Image REs and Line screen

[orka81]

Hi all,

Got into a discussion with the bosses earlier about image resolution vs. line screen. I know that the general rule of thumb is that the image res should be twice that of the line screen ruling.

We have Rampage 10.5 and run a CoRes Square dot. Our output res is 2438 with a line screen ruling of 200.

My question: does that rule still hold fast?

We occasionally get phone calls from photographers asking for specs. In the past I have told them 300 dpi, 100% to size for optimal results. The bosses want me to tell the designers 400dpi.

Thanks in advance,

orka81

 

[buckeye]

And that's just what it is, a general rule. Your bosses will never see a difference between 300 and 400 dpi. See previous thread on this subject.

http://printplanet.com/forums/prepress-workflow-discussion/16067-genuine-fractals

Erik

 

[gordo]

We occasionally get phone calls from photographers asking for specs. In the past I have told them 300 dpi, 100% to size for optimal results. The bosses want me to tell the designers 400dpi.

Just adding a bit to the thread buckeye pointed you to.
You should check your RIP (and PDF creation) settings to make sure it is not resampling images. Many (most?) RIPs are set by default to resample all images down to 300 dpi - so even if one wanted to use a higher dpi image, e.g. for FM screening, the extra dpi could get lost in the RIP unless that setting is changed.

BTW "seeing is believing" If you're a commercial printer you might want to run a test on the next job that has some waste/unused area or run a dedicated test. Take a high res image and resample it down to these resolutions in PShop 100 dpi, 200 dpi, 300 dpi, 400 dpi, and 600 dpi. Output at your 200 lpi and see if you/your bosses can see a difference. Ideally the image would be of a building - an image that contains lots of fine lines at different angles.

best, gordo

 

[orka81]

Gordo,

Thanks for the reply. Testing an image is a viable solution.

 

[gordo]

Gordo, Thanks for the reply. Testing an image is a viable solution.

Just remember to always resample the original high res to each of the lower res versions (i.e. resample the 600 dpi to 400, 300, 200, 100 NOT 600 dpi to 400 dpi then 400 dpi to 300 dpi, etc.)

BTW, If you need images to test with, you could go to a free stock photo agency like: http://www.morguefile.com

Good luck!

gordo

 

[meddington]

As gordo pointed out, some will opt for a higher res when printing with FM screening, as a 300 dpi image might actually look softer when printed with FM as opposed to say, 175 lpi. You might find that FM screening can be "improved" with higher res images or images with increased unsharp masking.

 

[gordo]

As gordo pointed out, some will opt for a higher res when printing with FM screening, as a 300 dpi image might actually look softer when printed with FM as opposed to say, 175 lpi. You might find that FM screening can be "improved" with higher res images or images with increased unsharp masking.

That is basically correct. You would use a higher res image in order to take advantage of the higher res FM screen. The slightly softer look with FM usually happens in two cases. First, when comparing a 25-35 micron FM screen against a coarse AM screen - e.g. 85-110 lpi as used in newspaper printing (the coarser the screen the greater the apparent contrast). The second, and very common cause of a softer look with FM results from the resampling of the image in the RIP. If the RIP is set to honor the res of the image and not resample, then, as an example, a 175 lpi image and a 20 micron FM image should look virtually identical (from a "softness" point of view) in the final presswork.

best, gordo
my print blog here: http://qualityinprint.blogspot.com/

 

[bmor]

Back when photoshop came with a printed manual they actually devoted afew pages to address this. They (Adobe) printed samples of something like 3x line screen, 2x line screen, 1.5x line screen etc. You could not see any image degradation above 1.5x line screen ruling.

 

[[email protected]]

DPI vs LPI

DPI vs LPI

Before fm screening et al, imagesetters used a halftone cell of 16 dots to reproduce the 256 shades of grey available in PostScript. Ergo divide your dots per inch by 16 and you get your lines per inch e.g 2540 dots per inch divided by 16 gives 158.75 lines per inch.

Now if your imagesetter has a fixed cell size it will not make any difference, but if it is able to vary the halftone cell size then you can set 175 lines per inch at 2540 dots per inch, but you reduce the number of shades of grey. If you cell size reduces from 16 dots down to 12 dots then you only get 144 shades of grey, but you can get 211.6 lines per inch from 2540 dots per inch.

Of course all of the above is irrelevant if you are using an alternative screening method such as frequency modulated.

If you are creating PDFs for print they are designed for output at 150 lines per inch using 2400 dots per inch assuming a 16 dot halftone cell - hence photographs (which need to be between 1 times and 1.5 times the line screen) should be 150 dpi minimum and 300 dpi maximum - anything above this is wasted in a standard PDF. Adobe for a fuller explanation.

 

[gordo]

To Anthony Minchinton -

You are mixing up gray levels capability and original image dpi/ppi requirements relative to halftone line screen ruling.
For the gray levels capability, you are describing screening/imaging device resolutions that existed prior to about 1987. A full explanation is available here:
Quality In Print: Halftones and gray levels • Part One
and the second part here:
Quality In Print: Halftones and gray levels • Part Two
The gray level limitations relative to output device dpi was overcome some 20 years ago and is implemented in most, if not all, modern RIPs.

As far as image dpi requirements relative to halftone line screen ruling is concerned - the issue is primarily one of the the frequency (ppi) of the image pixels vs the frequency (lpi) of the halftone dots. If the lpi is greater than the ppi, then the halftone will resolve the individual pixels of the image. You can test this yourself. Here's an example:

In the center of the image is the original. On the left is the image with a resolution (ppi) of 1/2 the lpi used to screen it. You can see the pixels that make up the image clearly. On the right is the same image but with a resolution equal to the lpi. You can barely make out the image pixels. The dpi of the RIP and lpi of the halftone in both cases is the same.
The 2x lpi for establishing image dpi comes from and old data sampling rule which doesn't quite apply to halftoning. In many cases you can use an image dpi that equals the lpi and you'll be fine. 2x gives you a safety margin. With FM screening, it's not uncommon to use the AM equivalent to determine image ppi. E.g. a 20 micron screen is about the same as a 385 lpi AM screen so use a 400 ppi image.

Hope this clarifies, gordo

my print blog here: Quality In Print

 

[gordo]

So if your printing at 2438 DPI, than LPI should be about 150 LPI. A higher line screen frequency can be problematic if you print on uncoated paper.

Really? So, what's problematic?

Gordon P

my print blog here: Quality In Print

 

[[email protected]]

Hi Guys,
some of the replies to the original post do not make sense here in England and I was wondering if this is due to a different technology being employed. What follows is a short history of imagesetter technology sold here in england:
1st GENERATION IMAGESTTERS
Used a grid of 4x4 (=16) dots in each halftone cell. Each dot had one fixed size. Each dot could be either on or off. Therefore you have 16x16=256 different combinations or shades of grey. Which just happens to be the same number of shades in an 8 bit per channel image.
2nd GENERATION IMAGESETTERS
Used a grid of 4x4 (=16) dots halftone cell. Each dot had two fixed sizes. Each dot could be off, on at size 1 or on at size 2. Therefore you have 16x16x16=4096 different combinations or shades of grey. This technology also proved popular with desktop laser printer manufacturers an example being the Apple Colour LaserWriter; for those of you who go back that far.
3rd GENERATION IMAGESETTERS
Used a grid of 4x4 (=16) dots halftone cell. Each dot had three fixed sizes. Each dot could be off, on at size 1, on at size 2 or on at size 3. Therefore you have 16x16x16x16=65,536 different combinations or shades of grey. Which just happens to be the same number of shades in a 16 bit per channel image.
4th GENERATION IMAGESETTERS
Used a 4x4 (=16) dots halftone cell. Each dot had four fixed sizes. Each dot could be off, on at size 1, on at size 2, on at size 3 or on at size 4. Therefore you have 16x16x16x16x16=1,048,576 different combinations or shades of grey.
AMPLITUDE MODULATION & FREQUENCY MODULATION
Marketing terms to describe (obfuscate) how the dots are placed in the 4x4 grid. Try it yourself: draw a 4x4 grid and then try to fill in 50% of the squares. You will notice that there are many ways to do this: you could work from the centre out, you could work from the outside in, you could start in a corner or you could create a checkerboard effect (fm screening)
5th GENERATION & BEYOND
We now have imagesetters with multiple sized dots or variable sized dots along with multiple sized halftone cells; at which point the imagesetter manufacturers are now just showing off.
PHOTOSHOP
So Photoshop tells you to make your image twice the value of your LPI simply because each line in the Lines Per Inch is made up of a grid of 4x4 (=16) dots. Divide 2400DPI by 300DPI and it goes in four times in each direction ergo each pixel maps directly to each halftone cell.
COMMENTS PLEASE
Does this make sense to you guys on the other side of the pond?

 

[gordo]

[Big Snip]
Does this make sense to you guys on the other side of the pond?

Nope.

And I believe the technology is the same here as over there (except that over here 2400 dpi is the most common resolution whilst 2540 dpi is most common over in Europe). Also, it's not the imagesetter that builds the screens - it's the RIP.

The cell size in which the halftone dot is formed is variable - it is not fixed at a 4x4 pixel matrix as you suggest. On a fixed resolution output device it can range, depending on the requested lpi, from about a 3x3 pixel matrix (capable of 10 possible levels of gray) to a 23 x 23 pixel matrix (capable of 530 possible levels of gray).

Let's take a simple example to make the math easier. We'll make our halftone dot cell a 2x2 pixel matrix. By turning on pixels within that cell we are only capable of simulating 5 levels of gray:

To increase the number of gray levels we could divide the halftone cell into a grid of smaller pixels:

What we've done is increased the dpi of the device - e.g. gone from 2400 dpi to 4800 dpi.
However if we can't, or won't, increase the device resolution to achieve more gray levels, there is another solution.
The way that virtually all modern halftone screening gets around the gray levels limitation imposed by AM screening and device resolution is by using a process called "dithering" - which is what I think you are trying to describe. "Dithering" leverages the fact that we see tone areas rather than individual dots.
Rather than using one halftone dot to generate a tone, dithering uses a "supercell" made up of a number of individual halftone dot cells. Then by alternating (dithering) between the different halftone dots that can be achieved, we create a tone area representing the average tones in the supercell and therefore a tone value we could not otherwise represent.
So, taking our original 2x2 halftone cell, we might define a Supercell made up of, for example, alternating 25% and 50% dots. The result will be a tone area that simulates a 37.5% value. A tone value we could not otherwise represent with our 2x2 pixel halftone cell.

There are other techniques - but the Supercell method is by far the dominant one.
It has nothing to do with "generations of imagesetters" Supercell screening was introduced around 1990 or so. It was adopted, and very well marketed, by Agfa (in their Agfa Balanced Screening) as a viable alternative to, for example, the likes of Heidelberg's IS hardware based screening.

Regarding: "AMPLITUDE MODULATION & FREQUENCY MODULATION Marketing terms to describe (obfuscate) how the dots are placed in the 4x4 grid."
IMHO, these are very good terms that describe the basics of these two approaches to screening. In one, the amplitude is indeed varied and not the frequency. In the other, the frequency is indeed varied and not the amplitude. Both AM and FM have hybrid variants - where the technologies are mixed, however using the AM and FM models is still useful for understanding a specific screen's underlying structure.

Regarding: "So Photoshop tells you to make your image twice the value of your LPI simply because each line in the Lines Per Inch is made up of a grid of 4x4 (=16) dots."
IMHO, nope. I think the PShop programmers simply implemented the industry convention that ppi should be 2x lpi. It has nothing whatsoever to do with gray levels. It has to do with the spacial resolution of the frequency of halftone dots as illustrated in my previous post with the sailboat image. Both halftones have the same resolution, both have the same number of gray levels. One resolves the pixels of the original image because the screen has a higher resolution than the image. The other does not resolve the pixels of the image because the screen has a lower resolution than the image.

hope this clarifies, gordon p
my print blog here: Quality In Print

 

[bharatk]

Gordo,

Thanks for the clearification.

All my career while outputing AM I always made sure the pics were of 300dpi. Good to be reminded about the 2x principal especially regarding FM screening and its AM screen equivalent.

Bharat

 

[gordo]

Don't forget that there are at least two places where the resolution of the image can be altered and destroy your hard work. First is when the document file is converted to PDF either by the document creator or by the workflow during the refine process - if the settings are left to default then the images will be resampled to a lower resolution. And second, when the image is RIP'd - most, RIPs are set up by default to resample images to 300 dpi.
Needless to say, in both cases, the option to resample should be turned "OFF"

best, gordo
my print blog here: Quality In Print

 

[claude72]

Really? So, what's problematic?

The dot gain is higher with uncoated paper... and with a 150 lpi or higher screen, gray and CMYK pictures that are not correctly prepared for uncoated paper are printed too dark on uncoated paper (and it's worst with recycled paper).

I think the PShop programmers simply implemented the industry convention that ppi should be 2x lpi.

Industry convention or "the 300 ppi urban legende"?

Not all the industry: I have seen some softwares having an 1.5 quality factor as default set, and I read on Adobe's web site an advice to use an 1.5 quality factor to reduce the size of the pictures files.

All my career while outputing AM I always made sure the pics were of 300dpi.

When I got with my first Agfa scanner, I read in its manual that for offset printing:
• line-art pictures have to be scanned at 1200 ppi,
• contone picture for FM "Crystal Raster" screens have to be scanned at 300 ppi...
• contone pictures for AM screens have to be scanned with a resolution:
- twice the screen ruling at 133 lpi and below 133,
- 1.5 x the screen ruling over 133 lpi...

... so, working with AM screens, the only case where you really need 300 ppi pictures is when you output 200 lpi screen!!!



In France, before the CTP platesetters, and still now for those who still work with films, the standard screen ruling is 150 lpi, and then everybody uses the basic standard 300 ppi resolution, applying (more or less knowingly) a 2x quality factor (or applying the urban legend)...

... but if you ask the question "Why 300 ppi?" to professionnal printers and designers, many of them will simply answer "because it's that" or "because I always did this way"...

... but some guys (a little bit more aware) will explain that using a resolution twice the screen ruling is a safety, made to allow an expansion of the picture up to 130% (in fact, 133,33 %) without any trouble, and without any loss of quality...

And if you calculate the real output resolution of a 300 ppi picture after a 133,33% expansion, you will see that its resolution drops to 300/133x100 = 225 ppi...
... and 150 x 1.5 = 225... so, 225 ppi is the right normal resolution to be used with a 150 lpi screen and the normal 1.5 quality factor which match with the 150 lpi screen ruling.


The funny thing, is that today these same guys carry on using the same rule of 300 ppi pictures expandable up to 133%, but they now work with 175 lpi screens, having then an only 1.29 quality factor... and they are still perfectly satisfied with the quality of the printed jobs!!!

 

[[email protected]]

=>257 grey levels

=>257 grey levels

Hello Gordo, sorry for the delay in replying (I have been unwell). I have tried reading your blog (which is excellent by the way) and somewhere in it you mention about supercells being able to give more than 257 levels of grey.
My apologies for my lack of knowledge in this department, but how does one employ these greater than 257 levels of grey?
Whenever I convert a Photoshop image to CMYK it only produces 257 levels in each of the CMYK channels.
If I use an 8 bit RGB image again I can only get 257 levels per RGB channel.
If I place a 16 bit RGB image in InDesign it gets output with only 257 levels per channel.
Again the same is true in Illustrator.
Am I misunderstanding what you are saying - if so my apologies in advance - or is it possible to produce more than 257 levels per CMYK channel when making plates? If so could you please point me in the right direction to achieve this?
Finally may I say how informative your blog and comments in PrintPlanet are, many depend upon your extensive experience which you are happy to share without charge, a truly commendable and appreciated person.

 

[gordo]

Hello Gordo, sorry for the delay in replying (I have been unwell). I have tried reading your blog (which is excellent by the way) and somewhere in it you mention about supercells being able to give more than 257 levels of grey.
My apologies for my lack of knowledge in this department, but how does one employ these greater than 257 levels of grey?
Whenever I convert a Photoshop image to CMYK it only produces 257 levels in each of the CMYK channels.
If I use an 8 bit RGB image again I can only get 257 levels per RGB channel.[snip]
Am I misunderstanding what you are saying - if so my apologies in advance - or is it possible to produce more than 257 levels per CMYK channel when making plates? If so could you please point me in the right direction to achieve this?

Welcome back!

You are mixing up two separate issues.

An 8 bit image (e.g. a greyscale photo or one of the color channels in a 4/C image) has 256 grey levels and you can't generate any more.

In order to be printed on press that image needs to be converted from a continuous tone greyscale into a bilevel (black and white only) halftone. The halftone dots simulate the grey level tones in the original image. Big dots representing dark tones while smaller dots represent lighter tones.

Now, the output device will generally have a fixed resolution - 2400 dpi, 1200 dpi or whatever.
That fixed resolution limits how many grey levels you can create with your halftone.
This principle is captured by the classic formula:

(dpi/lpi) squared + 1 = number of possible grey levels

For example let's say you have a 1200 dpi CtP device and you want to use a 100 lpi halftone screen. This is what the formula gives you:

(1200/100) squared = 144 + 1 = 145 possible grey levels.


So, because in this example you can only generate 145 grey levels then you're missing 111 grey levels (256-145 = 111) in reproducing the image. So there will be shade stepping (jumps in tones where they should be smooth) in smooth gradations because you do not have enough grey levels in your halftone screen to represent the 256 grey levels in your original image.

However in the late 1980s a work around to this problem was introduced called "supercell" screening.

AFAIK all modern RIPs today use supercell screening so there is no real relationship anymore between halftone lpi and imaging device dpi.

For example, the highest lpi on a 2400 dpi CtP device that I'm aware of was 1697 lpi on a poster printed with plates imaged on a Creo CtP device in 2000 by Metropolitan Fine Printers in Vancouver Canada. It won a "They said it can't be done" award at GrapExpo in Chicago.

What supercell screening does is based on the fact that the dpi/lpi formula is for a single halftone dot. But that's not what we see when we look at a printed image. Instead we see areas of tone. So, supercell screening looks at an area of tone and creates the requested unavailable tone by averaging tones that can be achieved in that area. For example, if the requested tone is 11% and we can't create a true 11% halftone dot, we use instead a mix of 10% and 12% halftone dots. That mix averages out to an 11% tone. So we see an area of tone that looks like 11% even though there aren't any actual 11% halftone dots in that area.

Make sense?

best, gordo

 

[Lukas Engqvist]

@claude interesting, but one thing is missed with the 1.5x rule and that is that some times there are details that have similar characteristics as do line art. So again it boils down to looking at the image...*I know we had this discussion in another thread about rasterised text, but only have the discussion in my peripheral memory.

 

[RickLee]

I have never liked supercell.

if you are mixing the percentages like 10% and 12% to get a pseudo 11% this will only look goo with one or two steps available.

as you climb higher in the LPI the fewer levels of grey are available. If you start mixing 1% and 4% to get 2% or 3% there is some artifacting that occurs. The higher the worst the artifacting. I know that the layman may have a hard time distinguishing this but try telling that to the agency that is looking at a silver metallic car that is complaining about the "feel" of the tone not being right but cannot figure out the problem.

supercell does not solve all of the problems. but no one is offering a way to optimize the supercell with the proper LPI/DPI/supercell tech.

for example using 2400 DPI can I create a smooth gradient at 600LPI from 1% to 10% over a 6" area? what if this is a photo with the same type of gradient?