Now, understand that there's "good enough" and then there's "perfect." The former is achieved when most discerning viewers will look at a photograph and say, "That's real nice!" The latter is achieved when improving the photograph further makes no visible difference whatsoever.
The gulf is huge. Most viewers consider an 8 x 10-inch print to be reasonably sharp when it conveys 3–5 line pair per millimeter of fine detail. But, a print won't look perfectly sharp until it conveys around 30 lp/mm. That is, if you put a 15 lp/mm print next to a 30 lp/mm print, a high percentage of viewers will select the 30 lp/m print as being sharper, although most of them won't be able to tell you why they did. But, if you put a 30 lp/mm print next to a 60 lp/mm print, they won't be able to see any difference.
It's technically impossible to produce such perfectly sharp prints from 35mm film. In fact, it is very difficult to do until you get to 4x5 film and some would argue that you actually have to go to 8x10 film to easily produce a perfectly sharp print. Most photographers, even professional photographers, were entirely happy with 35mm quality. Only a few percent felt an overwhelming need to go to sheet film. Similarly, most of you won't need (nor be willing to pay for) digital perfection. Still, it's worth finding out just where perfection lies.
How many pixels is 30 lp/mm in an 8 x 10-inch print? About 180 million. Jeepers!
But, what can a printer actually deliver, and what could a camera capture?
Can any printer out there deliver 30 lp/mm? We may be getting close; I tested one several years ago that could render 15 lp/mm. That cuts the number of pixels down to a "comfortable" 45 million. Except...
...Bayer array cameras don't deliver one pixel of resolved detail for each sensor pixel. Typically they are substantially less than 50% efficient. So we'll have to at least double that number to get us as much resolution as the printer can handle. Call it 100 megapixels. If we ever get those 30 lp/mm printers, we'd be talking about 400 megapixels. Double-jeepers!
Can a camera lens possibly acquire that kind of detail? Well, the very best 35mm lenses, used at optimum aperture, can deliver something on the order of 3 µ blur circles. At one pixel per blur circle the answer is around 100 megapixels. So, that's the practical limit, right?
Well, um, maybe not. Trying to record a blur circle with a single pixel is hardly ideal. In a mosaiced sensor, it's not even possible. To be sure of correctly recording that blur circle, you need at least four pixels. There we are back at 400 megapixels again.
Do we actually need 400 megapixels? I honestly don't know; I haven't run the experiments. At the vastly lower resolutions around today, we know that trying to capture one pixel's worth of fine detail with a single pixel produces poor quality fine detail (see illustrations). Edges and points look ugly; artifacting occurs when you try to extract that kind of pixel-level information that is very difficult to avoid. Multiple pixels per blur circle unquestionably produce perceptibly better quality fine detail. The $64,000 question is whether that will be true when the resolution itself is at the limits of human perception. I don't yet have an answer to that.
So maybe 400 megapixels is truly extravagant. Maybe going beyond 100 megapixels will never matter. But without real tests, all one can say is that it might. And, certainly, up to 100 megapixels there will be perceptible improvements in fine detail quality in modest-sized prints, even with current printer and lens technology.
Most of you won't need this, any more than you needed 8x10 view cameras. But for some of you, those won't be pointless pixels.
Featured Comment by Bahi:
1. 'There are real-world situations in which a poorer signal/noise ratio produces MORE visual sensory information, not less.'
2. 'Noise may actually be a GOOD thing. And even if it's not, will a higher level of noise be at all perceptible if the pixel size is sub-resolution?'
3. 'Mebbe it'll gel into a column, mebbe not.'