Please correct me if I'm wrong. Here's how I see it.
Both the 5d and the 1Ds3 have full frame sensors. This means that with the same lens, the image projected onto the sensor will be identical, including diffraction effects. When printing the same size, the image will have identical diffraction blurring. The higher resolution of the 1Ds3 will show some advantage because it will show finer detail.
At higher f-stops you will be experiencing diminishing returns with the higher resolution sensor. Eventually the image will become so blurred that the 1Ds3 sensor will not be able to resolve more fine detail than the 5d. The 1ds3 should still render the blurred image more accurately, but not noticeably so.
An excellent analysis, Phil. As you correctly point out, the effects of diffraction as projected on the sensor will be exactly the same for both cameras. The term "diffraction limited" can be confusing, since diffraction always sets a limit on resolution. In practice, the term is used when system resolution is limited by diffraction, rather than lens aberrations or the resolution of the camera sensor. In his excellent diffraction tutorial, [a href=\"http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm]Sean McHugh[/url] assumes the system is diffraction limited when the diameter of the Airy disk exceeds the allowable circle of confusion (COC). COC also comes into play with depth of field, and is relative, depending of the visual acuity of the observer and the distance at which the print is viewed. The same considerations come into play with Sean's diffraction calculator.
If you plug in values for 35 mm camera sensor size, default visual acuity (not stated, but apparently considerably less than 20/20), 25 cm viewing distance, and 10 inch maximal print size, the system becomes diffraction limited between f/22 and f/32. Megapixel count does not affect the calculation. If you change visual acuity to 20/20, maximal print size to 20 inches and keep the other parameters the same, the system becomes diffraction limited between f/4 and f/5.6.
These calculations involve what is "good enough" for a given print size, viewing distance, and visual acuity. Rather than trying to get an image that is good enough, one might strive to get the best possible image. This is the approach that Nathan Myhrvold
took in his somewhat controversial thread posted earlier on the LL. Details aside, I think Nathan's approach is valid.
When the size of the Airy disk is larger than the pixel size of the camera, the disk will extend over adjacent pixels, blurring the image. The size of the Airy disk as related to pixel size is demonstrated nicely on Sean's web site. The 12.7 MP 5D has a pixel size of 8.2 microns and the 21 MP 1Ds MIII pixel size is 6.4 microns. With green light, the Airy disk is 5.4 microns at f/4, 7.5 microns at f/5.6 and 10.7 microns at f/8. If you want to make full use of the resolution of the 1Ds MIII with a diffraction limited lens, you probably should keep aperture at f/5.6 or larger, whereas a somewhat smaller aperture would not limit the 5D. Since real world lenses may not be diffraction limited at those apertures, your results in practice may differ.
In my own tests using Imatest, a Nikon D200 (10 MP, pixel spacing 6.2 microns) and 50 mm f/1.8 lens, I noted peak resolution at f/5.6. Many similar test results are available at Photozone.de. For example, the Canon 85 mm/ f1.2
, where resolution peaked at f/4-f/5.6, and was most likely limited by the resolution of the 8 MP (pixel spacing 6.5 microns) camera used for the test. It would be interesting to see the results of this lens on the 1Ds MIII. The resolution wouldn't be that much increased in terms of lp/mm, since the pixel spacing is 6.4 microns. However, the image would require less magnification for a given print size (its resolution in terms of lp/picture height is greater).
Under these circumstances, the 21 MP camera becomes "diffraction limited" when stopping down earlier than the 12 MP device. However, as Sean points out, this does not mean that the image at a given aperture will be worse with the 21 MP camera; indeed, the higher resolution camera will have fewer artifacts such as color moiré and aliasing.