More digital camera PPI confusion

Yes, an image of 280ppi will print nicely. But at what physical size? The Kodak has dummied down settings, not very precise. Just shoot at the highest quality it gives – that will give you the most pixels to work with after the fact, simple as that.

PPI only has meaning at a given output size, as to a printer.

PPI no meaning viewed on screen. On screen, at full size, the size of the image is determined by how many pixels it contains vs the physical size of a monitor and pixel dimensions running on that monitor.

Mac

Marcia,

A pixel has no dimensions. It is a triad—three numbers representing the tonal values in each of three color channels.

You could be an Egyptian artist in the year 2000BC, laying out an image for the gods to view in the desert sands composed of 1-ft squares, each a different color corresponding to a modern-day digital camera’s pixel array output. Or you could be the queen of a beehive with the god-given ability to fill each honeycomb cell with individual colors also corresponding to the camera’s output. Or you could be a 6-year old child filling the squares in a coloring book to correspond with the camera data.

The Egyptian and the queen bee and the child receive the same input with the same inherent resolution. The image they are creating is broken down into the same total number of elements. But they each construct the image in their own way. They end up with the same number of pixels but different numbers of "pixels per inch or per foot or whatever the ancient Egyptians used for linear measurements.

You would never be confused by what each is doing with the same input. And you could measure their "ppi’s" and not be confused. Printers and monitors are no different. They take the same input and lay it out their own way. Why should this trouble you? 🙂 🙂

George

Interesting example George. I like it.

Marcia,

In order to print really crisp text in a brochure I go for 600PPI.<<

There is a big difference between printing rasterized text and printing images. Text prints as a solid (it isn’t screened) and therefore ideally needs to be at the printer’s native resolution if you want the output to match that from vector text.

Image data is screened, and therefore the required image resolution is a function of the screening algorithm and not the printer resolution.

Len,

that´s only partly correct. Gray text or arbitrary process color text (both vector) is screened.
But the screening algorithm tries to retain the crispness of the contours. These contours are available on device level with "infinite" accuracy and reproduced with device resolution.
Inspecting gray text by a magnifier shows the raster cells.

Best regards –Gernot Hoffmann

Gernot,

Gray text or arbitrary process color text (both

vector) is screened.<<

Granted. Only solids will not be screened

Len,

thanks for the confirmation. It´s not a trivial question. For a while I thought that the screening program would average the source pixels of a raster image in the raster cell (like a scanner). A long discussion in the Google PostScript Forum led to the conclusion that my opinion was wrong.
The screening program uses the source pixels as they are (virtually mapped onto the paper). Therefore raster images and vector graphics are handled just the same (and it´s possible to draw gray lines which look sharp).

A simulation of the screening process is here in the chapter Raster Halftoning : <http://www.fho-emden.de/~hoffmann/hilb010101.pdf>

Best regards –Gernot Hoffmann

Gernot:

Your PDF is interesting though the math is beyond me, as it is with most I fear.

Yet it speaks of an important issue, namely the screening of very high resolution raster images, specifically as it relates to rendering very fine type. And here I am thinking: product shots.

Conventional wisdom states that you should work in 300ppi for offset printing, or 2x the screen-ruling for optimal results. One also hears that smaller amounts like 1.8x screen ruling is sufficient.

And the same rules would have you work as close to final size as possible, avoiding scaling in a layout application (a rule that not many actually follow).

So if you worked in 4x your final size @300ppi and reduced everything 25% in your layouts, you would effectively send 1200ppi to the printer. A real no-no right?

In the case of very small type, the digital screening process would map this type between the halftone dots so to speak, and said type would hold up under magnification with a loupe very well. The same size type prepared at 300ppi final size would probably be comparable to the naked eye, but upon magnification would not be readable at all, just as zooming up in Photoshop would reveal mush.

Similarly, the same size image produced the old way through a halftone screen, would also be way-broken-up by the halftone dots, so this illustates the advantage of digital screening.

Of course, the type of rip which handles this type of imagery has a lot to do with the outcome. A CT/LW (ScitexCreo) type of rip, which reinterpolates all rasters, would destroy all that extra resolution, so you would be no better off than if you worked 300ppi at final size. But a rip which generates screened plotter files from the print requests would show this benefit.

Now don’t get me wrong. I’m not advocating working at an effective rez of 1200ppi for everything. That would be idiotic. Just small product shots, and books… things that you scan or shoot and you have a lot of them, and the idea of replacing type elements with vector type is prohibitive.

Then there is a bigger question that needs asking: Is it worth it to store, and print all that extra data to improve the readability of type which can only be read under high magnification? Answers will vary on that one. My answer has always been, "No, it’s not worth it, that type is not meant to be read anyway". But now I don’t know, it’s kind of like keeping everything in 16bit. Even though nobody else can easily see the difference, you somehow feel better holding on to that difference.

John,

you are right – sending too many source image pixels will spoil the "spots" (please let me call these clusters spots, opposed to dots (smallest device pixel)). The spots are fragmented then.
This can be observed in PhS if you convert a normal image to Bitmap with very low Lpi settings, e.g. 12 Lpi. I thought this were a bug but it isn´t.

Your comment "Scitex reinterpolates all rasters" is interesting. I had printed 1000×1000 pixels in stamp size and by 20 inch width using a large format RIP. Both prints are OK – no artifacts. There seems to be a correct handling in the program:
a) Reduce too much information by blurring (small print).
b) Interpolate if not enough pixels are available (large print) .
The recommendation 1.5…2 source pixels per raster cell width is probably still valid, but it´s not critical, IMO, depending on the RIP quality.

Of course, the high resolution of the imagesetter or printer IS necessary for all kinds of vector graphics even if these are rastered.

Another interesting question: what´s the thinnest printable line ? This is not a line printed by dots (1/dpi width) but printed by spots (1/Lpi width). One can print single dots by an imagesetter, but a gray line consists of spots.

Best regards –Gernot Hoffmann