Sharpness and resolution: Drum scan compared to enlarger

[Ron Marshall]

How does the sharpness and resolution of detail of a conventional chemical print compare to that of an inkjet made from a high quality drumscan (at "normal viewing distance" not with a loupe)?

[paulr]

in my experience the scan is capable of much greater sharpness. Even from a carefully made scan from a flatbed scanner (wet mount, etc.) I can make 3X enlargements that look like contact prints when compared to darkroom enlargements from the same neg.

The darkroom enlargements were made with an aligned enlarger, glass carrier, and Apo Componon HM lens.

At much bigger enlargements, the optical resolution of the scan is going to be a limiting factor. I have a 40x50 darkroom print of one image; it looks better than a digital print would look from one of my flatbed scans. But I suspect a print from a high end drum scan would look much sharper than the darkroom print.

There are two factors--one is optical degradation from the enlarging lens, which is significant even with the best lenses. The other is the ability to sharpen a digital file to essentially undo the softening from the scanning process. If you have a good workflow and a good scan, you'll be able to redifine your idea of a sharp enlargement.

[Bruce Watson]

There are two factors--one is optical degradation from the enlarging lens, which is significant even with the best lenses. The other is the ability to sharpen a digital file to essentially undo the softening from the scanning process.

I suspect that there is a third factor - the ability to mechanically align an optical enlarger so that sharpness in the darkroom print is the same corner-to-corner as it is in the center. This level of corner-to-corner sharpness is inherent in a drum scan. In an enlarger the alignment becomes more critical as the level of enlargement increases. So while the alignment might be fine for a 4x enlargement (a 20x16 from a 5x4 negative) you might see loss of sharpness in one or more corners at 10x.

And a fourth factor is the (usually slight) loss of image contrast, mostly in the more dense highlights of B&W negatives from optical enlarging. This comes from light scatter (Callier Effect). With drum scanning, you illuminate the negative with a tiny spot (one pixel at a time) which minimizes Callier Effect.

A fifth factor is the physics of inkjet printing. The image is on the surface of the print. In a darkroom print the image is embedded in the emulsion below the surface of the print. This makes the darkroom print more physically robust (doesn't scratch or scuff so easily) than the more fragile inkjet print. But not having to look through the gelatin makes the inkjet print sharper.

I've got an inkjet print that's 150cm wide made from 5x4 160PortraVC using an 80mm SS-XL that is just scary sharp. It contains sequoia trees and was taken looking across a meadow. In other words, the trees were pretty far away. Every branch and twig is clearly visible in the print. Not bad for a 12x enlargement from a short lens ;)

That said, the two (darkroom vs. inkjet) are very different media. They each have their own look and feel - there's more to a print than just sharpness and resolution.

[paulr]

I agree with everything Bruce added ... though I never thought about the issue of gelatin influencing sharpness. It's so thin I wonder if it's really a factor.

Definitely true about the different media having diffferent merits. Some of my work looks better in ink, some in silver.

Although in the color world, those arent the only choices. You could even do a direct comparison of methods by looking at a darkroom c-print next to a digital c-print of the same negative.

[robc]

Bruce,

Please could you explain aperture in drum scanner terms and is it really sample/dot size and does it determine dpi of the scanner?

The latest ICG scanners can scan at 12000dpi. Does that mean they would be use approx 1.5 micron aperture size?

Thanks

[Gordon Moat]

Hello robc,

You might find the second article in this link of interest. It discusses aperture in one particular drum scanner:

http://fb42.s6.domainkunden.de/kunden/hamann/Artikel/EN_Artikel_Main.htm

Ciao!

Gordon Moat
A G Studio (http://www.allgstudio.com)

[robc]

Sorry Gordon but that links worse than useless. It's impossible to make any sense of it since it is using interpolation, scaling and USM without any details of software algorythms used or the settings used. i.e. I have no clue what I'm looking at. And it doesn't address my question of how aperture relates to dpi.

[Bruce Watson]

Please could you explain aperture in drum scanner terms and is it really sample/dot size and does it determine dpi of the scanner?

The latest ICG scanners can scan at 12000dpi. Does that mean they would be use approx 1.5 micron aperture size?

Argh... (that means, this is going to take a while). OK. I'm a drum scanner operator. Retired Mechanical Engineer. So I understand the beastie and what it wants from me. But I've never studied the math behind it. I wanted to use the machine, not build a new one, if you see the difference.

The best I can do is tell you what I see and the results of tests I've run (not scientific, that is, not double blind studies, just tests to expand my own understanding) and my conjecture on what I think is the cause, without having done any "serious investigation." Fair enough?

So, how's a drum scanner work? It works by scanning the film and creating pixels, one pixel at a time. It does this in two axes. First, the drum spins the film past the light source and optics. The scanner measures the film (in three or four channels, RGB and sometimes infrared) in a line of spots. The distance between samples on this line is a function of the output ppi the operator specified before the scan started. When this line is completed, a stepper motor advances either the drum or the optical system forward a distance that is also a function of the specified output ppi. Then the scanner scans the next line.

The scanner's response to output ppi is to take samples every few arc seconds of rotation of the drum, and to step a stepper motor to move along the axis of the drum. Notice that I have yet to mention aperture. That's because it's not a consideration in the spacing between samples.

The result of all this effort is a pixel. That is, a tiny square. The scanner assigns an RGB value to the pixel which is an average of the attenuation of the light shown through the drum, mounting fluid, film, more mounting fluid, and overlay (usually an acetate like mylar). This attenuation is the average measured from a round spot the size of which is determined by the aperture.

For most scanners the aperture is a series of high precision holes drilled in a disk. It's function is different than the aperture you set when you exposed the film in the camera however. In the camera you limit how much light gets through. In the scanner you limit the diameter of the light that gets through. Subtle difference.

A typical series of aperture sizes is 3.125, 6.25, 12.5, 25, 50, 100, 200, 400, 800, all measured in microns. That is, you typically get eight or nine choices. Only one drum scanner I know of has a continuously variable aperture, and that's the current Screen. I've never seen it and don't know anyone who has, but I've seen the spec. sheets for it.

So what's the relationship between the pixel size and the spot size? Well, ideally, they would be the same. But they typically aren't the same. What happens is this. Say you have a 6 micron spot size and you are making pixels that are 4 microns on a side. The spot size overlaps the pixel completely, yes? The scanner reads the spot and assigns a value to the pixel. It then steps 4 microns to the next pixel location. You can see that the spot then overlaps with the pixel we just created, yes? When we average this new spot, we are reading some film that we just read because of the overlap.

The immediate effect is that we loose a tiny bit of sharpness, but we gain a tiny bit of smoothness. Maximum sharpness occurs when the spot size and the pixel size are nearly the same (in LF terms you want "coverage" that just reaches the corners of the pixel). This implies that for maximum sharpness there are only a handful of appropriate ppi settings.

In practice however, it's very difficult (if not impossible) to see any changes in resolution as you change the ppi setting on a drum scanner. One can measure it, yes, when scanning a calibration target like the old USAF 1951 resolution target. But in practical scanning of real film and real scenes, the small changes is resolution can't be seen.

As to whether your ICG scanner can make a 12,000 ppi scan file, of course it can. But can it resolve detail at that level? Of course it can't. It's going to be limited to it's minimum aperture size, and it's going to have some system attenuations added (optical losses, electrical losses, vibration that isn't zero, light source fluctuations, etc.). The actual limit is going to be something closer to 6000 ppi optical resolution. Going beyond that is generating more data in your file, but it's not more information -- another subtle difference.

And we haven't talked about that big variable -- film grain or dye clouds (depending on the film, and for simplicity I'm going to call it all grain). The detail in the film is carried in the grain, and in the spaces between grain. We see it as the way it attenuates light that is shown through it. That sad thing is, the size of the grain is a variable, and the amount of variance is quite large.

What this means to scanning is just this. A scanner uses a fixed aperture to measure film that has a more or less random grain size (at least it's random to the scanner). If your aperture is 6 microns and the spot you are measuring is "thin" then you might be measuring several small grains at once. If on the other hand you are measuring a spot that is "dense" then you might be looking at a grain that is huge compared to the spot size.

And this is why it doesn't matter nearly as much as you'd think. And this is why you can't see the fluctuations in optical resolution of the scanner. It's much smaller than other variations that are going on at the same time.

Scanning is nothing more and nothing less than an approximation of the original film. In the case of drum scanning, a very, very good approximation, but an approximation none-the-less. Remember that, and you'll be fine.

Now, if anyone is still awake and still reading, I congratulate you on your iron will and your stubborn streak. I made that as short as I could believe it or not. But it's far from complete. If you want clarification ask away. I'm sure someone has the answers.

[robc]

Thanks Bruce,

I'm still awake and the Iron will is performing well. So in a nutshell sample size is altered to suit ppi setting or at least there is a close link. So my next question is:

When doing a drum scan for say a 20x16 print, I assume you set scan ppi based on what print dpi you want to use and then you set aperture size to suit the scan ppi. Is that correct?


and just for example:
so if I want to print at 720dpi I would need 20x720 = 14400 so from a 4x5 neg I would need a scan at 2880ppi and an aperture approx 13microns ???

[Gordon Moat]

Hello robc,

In practice, since I have had many drum scans done for work, you get a drum scan that best matches what you need for output. In other words, when I have needed a drum scan for publication, I give the scanner operator the print size (with bleed) at the requested ppi (often 300, sometimes greater), then either tell them RGB or (usually) CMYK specifications (including Total Ink).

There have been times I have had drum scans done at a larger than needed file size. Such situations are when I want to manipulate the image. In practice, I have requested twice the ppi I needed for such files, then downsized in PhotoShop or LivePicture as a final step. The idea is that if I would heavily manipulate the file, I have more information than I need, sort of an insurance policy. Unfortunately this usually costs more, though in publication work that extra cost would be passed on to the client.

If the scanner operator is good, and you have all the output specifications, the work idea has been to use drum scanning as a one shot process. In other words, you just have the scan done to the exact specifications you require. While scanning pixels might be a physically different size than output dots, thinking of this as one-to-one relationship is not a bad idea. If you find you are not happy with the results, take the film back to the scanning place and have them re-do the scan. Best of luck.

Ciao!

Gordon Moat
A G Studio (http://www.allgstudio.com)

[Bruce Watson]

So in a nutshell sample size is altered to suit ppi setting or at least there is a close link. So my next question is:

When doing a drum scan for say a 20x16 print, I assume you set scan ppi based on what print dpi you want to use and then you set aperture size to suit the scan ppi. Is that correct?

Pretty much. Almost all drum scanner software will automatically set the optimum aperture for you. There are some reasons to override the default to use bigger apertures on occasion but not really all that often.



and just for example:
so if I want to print at 720dpi I would need 20x720 = 14400 so from a 4x5 neg I would need a scan at 2880ppi and an aperture approx 13microns ???


Pretty much. As long as you don't confuse the output file's resolution in ppi with the printer's use of actual print dots you'll be fine. That is an Epson printer printing at 2880 dpi will use eight ink drops on the paper for each pixel from the file regardless of the pixel density (ppi) of the file. PPI and DPI are not the same thing and are not really interchangeable although people do use them interchangeably. Sigh...

It's sort of a moot point anyway. All most drum scanner software wants to know is output resolution and the amount of enlargement. That is, it wants to know that you need 720ppi and 4x enlargement. The software does all the math for you. With the software I'm using with my scanner, you could alternately tell it the output resolution and the target print size (that is, 720ppi and 20x16 inches) and it will do the rest. And of course you'll also have to specify 8 bit or 16 bit file saves. I don't know of any drum scanner software that would let you set your scanner resolution directly (say, at your 2880 ppi). The software wants to do the math so you don't have to I guess.

[robc]

well I think thats one huge benefit of a drum scanner in that the ppi and aperture are truly variable whereas with a flat bed the hardware resolution and sample size is fixed so unless you want to print at the hardware resolution then you have to alter resolution with software which is a destructive process.

[robc]

Thanks again Bruce, I suspected that was how it worked but wasn't sure. Doesn't look too complicated at all...

Shame I can't justify the cost of one?

[robc]

Back on topic of original question.

One thing that is often forgetten when printing with an enlarger is that all enlarging lenses are optimised for a certain enlargement factor. My Rodenstock 150 APO Rodagon N is optimised for 6X enlargement. Print smaller than that and because of less enlargement it will still be very sharp but take it down to only 2X enlargement then you are pushing your luck. That means for optimum 10x8 prints from a 4x5 neg you should be using a lens optimised for 2X enlargement. My guess is that most 150mm lenses are optimised for 4X to 6X enalargement.
And then again if you push the enlargement way over the optimum of the lens, results will not be as good as using a lens designed for that enlargement factor.