How Big Can You Go?
By Carl Weese
There's a certain kind of landscape I've done, just once in a while, but for a long time, going all the way back to the 1960s when I did them with Kodachrome II loaded into a Pentax H3v with a 105mm Takumar. These landscapes are unlike my usual practice of working close to subjects using short lenses. They're made from moderate or long distance with moderate or really long telephoto lenses.
This fall I got around to buying a lens for my Micro 4/3 cameras that would be suitable for this kind of picture—a Lumix 45–200mm. Many of the pictures in the Winter Light Series that I've recently been posting at my Working Pictures Blog were done with this lens using the Lumix G3.
These landscapes can be a bit abstract, more about light and atmosphere than detail, but a certain level of resolution has to be there or I'm not happy with prints from them.
Obviously a picture like "Foggy Morning" makes very little demand for resolution.
A picture like "Willow" does make enormous demands on resolution, and doesn't work for me unless, no matter how big the print is, examining it at reading distance reveals a wealth of detail with convincing description of the subject matter. A print that looks OK at "normal viewing distance" but falls apart if you move closer to view it at reading distance doesn't cut it. Why entice me to come closer only to disappoint?
Large prints, 20 inches or more wide (I'm not talking about the monster mural-size prints that are popular in the photo art world today), seem to present these pictures, this particular aesthetic, really well. So that brings up the issue of enlargement, both traditional and digital. I began serious digital capture work around 2006, using an Olympus E-1. I was astonished at the quality of 5x7-ish uninterpolated prints, which looked like slight enlargements from medium format film, even though the 4/3 sensor’s physical size was about a quarter that of 35mm film.
However, I noticed I couldn't make them a lot larger than that. I found that I totally disagreed with the common wisdom that digital captures could be up-interpolated by a factor of two with no loss of quality. Nonsense! To my eye a print from a 2X upres file looked like hell.
Understand, most of my serious work is in large* and ultra-large format**, so I'm used to making contact prints from very large film negatives, or modest digital enlargements from those same negatives—so admittedly my standards are a little lofty. For digital files, 1.5X usually worked OK, 1.75X maybe, sometimes—but 2X, never. I can make much bigger prints now, but it's not because enlarging digitally has gotten any better; it's just because I've got more pixels to work with. The physical size of the Micro 4/3 sensor in my Lumix cameras of course is exactly the same as the old 4/3 E-1.
I was astonished again a year and a half or so ago when I saw the quality of the 20-inch wide prints I could make from Panasonic Lumix GF1 captures. But then, the uninterpolated GF1 file, at the native 300 ppi resolution for my HP Z3200 printer, is 13.333 inches. So that 20-inch print is almost exactly the same 1.5X up-res that I found was generally of good quality from the old 5-MP E-1. It's just a whole lot bigger.
When I began to do some of these distant-twigs-and-branches pictures with the new lens, the results were frequently disappointing. Some were fine, but many just weren't sharp enough. The pictures looked fine for online use or even for an 11x14-ish print, but completely fell apart with the very minor up-interpolation of the 16MP file to 20.5 inches, although my usual quick tests of the new lens showed its (computer-enhanced) optical quality to be excellent, at least at short to middle zoom settings, after which an annoying amount of vignette sets in. Amazingly good in fact, considering the $255 price tag. So I wanted to look into this and see what was going on. Obviously distant branches are an extreme problem, but there are similar issues in my more usual short lens pictures like, say, the detail and texture of a brick wall in one of my urban landscapes. If the bricks turned to mush in a print I'd find that just as unacceptable as what the distant branches were looking like. In fact, I gave up on a fairly pricey wide angle lens because I couldn't stand the never-really-sharp corners.
So late last week I had chores to do at a place where there are plenty of test subjects for this problem. My earlier experiments were done at reasonable shutter speeds, hand-held, with the default ON setting of the in-lens stabilization. This time I put the camera on a tripod and turned off the stabilization (Panasonic—and everyone else who offers some form of anti-shake--instructs us to turn it off when using a tripod). Bingo! As soon as I looked at the captures on screen, I could see they were crisper. On critical examination at 100% view, there was a surprising amount of variation, but the general level was a big improvement, and a lot more of them than before were just plain sharp!
This detail is a 100% crop (after you click on it to see the full size upload) showing that even all the way up in pixel peeping territory, there is good definition in the tree branches and surprisingly nice tonal variation in the deep black values of the crow sitting in the tree. It isn't necessary (or really possible) for a 100% view to look "perfect," but with some practice you can learn to judge how much definition is enough to make a convincing print, and how little definition means the print will fall short.
The variation, I think, is from focus. With shallow depth of field (even at 4/3 sensor size) a thicket of sticks or branches gives a difficult focus target, whether you use the AF or go manual—where in the depth of the thicket will you focus? Probably, most of the time, the front, but that's not so easy to find either by the eye or the AF. I chose a nice looking picture and used the up-interpolation function in ACR to make a file that will print 20.5-inches wide at 300 ppi, then sliced out a 7x20.5-inch strip through a critical area and sent it to the printer. The result is really solid. It's just as good as the enlargements from my short lens street shooting. Which is to say, better, at 20.5 inches, than my 1980s 15-inch square darkroom RA enlargements from Hasselblad negatives.
What may be going on here is primarily mechanical, not digital. After all, this is a very high degree of enlargement of the optical image projected by the lens. I never enlarged my 35mm negatives beyond 12x18" image area in the darkroom, and these 20-inch prints are from an optical image on the sensor that's about a quarter the size of 35mm film.
The 'Higher Expectations' factor
There is an ancient photographic rule of thumb that, with practice, you should get good results hand-holding the camera at a shutter speed that is roughly the reciprocal of the focal length of the lens in use. So, 1/60th for a 50mm, 1/30th for a 35mm, 1/250th for a 200mm, and so on. This is specific to 35mm classic format. A larger hand-held camera would allow slower speeds for enlargements to a given print size, which was convenient since their lenses were usually slower. It's pretty commonly understood that for sub-35mm sensor sizes this rule has to be modified by the "crop factor" of the sensor size. In fact I've noticed that manufacturers incorporate this into the software. With several different cameras I've found that if you set aperture priority or full program mode along with auto ISO, as the light gets dimmer the camera will stop lowering the shutter speed and begin raising ISO right about the point of "reciprocal of shutter speed multiplied by crop factor." The reason this is needed is that the optical image projected by the lens is subjected to more enlargement with a smaller sensor to reach a given print size, and so any degradation due to camera shake will also be magnified by that factor. But that's not enough, because the quality of digital sensors has made us greedy.
I'm thinking that the old rule of thumb needs to be multiplied not only by the crop factor, but also by a "higher expectations factor" for the bigger prints that can be made. I suspect that the old "reciprocal of the shutter speed" rule was envisioned for prints on 8x10" or 11x14" paper. So it wouldn't even apply to 12x18-inch enlargements from Leica negatives, and, come to think of it, it didn’t.
This means that even if you use a full frame DSLR, unless you are only posting online and have no interest in making prints—big prints—the old rule of thumb needs to be modified. I'd estimate by a factor of two, because I think current digital sensors can deliver good results at twice the magnification of the optical image that you could do with film. So the quality of the lens, and the steadiness of the camera, have to conform to a doubled, or more, standard of quality.
The good news here is that even difficult, detailed subjects shot with longer lenses can yield really good prints at 20 inches or more from 4/3 or Micro 4/3 sensors. The caveat is that there is no margin of error. The mechanical stuff has to be right on the money: perfect critical focus, complete suppression of camera shake. The optical quality of the lens, whether digitally enhanced or not, has to be up for the degree of enlargement. There's nothing like a tripod for holding the camera steady. Still, it's nice to be able to respond quickly with a hand-held camera, so I'm planning some tests at longer focal lengths and higher shutter speeds, with and without stabilization. If I get any interesting results, I'll report them.
*4x5-inch to 8x10-inch view cameras
**View cameras larger than 8x10
Note: Links in this post may be to our affiliates; sales through affiliate links may benefit this site. More...
Original contents copyright 2011 by Michael C. Johnston and/or the bylined author. All Rights Reserved.
Featured Comment by Bill Pierce: "Boy, do I think you are dead on. Most digital photographers routinely look at screen magnifications of sections of their photographs that have no parallel to what we did in the film world. And after a while, we realize we are looking at a lot of camera shake. It's gotten to the point where I'll use a larger than optimum aperture or higher ISO before I'll drop that shutter speed. And in lower light, a tripod. I feel pretty silly with a little tiny camera on a big tripod, but it works. I think you just wrote the most important 'tech column' on the web in quite a while."