Examensarbete

LITH-ITN-MT-EX--02/001--SE

InkKey Presettingin Offset

PrintingPressesUsing Digital
Imagesof the Plates

Linus Lehnberg
2002-05-21

Department of Science and Technology Institutionen för teknik och naturvetenskap

Linköpings Universitet Linköpings Universitet
SE-601 74 Norrköping, Sweden 601 74 Norrköping
 LITH-ITN-MT-EX--02/001--SE

Ink Key Presetting in Offset

Printing Presses Using Digital
Images of the Plates

Examensarbete utfört i Medieteknik

vid Linköpings Tekniska Högskola, Campus Norrköping

Linus Lehnberg

Handledare: Peter Dahlén, Reidar Larsen

Examinator: Björn Kruse

Norrköping 21 maj 2002

 Datum
Avdelning, Institution Date
Division, Department

Institutionen för teknik och naturvetenskap 2002-05-21

Department of Science and Technology

Språk Rapporttyp ISBN

Language Report category
_____________________________________________________
Svenska/Swedish Licentiatavhandling ISRN LITH-ITN-MT-EX--02/001--SE
Engelska/English Examensarbete _________________________________________________________________
C-uppsats Serietitel och serienummer ISSN
D-uppsats Title of series, numbering ___________________________________
_ ________________ Övrig rapport

_ ________________

Titel
Title
Ink Key Presetting in Offset Printing Presses Using Digital Images of the Plates

Författare
Author
Linus Lehnberg

Sammanfattning
Abstract
During a make ready in a web offset press it is important to produce as little waste as possible. Reducing the amount
of waste yields savings of both time and money. One way to do faster make ready is to preset the ink keys of the press
before it is started.
This diploma work, carried out at Sörmlands Grafiska Quebecor AB in the city of Katrineholm, Sweden,
examines how the preset may be done using low-resolution digital images stored in the vendor independent data
format CIP3 PPF. The press that has been used has a control interface that is not accessible from the outside. This
feature is shared with a lot of older presses. Therefore several methods of how to present and collect ink key settings
using offline methods have been tested.
To investigate the relationship between mean coverage over one ink zone and its corresponding ink key
opening data from a 32-page web offset press has been collected. The mean coverage was taken from the CIP3 PPF
files that were related to the collected print jobs.
The relationship that was found between the coverage and the opening can be described with a transfer curve
(one curve per printing unit and side). Using as few as three print jobs of high quality (density and dot gain within
given tolerances) a first set of transfer curves may be created. These are close to the real ones and using print jobs
where the ink key presettings have been calculated the transfer curves may be calibrated to perform better and better
presetting calculations. To generate and calibrate the transfer curves and to extract the mean coverage values from the
CIP3 PPF files and recalculate these to presetting values a computer program called IKPS (Ink Key Presetting
System) was made. IKPS was made using MATLAB from MathWorks INC.
IKPS have been tested for ink key presetting during a number of print jobs. Even though the transfer curves
were uncalibrated the system performed well. As comparison the results from a plate scanner was used. Even if online
transfer of the presetting values is preferable the big advantage with the IKPS is that it is an offline system and
therefore it is possible to implement it on any kind of offset press, old as well as new. In order to generate reliable
transfer curves the print jobs used for calibration must be of high printing quality and representative for that particular
press. How the ink key presettings are presented in the press control room depends on what kind of press it is. IKPS
works with CIP3 PPF files as well as low-resolution cmyk tiff files.

Nyckelord
Keyword
Printing, make-ready, ink-key, preset, CIP3, PPF
Abstract
During a make ready in a web offset press it is important to produce as
little waste as possible. Reducing the amount of waste yields savings of
both time and money. One way to do faster make ready is to preset the ink
keys of the press before it is started.

This diploma work, carried out at Sörmlands Grafiska Quebecor AB in the

city of Katrineholm, Sweden, examines how the ink key preset may be
done using low-resolution digital images stored in the vendor independent
data format CIP3 PPF. The press that has been used has a control interface
that is not accessible from the outside. This feature is shared with a lot of
older presses. Therefore several methods of how to present and collect ink
key settings using offline methods have been tested.

To investigate the relationship between mean coverage over one ink zone
and its corresponding ink key opening data from a 32-page web offset
press has been collected. The mean coverage was taken from the CIP3 PPF
files that were related to the collected print jobs.

The relationship that was found between the coverage and the opening can
be described with a transfer curve (one curve per printing unit and side).
Using as few as three print jobs of high quality (density and dot gain
within given tolerances) a first set of transfer curves may be created. These
are close to the real ones and using print jobs where the ink key presettings
have been calculated the transfer curves may be calibrated to perform
better and better presetting calculations. To generate and calibrate the
transfer curves and to extract the mean coverage values from the CIP3 PPF
files and recalculate these to presetting values a computer program called
IKPS (Ink Key Presetting System) was made. IKPS was made using
MATLAB from MathWorks INC.

IKPS have been tested for ink key presetting during a number of print jobs.
Even though the transfer curves were uncalibrated the system performed
well. As comparison the results from a plate scanner was used. Even if
online transfer of the presetting values is preferable the big advantage with
the IKPS is that it is an offline system and therefore it is possible to
implement it on any kind of offset press, old as well as new. In order to
generate reliable transfer curves the print jobs used for calibration must be
of high printing quality and representative for that particular press. How
the ink key presettings are presented in the press control room depends on
what kind of press it is. IKPS works with CIP3 PPF files as well as low-
resolution cmyk tiff files.
Sammanfattning

Vid ett intag i en offset press är det önskvärt att mängden makulatur hålls
så låg som möjligt. Att producera makulatur kostar både tid och pengar.
Ett sätt att hålla antalet kasserade ark nere är att göra en förinställning av
färgskruvarna innan tryckpressen startar.

Detta examensarbete, som utförts vid Sörmlands Grafiska Quebecor AB

(SGQ) i Katrineholm, undersöker hur denna förinställning kan göras med
hjälp av de lågupplösta bilder av tryckplåtarna som finns lagrade i det
tillverkaroberoende dataformatet CIP3 PPF. Den tryckpress som använts
har haft ett styrinterface som inte varit åtkomligt utifrån, en egenskap som
den delar med mängder av äldre pressar. Därför har ett flertal metoder för
att presentera och samla in färgskruvsinställningar offline testats.

För att undersöka sambandet mellan medeltäckning i en färgzon och

motsvarande färgskruvsöppning samlades data in från ett antal tryckningar
i en 32 sidig rulloffset-press. Medeltäckningen utlästes i de CIP3 PPF filer
som hörde till respektive jobb.

Det samband mellan medeltäckning och färgskruvsöppning som

upptäcktes kan beskrivas av en överföringskurva, en per tryckverk och
sida. Med hjälp av kännedom om så få som tre trycktekniskt bra
tryckningar (densitet och punktförstoring inom givna toleranser) kan en
första uppsättning överföringskurvor skapas. Dessa ligger nära de verkliga
och med hjälp av tryckningar där förinställningar beräknats kan
överföringskurvorna kalibreras ytterligare. För att kunna generera och
kalibrera överföringskurvor samt ta fram medeltäckningar från CIP3 PPF
filer och omvandla dessa till förinställningsvärden har ett program kallat
IKPS (Ink Key Presetting System) skapats med hjälp av MATLAB från
MathWorks INC.

IKPS har använts vid ett antal tryckningar och har, trots att inte
överföringskurvorna kalibrerats, gett ett bra resultat. Som jämförelse har
en kalibrerad plåtskanner använts. Även om direkt överföring av
inställningarna till pressen är att föredra så är IKPS största styrka just att
det går att använda på vilken offsetpress som helst, gammal som ny. För
att överföringskurvorna skall bli bra är det väldigt viktigt att jobb som
används för kalibrering är representativa för pressen samt av god
tryckteknisk kvalitet. Hur färgskruvsinställningarna presenteras i pressen
beror på vilken press det rör sig om. Viktigt är dock presentationen sker på
ett sätt som tryckarna är bekanta med. IKPS har också visat sig fungera väl
med lågupplösta cmykade tiff-bilder av tryckplåtarna istället för att
använda CIP3 PPF.
Preface
This diploma work was carried out at Sörmlands Grafiska Quebecor AB
(SGQ) as the last part of my studies to obtain a Master degree in Media
Technology at Linköping University, Campus Norrköping. The diploma
work has also been a part of the Swedish research project T2F
(TryckTeknisk Forskning). There are a few people besides my examiner
professor Björn Kruse that I would like to thank.

First I would like to thank Peter Dahlén at SGQ for giving me the
opportunity to do this diploma work at SGQ. I also thank Reidar Larsen at
SGQ for all help and assistance and for letting me share his great practical
experience and knowledge about printing and web offset presses. A big
thanks goes to the SGQ printers that works at press 57. Thank you guys for
helping a theorist in the reel world.

I would also like to thank my fiancée Lillemor Eriksson for her patience
and support and for giving birth to our wonderful daughter Alva. I love
you.

Finally I would like to thank my friends from the Secret Room.

Katrineholm, August 2002

Linus Lehnberg
Index
ABSTRACT

SAMMANFATTNING

PREFACE

1. INTRODUCTION 1
Background 1
Aim 2
Goal 2

2. METHOD AND LIMITATIONS 3

Method 3
Limitations 5
Hard- and software 5

3. BASIC PRINTING THEORY 7

Offset printing 7
Colour printing 11
Halftone screening 11
Raster image processor 12
Density measurement 12
Computer to Film vs. Computer
to Plate 12
Plate scanning 13
Heidelberg Harris 850 A 14

4. CIP3 AND THE DATA FORMAT PPF 15

Background 15
CIP3 PPF-file format 15
New name and a new format 16
CIP3 PPF at SGQ 17

5. PREVIEW IMAGE EXTRACTION 18

6. MEAN COVERAGE CALCULATIONS 21

7. CALCULATE THE INK KEY SETTINGS 25

8. USING THE SETTINGS IN THE PRESS 29

9. TRANSFER CURVE CALIBRATION 31

Collecting approved data in the press 31
The method used for data collection 32
Processing of the collected data 34
First steps toward semi automatic
data collection 34

10. DISCUSSION AND EVALUATION 36

Testing the IKPS 36

11. CONCLUSIONS 42

12. RECOMMENDATIONS FOR FURTHER WORK 43

13. WORDLIST 44

14. REFERENCES 45

15. APPENDICES 46
1. Introduction

Background
In today’s printing industry with its tough competition and low prices it is
important to cut costs wherever it is possible. Along with the introduction
of fully digital workflows from customer to printing plate, the
development goes toward a higher grade of automation throughout the
whole printing process. In the printing-house it is important to reduce the
time for make-readies, i.e. the time between one print job stops and the
next job starts producing printed sheets that look good should be as short
as possible. By reducing the make-ready time the press can produce a
larger number of approved sheets per unit of time and the result is a lower
cost per printed sheet.

When a new print job is started in a web offset press (or in any press that
not uses a digital printing method), there is always a certain amount of
wasted paper before the printed result looks as it was intended. How much
paper that is wasted before the result is acceptable depends in a web offset
press mainly on three things:

1. Register – how well the four different colours print on top of each
other
2. Folding – the printed sheet must be folded in a correct way according
to how it is about to be finished
3. Presetting of the ink keys – how well the ink keys are set to give a
good colour reproduction across the printed sheet

Register and folding are in some meaning mechanically dependent and

impossible to do while another job is running in the press. Presetting of the
ink keys is however possible to determine in advance and as soon as a
make ready starts, the keys could be adjusted to their new settings. By
focusing on how to make the presetting of the ink keys as good as
possible, at least one of three error sources could be, if not eliminated, at
least minimized. The problem to be solved is how to calculate this
presettings using digital images of the printing plates, and how to
implement the method in the print shop’s workflow.

Aim
The aim of this diploma work is to show how the presetting of the ink keys
could be done as general as possible by the use of digital images of the
printing plates.

1
Goal
The goal is to make a system that includes all necessary components that
are needed for calculation of the ink key presettings for the printing press
Heidelberg Harris 850 A at Sörmlands Grafiska Quebecor AB.

2
2. Method and limitations

Method

Report
This report is divided into four main sections. First there is a section about
the diploma work with background, aim, goal, and method. Next there is a
part that sheds light on different theories, data formats and techniques that
in some way relate to the problems that are about to be solved. The third
section covers what actually have been done. To get a logical structure, the
chapters that belong to the third section will follow the workflow plan in
figure 1. The fourth and last section contains an evaluation of the work
done with results, conclusions and a proposal for further improvement.

Figure 1. The flow chart for the ink key presetting system.

Data collection
Information from at least twenty different print jobs has been collected to
get enough data for an evaluation. From every print job that became a part
of the data set a printed sheet was collected and measured with respect to
density and dot gain. The density measures have been taken using the
KBA Densitronic S 2000 from LITHEC Gmbh.

The printing press used in this diploma work, Heidelberg Harris 850 (in
house name: press 57), has not a data interface that was accessible without
special knowledge and/or equipment. Therefore, to gather information
from the press, a digital camera, Canon IXUS, was used to capture the ink
key settings in the press control room as digital images. Different methods
to get the ink key settings into the computer as numerical values were also
examined.

To get information about how well a plate-scanner performed the ink key
presetting, the digital camera was used to capture ink key data from a
number of different print jobs in another press than press 57. After a

3
discussion with SGQ’s technicians a MAN Roland Rotoman (in house
name: press 54) was chosen. This choice was made since the plate scanner
has been calibrated regularly.

Programming and data processing

All the programming in this diploma work has been done using the
technical computing language MATLAB from Math Works INC. Besides
powerful mathematical and plotting functions this program has a very
good I/O interface when reading and writing from and to disks. The
language comes with a program called GUIDE that makes it easy to create
graphical user interfaces (GUIs). It is possible to add something called
toolboxes to MATLAB depending on what special fields of engineering or
calculating the user is interested in. In this case, the Image Processing
Toolbox has been used.

One drawback with MATLAB is that in order to run the programs created
within MATLAB on a computer, the entire MATLAB program package
has to be installed on that computer. To solve this problem, a special plug-
in that generates stand-alone applications has to be bought from Math
Works. This plug-in has not been tested or used in this diploma work.

The collected data was examined using MATLAB. Focus was on finding
the relationship between mean coverage in the digital image of the printing
plate in one ink zone and its related ink key opening.

The data exchange format PPF (Print Production Format) from the CIP3
(Co-operation in Prepress, Press, and Postpress) group was examined and a
program that extracts necessary data from the PPF-file was made. The
most important parts of the PPF-file are the digital preview images of the
plates and the data that relates to these images. This program was also used
during the data processing phase.

The results from the data processing were used to construct a system that
performed the re-calculation of mean coverage to ink key opening. An
important part was the collection of approved ink key settings from the
press in order to teach the system how to minimize the presetting error. A
program that facilitated the collection of data was constructed. Alternative
ways of how to collect the approved ink key settings was also tested.

Verification
When it comes to verification of the result, absolute mean error and
standard deviation was used to evaluate how well the system performed
the ink key presetting. Success was measured in the distance in percent
units the ink key had to move in order to produce a good result on the
printed sheet. The calibration algorithm was tested theoretically with
mathematical functions as transfer curves. The system was also compared
to the plate-scanners at SGQ in order to see how well the presettings are
done. Finally the system was used during a number of print jobs in press
57 at SGQ.

4
Limitations

Theory part of the report

The readers of this report are supposed to have knowledge about the
graphical industry and its processes (prepress and printing). However, to
establish a terminology and to give a reader with only some knowledge
about the subject a chance to understand, some basic printing and prepress
theory is presented.

Data collection
The collected data consists of printed sheets that have been taken from the
printing press within an hour after the balanced sheet counter had started.

Method of working
The development of a system and its algorithms, like the ones in this
report, is a strictly iterative procedure with a constant improvement of
what has been made. An iterative cycle may look like this:

1. Basic idea of how a problem shall be solved.

2. Implement an algorithm that solves the problem.
3. Evaluate the implementation.
4. Use the knowledge achieved in step 3 to make improvements in
either step 1 or step 2.

As well as it is hard to say whether the hen or the egg came first, it is also
hard to draw strict lines between the different parts of the system and in
what order they have been developed. For example, it is hard to make
presetting calculations without any possibilities to collect data from the
press and vice versa. In this report efforts have been made to present the
development of the system in the same way as it is intended to work when
it is finished.

Due to the fact that only PPF files from the Prinergy prepress system was
available, the functionality of the program that reads CIP3 PPF files is
only guaranteed for PPF files generated by Prinergy. The program is
however constructed in a way that makes it easy to implement reading of
PPF files generated by other prepress systems.

Hard- and software

Hardware
Canon IXUS (digital camera)
Compaq 1 GHz Intel, 512 MB ram, USB
Heidelberg Harris 850 A
Harris Graphics platescanner
iMac 500 MHz, 256 MB ram, CD/RW
KBA Densitronic S 2000 (density measurment)
MAN Roland Rotoman D

5
MAN Roland Rotoman platescanner
Philips AZ8267, remote, digital tuner, 3 band eq.

Software
Adobe Acrobat 4.0
Adobe Illustrator 9.0
Adobe Photoshop 6.0
MATLAB version 6 from Math Works INC.
Prinergy 2.0.7.0 (prepress system from Creo)
Microsoft Word 2000
ZoomBrowser EX Version 2.2 (software for the digital camera)

6
3. Basic printing theory
The idea of printing is to reproduce an image on a printing substrate. A
more limited definition could be to define printing as a process that
mechanically transfers an image from a permanent image carrying media
on to a substrate. Depending on what material this substrate is made
(paper, plastics, metal, cloth etc.) the printing technique used varies. The
most common printing substrate is paper and the most common technique
for printing on paper is offset printing, which is the technique used at
SGQ.

Offset printing
When printing in an offset press, the paper is not in direct contact with the
printing plate. Instead, the ink is first transferred to a cylinder with a
rubber blanket and then from the blanket to the paper. Most common as
permanent image carrying media in offset printing are printing plates made
of aluminum, covered with a thin layer of an emulsion that is either
photosensitive or heat sensitive. It is the emulsion that makes it possible to
reproduce images on the plate.

Figure 2. Offset principle.

Offset printing is a lithographic technique where the printable and non-

printable areas of the printing plate are in the same plane. This can be
compared to two other common groups of printing techniques that uses
printing plates. First is gravure printing where the printing areas consist of
very small cavities. The volume of each cavity is related to the amount of
ink at that point in the image. Second is flexographic printing where the
printing areas are higher than the surroundings.

7
 Figure 3. Schematic side view of three printing plates that adopts three
different printing methods.

Offset printing is divided into three different areas depending on the

principle of how the ink settles and dries. First is cold set web offset
(CSWO) where the ink settles by loss of oil from the ink. Web means that
the paper is fed to the press from big paper rolls. In CSWO the ink really
never dries. CSWO is mainly used for print jobs with a short life cycle,
typically newspapers. Next is heat set web offset (HSWO). When printing
HSWO print jobs, the ink settles and dries partly because it loses oil to the
printing substrate, and partly by the provision of external energy. One
example of energy source is warm air. The last method is oxidative sheet
fed offset (OSFO). Sheet fed implies that the paper is fed into the press as
single sheets. In OSFO the ink settles and dries because of loss of oil to the
substrate together with a chemical oxidation process. At SGQ the two
latter methods are used.

The difference between the printing and non-printing areas in offset

printing is created by different surface properties. That is, the parts of the
printing plate that should attract ink are treated so that they repel water
(made hydrophobic), and the non-printing parts attracting water (made
hydrophilic). When a small amount of water, moist, is applied on to the
plate it only sticks to the hydrophilic areas and thereby helps the ink,
which has characteristics that are similar to those of oil, to only stick to the
hydrophobic areas. The water also keeps the process cool and free from
paper particles. It is fundamental to get a good and stable balance between
ink and moist when it comes to achieving a good printing result. There
also exist offset presses without water. This technique is called dry offset.
In this process the non-printing areas consists of a material that is strongly
repellant to the printing ink. Because the lack of water, other techniques
have to be used when it comes to cooling and paper dust removal during
dry offset printing.

To get the ink in contact with the printing plate, via a rubber blanket, to the
paper, printing units are used. Each colour has its own unit. A traditional
offset printing unit consists of an inking device, a dampening device, a
plate cylinder, a rubber blanket cylinder, and a back-pressure cylinder.
The width of the cylinders sets the limit of maximum printing width and
the circumference maximises the printing length. In figure 4 a schematic
side view of a typical inking unit is shown. Remark that this printing unit
prints on one side only. When both sides are printed at once, the inking
unit is said to be perfecting. In a perfecting inking unit there is no back-
pressure cylinder. Instead, the part of the printing unit that prints the back
side uses the rubber blanket cylinder of the unit that prints the front side to
get a back pressure, and vice versa.

8
 Figure 4: Schematic side view of a typical offset printing unit.

The ink fountain is the part of the inking device that holds the ink (named
ink in the illustration above). In the ink fountain a fountain roller picks up
the ink so it can be distributed via the drop roller further in the ink unit.
The drop roller, symbolized with a double arrow in the figure above, is a
roller that moves back and forth, transporting ink between the fountain
roller and the next roller in the ink unit. To control the amount of ink that
is about to be transferred to the paper, the amount of ink that is picked up
by the fountain roller has to be adjusted in some way. This control is made
by adjustment of the ink keys. Each ink key controls an equal fractional
part of the width across the ink fountain. This fractional part is known as
an ink zone. The width of each ink zone is simply the total controllable
width divided by the number of ink keys. Below is an illustration of a
sheet with the ink zones marked as vertical lines. The number of ink keys
of the press that printed this job was 25 with a width of 40 mm each.

9
 Figure 5. A printed sheet from a Hedielberg Harris 850 press. The press has
25 ink keys, each with a width of 40 mm.

The ink keys are operated by push buttons from the press control room via
the press’ computerized control system. This makes it possible to store the
settings of the ink keys if the print job is interrupted. Mechanically, the ink
keys control a thin metal blade, the ductor knife, that is fitted across the
fountain roller. The ductor knife is cut up in slices to match the width of
the ink zones. There also exist ductor knives that consist of a blade that’s
not cut up into slices, even if that is not common in modern presses.

Figure 6. The image to the left represents a control panel from a sheet fed
KBA press. This machine has 32 ink zones, each with a width of 30 mm. The
image to the right is the corresponding ink fountain. Notice that it has
numbers on the front, representing the ink keys. It also possible to
physically see on the fountain roller that the ink keys to the left are more
open than the ones to the right. Worth mention is also that this press doesn’t
have a traditional ductor knife, instead each ink key is adjusting a precision
milled steel knife. This results in a better control and higher precision of the
amount of ink that the fountain roller collects.

10
As a comparison, gravure printing does not have ink keys. Instead, the
amount of ink that is to be transferred to the paper is determined by the
depth of the screen cells that builds the image.

Colour printing
When printing with a gray scale obviously only one colour (black) of the
ink is needed. Colour printing is based on subtractive colour mixture, and
the colours used are cyan, magenta, and yellow (often referred to as
process colours, or c, m, and y, or cmy). The opposite to subtractive colour
mixture, additive colour mixture, is used by media that transmits light such
as TV and computer screens. Additive colour mixture uses the colours red,
green, and blue (r, g, and b, or rgb).

Subtractive colour mixture Additive colour mixture

Figure 7. Subtractive colour mixture vs. additive colour mixture.

The result when printing cmy together should theoretically be black (i.e.
all the incident light should be absorbed). This is not true due to problems
with creating printing inks with perfect spectral distribution and therefore
black is added as a fourth colour to improve the printing contrast. To avoid
conflict and mix up with blue in additive colour mixture, black is
abbreviated k. It should also theoretically be possible to reproduce all
visible colours with the use of only cmy but again due to the imperfection
of the spectral distribution of the inks this is not possible. In order to be
able to reproduce more colours, that is getting a larger colour space, more
than four colours can be used when printing.

Halftone screening
It is impossible to reproduce continuous tones from a printing plate due to
the fact that printing is based on areas that are printing or non-printing.
Instead all images that are about to be printed has to be converted to
halftones. The name of this process is halftoning. When printing a colour
image, the image is first separated into the four process colours and each
separation is then screened individually.

11
 Figure 8. To the left is a simulated (because all printing methods print half
tones) continuous tone original. The right picture is the corresponding half
tone reproduction (this picture may not be printed out nicely, because of
interference between the printer’s screening and the screening of the image,
but it looks nice on the computer screen).

There exist two classes of methods how this may be done. The methods
are either that continuous tones are represented by dots of varying size,
amplitude modulation, but with the same distance from each other, or dots
of the same size but with a varying number, frequency modulation. The
names of these methods are AM-screening respectively FM-screening or
stochastic-screening. Each method has its pros and cons and the latest
algorithms mix AM and FM screening to get the best from two worlds.

Raster Image Processor

The calculation from continuous tones in the computer to printable half
tones is made in the raster image processor (RIP). It is the RIP that
provides the image of each separation that is about to be transferred to the
printing plates.

Density measurement
To measure how much ink that has been put on the paper, a densitometer
is used. Density is defined as the logarithm of the opacity (quotient
between the incident light and the reflected light). The more ink that is put
on the paper, the higher the density value:

 Incident light 
Density = log10  
 Reflected light 

Computer To Film vs. Computer To Plate

The transfer of the images that are about to be printed to the printing plates
could be made in two ways. Either via film to the plate (computer to film,

12
CTF) or direct to the plate (computer to plate, CTP). SGQ is only using
CTP technology because CTP involves fewer steps in the production chain
than CTF, and also because CTP gives a higher accuracy and a more
predictable result.

Platescanning
Trying to make as correct ink key presettings as possible by using
knowledge about the printing plates is not a new idea. If no presetting
systems are available the printer can look at the plates and make a rough
manual presetting. A technology well established is the use of plate-
scanners. This is exactly what it sounds like, a scanner that physically
scans the plates used for printing. A modern plate-scanner uses a digital
camera to measure the mean coverage. The scanners at SGQ use an older
technique that measures the reflection from the plate over the ink zones
and then calculates the ink key settings. Properly taken care of, plate-
scanner systems can give quite accurate and predictable presettings
compared to making the settings by hand. However, there are some big
disadvantages with the use of plate scanning. One of them is that the
surface structure of the plates sometimes is inhomogeneous which results
in bad contrast. This gives inaccurate mean coverage calculations. Another
problem is that it is a time consuming process to perform the scanning.
The possibility to damage the plates during the measurement may also be
thought of as a disadvantage.

Figure 9. Platescanners. The most leftward one is the scanner used for
press 57.

13
Heidelberg Harris 850 A heat set web offset press
The Heidelberg Harris 850 A is a 32 page HSWO press built in 1988. It
was bought second hand from France and installed at SGQ during the
summer/fall 2000. The press has an overall length of approximately 45
metres, and it is operated by four personnel: two printers and two print
assistants. In house, the press’ name (or rather number) is 57.

Press 57 is used for four colours magazine printing on different paper

qualities with a grammage substance ranging from 40 grams to 100 grams.
The ink fountains are equipped with 25 ink keys, each with a width of 40
mm.

14
4. CIP3 and the data format PPF

Background
The initiative to CIP3 was taken by Heidelberg
Druckmaschinen AG assisted by the Fraunhofer
Institute for Computer Graphics. CIP3 stands for
“International Cooperation for Integration of Prepress, Press, and
Postpress” and the abbreviation is pronounced “sip three”. The CIP3 group
was formed 1995 and is an international cooperation of well-known
manufacturers from different areas of the graphical industry (Agfa, Adobe,
Heidelberg, MAN Roland, just to mention a few of over 35 companies).
The main problem for the group to solve was the problem with data
exchange between equipment and systems from different manufacturers.
The solution to this problem became a vendor-independent interface called
Print Production Format (PPF). The goal of PPF is the computer-integrated
manufacturing of print products. With PPF, a digital link spanning all
production processes from prepress to press and postpress can be realized.

History of the CIP3 Print Production Format

Dec. 1993 Development of a concept
Sep. 1994 First draft specification
Dec. 1994 Development of a functional prototype
Feb. 1995 Foundation of the CIP3 group with 15 members
May 1995 Version 1.0 of the CIP3 PPF, presentation at DRUPA´95
Aug. 1995 Prepress station creates first CIP3 PPF-file
Aug. 1996 Version 2.0 of the CIP3 PPF
Dec. 1996 26 manufacturers as members in the CIP3 group
June 1997 Version 2.1 of the CIP3 PPF, presentation at IMPRINTA´97
June 1998 Version 3.0 of the CIP3 PPF is released

CIP3 PPF-file format

The CIP3 PPF may be seen as a container format for the exchange of data
between the different stages of a print job. That is, a PPF file does not
contain the high-resolution separations used for printing, only information
about them and the sheet that they are about to be printed on. The
information in the PPF file may consist of:

- administrative data (i.e. job name, copyright, etc)

- low-resolution preview images for each colour separation
- transfer functions (original to film, film to plate)
- colour and density information
- register marks
- cut data
- folding procedures
- private data (vendor or application specific)

15
At the moment, the printing industry mostly uses PPF files for generation
of ink key presetting data. There for a PPF file usually only contains image
previews along with some administrative data. Even if most of the existent
applications that use PPF files are press related, there exist applications
and equipment for the post press area. The ideal case, even if that is not
common today, would be a PPF file with all the necessary information to
perform a print job with respect to both the printing phase and the further
processing such as folding and cutting.

Semantics, syntax, and logical structure

The information in the PPF file is stored in a treelike way that is both
logical and structured. This allows easy and flexible access to the
information stored inside. One PPF file can contain everything from only
the information about a single sheet, up to multi sheet files with product
definitions.

Instead of inventing a new programming language and data structure,

Adobe’s page description language PostScript is used as the basic format
for the CIP3 PPF. The files are saved in ASCII source text form i.e. they
are fully readable and editable. Compression is supported as an option only
for the included low-resolution image data. The PPF also follows the
syntax rules of the PostScript language. This is especially true for the
coding of numbers, names, character strings etc. To reduce the complexity
there is a big difference between “real” PostScript and the CIP3 PPF. That
is the CIP3 PPF file format does not allow any kind of calculating inside
the PPF file using the PostScript language. The content and structure of the
CIP3 PPF will be explained more in detail in another section of this report.

New name and a new format

In July 2000, the CIP3 group changed its name to CIP4
(International Cooperation of Processes in Prepress,
Press, and Postpress). The reason for this change of
name was that the former CIP3 group were given the
rights to a new data format called JDF (Job Definition
Format). JDF enables the integration of commercial and planning
applications into the technical workflow and it builds on PPF and Adobe
Systems' Portable Job Ticket Format (PJTF). Although JDF is able to
contain all data that is in a PPF file, PPF will however still exist as a data
format.

16
CIP3 PPF at SGQ
At SGQ’s prepress department, Prinergy from Creo is used as a complete
digital workflow management tool that organizes page processing,
proofing, and CTP. The system works with pdf files and has a lot of
possibilities when it comes to automation of the prepress work.

The output of CIP3 PPF files from Prinergy is a feature that is not included
in the standard installation. Instead, a plug-in with the name PrintLink has
to be bought and installed to make it possible to create PPF files.

When it comes to output of the preview images, PrintLink offers three

choices: each separation is stored as a single file, each side of a sheet is
stored as a single file, and finally the entire sheet is stored as a single file.
Storing separation by separation results in a total of eight files. The next
choice yields two files with four images per file. The last option gives one
file with eight images stored within, and it is this way to store the images
that is selected because it is convenient to deal with one file instead of four
or eight.

The generated files may be stored job relative on a file server in the same
folder as the rest of the files that belongs to that specific job, or somewhere
else on a computer or server that is attached to the local area network. It is
also possible to specify user name, password, and an address to an ftp
account to where the files can be sent. The last alternative is of interest if
the printing plates are not made at the same location as where the press is
located.

PrintLink are able to generate preview images with three different

resolutions: 12.5, 25, or 50 dpi. These resolutions are only fractions of the
original resolution that typically is set to 2400 dpi.

17
5. Preview image extraction
The preview images contained in the ppf file are low-resolution variants of
the high-resolution images that are used for plate making. There is one
image per separation and side of the sheet that is about to be printed. This
gives a total of eight images: front c, m, y, and k and back c, m, y, and k.
To get a convenient size of the ppf files, the resolution of 25 dpi was
chosen for the preview images. With this resolution, each ink zone is
approximately 40 pixels wide. How this “approximately” problem is
solved will be explained in the next chapter.

To collect data from the ppf file, it is opened and read line by line as a text
file. Due to the logical structure of the ppf file it is easy to find the data
that is related to the preview images. As an example, a ppf file generated
by PrintLink containing eight separations is shown in appendix 9.

The logical structure in the file is built by eleven different kinds of

structure types. In this example file, four different kind of structure types
of the form CIP3BeginUnit and CIP3EndUnit are represented. In general
Unit specifies both type and name of the structure. The structures in the
file above are Sheet, Front, PreviewImage, and Separation. These four
structures have to be represented in a ppf file that holds one or more
preview images.

Each structure type has a number of attributes that holds information about
the job or sheet depending on what kind of structure type it belongs to. The
name of the attribute relates to which structure it belongs, e.g.
CIP3AdmJobName and CIP3PreviewImageWidth. The word def that ends
the attribute lines is a PostScript command that tells that the word that
precedes the data held by parenthesis should be used to define that data.

First in the file are two lines, where the first one tells that this file should
be interpreted as a PostScript file, and the second gives what version of the
ppf file it is. The double percent sign is, as in PostScript, used for
comments. The last line in the file, %%CIP3EndOfFile, denotes where the
ppf file ends.

After the file type information comes general administrative data about the
sheet such as CIP3AdmJobName and CIP3AdmCreationTime. Most
important is however the CIP3AdmPSExtent attribute which gives the
exact size of the preview image. In addition to the native PostScript
language, where there exists only one measurement unit (that is points
with one point = 1/72 inch), it is possible to use a total of four predefined
measuring units within the ppf file. The available units with corresponding
abbreviations and conversion factor from points to the unit are: millimetre
(mm, 75/25.4), centimetre (cm, 72/2.54), inch (inch, 72), and point (point,
1). In the example above the numbers are given without units. If no
number is given, as in the example file, the default unit is points.

18
The CIP3AdmSeparationNames attribute gives the name of the separations
that is included. In this example the file only contains two separations,
cyan and black. The order of the separations is stated by the order they
appear in this attribute.

CIP3BeginPreviewImage holds the separation structures. Each separation

structure holds the same information about the preview images except the
actual image which is contained between CIP3PreviewImage and the >
sign. CIP3PreviewImagesComponents tells whether the preview image is a
composite image (attribute value is set to 4) or a preview image with
separations (attribute value set to 1). The size of the preview image is
given by the attributes CIP3PreviewImageWidth and
CIP3PreviewImageHeigth. In order to be a valid preview image, these
values shall match the ones given in CIP3AdmPSExtent +/- 1 pixel. With
the values given in the example file, (image width = 968 pixels, and image
height = 1288 pixels) it is easy to verify that this holds:

x ps extent 2786.43
⋅ x resolution = ⋅ 25 = 967.51 ≈ 968 pixels = image width
72 72
y ps extent 3710.55
⋅ y resolution = ⋅ 25 = 1288.39 ≈ 1288 pixels = image height
72 72

How the preview image data is stored is described by the

CIP3PreviewImageMatrix attribute. Standard orientation for PostScript
images is from left to right and from bottom to top, but the ppf allows
eight different orientations. The images generated by PrintLink are stored
from bottom to top and from right to left.

Figure 10. How the preview images’ data are stored in the ppf file.

The geometric resolution in dpi is found in CIP3PreviewImageResolution,

and the photometric resolution is given by the
CIP3PreviewImageBitsPerComp attribute, and it is either 1 or 8. When the
image is coded with one bit per component, 0 represents full ink and 1
represents no ink. In the eight bit case, 0 is full ink and 255 is no ink. This
may sound “reversed” but with subtractive colour mixture in mind “no
ink” makes sense because “no ink” yields white (unprinted paper), and is
represented on the computer screen as rgb value [255 255 255].

19
Whether the images are compressed or not is given by the
CIP3PreviewImageCompression attribute. The four compression methods
that are possible to use in a ppf file are: run length decode, DCT decode,
CCITT fax decode, or no compression. Prinergy’s PrintLink does not
support compression and there for this attribute is set to “None”.

The CIP3PreviewImageEncoding attribute tells how the information in the

preview images’ colour components is encoded. If the image has one bit
per component the encoding is binary, and if it has eight bits per
component it is either ASCII hex encoded or ASCII85 encoded. The
images from PrintLink are ASCII hex decoded, and this gives that the
colour information in one pixel is stored in two characters.

To view the preview image on the computer screen it has to be

recalculated from a subtractive cmyk colour space to an additive rgb
colour space. This is made with the following equations
R =(1-C).*(1-K)
G =(1-M).*(1-K)
B =(1-Y).*(1-K)

where .* is element-by-element array multiplication. Note that this

conversion only is a mathematical conversion based on the theory of
complementary colours. It is important to remember that rgb and cmyk
data always are device dependent and therefore will a conversion like the
one above never convert the cmyk values to the original rgb ones. This is
however no problem at all because the purposes of the preview images are
only to get a look of the content, not to view the correct colours.

20
6. Mean coverage calculations

To calculate the mean coverage for the ink zones of one separation, the
corresponding preview image is divided up into columns, each one with a
width that is matching the width of the ink zones in the press. When the
image has been divided, the mean coverage of each column is calculated.

Figure 11. Relation between image width and total width of the ink zones.

In reality it is not as simple as this, mainly because of three things. First,

the screened tint blocks at each long side of the plate have to be removed.
The mission of the tint blocks is to trap superfluous ink at the edges of the
plate, thus preventing the ink to leave the plate cylinder. Second, it is not
likely that the width of the preview image matches the width of the total
numbers of ink zones. The third difficulty is that the number of pixels per
ink zone is probably not an integer. To solve these problems, the approach
is to divide the mean calculation in four steps: i) find and remove the ink
traps, ii) add paper white to both ends of the image to make the image
width match the width of the total numbers of ink zones, iii) make the
number of pixels per ink zone an integer, and iv) divide the image into
columns corresponding to the ink keys and calculate the mean over each
key.

To find and remove the ink traps, the fact that there is always a small gap
of paper white between the ink trap and the real content of the plate is
utilized. This becomes obvious if looking in figure 12 where the ink traps
can be seen as steps with a mean coverage of around 50%.

21
 0.6

0.5

Mean coverage
0.4

0.3

0.2

0.1

0
0 100 200 300 400 500 600 700 800 900 1000
Pixel column

Figure 12. Mean coverage in each pixel column for one separation (black).
The ink traps are clearly seen as sharp transitions at both ends of the plate.
Notice also that the white paper between the magazine pages is seen in the
diagram as sections with near zero coverage.

To find where the ink traps end (seen from left) and start (seen from right),
look position by position in the array that holds the mean values. If, in the
left case, the differences between the current position and the three next
positions are higher than some threshold value, then the current position is
the end position of the left ink trap. Opposite holds for the right ink trap.
Figure 13 shows the same diagram as in figure 12, but with the ink traps
removed.

0.6

0.5
Mean coverage

0.4

0.3

0.2

0.1

0
0 100 200 300 400 500 600 700 800 900 1000
Pixel column

Figure 13. The same diagram as in the figure above, but with the ink traps
removed.

To match the widths, simply calculate the total width of all the ink zones,
and then add a number of pixel columns of white to both ends of the
image. If the number of pixel columns to add at each side of the image is

22
NC, the number of ink zones is NZ, the resolution in pixels per mm of the
preview image is imres, the width of an ink zone in millimetre is inkwidth,
and the width of the image in pixels is imwidth, then

1 
N C = floor  ( N Z ⋅ inkwidth ⋅ res − imwidth ) 1
(1)
2 

(1) always holds because the width of the largest paper that is printable in
the press is always smaller than the total width of the ink zones and as a
result of this, the total width of the ink zones is always bigger than the
width of the image.

If the number of pixels per ink zone is not an integer, the width of the
image has to be interpolated. If the actual image width in pixels is
actimwidth, the number of ink zones is NZ, and the new image width in
pixels is newimwidth, then

 actimwidth 
newimwidth = ceil   ⋅ N Z 2
(2)
 N Z 

The new image width is used as a target value when interpolating the
image to its new size. To save time, the interpolation is not made on the
entire image. Instead it is made on a vector that is the result from taking
the mean of each pixel column in the image. In order to achieve a good
result, bicubic interpolation is used.

Finally the image is divided into columns representing the ink zones, and
the mean value of each zone is calculated. These mean values are stored in
a data structure that holds all data that is needed by the IKPS. In figure 14
the complete data structure is shown.

1
Floor means that the result should be rounded towards the nearest lower integer. With,
NZ = 25, inkwidth = 40 mm, and imwidth = 968 pixels, the number of columns to add is

1
(
N C = floor  25 ⋅ 40 ⋅ 25 ⋅ 1
2 25.4
)  1 
− 968  = floor  (984.252 − 968) =
 2 
= floor (8.126 ) = 8

2
Ceil should be read out as ceiling, and the result inside the parenthesis should be
rounded upwards toward nearest integer. Continuing the example in footnote 1, the image
width when the white pixel columns have been added is 984 pixels. If this number is put
in equation (2) together with the number of ink keys as in footnote 1, then should the new
image width after interpolation be

 984 
newimwidth = ceil   ⋅ 25 =ceil (39.36) ⋅ 25 = 40 ⋅ 25 = 1000 pixels
 25 

23
Figure 14. The data structure that holds all necessary data for the IKPS.
The letters p and s should be read as f (front) and b (back). Notice that the
preview images not are stored.

24
7. Calculate the ink key settings
The problem to solve when calculating the ink key settings is: How much
should the ink keys be opened to get sufficient amount of ink in the
corresponding zones on the paper? This could also be thought of in terms
of volume.

The thickness of the ink layer on the paper is approximately 1 µm. This
thickness together with a mean coverage in an ink zone and area of the
zone gives the volume of ink that is needed in that zone. In order to
support this zone with its need of ink the corresponding ink key has to be
opened enough. The volume of ink that is available on the fountain roller
is given by the ink key width, ink key opening and radius of the fountain
roller.

Figure 15 shows a schematic side view of a fountain roller together with a

plot of how the ink area grows when the ink layer gets thicker. The area
function is

A = π (h 2 + 2rh) (1)

where h is the ink key opening and r is the radius of the fountain roller. (1)
integrated over the ink key width gives the ink volume.

Figure 15. Theoretical volume growth on fountain roller

This volume is only a theoretical one mainly because properties of the ink
affect how well it will build on the fountain roller. Mechanical properties
and status of the inking device will also affect how ink transports from the
ink fountain to the printing substrate.

25
 The idea is that this relation between area coverage and ink key opening
may be described with a transfer curve with a shape similar to the curve of
the square root of (1). To verify this idea on transfer curve shape, a number
of print jobs were collected and the mean coverage in a zone was plotted
against its corresponding ink key opening.

Zone coverage vs. ink key opening, unfiltered Zone coverage vs. ink key opening, filtered
Number of jobs: 6 Number of jobs: 6
0.5 0.45

0.45 0.4

0.4
0.35

0.35
0.3

0.3

Ink key opening

Ink key opening

0.25
0.25
0.2
0.2

0.15 [diagram: coverage vs. ink key opening] 0.15

0.1
0.1

0.05
In diagrams XXX and XXX, the curves are made of data from six different 0.05

0
print jobs in press 57. No respect has been taken to density or to what side 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
or colour the data represents.
Zone coverage
0 0.05 0.1 0.15 0.2 0.25
Zone coverage
0.3 0.35 0.4 0.45 0.5

Figure 16. Zone coverage vs. corresponding ink key opening. The diagram
to the right contains mean filtered data.

The figures above show that there exists a relation between zone coverage
and ink key opening after only six jobs examined. If the same kinds of
diagrams are plotted for the entire data set (24 different print jobs), but
with one diagram for each printing unit and side, the curve shaped relation
is even more obvious. In figure 17 below, two diagrams for the printing
unit that prints front cyan are shown. The data have been rounded toward
nearest whole number.

Figure 17. Zone coverage vs. corresponding ink key opening for the front
cyan printing unit. The data shown in the lower diagram is the mean value.

To build the initial eight transfer-curves, one per colour and side, a number
of jobs from the press that shall adopt the IKPS has to be collected along
with the CIP3 PPF files that relates to these jobs. The mean coverage

26
values over each ink zone are calculated from each preview image in the
PPF file and stored together with the corresponding ink key opening.
Because of the variation in density and ink fountain roller speed between
the different jobs some kind of smoothing has to be applied to curves.
Below in figure 18 are three examples of how these curves may look when
built from three, nine and fifteen jobs respectively, with and without
smoothening of the curve.

Figure 18. Transfer curves generated with 3, 9, and 15 jobs respectively.

27
The press dependent transfer-curves are stored in a data structure called
machinedata. The idea is that this structure shall contain all presses that
use the IKPS. The data fields ‘olderror’ and ‘oldolderror’ are explained in
the section Discussion and evaluation.

Figure 19. The machinedata structure that holds the data about the presses
that utilises the IKPS.

28
8. Using the settings in the press
Now, three steps out of five have been made: preview image extraction,
mean coverage calculation, and calculations of the ink key settings. Next is
to get the calculated presetting values to the printing press. It would have
been easy to do this if there had been a way to access the press’ control
system so that the calculated values could have been fed directly to the ink
keys like the plate-scanner does. This is as mentioned earlier in this report,
hard to do without special knowledge about the system and/or the right
equipment.

The simplest method to present the presetting values is to print the curves
on paper, one-by-one, and hand them to the press operator. What is
important is that the presetting curves are presented in a way that is
familiar to the persons that are about to use the settings. To use the
printouts to perform the presetting turned out to be difficult because it was
hard to relate the paper curve with the curve on the press’ computer
monitor. This held even if the scale and curve type (bars, dots etc.) was as
similar to the computer screen as possible. Printing the curves on paper
will however be the only way if the ink key settings are shown in another
way than on a computer monitor. This is the case for example the MAN
Roland Rotoman presses. The ink key presentation of the Rotoman is
shown in figure 20a.

Figure 20 a. The ink key presentation in a MAN Roland Rotoman D. Every

red dot is a lamp that indicates the opening of the ink key. The total width of
the table is approximately 1 meter. This width makes it hard to print the
presettings on transparent film. Printing the presettings on a large format
ink-jet printer may however be a solution.

29
Instead of using printed curves on paper, the curves were copied onto
transparent film. The transparent film was then attached to the computer
screen in the press, matching the diagram showing the ink key settings.
Now there was no problem to adjust the ink keys to match the suggested
presetting curve.

Figure 20 b. In order to facilitate the ink key presetting a transparent film,

with the pre-setting curve printed on it, is attached to the computer monitor
in the press. The bars, representing the ink key settings, have not yet been
adjusted to their suggested positions.

30
9. Transfer curve calibration

In order to improve the quality of the presetting calculations, calibration in

some way has to be made to the transfer-curves. The calibration part is
divided into two steps: collecting data in the press, and processing of the
collected data in order to improve the transfer curves so they perform
better presettings. It is important that the ink key data that is about to be
collected represents ink zones with density and dot gain within given
tolerances.

Collecting approved data in the press

Starting with the data collection in press 57, the same conditions hold as
for the presetting of the ink keys in the press. This means that there is no
way to get the ink key settings by numbers as a data file, instead other
techniques have to be used to collect the data. The different techniques
divide into three main groups depending on the level of user interaction:
manual methods, semi-automatic methods, and automatic methods.

Figure 21. Problem to solve when collecting data in the press: computer
screen diagrams to numerical values.

Manual methods
The most basic and straight forward manual method is to simply look at
the approved key settings diagrams on the computer screen in the press
control room and take down, on paper or directly into a data file, the
estimated settings, colour by colour for both front and back.

Another manual method that facilitates the estimation and input of data is
to use a computer program, in which the values may be inputted via a
graphical user interface (GUI). Instead of doing this on location in the
press, reading directly from the press’ computer screen, photographs of the
approved key settings may be taken with a digital camera. This is the
method used and it will be explained below.

31
Semi-automatic methods
Methods that belong to this category are methods that needs human
guidance of some kind to be able to extract data automatically. The
photographs in the manual example above are suitable for this purpose.
This method is described later in this chapter under the heading First steps
towards semi automatic feedback. Due to time shortage, this method has
only been tested briefly.

Automatic methods
If the key settings data (or other press data) were available via some kind
of data interface as formatted data, the automation of key settings data
collection would be as simple as extracting information from a text file.
The sheet fed presses from KBA at SGQ has this possibility. Information
about a print job may be stored in the press’ control computer as a
formatted text file. This file may then be saved to a PC-formatted 3½”
floppy disc for further processing.

If the computer screen images could be captured directly from the graphic
card and saved to a computer as images, the semi-automatic method
described above could be turned into an automatic method.

The method used for data collection

The method used is using digital photographs of the approved key settings
together with a GUI for the input of data. In the press’ control room,
pictures were taken of the approved settings of the ink key values. The
resolution of the digital pictures was 1600 x 1200 pixels and the pictures
were jpeg compressed in the camera. Each image was roughly 600 kB of
size. On the computer screen in the press, see figure 22 below, the ink key
settings are displayed as bar graphs in two different scales, 0 to 33 1/3 % or
0 to 100 %, depending on the value of the ink key with the largest opening.

32
 Figure 22. Photographs of the ink key settings, taken with a digital camera,
as they appear on the computer screen in the press’ control room. The top
picture has a scale ranging from 0 to 33 1/3 %, and the scale opof the bottom
one is 0 to 100 %. Two bars belongs to each ink key: the blue(left) one is the
current key setting, the green(right) one is the presetting value as estimated
by the plate-scanner.

From digital photos to numbers

To make the transfer of the bar graphs into estimated numbers easier, a
program in which it is possible to use the mouse to feed the values into the
computer was made. The GUI of this program is shown in figure 23. When
all the values have been fed in, the job to which this values relate is picked
in the list box, then the save button is pressed, and the feedback values are
stored in the data structure shown in figure 14.

33
 Figure 23. The GUI of the ink key feedback program. The mouse is used to
point and set the values in the graph. Notice the scale button which makes it
easy to change the scale to make the graph look the same way as in the
press.

Processing of the collected data

Basically, the error between the presetting and the approved setting for ink
key n for one collected job can be described as:

errorn = approved setting n − pre setting n , n = 1, 2, ..., N where N = number of ink keys (3)

It is important to understand that this error calculated in (3) theoretically

not is ink key depending, instead it is dependent of the coverage of the ink
zone that corresponds to ink key n. The ink key number is of use when it
comes to keep track of what values of a job that belong together: mean
coverage, presetting, and approved setting. Mean coverage together with
the transfer curve that gave the presetting gives what value in the transfer
curve errorn shall affect. The problem that arises is how much of the error
that should be fed back in order to improve the transfer-curves. Transfer
curve calibration is discussed further in the section Discussion and
evaluation.

First steps toward semi automatic data collection

With use of digital images of the ink key settings it is possible to perform
semi automatic extraction of their values. The foundation on which this
method rests is the use of digital image processing together with
knowledge about the image content. By applying bright red diamonds at

34
strategic places on the computer monitor in the press control room it is
possible to compensate for image rotation and scale. Knowledge about the
digital camera and the control room monitor makes it able to correct the
images for barrel distortion. Manually selecting the threshold level makes
it easy to avoid unwanted parts of the images to survive after the threshold
step.

This has only been tested very briefly. Some examples of how the images
may be looking after correction and threshold are shown below. The next
step after threshold, which has not been implemented yet, is to extract
what numerical values the bars represent.

Figure 24. Top right is the original image. Top left is the original image
after geometric correction. The bottom threshold picture is the blue channel
of the original rgb image.

35
10. Discussion and evaluation

Testing the IKPS

One way to test the IKPS would have been to set it up to work with a press
that makes ink key presettings with a plate scanner. When the print job is
running and density and dot gain are within the given tolerances, the final
key settings could be captured and compared with the suggested settings
from the IKPS. This is however not possible because it was obvious quite
early during this diploma work that the resulting key positions are
dependent on their initial settings.

Plate-scanner performance
To have something to compare the behaviour of the IKPS with, data was
gathered from print jobs in another press. This press, a Rotoman D built
1993 (in house named press 54), uses a plate-scanner to perform the ink
key presettings. The reason why the plate-scanner that is used for press 57
not was chosen is because that scanner never had been optimised in order
to improve the presettings. Improvement of the presettings, that is making
feedback from the press to the plate-scanner, is something that is made on
regular basis in press 54.

The procedure of data collection was as follows: The key settings made by
the plate-scanner were captured right after the keys had been set (before
the press had started). The final key settings were then taken when the
print job was both approved and had the density and dot gain within the
given tolerances. The speed of the ink fountain roller was also recorded for
both the presettings and the approved ones. Because the density and dot
gain were within the tolerances no printed sheet was taken.

In appendix 1 it is clearly seen that the scanner makes a good job.

However, the scanner may sometimes fail. This is seen in appendix 2.

IKPS performance with the data set used for development

The first step in order to use the IKPS was to construct the eight different
transfer-curves that were used to perform the transfer from mean coverage
to ink key opening. The initial curves were made up from three different
print jobs simply by connecting the mean coverage values with their
corresponding ink key openings. The curves that were generated initially
gave a rough hint of the shape of the transfer-curves.

It was hard to make good presettings with the collected data set. With
some print jobs the results were really good whilst other jobs only showed
a rough similarity between the presetting suggested and the approved ones.
The reason to this is mainly the variation in density across the printed
sheet for some print jobs that was collected. Some of the jobs have all their
density values near the target ones, while other jobs show big density

36
variations. The idea when the data was collected was that if the data was
collected as soon as possible after the balanced sheet counter had started it
should be easier to make a transfer curve that made presettings that quickly
took the press to this state, thus reducing both time and wasted paper. It
showed however that this was not a good idea because the press had not
reached a stable balance between ink and water at such an early time in the
print run. Instead of collecting data at such an early point during the print
run, the ink key settings with its corresponding reference sheet should have
been taken no earlier than after 20,000 impressions. After this point the
press is considered to be in a stable state during the rest of print run. This
fact was not known during the data collection part of this diploma work.
Some examples of presetting suggestions made in the collected data set are
seen in appendix 3 and 4.

IKPS performance when the system was used with press 57

The IKPS was used to preset the ink keys for four jobs in press 57. To
make the transfer curves the collected data set was used together with
experience from the development of the IKPS. Note that these transfer
curves not are calibrated. The result from two of the jobs is shown in
appendix 5a, 5b and 6a, 6b. It is clearly seen that the shape of the
presettings are almost perfect but in some cases the curve is translated
(appendix 5b, and 6b, back magenta). This may depend on the transfer
curve making to high presettings. It may also depend on to high ink
fountain roller speed.

One problem as mentioned above is that when making the presettings in

the press no value for the speed of the ink fountain roller was available.
During the tests of the system the printers suggested the fountain roller
settings. The collected data set was of no use when it was processed in
order to find a relation between mean ink key opening and the speed of the
fountain roller. A reason to this may be that there are several ways to run a
press when it comes to balance between fountain roller speed and total ink
key opening.

Testing the transfer curve calibration

Due to time shortage it has only been possible to test the IKPS in press 57
with four different print jobs. Because of the low number of tested jobs it
has not been possible to read back ink key data to the IKPS in order to
check the behaviour of the calibration algorithms. Instead the the
algorithm has been tested theoretically as described below.

The idea was to construct two transfer-curves. The first curve, A, simulates
how the press converts mean coverage into ink key settings. This is the
target curve for the transfer curve used by the IKPS to preset the ink keys.
The second curve, B, is a curve that is an estimation of curve A. Curve B
is made from curve A, but translated and with added noise. Equations for
curves A and B are shown below along with plots of the curves in figure
25.

37
 1
A = 0.05 + sin( x )
2
1 11 
B = sin( x ) +  − r 
2 52 
 1 π 2 π 3 π 99 π π 
x = 0, ⋅ , ⋅ , ⋅ , ..., ⋅ ,
 100 2 100 2 100 2 100 2 2 
r is a vector with 101 positions, each positions holds a random number between 0 and 1

Transfer curves for verification

of the calibration

A
B

0.5

0.4
Ink key opening * 100%

0.3

0.2

0.1

0
0 10 20 30 40 50 60 70 80 90 100
Coverage %

Figure 25. Plot of the curves used for testing the feedback algorithms.

Curve B was used to calculate preset values from mean coverage values,
21 at a time, simulating real print jobs. The number 21 may seem odd
because the press have 25 ink zones but the mean coverage values in the
collected data set had the ink traps included. In order to remove the ink
traps afterwards the two outmost ink zones at both sides were omitted
which yields 21 valid zones. The same mean coverage values were then
used together with curve A to simulate approved key setting values. A
number of six jobs were simulated and then used for calibration. Another
six job were then simulated with new mean coverage values together with
the new transfer-curves that were given from the calibration step.

This test is quite far from the real situation because of the perfect feedback
values. In reality there will always be variations in density across the
printed sheet. It is also possible that the press has minor mechanical
problems (i.e. with the rollers and/or the ink fountain) that affect the
behaviour of the transfer-curves without causing any problem with the
printing process. In order to simulate the variations in the feedback values
a small amount of noise was added to curve A. Equation of the new curve,
A', is shown below. Worth noticing is that the noise in curve A' differs

38
between the simulated jobs while the random noise that was added to
curve B was the same for all the first six jobs. Curve A is still the known
target transfer curve.

1 1 1 
A' = 0.05 + sin( x ) +  − rn 
2 15  2 
 1 π 2 π 3 π 99 π π 
x = 0, ⋅ , ⋅ , ⋅ , ..., ⋅ ,
 100 2 100 2 100 2 100 2 2 
rn is a vector with 101 positions, each positions holds a random number between 0 and 1,
the subindex n goes from 1 to the total number of jobs

As mentioned in the previous chapter Transfer Curve Calibration, the

error between the suggested settings and the approved ones can be written

errorn = approved setting n − pre setting n

n = 1, 2, ..., N where N = number of ink keys

In a perfect world without any variations in the approved values errorn

should be the true difference between the transfer curve used for presetting
and the true transfer curve that is unknown. When curve A' is used to
simulate the approved values it is no longer true that errorn shows the
correct error. In order to protect the transfer curve from putting to much
trust in incorrect differences between approved settings and presettings
only a fraction of the error is used to calculate the new transfer curve. The
equation above then turns into

errorn = K p (approved setting n − pre setting n )

Kp is a factor that tells how much of the proportional error that should be
used for correction. A large value on Kp gives a fast but instable
calibration and opposite a low value gives a correction that moves slowly
but stable towards the unknown target curve. To avoid instability Kp is
often smaller than 1. This creates another problem that is related to
proportional correction. The problem is that a proportional correction with
Kp smaller than 1 corrects the transfer curve to a position slightly below
the unknown target curve. A solution to this problem is to rely more on
older errors. Now the error to correct becomes

error = K p (approved setting − pre setting ) + K I old olderror + K I oldold oldolderror

Olderror and oldolderror are simply vectors of the same length as the
transfercurve vectors but instead of containing ink key opening values for
a specific mean coverage an error vector hold the errors between the
presetting and the approved values at a specific mean coverage. At the first
calibration round olderror has some values but oldolderror is still zero.
After the second calibration round both vectors contain values. KI old and
KI oldold are factors that tells how much of the old errors that shall be used

39
during calibration. If the K factors are too large the system will start to
oscillate. This is shown in figure 26 below.

Calibration simulation with self oscillation.

All K factors larger than 1.
6 calibration rounds with six jobs at each round.

0.5

0.4
Ink key opening *100%

0.3

0.2

0.1

0
0 10 20 30 40 50 60 70 80 90 100
Coverage %

Figure 26. The black curves are the red transfer curve after each
calibration round.

If instead more cautious K values are chosen the calibration behaves, as

seen in figure 27, more civilized.

Calibration simulation.
Kp = 0.7, Kiold = 0.1, Kioldold = 0.1

0.5

0.4
Ink key opening *100%

0.3

0.2

0.1

0
0 10 20 30 40 50 60 70 80 90 100
Coverage %

Figure 27. The black curves are the red transfer curve after each calibration round.

40
Even if the feedback algorithm corrects the transfer curve towards a better
one, it is hard to create a perfect transfer curve. A perfect transfer curve is
a curve that is perfectly smooth and increasing. This means that the curve
should not contain any local extreme points (minima, maxima, or
horisontal parts) and as a result of this the curve is one to one. It is
impossible to avoid defects (local maxima or minima) due to variations in
density during feedback. Even if it is possible to avoid some irregularities
by using different kind of smoothing techniques on the curves, most of the
problem consists. Using too much smoothing on the curves also gives
problems with accuracy.

A way around the problem with irregular transfer-curves is to introduce

the possibility to smoothen the curves by hand. A person with knowledge
about how the transfer-curves work and are supposed to look like will have
no problem to adjust the curves in order to avoid distorted parts. The result
of this will be curves that give high accuracy after a minimum of
feedbacks.

It is really important that the jobs that are used for feedback are relevant.
In this case relevant jobs are jobs with density and dot gain within the
given tolerances for the press that are using the presetting system. If there
exists different sets of target values, the set that is most frequently used is
chosen. As seen in appendix 7 and 8 the height of the transfer curve
depends on the target density used for feedback. In appendix 7 transfer-
curves derived from values below density - tolerance, within target density
± tolerance, and over density + tolerance respectively are plotted. With the
data set used for this report the relations between the derived transfer-
curves are most clearly seen for the printing unit that prints front cyan.
Even if this relation is not so clear for the other printing units it is still, as
seen in appendix 8, a difference between using too low density and using
too high. The reason why the difference is so small between using too low
respectively too high density may depend on the reason that there exist a
lot of values that are just a little under or over the target density ±
tolerance. It is also very important that the ink fountain roller speed is
unchanged from the speed that was given from the preset system.

41
11. Conclusions
The IKPS works and makes as good ink key presettings as the plate-
scanners. The biggest advantage with the IKPS is that it is an offline
system and therefore it is possible to use together with any offset press, old
as new.

It is very important that jobs used for calibration of the transfer curves
have a high printing quality with respect to density and dot gain.
Calibrating the curves with jobs that have to large density variations across
the sheet may result in transfer curves that not are one-to-one. Such defects
are edited by hand but with repeated calibrations with bad data it may be
hard to figure out what the transfer curve should look like.

Even if the system makes good offline presettings, the system will
however not be really user-friendly until it can deliver the presetting
values directly to the ink keys, and when the settings are approved,
automatically perform collection of the current key settings. Making
transparent films with curves representing the key settings works but
includes to many steps to be really useful when it comes to real
production, at least at SGQ where the printers are used to plate scanners
that automatically presets the ink keys.

Instead of using PPF files as preview image carrier it is possible to use

other image formats. Quick tests with 25 dpi eight bit cmyk-tiff files
generated by Prinergy have been made with good result. The idea is to
extract the cmyk separations from the tiff files and store them separation
by separation inside a PPF file. The use of PPF in this case is only because
IKPS is programmed to work with this format.

Further investigation is needed in order to find a way to preset the ink

fountain roller. The intention was that this should have been implemented
in the system but the data set was of no use when it came to find a relation
between mean ink key opening and the fountain roller speed.

42
12. Recommendations for further work
The first thing to do is to make further investigations on how to preset the
speed of the ink fountain roller. Second is to establish a contact with the
major press manufacturers in order to get information on how to access the
control systems of the different press types that are to adopt the IKPS. The
system should however keep the possibility to work offline in order to
support older presses without an online interface for ink key presetting.

If it not becomes possible to collect data automatically from the press, it is

of interest to develop data collection by the use of image analysis. Quick
tests have been made with the digital pictures taken during the work with
IKPS and it has been seen that image analysis have the possibilities to
work well.

How to make the presettings of the ink keys by hand if no access to the
press’ control system is given is also interesting to improve. One
suggestion is to use some kind of video mixer in order to show the
presetting curves directly in the press’ monitor along with actual ink key
positions.

Some kind of compensation for different kind of paper qualities maybe has
to be implemented. This is a function that is available in many plate
scanners.

43
13. Wordlist
Back The side on a printed sheet that contains the second page of
that sheet

CIP3 International Cooperation for Integration of Prepress, Press,

and Postpress

CIP4 International Cooperation of Processes in Prepress, Press,

and Postpress

CMYK The names of the colours used in subtractive colour mixture

(printing). It stands for Cyan, Magenta, Yellow, and blacK or
Key-colour.

Front The side on a printed sheet that contains the first page of that
sheet

GUI Graphical User Interfaced

HSWO Heat Set Web Offset

IKPS Ink Key Pre-setting System, used as an abbreviation for the

system that is described in this diploma work

JDF Job Definition Format

PPF Print Production Format

Press 57 The in house name at SGQ for the Harris Heidelberg 850
printing press.

Prinergy A prepress system developed by Heidelberg and Creo Scitex.

PrintLink A plug-in for Prinergy that generates the CIP3 PPF files.

RGB The name of the colours used in additive colour mixture

(used in TV-displays and computer monitors): Red, Green,
and Blue.

SGQ The abbreviation used in this report for Sörmlands Grafiska

Quebecor AB.

44
14. References
M. Sonka, V. Hlavac, R. Boyle, Image Processing, Analysis and Machine
Vision, Crawfordsville: PWS Publishing, 1999, ISBN 0-534-95393-X

T. Glad, L. Ljung, Reglerteknik. Grundläggande teori, Lund:

Studentlitteratur, 2:dra upplagan 1989, ISBN 91-44-17892-1

K. Johansson, P. Lundberg, R. Ryberg, Grafisk kokbok 2.0 – guiden till

grafisk produktion, Bokförlaget Arena AB, 2:dra upplagan 2001,
ISBN 91-7483-161-1

Building GUIs with MATLAB,

http://www.mathworks.com/access/helpdesk_r11/help/pdf_doc/matlab/gui/
buildgui.pdf (Available 2002-05-16)

S. Daun, G. Lucas, J. Schöntut, Specification of the CIP3 Print Production

Format version 3.0,
http://www.cip4.org/documents/technical_info/cip3v3_0.pdf (Available
2002-05-16)

Instructional Package 506/6-13: Inking System,

http://graphics.tech.uh.edu/Press%20I/Inking_System.html (Available
2002-05-16)

TIFF Revision 6.0,

http://partners.adobe.com/asn/developer/PDFS/TN/TIFF6.pdf (Available
2002-05-16)

45
15. Appendix

Appendix 1. Good scanner performance

46
Appendix 2. Less good scanner performance

47
Appendix 3. Preset suggestions made on the collected data set

48
Appendix 4. Preset suggestions made on the collected data set

49
Appendix 5a. Results from using IKPS in press 57

50
Appendix 5b. Results from using IKPS in press 57

51
Appendix 6a. Results from using IKPS in press 57

52
Appendix 6b. Results from using IKPS in press 57

53
Appendix 7. Relation between coverage and ink key opening.

54
 Ink zone coverage vs. ink key opening
side: front, colour: yellow
40
Under target density − tolerance
Within target density +/− tolerance
Over target density + tolerance

35

30

25

20

55
Ink key opening %
15

10

5
Appendix 8. Relation between coverage and ink key opening.

0
0 5 10 15 20 25 30 35 40 45 50
Coverage %
Appendix 9. Example CIP3 PPF file.
%!PS-Adobe-3.0
%%CIP3-File Version 2.1
CIP3BeginSheet
/CIP3AdmJobName (ark5_7 SIG005) def
/CIP3AdmJobCode () def
/CIP3AdmMake (Creo) def
/CIP3AdmModel (Prinergy) def
/CIP3AdmCreationTime (Thu Aug 09 07:39:45 2001) def
/CIP3AdmSheetName (Sheet 1) def
/CIP3AdmPlateType (CutSheet) def
/CIP3AdmPSExtent [2786.43 3710.55] def
/CIP3AdmPlateTrf [1 0 0 1 0 0 ] def
/CIP3AdmPlateExtent [3710.55 2786.46] def
/CIP3AdmPressTrf [1 0 0 1 0 0 ] def
/CIP3AdmPressExtent [3710.55 2786.46] def
/CIP3AdmPaperTrf [1 0 0 1 -82.205 420.945 ] def
/CIP3AdmPaperExtent [3571.65 3628.35] def
/CIP3TransferFilmCurveData [0.000 0.000 1.000 1.000 ] def
/CIP3TransferPlateCurveData [0.000 0.000 1.000 1.000 ] def
CIP3BeginFront
/CIP3AdmSeparationNames [ (Cyan) (Black) ] def
CIP3BeginPreviewImage
CIP3BeginSeparation
/CIP3PreviewImageComponents 1 def
/CIP3PreviewImageWidth 968 def
/CIP3PreviewImageHeight 1288 def
/CIP3PreviewImageMatrix [0 1288 -968 0 1288 0 ] def
/CIP3PreviewImageResolution [25 25] def
/CIP3PreviewImageBitsPerComp 8 def
/CIP3PreviewImageCompression /None def
/CIP3PreviewImageEncoding /ASCIIHexDecode def
CIP3PreviewImage
928c8d8b8c8b8c8c8a8c8c8a8c8b8c8c8a8d8c8c8b8b8c8b8b8b8c8b8b8d8b
8b8b8c8c8b8b8c8c8b
8a8c8c8c8b8c8c8b8b8c8b8b8c8c8b8b8b8c8b8c8b8c8c8a8c8c8a8c8b8c8c
8a8d8c8b8c8b8c8b8b
<-- Lots of image data omitted -->
47474747464747464846484846484646484647474748464647464746474847
474746484646484648
474648464847464847474747474747474747484748484648
>
CIP3EndSeparation
CIP3BeginSeparation
/CIP3PreviewImageComponents 1 def
/CIP3PreviewImageWidth 968 def
/CIP3PreviewImageHeight 1288 def
/CIP3PreviewImageMatrix [0 1288 -968 0 1288 0 ] def
/CIP3PreviewImageResolution [25 25] def
/CIP3PreviewImageBitsPerComp 8 def
/CIP3PreviewImageCompression /None def
/CIP3PreviewImageEncoding /ASCIIHexDecode def
CIP3PreviewImage
928c8d8b8c8b8c8c8a8c8c8a8c8b8c8c8a8d8c8c8b8b8c8b8b8b8c8b8b8d8b
8b8b8c8c8b8b8c8c8b
8a8c8c8c8b8c8c8b8b8c8b8b8c8c8b8b8b8c8b8c8b8c8c8a8c8c8a8c8b8c8c
8a8d8c8b8c8b8c8b8b
<-- Lots of image data omitted -->
47474747464747464846484846484646484647474748464647464746474847
474746484646484648
474648464847464847474747474747474747484748484648
>
CIP3EndSeparation
CIP3EndPreviewImage
CIP3EndFront
CIP3EndSheet
%%CIP3EndOfFile

56