You are talking about the effect of sensor size, and by implication viewing magnification, rather than photosite density. The lesson to be drawn is if you want to minimise the effects of diffraction get a camera with a big sensor with lots of pixels. Unfortunately, a lot of people are under the misapprehension that they need to buy a camera with a sensor with not many pixels, which is not the best choice.
The formula I offered makes it clear that the f-Number at which diffraction will begin to inhibit a desired print resolution is affected only by enlargement factor, and thus sensor size indeed seems to be far more important than photosite density.
I'd rather put it this way: If you want to maximize the number of diffraction-free stops you can shoot with, select cameras having enough Megapixels to produce the print sizes you are seeking at the print resolutions you desire, but avoid cameras having sensors with high photosite densities. (Notice I've not said a word about sensor size here.)
.. not that many large format lenses. Of course, lenses designed for 1/2.5" sensors routinely resolve at that resolution.
200MP cameras could be designed to give the user whatever the user wants in terms of spatial resolution and bit depth and/or compression in the output. When you start oversampling your lenses, you really don't need the kind of critical pixel-level accuracy that is needed with low pixel counts. JPEG-like compression could be done on the full-res RAW data without a hint, viewed at standard print sizes. 14 bits is overkill for current sensors at their best ISOs; a 200MP 2-micron sensor could get away with a few bits and a LUT.
With 200MP, the big factor (other than technique and optics, of course, which can always ruin everything) is the light. If the exposure is high or the gamma is low at base ISO, those pixels can be pretty accurate and withstand fairly high magnification.
What if you need to crop? What if you need to rotate or correct perspective? Oversampling the lens makes those operations more artifact-free.
The pixel-centric thought virus has a very powerful lobby. The lobby is two-faceted; there is the "100% pixel zoom" test, and the "downsizing with software that drops or unevenly weights input pixels" test.
[snip]
Your failure to include the phrase "enlargement factor" makes the above statement completely irrelevant to the argument I've been making in this thread.
This is true to an extent, but as John Sheehy points out, the extra pixels may benefit you in many other ways than making bigger prints. In particular, because you have more information in the original capture (i.e. you know to a greater precision where the photons landed, as well as how many landed in a region) you can apply a lot more thought and intellingence to post processing to gain ultimately better results.
That's not a good way of putting it. The phrase 'diffraction free' is entirely misleading and wrong-headed. The stops are not diffraction free, it is simply that the sensor has cut of the resolution that even the diffraction limited lens can offer. If you look at a lens resolution chart, you will see there are broadly speaking two factors at work. Lens aberrations limt resolution. As a rule these become resolved as the lens stops down, so resolution rises as you stop down from wide open. At the same time, diffraction limits resolution as the lens stops down, so resolution decreases as you stop dow. put these two together, and you get the familiar hill shaped curve which peaks at f/4-f/11 depending on the lens. All the downhill part of that curve past the peak is diffraction limiting. All that happens with a coarse pitch sensor is you are slicing the top off the hill.
I don't think that there is any good reason to place the line there. It is an entirely arbitrary limit. The point about being 'limited to using half of the stops available' is completely bogus. Part of the art of photography has always been trading DoF against peak resolution. Using a camera which can utilise the entire resolution curve of the lens gives you more options, not less.
Sensor size arguments, not pixel pitch ones. Larger formats have always given more options wrt usable apertures. On a large format lens, you'll find it allows you to stop down to f/45. Of course, in small formats one tends to gain at the other end, simply for practical considerations of size and expense, an f/1.0 or a 15x zoom for a 10x8 camera may be technically feasible but you wouldn't want to pay for one or carry it around. Digicam lenses are huge in relation to sensor size and have incredible peak resolutions, but still provide a compact end result. You pays your money, you takes your choice.
As I said, the concept of 'diffraction free' as you put it is ill found, and therefore so is your arbitrary limit of pixel density.
This has been discussed elsewhere, it woud be an advantage for a nuber of reasons.
Where do you get the 70lp/mm figure from? Even Canon kit zooms will outresolve that.
Using state-of-the-art pixels on a DSLR (or DRF) sensor is an increasingly interesting notion. The amout of digital processing in DSLR's is still quite small (they are using mobile phone processors clocking at a few hundred MHz). Part of the reason for that is power considerations, the other is cost. However, if you traded the money saved on an OLPF for a suitably designed ASIC front end, the net result, as you point out, could be something much more flexible and higher quality than current products. Why bother to choose between a 12MPix D3 and a 24Mpix D3x, when you could have a camera which offers you any output resolution you want, down from 200MPix? What could Iliah Borg do with a 200MPix raw file to play with?.. not that many large format lenses. Of course, lenses designed for
1/2.5" sensors routinely resolve at that resolution.
200MP cameras could be designed to give the user whatever the user
wants in terms of spatial resolution and bit depth and/or compression
in the output. When you start oversampling your lenses, you really
don't need the kind of critical pixel-level accuracy that is needed
with low pixel counts. JPEG-like compression could be done on the
full-res RAW data without a hint, viewed at standard print sizes. 14
bits is overkill for current sensors at their best ISOs; a 200MP
2-micron sensor could get away with a few bits and a LUT.
The decisions could be automated, too, and the firmware could
determine, based on user settings, what level of output
resolution/compression is needed for the image as a whole, or even
broken down into areas (out-of-focus areas can stand far more spatial
loss than in-focus areas). There is no reason for anyone to say nay
to high-MP sensors because they force high-MB or GB output; high
oversampling is good no matter what output resolution you want. You
can even have reduced-resolution RAW, which is still superior in
flexibility to reduced-resolution JPEGs.
With 200MP, the big factor (other than technique and optics, of
course, which can always ruin everything) is the light. If the
exposure is high or the gamma is low at base ISO, those pixels can be
pretty accurate and withstand fairly high magnification.
--
John
And even then, this is a worst case scenario. I judge the resolution needs of a sensor based on something like a Tamron 90mm f/2.8 macro or a Canon 300mm f/2.8 which can probably use 10 to 15x the density a kit lens can.
Get a good night's sleep while RPP worked on it ;-)
The last two sentences (one parenthetic) are what I was taking issue with. Earlier, you correctly attributed the main difference as far as diffraction goes, has to do with sensor size. But then to sit at fixed sensor size and limit the number of pixels because of some concern about diffraction, is simply the wrong way of thinking about the issue. Even well into the diffraction-limited regime, decreasing pixel pitch improves MTF; it just doesn't increase it by the ratio of the pixel pitches, but rather by something less, as the example I discussed in the link demonstrates.
From the article...
Again from the article:
This came out a bit more garbled than I would have liked. Let me try again.
