NAGASA - Channel Links: Information Requests and Reports

Topic: CTP - Why densitometers do not work

Date: May 30, 2000

Steve Colthorpe * & Greg Imhoff **

Keywords

CTP, Lithographic, Dot Area, Densitometer, Plates

Abstract

The advent of CTP has highlighted a problem with quality control, namely that
conventional densitometry is not reliable enough for consistent readings of half
tone values, making calibration difficult to nearly impossible.

Fortunately there is a solution which has been used in laboratories for several
years based on video frame grabbing techniques, more recently using a CCD
camera. This technology has now been implemented in a hand held unit known
as a digital DotMeter, and a description of some of the problems encountered
and the solutions will be explained, together with a comparison with analogue
densitometers.

Why densitometers do not work

Densitometers have been with us in the Graphic Arts industry for many years
now. It has been the one tool available to us that has the ability, though in
practice seldom used, of putting some science into what is still considered by
many to be an art form. Its use for the stable production of film output, and
calibration thereof cannot be ignored. Neither can its use in the pressroom for the control of ink weights and associated factors be understated. However when
it comes to the area of checking quality of printing plates from a CTP system,
this is one area that the densitometer cannot be applied.

Who says?

3M-Imation in a presentation at these proceedings in 1997. ¹

Another important body of opinion to share this view is that of other plate
manufacturers, all of whom rely on planimeter and / or microscopy techniques
to obtain accurate results.

Thirdly, densitometer users themselves - in many respects THE most relevant
witnesses in the case - have found that these devices are less than helpful in a
CTP environment.

And, last and probably least, I am saying so right now!

Why don't they work?

Before the advent of CTP the production of printing plates was a reasonably
predictable process with various control mechanisms in place to ensure faithful
transfer of known film to plate. The film is of course 'known' thanks to the good
ol' densitometer. These transfer techniques of film to plate usually include the
reproduction of a grid / wedge to give the correct exposure level. Such systems
have been emulated quite successfully ² on CTP systems. However exposure
level alone is not the only requirement when it comes to CTP. Linearisation of
the percentage dot on the plate is also a function of the recorder aperture and
size ³ and not just exposure level.

The grain profile of the plate will also have an effect on the accuracy of the
densitometer readings, as will variations in plate material and emulsion. On
plate the actual density of the emulsion is generally non-critical and certainly on
thermal plate can only be of interest to the offset lithographic plate
manufacturer. In general the density (solid tone) on plate bears no relationship
as to its ability to hold or reject ink and transfer to paper. What is critical is the
percentage dot areas on the plate that will transfer an image to paper. That is
what this business is all about

Another important aspect is the human factor. Operators may have specified an
incorrect tint value. Hopefully this would get noticed on film, either by eye or
densitometer, before the press was run. This is far more difficult with CTP
where the densitometer cannot be relied upon. For example, how can you check
a high-light dot in the middle of a plate, away from manual calibration areas.

Reflectance itself brings further problems. If the surface being measured was a
perfect mirror, then the reflected density would be infinite, because all of the
light would get reflected straight back to the source with none ever reaching the
sensor (light detector). If the surface was totally randomly rough then a random
proportion of light would get into the sensor.

This phenomenon is, of course, only due to the surface reflectance and has
nothing to do with measuring the reflected density, other than make it harder. To
over-come this problem a well known modification to the famous Murray-Davies
equation - the Yule-Nielsen equation - was developed, originally for the
specific application of ink on paper. When applied to plates it can only be
effective provided the user knows the 'n' factor for the plate for each type of
plate.

Finally there is also a potential problem with electronic stability which will
become more and more critical as the density range to be worked with gets
smaller.

Hence my opening comment that densitometers 'do not work' on plates and
therefore cannot be relied upon. To justify this we need to look at the workings
of a reflection densitometer.

How densitometers try to work

Fig 1 depicts a typical model which complies with the ISO specification.⁴ for
such devices. A collimated light source is used to illuminate the sample and
detected by a sensor at 45 degrees. This is then converted to density via a log
amp and displayed on a meter, perhaps with other compensations applied.

When attempting to use a densitometer to measure percentage dot, it is
necessary to know the white and black level of the plate. Here in practice is the
first problem. Namely that any variation in the white or black level across
the plate will have a direct effect on any reading taken. In practice variations of
+/-13% have been seen across the plate background (white area). Again the only
way that a densitometer can be used as a DotMeter for plates is by very careful
calibration of the black and white levels within very close proximity of the
target area. This also assumes that you know the 'n' factor for the plate.

The accuracy of the result obtained is the next issue to address. If the 'black' of
the plate is say 1.0 D, this means that in reflectance terms 10% of light is
reflected, and the 'white' by definition is 100% reflected. A 50% tint area is
made up of equal areas of 'black' and 'white' so that in reflectance terms a total
of 55% (i.e. half of the white plus half of the light from the black) is reflected.

As density = -log (Reflectance) this equates to a density of 0.2596 as compared
to the 0.3 one would expect in a perfect world. A chart showing the effect of
Dmax is shown in Fig 2 below.

Though this is of course taken care of by the densitometer manufacturers as far
as the user is concerned, the fact is that the smaller the density range to work
with the more susceptible the unit is to errors due to electronic drift from noise
and temperature.

Where densitometers go wrong

A comparison between the arrangement of dots on a piece of film and the
equivalent arrangement on an average printing plate makes it easy to understand
why a densitometer has little trouble giving a reading from the former, whereas
it encounters enormous problems in interpreting the latter.

The distribution of dots on the smooth surface of the film is easily calculable
and hence represents something very much like an 'Ideal World'.

In the 'Real World' of the printing plate, however, there is no such comfort. The
'noise' from the plate material makes it effectively impossible to obtain trust-worthy
results with traditional methods and devices designed to operate on film.

Densitometer - image view	DotMeter - image view
Which one is 50% ?

What are the options?

So how can we measure percentage dot area on CTP and conventional plate
systems?

There is the age old 'eyeball' method which, by definition, implies a significant
element of what we must call 'guesswork'. Following this route, you could
expect a level of accuracy that has a potential error margin of +/- 20%. At some
time the skilled 'dot etcher' will leave the company, then the delta may become
+/- 30%

A good loupe and experience will bring it down to +/- 10%.

You could use a densitometer and hope for the best and you should get to within
+/- 5% of the correct reading.

For many years plate manufacturers and other research laboratories around the
world have used video techniques involving either planimeter or computed
results. The planimeter in this application should really be called a 'cyborg' as it
is part computer and part human. A photomicrograph is taken of the dot and is
then put on a digitising tablet and traced around by hand. The area of the dot is
then calculated by the computer. The application of these techniques leads to a
dependable accuracy to within +/- 1%.

The important part of this process is that the human operator decides where to
'threshold' the image, that is define the border between the black and white
levels of the plate. Clearly, however, such an instrument is not particularly
portable or quick to make readings. For many years now there have also been
all-electronic planimeters using a microscope, video camera and computer
system ⁵ . Using off the shelf components has again meant a lack of portability.

Within the last few years advances in electronics have enabled the realisation of
a portable system, and Centurfax was the first company to realise this concept in
the commercial world when a working device was openly demonstrated at
Imprinta 97.

Many of the practical problems of using such a system have previously been
highlighted and include:

1. Focus of image
2. Exposure control
3. Thresholding
4. Aperture errors.

These problems have been addressed by the development of a commercially
viable DotMeter.

What is a DotMeter?

A DotMeter works via the principle of combining a CCD camera with a
microscope. The camera takes a 'snap-shot' of the area being measured and
literally counts the black and white pixels in the image. Rather than taking an
average of dot density (as with a densitometer), the DotMeter is actually
measuring image area and providing an absolute value of dot coverage.

Other key features of the best DotMeters are that it can offer automatic
calibration in a single shot (rather than having to be re-calibrated against a
known value area before each new reading); it provides additional data on
screen ruling and screen angle; and it furnishes users with the ability to read
film, plate and paper with one instrument.

It should be stated at this point (April 99) that - to the best of our knowledge -the
above mentioned facilities are currently only offered by the CCDot from
Centurfax.

How a DotMeter works

The design of a high quality DotMeter includes a glass disc of approximately
25mm diameter which is placed on the sample plate. This design makes is easy
to keep the medium flat, thus maintaining good contact with the sample. It is after all a camera and focus is critical. Depth of focus is typically less than
0.2mm for any such system.

A key design criterion of a high quality DotMeter was to ensure that errors in
focus would not be introduced into the unit as a result of movement, as in the
case of the typical 'stapler', 'clam shell' or pressure sensitive 'anvil' units based
on traditional densitometer design. Instead, a high quality DotMeter uses a
partially silvered mirror to give a genuine WYSIWYG viewing system with no
moving parts which makes it virtually wear and maintenance free for the user.

Exposure control is another very crucial area in setting up any image analysis
system. Most video systems have automatic gain control built in. This will tend
to reduce the contrast when placed on, say, a 10% tint as it tends to take the
average and not the peak density. The other possibility is to set exposure level
dependant on the white and black levels in much the same way as a conventional
densitometer. This approach used by some suppliers does make the system
susceptible to variations in density, just like a conventional densitometer. A high
quality DotMeter on the other hand always finds the best contrast between
printing and non-printing areas irrespective of the tint value being examined. A
complementary coloured light source guarantees optimum contrast and ensures
that process colours are measured accurately. It can take a little while to
determine the optimum value as the nearer the value is to a limit, the more
difficult it gets to distinguish signal from noise.

Providing that the exposure level has been set correctly, and that the image is in
focus, then the image may be thresholded at 50% value with very minor errors.
The biggest single problem is the level of optical noise from the medium,
however using DSP (digital signal processing) techniques this noise can be
overcome. Assuming that a high quality DotMeter is used within its working
range (typically 85 lpi - 215 lpi) then errors due to the aperture can be kept to
within 0.5%.

Where do you need a DotMeter?

The answers to this question are simple and relatively obvious. A DotMeter is
vital for linearising any CTP RIP and is the only viable tool for professional
Quality Control to within any acceptable level of accuracy.

This is all very well if the object of the exercise is only the pursuit of technical
excellence. But is this merely the beauty of the abstract? Is there really any
commercial benefit in being sure that you have got it 'nearly right' or is this just
a lot of fuss attempting to seek unnecessary levels of near perfection? Surely
there is more to it than this.

A recent submission by a major US print corporation offered the results of
research and analysis into the actual cost of errors in the printing industry. This
high-profile group produced an estimate that the average direct cost incurred is
approximately US$650 per error, which makes a device to avoid these problems
excellent value for money.

This in itself is significant enough. But, to the best of our knowledge, this figure
does not take into account the enormous hidden cost of repeat business lost as a
result of the errors in question.

What we have to ask ourselves is: "What is the 'n' (nuisance) factor for these
hidden costs?"

I offer you the following equation as an expression of this potential commercial
catastrophe.

where 'n' = 'no profit'

What's the 'n' factor? Well, working it out exactly is of course only marginally
relevant, and the equation like the real solution is impossible to solve, but would
probably feature a very large number ending in zeros!

TOP