How do you work with these large file sizes

[Bob McCarthy]

I'm not refering to the size in megabites but resolution. I've been scanning at 2400 dpi to maximize quality. File are typically 10,000+ by 7500+ pixels. When printing I've been setting printer resolution at 600 ppi (native on HP DJ), I'm still having to downsize with bicubic "sharpener". on many prints

Should I be using qimage or am I on the correct path?

Is 2400 dpi overkill?

bob

[Jeffrey Sipress]

2400 may not be overkill for the scan, but you don't need to print at more than 200-300 dpi. You will see no difference. And, I too, often need to downsize for my final output. I use the Fred Miranda WP Pro utility, which basically does the reduction in steps, with optional sharpening. When you get to the final size, then sharpen to print.

[Michael Gordon]

Bob: you need to conduct your own tests, but mine have proven that 600dpi vs. 300dpi on the DJ130 provides only a micro gain, and it's only visible with a loupe. I print everything on the DJ130 at 300dpi.

[Bob McCarthy]

Maybe I've been sniffing the QImage marketing line too long. Great program for really big (size) jobs. We do banners around here occasionally.

Most everything else we do at 300 or 360 dpi. I have read that HP and Epsons run at higher resolutions internally (600 dpi-HP & 720 dpi-Epson). One can bypass the internal upresing by printing at native. Probably a bunch of bunk.

I use PK sharpener pretty much exclusively for the sharpening side at all three levels, (capture, creative, output).

How destructive is bicubic I think is my question?

Bob

[Ken Lee]

Scanning at 2400 ppi and sending 360 dpi to the printer, means you can make an "enlargement" of 2400/360 or 6-7X, and the results should be critically sharp. (By that, I mean that someone can stick their nose right up to the image and it will look rather sharp, as long as the image itself is sharp).

This means that if you use 6x7 film, which is 2 1/4 x 3 1/4, you can expect a very nice 11x14.

Given that many scanner manufacturers over-state their resolution, you might want to limit your enlargment to 5X or less.

To print a 4x5 image at 12x15 at 360dpi, you can downsize to 4320x5400.

[Ed Richards]

> Scanning at 2400 ppi and sending 360 dpi to the printer, means you can make an "enlargement" of 2400/360 or 6-7X, and the results should be critically sharp.

Unless you are scanning with a drum scanner or a Creo or the like, you are not really getting 2400 DPI. I scan at 4800 and then downsample to about 1800, which is closer to the real resolution. The oversampling reduces noise, and the files are about 1/2 the size of 2400 DPI. I then use Quimage to handling the printing. Its biggest value for me is not having to resharpen images for different print sizes.

[MJSfoto1956]

you are confusing resolution with dots per inch (DPI). Since a tone of any color cannot be made up of one dot of pure color there needs to be a "spot" consisting of a stochastic (random) pattern of dots. In this manner, the number of "spots" per inch would be roughly equivalent to your resolution. A 1440dpi printer has to generate larger "spots" than a 2880dpi printer in order to reproduce the same tonality for any color. In the old days we would call this Lines Per Inch since the spots were in a regular pattern. Typical was 150 LPI. You can still see this today by looking at most any printed magazine with a loupe. The familiar pattern of cyan magenta yellow and black spots will reveal itself. However, some printed publications have completely moved to stochastic screening and thus are analogous to your inkjet output.

The rule of thumb is that the eye starts to blend things together around 150 SPI (or LPI). So in general, while you might be sending a 360 pixel per inch image to your Epson, my own tests reveal that you'd be hard pressed to see any difference if you sent a 300 pixel per inch image instead -- this is especially true if you choose 720 dpi or 1440dpi instead of 2880 dpi.

J Michael Sullivan
Editor/Publisher, MAGNAchrom

[Ken Lee]

I mention 360 because I have often read (here and elsewhere) that the Epson drivers deal with increments of 360 (720, 1440, etc.) as opposed to Canon, which is based on 300. According to this common wisdom, using integer multiples of the appropriate number will result in smoother rendering of diagonals, etc. Is this an "old wive's tale" ?

[Bruce Watson]

I mention 360 because I have often read (here and elsewhere) that the Epson drivers deal with increments of 360 (720, 1440, etc.) as opposed to Canon, which is based on 300. According to this common wisdom, using integer multiples of the appropriate number will result in smoother rendering of diagonals, etc. Is this an "old wive's tale" ?
No. The people on the yahoo group Epson Wide Format (http://groups.yahoo.com/group/EpsonWideFormat/messages) argued this over and over with many tests. The group seems to have found a consensus that the Epson wide format printer drivers want the input file to be at 360ppi, while the desktop drivers want 720ppi. The drivers will in fact up and/or down-rez as required before they apply the dithering algorithms.

That said, it seems to be a characteristic of the driver, not the hardware. Since most RIPs bypass the OEM driver and instead implement their own, if you are using a RIP all bets are off. My experience with StudioPrint is that it will use as much resolution as you give it. The more you give it the better the print - but it's a curve of diminishing returns as you would expect.

But, in general, Epsons want input resolutions that are multiples of 360ppi while HP, Canon, and Chromira drivers seem to want multiples of 300 ppi. Lightjets want some interesting sizes depending on the model of the Lightjet - that 304.8ppi number is really 120 ppcm for example. The higher resolution model wants 160 ppcm (406 ppi).

So, just like you would expect, one should size one's file based on how one is going to use it.

[Ed Richards]

> So, just like you would expect, one should size one's file based on how one is going to use it.

Or use Qimage or a RIP and do not worry about it. Qimage is not a RIP, hence the low price, but does handle talking to the printer and rescaling the file very well.

[mdd99]

What MJSfoto1956 said. It's right on.

[Ken Lee]

Bruce - Thanks for the tip about Epson printers: desktop versus wide format.

According to the QImage web site (http://www.ddisoftware.com/qimage/quality/):

"Epson photo printers have a native resolution of 720 PPI even though some of the printers are capable of 1440 DPI, 2880 DPI, etc. This simply means that the driver operates at 720 PPI but the ink droplets themselves are placed on the page at 1440 DPI, 2880 DPI, etc."

Apparently, QImage resizes input to 720ppi before sending to the printer.

I intend to run some experiments with this.

[David Luttmann]

Qimage Pyramid interpolation is poor.

[Bruce Watson]

Apparently, QImage resizes input to 720ppi before sending to the printer.

I intend to run some experiments with this.
One would think that it would simplify the math significantly if one could create a dither pattern from a known source. If the printer firmware knows that it's only going to see exactly one input resolution, the dither pattern can be almost static. This would certainly take the "strain" off the hardware and let the printer manufacturer really slim down the onboard electronics. That is, the manufacturer can decontent the printer and sell it at a lower price point. But what does this do to print quality?

Many third parties think it's not the best thing for print quality. This is why many (not all) RIPs bypass the software drivers and the firmware when possible and drive the printer directly themselves. The argument being that they can create a better dither pattern (make the print sharper, smoother tonality, whatever) than the manufacturer if they optimize on the fly for the incoming file - whatever resolution that may be.

Now, does it really make a difference? A difference someone buying a print can see? A difference the photographer can see? And if it does, which is better? All subjective questions; each of us has to form his/her own answers.

[Brian Ellis]

"That said, it seems to be a characteristic of the driver, not the hardware. Since most RIPs bypass the OEM driver and instead implement their own, if you are using a RIP all bets are off."

Is this also true of a color managed system in which you opt to let Photoshop (rather than the printer driver) determine the colors?

[robc]

From the epson website, the below figures should give a clue to why epson printers work with the numbers they do. Is it a coincidence that 180 x 8 = 1440 and that 1440 is one of the resolutions they work at? Knowing the distance between nozzles and how the printhead and paper movement is calculated would give more clues but I doubt epson are about to reveal that.


Epson Stylus Pro 4800

Nozzle Configuration
Color and Monochrome heads: 180 nozzles x 8

Resolution
2880 x 1440 dpi; 1440 x 720 dpi; 720 x 720 dpi;
720 x 360 dpi; 360 x 360 dpi; 360 x 180 dpi

[ishmael]

I do not think that scanning with 2400 dpi is overkill. In my experience, even 4800 dpi is not enough to resolve the grain of low to medium speed films. You will get enormous files (10000 pixels in the short dimension) and then downsize them to the printing dimensions. For that I use Photoshop's bicubic sharper.
A little hint: go do something else while sharpening those files.

Carl

[paulr]

The people on the yahoo group Epson Wide Format (http://groups.yahoo.com/group/EpsonWideFormat/messages) argued this over and over with many tests. The group seems to have found a consensus that the Epson wide format printer drivers want the input file to be at 360ppi, while the desktop drivers want 720ppi. The drivers will in fact up and/or down-rez as required before they apply the dithering algorithms.

This is true.

As far as printing resolution, you're dealing with three possible limitations: The driver's native resolution; The ink/paper combinations ink spread limited resolution; and your eyes, at the viewing distance of the print.

Your print driver's resolution is probably the only given.

You can determine what the ink/paper can handle with a loupe. In my experience with matte surface art papers, the difference between 360ppi and 720ppi is very small ... it shows up more in aliasing of angled lines than in fine detail. I work with mostly small prints, so for me the small advantage of 720ppi is worth the extra file size.

As far as your eyes, unless you're making prints that will be viewed at a distance much greater than 10 inches, 200 to 300 ppi will not give you the highest discernable quality. If your eyes are good, at 10" viewing distance you have a chance of discerning detail (if the contrast is quite high) at up to 11 line pairs/mm. Depending on what formula you use for translating ppi into useable lpi, this works out to somewhere between 370ppi and 560ppi. Which suggests that 360ppi is at least a bit lower than ideal, and 720ppi is at least a bit overkill.

As far as scanning resolution, this is a whole other can of worms. It's helpful to find the true optical resolution of your scanner. I like to use a simple factor of the ccd resolution (the number used in the marketing) that's at least as high as the true optical resolution.

If you use software that can downsample for you, you can get slightly improved noise results by having it scan at the ccd resolution, and then downsample automatically to the lower resolution.

Example: my epson 4870 has a ccd resolution of 4800ppi. Its maximum optical resolution, when I use the actual focal plane with a wetmounted negative, is about 2200ppi. So I have my software scan at 4800, and downsample to 2400 ppi.

This is for my raw scan. It's still too big to print at the sizes that i need. But I like to have the extra pixels there for when I do things like rotating the file to level it. After leveling, and the first round of spotting, I downsample to my final size--which is determined by printing. For me that's 720ppi at print size.

There's more to my workflow than this, but these are the basic ways I look at the issues.