This article was originated as several postings by Erik Nikkanen responding to Gernot Hoffman, in the Usenet new group sci.engr.color in January 2005. This set of articles had a great effect on me because it descrbes, in detail, some of the mechanical issues that affect the consistency of offset printed media.

The material is reproduced here by permission of the Author, Erik Nikkanen.

Gernot Hoffmann wrote:
>
> An offset press is adjusted for ink densities
> at 0%, 50% and 100% for each ink, measured by
> visual dot area.
> Will this guarantee neutral grays under D50?
>
> The background:
> My philosophy: adjust the press according to
> proof prints (assumed, these are pleasant).
> The alternative: adjust the press by formalistic
> measurements of test patches without having any
> look at images.
> -
> Best regards --Gernot Hoffmann

Hi Gernot,

I am not exactly sure what you mean but I will give some comments related
to printing on an offset press that might help you with your question.

In my opinion, right now we are going through a transition period where
craft practice and thinking are trying to be accommodated with the new
capabilities of computerization. So there are many terms and concepts that
are really historical craft based but are not scientifically or
mathematically based in a way that would fit better with the new
computerization efforts. Much of what people think about the offset
process is more myth than reality and eventually this will work its way out
of the knowledge base.

The offset press is not at all like an ink jet printer. One has no control
over local amounts of ink film thickneses applied to the plate and
therefore to the paper. The idea model of offset is that one is applying a
uniform ink film to the plate and but that one adjusts the size and maybe
placement of dots of the CMYK plates to result in a specific colour.

So the starting point, in concept is to have a specific ink film thickness
or SID (solid ink density) values across the plate. From this point, by
printing some kind of test form, a map of how the press should print is
made. From the data of this test form, prepress determines what amounts of
dots on the different plates are required to make a particular colour. The
proof is supposed to be made in a way that the press can match it.
Now there is a lot of confusion here about what the correct steps are. A
confusing use of dot gain curves and profiles. There are probably several
different routes that get to the same place so I can't comment on these but
in general the proofs should be a target that is obtainable by the press
with the right prepress steps. It is possible to have a proof that is not
obtainable. Very frustrating for the operators.

Now for some issues which make the ideal model fall apart a bit.

The assumption has been that the press has a specific fingerprint that is
described by the test form. This is true for ink jet printers but it is
not true for offset presses. The assumption that an offset press prints an
even ink film around the plate, in line with the SID patch is not true. An
offset press does not have a specific fingerprint of how it prints that is
independent of the image it is printing. This is due to the patterns in the
ink, where ink has been taken off the form rollers by the image on the
plate. These patterns do not get totally removed on the single rotation of
a form roller and the result is variations in the ink film applied to the
plate as the form roller comes back in contact with the plate. Sometimes
this may have a small effect and at other times this may be large. It can
affect the amount of dot gain in local areas and therefore affect colour.

The test form printed will produce a set of values. Move the test patches
to different places in the test form and one can get a different set of
values. All test patches are printed supposedly with the same SID (ink
film thickness) but the results are different due to the test form
configuration itself.

There is also an assumption that dots are just small solid patches and that
they have the same ink film thickness as the solid areas. This is not true
either and can have an effect on prepress if prepress models are too
simple. Dots actually change the hue slightly from a solid partially
because they do not have the same ink film thickness as the solid. Hue
changes with ink film thickness.

So there is the practical problem of what needs to be done at press.
Since the press does not print in a totally predictable way with respect to
the profile or fingerprint (map) then having a totally valid target such as
a proof can be a problem. Therefore the operator does have to look at the
print because he can not totally count on getting to the right colour by
going to preset density targets.

This could be improved but right now there are a lot of design faults with
how the printing press is designed and these faults have caused not only
these problems of predictability but have the problem of inconsistency
also. The development of the process has not matured. It has been refined
but it has also been limited to a narrow vision of what the technology
should be.

Now the other question regarding if one can get a neutral gray. Well with
CMY inks, one can not get a perfect black. I suspect that it is also hard
to get a perfectly neutral gray. I think this is because CMY inks are not
colours but imperfect filters of light and that no combination of these
imperfect filters will result in a perfect gray. Close but not perfect.
This is one reason why I think there are holes in colour gamut's when
dealing with real materials that produce very wide spectral curves on there
own. CMY each produce a spectral curve that goes full width of the visual
spectrum and therefore since they are not perfectly shaped, can not produce
a perfect gray in any combination. Then there are also issues regarding
wet trapping of inks but I think the main problem is their curves.

Anyhow, to get a gray that is close to neutral, different dot percents need
to be determined that will account for the particualar physical conditions
of the CMY ink dots printed by a press. Most likely the dot percent values
for each CMY dot combination to produce a gray will be different. It will
not likely be 50 50 50 %. The required combination of CMY dots will be
different for different ink sets and different papers etc.

I hope this made sense to you.

Erik

Gernot Hoffmann wrote:
>
> Erk,
>
> thanks, excellent explanation. Much more than expected.
>
> By 'proof prints' I meant inkjet proof prints.
> Adjust the press until the offset print looks very similar to
> the inkjet print.
> -
> Best regards --Gernot Hoffmann

Gernot,

Yes, I think I did understand what you meant by proof. It was the implied
thought that a press could be adjusted to match any proof in general that I
was concerned about. A press has limited ability to be adjusted so the
proof and the prepress must be compatible first, so that the limited
adjustment of what the press can do is capable of hitting the proof
target.

I am not so interested in Standards. I am more interested in capabilities
of processes. Many of the Standards, which offset printing is
statistically incapable of meeting, are actually aimed at the proof. Some
of these Standards use SID (solid ink density) and dot gain. This is with
the implication that the press should match the proof to those values. Here
is one area of confusion and an area that can cause errors. When one talks
about ink jet proofs, be they continuous or dot proofs, they can be
measured for SID and dot gain. But since these proofs are printed with
different ink sets, the density values of the proofs and the offset print
are NOT directly related. Since dot gain is calculated from density
measurements, they too are not related with two different ink/paper sets.

Of course there is now the effort to describe things with colorimetric
values and not only the traditional values of dot gain and density.
Personally, I think that the colorimetric description of printing is not
the right technology for printing but it is a start of the process that is
recognizing that dot gain and density are not sufficient to describe the
process. So historically we are in a transition phase. Dot gain curves,
IMO are one of those historical tools that is a weak spot. Right now, the
industry can not think of not using them in the process but they are
actually not needed and they cause error.

In principle, since one should be able to map directly from press
conditions to colorimetric or non colorimetric data from a test form (if
the press was designed better), one does not need dot gain curves. Dot
gain curves do not convey any colour information and are basically a
measure of the grayness of the print of screens, from 0 to 100%, of an
ink/paper combination. This does not mean that dot gain does not happen or
is not important but many other physical conditions are also included in
the mapping process. Dot gain and density are important for process
control at the press but not necessarily for the calculations in prepress
and proofing.

We are in an historical transition phase where printing is trying to be
more scientific and less craft based but we are not there yet. My view is
that the offset printing industry is actually anti science because they
have a lot of resistance to theoretical approaches to solving problems.
There is the impression that just because a lot of experimentation is done
and costly technology is used, that things are then scientific. I would
argue that the level of science an industry has is more related to their
level of theoretical descriptions of the processes. The printing industry
confuses theory with observation.

I am not saying that there is no valid science but that there is a mixed
bag of good and very poor knowledge that are a part of the accepted
knowledge base. I think it will change and I hope it will change. Recently
a press manufacturer successful tested a simple and low cost technology of
mine that corrects the generic press design fault that if the fundamental
cause of ink/water balance and its related density variation problem. It
took eight years to get a press manufacturer or anyone to be interested in
testing this technology. The technology potentially can be a commercial
improvement, which the press manufacturer is now working on, but it also is
a way to demonstrate the science of the problem. In eight years, no
printer, press manufacturer, graphic arts institution, university that has
graphic arts departments, was interested in investigating the scientific
aspects of this concept.

So if this press manufacturer is successful with a commercial version, it
will demonstrate the simplicity of the density control problem. It will
show that what people had believed was true, was not true at all and that
it might be time to reevaluate other issues. So we are in transition.

Sorry I have gone on a bit off the topic.

Erik

Gernot Hoffmann wrote:
>
> This link should work:
> http://tinyurl.com/527fc
>
> G.H.

OK, I did get to read the comments via the guest option.

Just a few comments.

On press, one does not want to have different densities across the sheet or
web. The aim is to have a constant SID across the press. It may be a
different density for each ink or print unit but for each unit it is
supposed to have constant (SID) density.

The operator has to make adjustments for several reasons.

One is that the press is not capable of going to a specific density on its
own and it can not be calibrated to do this at this time. This is a
function of the lack of capability of the press to meter ink insistently
into the press and the lack of accuracy in the ink feed presetting
calculations that are use today.

If the press is capable of going to a specific density and the colour is
still not right, then it is partially related to press design issues but
mostly to the situation where the prepress is not describing the image
properly for the conditions on the press. This is a separate issue from
process variation.

I look at the printed image as a structure that is built up by the way the
physics on the press work. Different presses will build this structure
slightly differently. But basically, one has three different conditions
that need to be met. One is solid density which is almost directly related
to the ink film thickness on the substrate. Because of the way the press
applies ink to the plate, SID is the foundation of the process. Dot gain
effects and overprint effects are then related strongly with SID.

So in an image, one can have all three of these conditions. Solids,
screens and overprints. If the prepress is done properly, then all of
these conditons will result in the correct colour values. When a press
operator can not get colour in the screen regions and has to modify the SID
to get the proper screen colour, that is an indication that the prepress
has not been calculated properly. Since many images in commercial print
are only screens, it is not such a problem for an adjustment to be made to
help bring it in line with what is desired.

But at times, there will be two or all three conditions in a printed
image. Packaging uses lots of solid areas. In this case, if the colour is
not obtained, then the operator must decide what has to be sacrificed.
Maybe the customer is more interested in the screened image than the solids
or maybe the customer is more interested in getting the company logo right
and then solid density or even overprint is more important than screened
image. Anyhow, if the operator must sacrifice one condition to be able to
match another, it is an indication that there is probably a problem in the
prepress calculations.

My view is that if the science is right and the equipment is capable, then
one should not take hours to obtain the match but that the press will
converge directly to the preset targets without opertor or closed loop
adjustments to ink. The process can be predetermined.

Now just a comment about ink adjustment or ink feed. It is common to think
that adjusting the ink keys adjusts the amount of ink that goes into the
press. This is a false assumption. The amount of ink that comes out of
the press on the paper is not the same amount of ink that is adjusted and
metered by the ink keys. Or the amount of ink metered by the ink keys is
not directly related to the amount of ink that is fed into the press. Just
about all modern offset presses do not have a positive control on ink
feed. Ink is the primary variable and it is not under positive control but
is allowed to be fed into the press in an inconsistent way that is affected
by changes in variables such as water, temperature, press speed, humidity,
air movement, etc.

The total ink feed is the combination of ink metering and ink transfer into
the roller train. Even if an ink fountain metered ink perfectly, the
majority of variation of total ink feed is due to the inconsistent transfer
of ink at the ductor. In North America, the ductor is the roller that
swings back and forth to take ink from the ink fountain roller to the
roller train. It is the ink transfer rate at this transfer point that is
affected by all those variables mentioned. Fix the transfer problem and
the process becomes changed and then potentially can be predictable.

The other problem is with presetting. The industry traditionally says that
the ink keys should be set based on the image area of the plate. This is
the basis of all existing presetting software. Unfortunately, there is no
direct relationship with ink consumption and image area of the plate and
this can result in errors of up to the 40% range when calculating preset
data. The factors that creat this error are basically known but the
industry has a history of not being able to bring two known relationships
together to realize a third relationship.

It has been a long time since the industry knew that wet trapping prints
less ink on the substrate than dry trapping. A wet trapping solid might
print only about 70% of what it would print if it printed directly on the
paper (dry trapping). If your presetting calculation assumes that the ink
consumption is the same and provides for 100% ink film when in actuallity
the wet trapping produces a 70% ink film, then the error in the calculation
is 100%/70% which is about 1.41 or about 40% out. There are also other
conditions on press that affect the ink consumption per ink key and should
be accounted for. So I hope you can see that a more accurate calculation
is desireable.

This is not new. I described this problem in a TAGA paper in 1997. It has
also been discussed on forums. With all the push for CIM and JDF one would
think that accurate data would be good to have but it seems that the people
that push these concepts are mostly interested in the ability to
interchange data and not in the accuracy of the data. They have been told
directly but to me this is part of the problem in the printing industry.
An interest to understand and take advantage the physical rules behind the
process is just not there yet. I don't understand why this is so, but that
is my experience. Logic is not a motivator.

Your comment about the density being affected after a while, after a key
adjusment is perfectly true. The basic problem of density control is a
mass transport and mass storage problem. Where mass means ink. The density
that is printed on the sheet is more directly related to the ink storage
conditions in the roller train. Transients in density occur when there is
difference in ink flow out and ink flow into the press roller train. So
steady state average density is NOT related to the design of the roller
train but is only related to the consistency of the ink feed rate.
Transients ARE related to the roller train design and how the roller train
stores ink.

There has been a lot of work to model presses and those models always
produce consistent density but real presses don't. The fault is that the
models assume consistent ink feed. I say modify the presses so that they
are mathematically definable. Make the presses perform like the models and
not fudge the models to perform like the real presses.

There is actually very good news. It means that existing presses and
preset software can be modified at low cost to obtain a great deal of
consistency and predictability. All that is lacking is the will to do it.

Erik

>
> 1. Once a press is adjusted to a particular setting and
> the adjustment is not changed, does it then at least give
> reproducible results?

Yes and no. The density control process on press is now a stable process
but it is not consistent. If one sets the adjustments to the ink and then
walks away from the press and just let it run, the density will be quite
stable. But if some disturbance happens, such as air movement or
temperature build up, then the density will change and find a new stable
equilibrium. In this way the press can be considered as stable but that
stability can be moved to a new stable point without the operator making
any adjustments.

If one runs a press and it gets to a nice steady state condition and then
stops the press for a couple of hours and then restart the press again
without adjusting the ink keys or other ink setting, the density can very
well run at a different value and then might not even get back to the
steady state condition that existed before.

The adjustments are mechanical but the actual ink feed is not directly
related to the mechanical settings due mainly to the equilibrium nature of
the ink transfer rate at the ductor. If the mechanical adjustments were
designed to be directly related to the ink feed, then there would be
reproducible results.

As an example. If a problem required the exact metering of a liquid for
some process, what would be the right approach? One might use a normal
centrifugal pump and try to calibrate the flow in the pipes or even put a
flow meter on the pipe system and close loop the control back to the speed
of the pump. These can work but what is done for the most critical
processes? For the most critical liquid metering needs, constant
displacement pumps are used. These need to be calibrated but once they
are, no closed loop control is required and the performance is better
because the flow rate is inherently fixed or predictable. The constant
displacement pumps provide higher performance at lower total cost. With
any process problem, there are several potential solutions and picking the
right solution can be better and cheaper if one understands what is
physically required.

So in general, presses do not give highly reproducible results if the
adjustments are the same.

>
> 2. Why do you think that ICC color management is rather not
> the right way to go? What's wrong if one does not attempt to
> adjust the press to meet particular printing standards, or to
> match a given proof (which does not work very well anyway, if
> I understand you correctly), but if one just takes a particular
> press "as is", and characterizes the colorimetric behaviour of
> the particular press in order to adjust the document before
> printing (by applying the profile) so that it will print on
> this particular press with the right colors?

Too many things mixed together here. I am a futurist in a way so when I
say I don't think colorimetry is the right technology, please don't jump on
this too much. It may be that ICC colour management will be the final
system because other methods can't do all the things that are needed. But
on the other hand, there are non ICC products for proofing that are
performing quite well now. I prefer to think in terms of physical and
measurable factors and that means I would prefer analyzing reflected light
and having a system that deals with reflected or projected light. Going to
colorimetry transforms that physical data into data which does not have a
physical meaning. I would prefer a spectral approach where the data keeps
its physical meaning. Some people in the industry are working on that now,
so it is not so crazy.

The press should match the proof but if the science is right and the
equipment is capable, why should the operator have to make an adjustment at
all.

Presses are not always run to standards. If one talks about dot gain one
can say that a press does not print to the standard dot gain at the
standard density quite often. So what happens? A fudge is done in
prepress so that the press looks like it is printing to the standard values
but physically it is not. This is a practical method since trying to
adjust a press to have a particular dot gain at a particular density can
require the reformulation of the inks with respect to strength, viscosity
and water pickup. That is not an easy thing to do.

In the end, it is not important what physical dot gain one prints to but if
the colour is matched. Therefore having dot gain in a standard can seem
like a good idea but it is also a step that in my opinion is not needed if
one can go directly from a press characteristic fingerprint to the desired
colour.

> Isn't PDF X3 an attempt to go into this direction?
> I.e. the designer creates a PDF X3 document, not in CMYK, but
> with device independent colors, and the the print service
> provider receives the PDF X3 document and converts it to CMYK
> separations using the particular color profile of the press
> on which the document is going to be printed.

I can't speak of that specifically. I am interested in the physical
capability of the process. What you say above makes sense in principle but
it is based on the assumption that the press has a specific profile for an
ink/paper combination. As I tried to describe in the earlier post, offset
presses do not have specific profiles the way ink jet printers have. So a
profile for an offset press is a rougher estimation of how the press will
print. My interest is in having presses modified or designed so that the
profile would be less rough and would be more independent of the image
being printed. Right now colour values are used in a profile but other
values such as reflectance might be used in a way that makes a system also
work. It is something for the future.

Erik

> do you mean by this
>
> 'This is due to the patterns in the ink, where ink has been taken
> off the form rollers by the image on the plate. These patterns do
> not get totally removed on the single rotation of a form roller
> and the result is variations in the ink film applied to the plate
> as the form roller comes back in contact with the plate.'
>
> that an image with heavy inks creates a faded copy on other positions
> on the page ?

Yes. These paterns can be cyclical and/or random. It has to do with the
two ink films coming together in a nip and then splitting. For simplicity,
one usually assumes a 50 50% percent split. This means that it is not
possible for the form roller to gain a uniform ink film with usually only
one split before it again comes in contact with the plate.

Mechanical ghosting is a very visible result of basically the same
problem. But these paterns are there all the time. Many presses try to
reduce these paterns by having several form rollers on the plate. The aim
is to reduce the variation by random averaging the ink film. One problem
is that for normal printing, all forms rollers are applying ink to the
plate as the images pass by. This means that all form rollers will have
this patern or hole in the ink film where ink went to the plate image from
the form roller. Most of the time, it is not a very big problem but it
does exist and one does not know for sure when a screen area is affected by
this situation.

Also printers can have a problem where they can not get the same SID at the
top of the printed sheet or repeat length as they get at the bottom of the
sheet. This is more related to the image printed and how ink is stored
near the form rollers. If one has these paterns in the printed ink film
and the situation where one can not always print the same density on the
same sheet, from top to bottom, this then implies that one can not have a
profile that describes how the press prints. This is a problem with the
capability of the particular press technology and no amount of software
will account and correct for this.

These conditions can be improved. One can not totally eliminate these
situations on most offset presses but the goal is to reduce the variation
to a level that is not a problem. A totally different offset press design,
such as the KBA presses, Karat, Genius, and Rapida 74G have a simple roller
train design where there is only one form roller and this form roller
surface is printing in register with the plate. This interesting design
prevents any such paterns from existing. These press designs have some
other weaknesses but they are for sure an interesting design. These
presses can not be adjusted to any great extent and any steps to change
colour must be done almost totally throught the prepress steps but in
principle, they should have a profile which is much more accurate and
independent of the image printed than with normal offset presses.

> In Dan Margulis' book 'Professional Photoshop' is an impressive
> example: a red parrot image has influence on green patterns below.
>
> If this should be so - why are the faded copies not visible on plain
> paper areas ?

??? If you mean by plain paper areas, the areas where there is no printing
it is because there is no printing there. These ink film variation paterns
can only be transfered to the paper via the inked image areas of the
plate. Since the plate, in the non image areas, does not accept ink, it is
not affected by any variations of ink film on the form rollers.

I hope that explains what you asked.

Erik