Helpful Posts: 0
15th April 2008, 02:53 PM
Larger camera pixels enlarge better
I want to offer a suggestion for the tutorials section. It has not addressed fully the ability to enlarge digital images.
Issue: photos taken with Canon D30 and D-1 cameras with L series lenses at
f/5.6 enlarge better than most digital images. Why?
Resolving power of lenses. The resolving power of digital cameras is determined largely by the sensor (unless the lens is very bad) (tests at DP Review tend to have resolving power equal to sensor size). Consequently these measurements are largely useless in evaluating interchangeable lenses. Lens tests are best done with film the old fashioned way. As you probably know, lenses have a peak resolution when stopped down 2-3 stops. For the Canon f/2.8 L series zoom lenses, the peak resolving power is at f/5.6 and
f/8. These lenses are diffraction limited from f/8 to f/32 and optically limited from f/2.8 to f/5.6. Typical Canon zoom lenses resolve a peak 40+ lp/mm (80+ pixels/mm) and L series zooms resolve a peak of 60+ lp/mm (120+ pixels/mm). Top quality normal primes can resolve a peak of 80+ lp/mm (160+ pixels/mm). Canon D30 has about 100 pixels/mm and the 20D about 160 pixels/mm.
Enlargement of digital images is done in many ways, but all are 100% mathematical algorithms (no optics). The ability of an algorithm depends primarily on the accuracy of the measurements (sensors). The accuracy of the sensors depends on diffraction (you have an excellent section) and on the lens quality. In general, if the lens quality and aperture are poorly chosen, the ability to enlarge will suffer. I would guess that 300
pixels/inch is for making prints from images with average sensor measurements (Airy disk = sensor size). Professionals submitting 300 ppi images for publication have typically enlarged the image from a lower ppi rating.
In summary, the Canon D30 with an L series exhibits extremely good sensor accuracy. Its ability to enlarge is at least a factor of 2 better than the 20D with a Canon regular lens. I enlarge both to about the same size, with the 20D slightly better. To create the same pixel quality, the 20D needs to be used with large aperture prime lenses (e.g. 50 mm f/1.8). The Canon 5D at 120 pixels/mm should work well with L series zoom lenses.
15th April 2008, 02:57 PM
Your email discusses a topic that this site has not touched on very much at all yet: optics.
I would generally agree with your statement: that you can typically enlarge a photo from the lower megapixel DSLRs more than would be expected (relative to a given pixels per inch criterion) compared to a camera with higher megapixels. This is mainly because the larger pixels are capturing only the highest contrast (best MTF) length scales of the lens. Any lens can more accurately translate contrast detail at 30 lp/mm compared to 90 lp/mm.
That being said, a higher megapixel camera will almost always enlarge better (with comparable noise levels and dynamic range...rarely the case) because the eye expects to see detail all the way to its resolving limits-- even if this smaller scale detail may not be of as high quality as the larger scale textures at the resolution limit of the lower megapixel camera. This would agree with your observation of diminishing returns with the D30 vs 20D/30D/40D: with the D30 one might be able to make a 200ppi print that looks just as detailed as say a 220ppi print on the 20D, but yet the 20D print can still be enlarged a bit more overall.
One bit of caution is that an 80 pixels/mm camera with a Bayer sensor cannot resolve 40 lp/mm-- in reality this is much less. We are therefore very close to the resolution limits of many lenses, but not quite there yet. This was part of the reason I made the plunge for the 5D, and I can arguably see improved microcontrast as a result (compared to the 20D) because I am not pushing my lenses as close to their resolution limits (at least in the center).