happening.  Having all the source code, you can see 
what is going on. Its simple and direct.

I just find it a bit frustrating that the last little bit of fine tuning is
fiddly and would like to figure out a way to make it less so.

best,
Costa

--- In QuadtoneRIP@yahoogroups.com, "ccolbertbw" <ccolbert@u...> wrote:
...
> For me, it is more difficult to get  a smooth well-linearized 
> transition from 95  or so up to that final dmax....

> ...Having all the source code, you can see 
> what is going on. Its simple and direct.

I wish I had the knowledge to work with it at that level.

> 
> I just find it a bit frustrating that the last little bit of fine
tuning is
> fiddly and would like to figure out a way to make it less so.

Regarding this and the first comment, I think more
measurement/correction points in that area would help. With
StudioPrint, I will see differences in linearizations between 2
different papers between 95% and 100% untill I do an 80 patch
linearization. Iterative linearizations would not help this if there
are no sample points in that area.
Obviously this is a rare problem, and 21 steps is working well for
most situations. I did run into very similar results once though with
QTR on a particular ink setup, and I'm going to start from scratch on
that one.
Otherwise, I'm not sure what the solution would be. Start with heavy
limiting so things are behaving, then creep up the K boost?

Tyler

That last 5% from 95 to 100 seems to be the most prone to difficulties.
The various paper and ink combinations seem to react differently when 
they
get close to overloading.  I haven't found a consistent way to predict 
what's
happening there.

As you know the three main variables that you can experiment with are
BOOST, LIMIT and OVERLAP.  I think there's a tendency to use too much
boost.  It brings the extra black in very steeply at the end and can 
give the
jumps in density.  I would typically use only 10% to 15% more that the 
ink
limit.  For instance if LIMIT=60 use between BOOST=66 and BOOST=69.
Something like 80 or 90 is going to be too much.   Another idea if you 
really
want to increase black ink is to up the LIMIT as well as the BOOST.  
What
this does is increase the amount of black for its whole range rather 
than
just at the end (which the BOOST does).  So in the example rather than
LIMIT=60 and BOOST=72 which will be steep at the end, use LIMIT=66
and BOOST=72.   It will spread out the black increase.

The other big issue is what happens to the gray inks at the end.   In 
general
the amount of gray plunges rapidly to allow the increase in black 
without
overloading the paper.  Sometime the extra gray seems to increase dMax 
and
sometimes decrease it.  While the OVERLAP effects the whole range it
has a particlar effect at step 100.  It basically gives the amount of 
residual
gray ink still being output.  OVERLAP=0 means no gray at all at 100,
OVERLAP=20 will mean 20% of gray ink is still output.   I've found that 
sometimes
keeping a little gray at the end controls the steep black -- i.e. 
improves the
linearity.  OVERLAPS of 10 to 20 seem appropriate.

You really have to see what's happening on the particular ink/paper 
combo
to see what helps.  None of these ideas are absolute recommendations.

Finally there's adding more steps in the linearization.  Usually 21 is 
fine, but
the 51 can be used.  You can actually use any number of steps as long as
they are evenly spaced.  The "evenly spaced" can be an issue once you 
get
higher up because of the 8-bit gray values.  E.g. 100 steps would have 
some
steps of 2 and some of 3.  You can fix this in two ways.  One is to 
pick the right
number of steps -- 86 steps is exactly 3 units at a time.  The other is 
to use a
16-bit gradient (PS CS only), posterize, and then convert to 8-bit.  PS 
will
dither the intermediate values.  The other thing you have to be careful 
about
with a lot of steps is that the MUST be increasing -- no dips allowed :)

Roy


On Saturday, May 7, 2005, at 08:27  PM, Tyler Boley wrote:

> --- In QuadtoneRIP@yahoogroups.com, "ccolbertbw" <ccolbert@u...> wrote:
> ...
>> For me, it is more difficult to get  a smooth well-linearized
>> transition from 95  or so up to that final dmax....
>
>> ...Having all the source code, you can see
>> what is going on. Its simple and direct.
>
> I wish I had the knowledge to work with it at that level.
>
>>
>> I just find it a bit frustrating that the last little bit of fine
> tuning is
>> fiddly and would like to figure out a way to make it less so.
>
> Regarding this and the first comment, I think more
> measurement/correction points in that area would help. With
> StudioPrint, I will see differences in linearizations between 2
> different papers between 95% and 100% untill I do an 80 patch
> linearization. Iterative linearizations would not help this if there
> are no sample points in that area.
> Obviously this is a rare problem, and 21 steps is working well for
> most situations. I did run into very similar results once though with
> QTR on a particular ink setup, and I'm going to start from scratch on
> that one.
> Otherwise, I'm not sure what the solution would be. Start with heavy
> limiting so things are behaving, then creep up the K boost?
>
> Tyler
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
-
Roy Harrington
roy@...
Black & White Photo Gallery
http://www.harrington.com

--- In QuadtoneRIP@yahoogroups.com, Roy Harrington <roy@h...> wrote:
...
> Finally there's adding more steps in the linearization.  Usually 21 is 
> fine, but
> the 51 can be used.  You can actually use any number of steps as long as
> they are evenly spaced....

I had no idea, I've been using the Windows version with the gui which
has a fixed number of input boxes. So I assume I'd edit the desriptor
file instead with the measurements the "old" way? Very good to know.
Tyler