Ellipse Vrs Dot shape benefit "example attached"

[yorkie]

 author=drdot link=topic=13.msg1979#msg1979 date=1304472169]
Finally, elliptical, ellipse, diamond, euclidian elliptical, etc are all variation of elliptical dots. The generic family of ellipse is simply dots based on different aspect ratios.
[/quote]

I believe that one the underlying problems is that the ratio isn't "really" changing, just the shape of the dot within the cell. Attached is shape i've created, but have not had the time to test.

This example is a round dot halftone at a 2:1 film ratio. This results is elliptical dots oriented to the horizon (regardless of angle). The horizontal elliptical dots are intended to counteract textile moire.

Rather than a 45 LPI or 55 LPI, the halftone might be 60 LPI in one direction, while 30 LPI in the other, or a 50/25 lpi.

[drdot]

[I believe that one the underlying problems is that the ratio isn't "really" changing, just the shape of the dot within the cell. Attached is shape i've created, but have not had the time to test.

This example is a round dot halftone at a 2:1 film ratio. This results is elliptical dots oriented to the horizon (regardless of angle). The horizontal elliptical dots are intended to counteract textile moire.

Rather than a 45 LPI or 55 LPI, the halftone might be 60 LPI in one direction, while 30 LPI in the other, or a 50/25 lpi. [/quote]

This is an interesting example. I don't know how you built this example, but here are a few things to consider. The actual LPI represents the individual halftone cell which is then made up of the spots or pixels of the output device at the output resolution.

For simplicity sake, let's assume the halfone line count is 50 lpi and the output resolution is 600 dpi. This means each halftone cell is 1/50". Each halfone dot then contains 600/50 or a grid on grid of 12 spots or for each halftone dot 144 individual spots from the output device.  Each tonal step would be 1/144 or .69 of one percent for each gradient division between white and black.

You are correct in describing a different apparent resolution in one direction over the other. The problem comes with all ordered array halftones, typically referred to as AM or amplitude modulated halftones, the size of the actual dot will vary directly with the tonal percentage.

This is why moire is so common in highlight and quarter tones. The dots are much, much smaller than they are at the midtone. In fact, the largest dot area for any lpi, as defined by the perimeter of the dot, is at 50%. Dot area mirrors on either side in positive/negative relationship. So, a 40% positive dot is the same size as a 60% negative dot.

The way the dot builds inside the individual cell is determined by the Postscript Spot Function. The halftone designer will create a matrix that assigns each pixel a threshold value for it to turn on.  As an example, a 16 x 16 pixel grid is typical to represent 255 steps from white (all off) to black (all on.) Each of the 255 pixels in that cell has an address that is then mapped to the exact change in tone. When that gray tone value presents, the pixel turns on and the dot builds.

The order and shape of the array is entirely up to the designer and the possibilities are almost limitless.  I should point out that it is rare for anyone to work directly in native Postscript. This is the software that drives the RIP. Some RIP software will allow access to the Spot Function, but it is not common.

[Dottonedan]

Can't wait to chime in but need to put it off for a while.  Exciting stuff. I love this kind of gab.

[squeezee]

Some RIP software will allow access to the Spot Function, but it is not common.

You can play with the spot function in the Wasatch RIP, it's in lib\basicspot.ps
Here's an elliptical function:

/Ellipse6040
   { exch dup mul 0.66 mul exch dup mul add 1 exch sub }

You can change the ratio of the axes by varying the 0.66 figure - 60/40 = 0.66, 70/30 = 0.42.
Do a backup before messing with any settings!

[yorkie]

The order and shape of the array is entirely up to the designer and the possibilities are almost limitless.  I should point out that it is rare for anyone to work directly in native Postscript. This is the software that drives the RIP. Some RIP software will allow access to the Spot Function, but it is not common.

I would be one of those rare individuals who speaks the postscript language fluently. I first learned postscript in about 1989 when i hacked freehand (1.0) to print to a hacked lasertwriter II ntx to print digitally onto attached removable platter hard disk, then swap the disks platter into a mac II i wrote a program to read the image data from a postscriot formated disk and converted the image files to VSAM onto an IBM 4381 mainframe (all new at the time). Then i spent a decade doing other fun projects involving postscript. Anyone remember "lasertalk"? 

As a hobby I also write my one device drivers for ghostscript.

Every RIP allows the access of the spot function, it is the application that limits its availability, except for some special purpose rips like the one epson sells for the large format printers, which replaces halftone dots with dithered dots.

[drdot]

@Yorkie  and anyone else who is deeply interested, Adobe Press has an excellent book on Postscript Halftoning written by Peter Fink way back in 1992 or so. I remember poring over it trying to understand halftone arrays, moire periodicity, super cells, irrational angles, and so on. It is still one of my core reference books on digital imaging along with Ulrich's Digital Halftoning.

Here is info on Fink's Book:
      Postscript Screening: Adobe Accurate Screens
     explains in detail all the issues and specifically covers Adobe
      Accurate Screens -- Adobes' own screening technology. Peter Fink
      is an expert on the subject of halftone screens as they relate to
      color issues.

      Author: Peter Fink
      Publisher: Adobe Press, Oct 1992.
      ISBN 0-672-48544-3

[yorkie]

The things which interest me, go beyond "accurate screening".

"in"-accurate screening was a byproduct of the limitation of the computers of pre-1992. It has been 19 years since 1992. Computers back in the day were 30 mhz, not dual or quad 3 ghz.

It is my belief, that screen printing on textiles needs its own book.

Please go back and review the not proportional ratio screen i posted. Is that in the book?

[drdot]

It is my belief, that screen printing on textiles needs its own book.

Please go back and review the not proportional ratio screen i posted. Is that in the book?

I think you're missing the point. Both of the references I cited are foundational material to build your skill set. As an example, the post on moire and "strobe effect" would be enlightening as what you describe is, indeed, a form of moire. The discussion in Fink's book on accurate screens, etc, has more to do with moire issues than on processing speed of the computers. Have you read it?

Your interest in  a "not proportional ratio screen" is nothing more than a hybrid screening option.  Mike Ware of Waatsch Technologies has developed a wonderful hybrid Stochastic/AM screening model available in their Soft RIP product. It is ideal for screen and flexo printers.  It's his approach to dealing with highlight and shadow issues inherent with AM halftones. It is his attempt to solve the same problem you are proposing to solve with your idea.

Postscript offers almost infinite possibilities on how to create a halftone model depending on the printing method, ink characteristics and behavior, and substrate. What you outlined is one possible solution.
Structurally it has some issues, but that also is not the point. Experiment. See what you get. That's how discoveries are made.

Any halftone approach is a balance between visual surface texturing (noise), accurate tone translation, and stable consistency of production.  The model you are proposing is still an ordered array model (AM) and any ordered array model will have challenges for screen printers.

For us, the ideal halftone model would be an intelligent mix of FM and AM halftoning based on the rate of change of surface detail over any defined area. What this means is that the algorithm would analyze the image and where the detail was low (eg sky, large areas of flat color) it would place an AM or ordered array dot. Where there is rapidly changing detail (hair, fabric texture, regular patterns like picket fence or screen door mesh) you would get an FM dot of a predetermined dot size.

This approach is processing intensive, but offers the possibilities to balance extremely fine detail with precise tone control. It's a way of losing the visual dot as part of the printed image.

[squeezee]

Have you seen Mike's latest wheeze?
Imagine that you can print from 10-90% quite happily but lose detail after that.  Mike has got a trick where the dots are postscript down to 10% and remain at that size so that you can print them.  To get a 9% dot he removes some of the dots and progressively to 1%.  It doesn't look perfect but it's a really good approximation.  It's a sort of stochastic approach to AM dots.

[Clark]

Have you seen Mike's latest wheeze?
Imagine that you can print from 10-90% quite happily but lose detail after that.  Mike has got a trick where the dots are postscript down to 10% and remain at that size so that you can print them.  To get a 9% dot he removes some of the dots and progressively to 1%.  It doesn't look perfect but it's a really good approximation.  It's a sort of stochastic approach to AM dots.

I think my old multi light exposure table used to do this on its own.....lol

[yorkie]

Have you seen Mike's latest wheeze?
Imagine that you can print from 10-90% quite happily but lose detail after that.  Mike has got a trick where the dots are postscript down to 10% and remain at that size so that you can print them.  To get a 9% dot he removes some of the dots and progressively to 1%.  It doesn't look perfect but it's a really good approximation.  It's a sort of stochastic approach to AM dots.

That is exactly the sort of thing i was discussing in my latest post.

http://www.norwb.com/index.php?topic=142.msg2263#msg2263

[yorkie]

It is my belief, that screen printing on textiles needs its own book.

Please go back and review the not proportional ratio screen i posted. Is that in the book?

I think you're missing the point. Both of the references I cited are foundational material to build your skill set. As an example, the post on moire and "strobe effect" would be enlightening as what you describe is, indeed, a form of moire. The discussion in Fink's book on accurate screens, etc, has more to do with moire issues than on processing speed of the computers. Have you read it?
[/quote]

No, i never bought that book. In 1992, i was a member of the adobe developers association. Everything that was anything to the postscript language was documented in the developer mailings, including floppies with containing code.

The term "accurate screens" involves the ability to select a specific angle/frequency, rather than the snap value to the nearest adobe screen angle/frequency. This was 100% computer speed related. The snap values match th memory layout of the ram. 1200 dpi does not divide evenly by 55. Either you divide 1200 by 21 pixels and get 57.14 lpi or 22 pixels and get 54.54 lpi.

Postscript offers almost infinite possibilities on how to create a halftone model depending on the printing method

Postscript 3 only implemented the framework for advanced screening, but left it to the manufacturers to implement in the rip. If i'm wrong, you should be able to reproduce a none proportional screen.

For us, the ideal halftone model would be an intelligent mix of FM and AM halftoning based on the rate of change of surface detail over any defined area.

For me, I just want to print pretty shirts. My pet peeve is moires. Beyond FM and AM is morphing shape of the spot and the shifting the phase of the spots. Processor intensive isn't much of an issue these days. A 60 meg raster files is trivial on todays computers.

[blue moon]

bump!

this is a great post that many of the new members have not seen. I thought it would good to bring it into the spotlight again.

pierre