dante stella stories photographs technical guestbook
|Oh man, the shit piled up so fast you needed wings to stay above it.|
|Same conceit, different day|
I am starting to think that I will never live long enough to speak out against all of the photographic pseudoscience circulating on the internet. This week, I have hit the breaking point with the "reviews" of lenses based on "1:1," "actual pixels," or "pixel-for-pixel" comparisons of images either shot on digital or scanned to digital. Let's call this "pixelography." While we all marvel at some of the tiny things you can seen in a 4000dpi negative scan or a 1:1 blowup of a D2x image, comparing two pieces of kit based on individual pixel blowups is a direct indicator of ignorance. I'll give you my top three reasons, and at some point I'm sure I'll stumble onto some nasty rebuttal on dpreview.com.
1. The final output rarely cares about individual pixels, because there is usually a lot of manipulation that leads to a print or what you see on a web page. Let's take the example of a Kodak 14n or SLR/n file (primarily because the resolution is a nice even number). At full size, the camera kicks out a 4500x3000 file. Suppose we use a dye sublimation or Fuji Frontier machine. Each has about a 300dpi final resolution. With either of those systems, and assuming a 1:1 correspondence from camera pixel to output pixel, you get a 40x26cm print (15x10 inches). At a normal viewing distance (for that size or any), individual camera pixels are neither visible nor relevant.
Next, consider that most pictures are never output to anything bigger than 4x6. Or 8x12. If you go smaller than a 1:1 pixel to output ratio (so smaller than 40x26cm), you end up downsampling and sharpening, whether you do it or the printer driver does it. Not only are the differences between lenses minimized by downsampling; they are also actively obliterated by unsharp masking. The Fuji Frontier, the mainstay of mass digital printing, normally applies very heavy sharpening from digital files.
Next, contemplate inkjet, where there is no continuous tone reproduction and the printer has to use a lot of its massive resolution to simulate continuous tones (I have to say I love seeing manufacturers sell 1400dpi printers to the credulous). Changing 16-bit individual pixels into dot patterns requires a lot of processing and some sharpening, so again, you're not getting the benefit of things on the individual pixel level.
Finally, ponder web-sized views, where a cutting the size of a file down by a factor of 10 to 15 is fairly normal. You 1,024x768 full-screen view isn't even one megapixel. Think of how many "actual pixels" you are losing by scaling to that size.
2. Expensive, fast lenses exist for a reason. My favorite conceit on the internet "I must admit that my [cheap slow zoom] is better than [someone else's] [ultrafast professional zoom lens] on a pixel-for-pixel basis." The subtext is then that only "suckers" buy expensive glass when some cheap plastic fantastic can do 80% of the job. This misses the purpose of fast professional glass. It is designed to shot wide open to get the 20% that the cheap lenses can't get. People who use them don't care if there is
— You can question all you want why people would buy a $3,000 Leica Summilux 35/1.4 for a rangefinder camera and point to the fact that at f/8, a cheap 35mm f/2.8 Russian lens might do equally well. Maybe under those circumstances (at an aperture where any slow lens is good and where the Summilux is starting to degrade), there is a comparison. But it doesn't tell the whole story. The Summilux can take the shot wide open in 1/4 the light. When you factor in the fact that no cheap, slow wideangle is even close at 2.8 to the Summilux at 1.4, you are now talking about taking an equivalent shot in 1/16 the light. That is an absolutely huge difference.
— You might look at a 17-35/2.8 Nikkor (designed for both film and digital) against a 18-200/3.5-5.6 Nikkor (digital only) and ask yourself, why would I ever pick a $1,600 lens when at x and y focal lengths, the $600 18-200 is just as good? There are two easy answers. First, in addition to the speed aspect, a wider maximum aperture means better autofocus operation and better chance of locking on. Second, 3-10x slow midrange zooms all exhibit distortion in various ways, particularly at the wide end. The 18mm setting on most digital-only zooms is nothing to write home about. Think of it this way: you can spend your life manually correcting distortion in architectural pictures - or you can get it right the first time.
— You might compare a 4000dpi scan of a 35mm negative to a 6x9 negative. Wow, you think, my 35mm Spotmatic is superior to my Super Ikonta C in resolution. While it may be true that the best medium format cameras may resolve (per unit of film), 1/2 to 2/3 what a 35mm does, their negatives get far less enlargement and need to record far less on film to get equivalent output resolution. Where medium format blows away 35mm is in tonal reproduction - which has very little to do with how small a detail a lens can see. This is lost on the pixelographers.
The bottom line to all of this is that expensive glass and large formats exist: for a reason: to extend capabilities. It's not cheap to make big, fast glass and in a lot of cases, it's not for everyone for every thing. Similarly, medium format can be a lot harder to deal with than taking your film to Costco. But those two things can deliver results that more economical products do not.
3. Aesthetics, not precision, is the goal of photography. I was at a Rodin-Claudell exhibit today and was reflecting on how little actual detail Rodin scupltures have. From a normal viewing distance, they look realistic. From closer up, you can see that they are highly impressionistic, and in most parts of most sculptures, the smallest detail actually recorded was the size of a human fingertip. Similarly, it's interesting to reflect on the fact that most of the world's most famous photography (Adams, Weston, Strand, etc.) was taken on equipment and using methods that had relatively low performance by today's standards. Look hard at an original Adams print from about 2 inches away and ask yourself if it would pass your standards. Millions of dollars in his prints have been sold; how many of yours? Adams knew enough to use a lens that was not the best in the center, but uniformly "good" all around. The key is management of the image; controlling the transitions between sharp and unsharp is far more important than the sharpness of any particular part of an image. Pixelography, that pernicious 1:1 viewing, does not advance management of the image because it is centered on particular points that are meaningless outside of the whole.