I had a few prints made recently by Bill Nordstrom at LaserLight using digital processes (drum scanning, photoshop, continuous tone printer). The prints are made from 5x7 inch color transparencies and are 22x30 inches. They look good. Now there is one thing which disturbs me. Here is the logic:

1. The files which were used to produce the final print are only around 4500x6000 pixels. Bill says that larger files would be overkill since he is printing at 12 lpm interpolated from 8 lpm.

2. One could easily produce a 4500x6000 file by scanning a 6x7cm slide.

3. This would imply that you could produce a 22x30 print from a 6x7cm with the same quality as a 22x30 print from a 5x7 inch.

4. In traditional printing, I thought that it wasn't that difficult to distinguish a 22x30 print made from a 6x7cm from a print of the same size made from 4x5.

So there is something which doesn't sound quite right there. Any comments ?

-- Quang-Tuan Luong (luong@ai.sri.com), December 16, 1998

Answers

This is a different technology and requires us to change how we think. In traditional printing we are directly enlarging the original into a continuous tone reproduction. Using digital technology, we reproduce the original image into a 300-400 dpi reproduction. The output resolution is fixed, and nothing you do in PhotoShop will change that, it is simply the limit of the current reproduction technology. We need to change our thinking to work backward from the final print to the output file to the original image. If you start at the fixed resolution and do the math you'll see that the final print size determines the necessary file size, which in turn gives you the requirements for the original image size.

The answer is yes, a 6x7 original will give a 22x30 print every bit as good as a 5x7 original using current digital processes. Whether that is bad or good depends upon your point of view, and whether you've already invested in a large format system or have your eyes open for a good used 6x7 system.

-- Darron Spohn (sspohn@concentric.net), December 17, 1998.

Digital processes are limited by the accuracy of the scanner and the printer. As you point out, one limitation of your print quality is the resolution of the printer. As far as it goes, your analysis is correct.

But other considerations can be important. In theory, a digital scan can clearly resolve almost 200 lines per inch if you scan at 400 dpi. In practice, you are testing the accuracy of your scanner. Small amounts of jitter in the mechanism can significantly lower the quality of your image. In addition, the scan results can be adversely affected by the sensitivity pattern of each photosensor. To accurately reproduce your image, the sensitivity pattern must blur out detail in your image that is too fine to be accurately represented in digital form. For example, a scanner operating at 400 dpi must blur out details finer than 200 lpi. Failure to do so corrupts the digital image (an effect called aliasing). Oher types of problems can degrade your digital image too, but you get the idea. Similar problems can also affect the printer. Each problem can introduce extraneous digital artifacts that degrade the quality of your image.

Now, it is significantly easier to build a scanner that scans a negative at 200dpi than at 400dpi. For example, the effect of jitter in the mechanism is automatically reduced by a factor of two. The result is that scanning a larger negative should place fewer demands on the accuracy of your scanner, assuming of course that the designer knew what he was doing. Similarly, it is easier to build a low-resolution printer than a high-resolution printer.

To get back to your analysis, the question becomes, can you make as good a 4500x6000 pixel scan of a 6x7cm negative as you can from a 5x7in negative? At least in theory, the answer is probably not. The 22x30in print from the 5x7 negative should look better because fewer digital artifacts were introduced when scanning the larger negative.

In practice, much depends on the design of the individual scanner. Or, perhaps nobody sells such a high quality scanner for large negatives because the demand is too small. The bottom line is you'll have to try it and see which looks better. In theory, the larger negative should help -- despite the limitations of the printer. In practice, the image qualtiy depends on the quality and design of both scanner and printer. I fully agree that a traditionally made 22x30inch print would be a most interesting point of comparison.

-- Michael Heal (kmheal@audiolab.uwaterloo.ca), December 17, 1998.

Michael your analysis holds true for a 400 dpi scan, but the drum scanning process captures images at 5000 dpi. As Bill Nordstrom pointed out, this is simply overkill for a 22x30 print from the Lightjet 5000 printer. In reality the skill of the person preparing the image is much more important than the size of the original, all other things being equal. True, a larger original will yield more detail, but at 22x30 inches reproduction those details will be lost. Using the current digital technology a 22x30 print from a 6x7 original is so close the same size print from a 5x7 that the two almost indistinguishable.

As the technology matures and output devices improve that will change, so I am not recommending anyone abandon their large format equipment. Besides, there are already excellent digital backs for 4x5 cameras on the market. When those prices come down out of the stratosphere we will be able to bypass the film stage and go straight from the camera to the computer.

-- Darron Spohn (sspohn@concentric.net), December 17, 1998.

Darron, I understand your comments and in essence Bill told me the same. But how would you explain that (if it is indeed the case) at 22x30, in digital 6x7 equals 5x7, whereas in traditional 6x7 is not quite as good ? either there is some magic (that i don't understand) accomplished by digital (so that the 6x7 digital is better than the 6x7 trad), either the 5x7 traditional is still better than the 5x7 digital obtained with the laserjet 5000, which is also mysterious since i am ready to buy that the naked eye doesn't resolve beyond the definition of the laserjet 5000.

-- Quang-Tuan Luong (luong@ai.sri.com), December 17, 1998.

Digital printing can be better than traditional printing if you start with a good enough scan. Then you correct the colors in PhotoShop, crop and resize the image as desired. Apply an unsharp mask and save at the desired output size and resolution. The key here is the output size and resolution. With a 5x7 original at this output size you are scanning and editing more information than the Lightjet can resolve. As Bill said, the files are overkill. A 6x7 original scanned at 5000 dpi can capture all the information you need for 22x30 reproduction.

Yes, you'll get more shadow and highlight detail from a 5x7 original than from a 6x7 original, but but by going digital you can compensate for this while reducing the grain size and retaining the sharpness. This is the tricky part, and takes a lot of experience to develop the judgment necessary to prepare the files properly. Used wrongly, unsharp masking will cause digital artifacts in the highlights that will make your skies turn mottled. Used correctly, unsharp masking will smooth the grain while sharpening the photo when viewed, just like it does in traditional processes.

The differences in original film size are less apparent through digital processes than they are through traditional processes. After viewing Lightjet output at Galen Rowell's gallery last week, and over the past year at the Ansel Adams Gallery, I'm convinced the technology has arrived and the time to go digital is now. It is the best way to retain complete creative control over your images without spending $25,000 on a color darkroom. See the "Digital Imaging and High Quality Prints" thread on photo.net. There is a more in-depth discussion of this technology therein.

I don't claim to be an expert on this, yet, but I'm getting there with the help of a friend who is an expert. I've watched him working on files for Lightjet printing, and he's explained much of this to me when I could get him to slow down enough that I can keep up. Give me six months and I'll be ready to teach other people how to do this.

-- Darron Spohn (sspohn@concentric.net), December 17, 1998.