or how and why pictures from medium format – and larger – camera will always look better.
Answer: it’s the lens MTF, stupid.
It always struck me how medium and large format pictures looked more life-like to me. I’ve realized that it’s not (just) a question of resolution and dynamic range, as even “low” resolution jpeg pictures on the web look great / “real”.
Great 3D / life-like feel © Leszek Kowalski @ photodom.com
Other cool fashion like pictures from Leszek Kowalski
Forget Film vs digital, forget CCD vs CMOS, forget anti-aliasing filter or not, forget megapixel race: what matters most – for image quality – is the sensor size… not because of the sensor performance but because of the lens!
- Lens don’t have infinite optical resolution: micro contrast decreases at high spatial frequencies (small details). c.f. Lens quality & Modulation Transfer Function
- For an identical image, if the format is larger, the same detail will also be larger (on the image plane) so that the lens will need less resolving power to resolve it. It’s a shift to lower spatial frequencies: The lens is used in a “better zone” of its MTF, meaning used in a better zone of its optical quality, giving a better micro contrast etc…
e.g. a 10um detail on a 24×36 mm sensor = a 18.3um detail on a 56×56 mm sensor. The lens will have to resolve a smaller detail on the 24×36 than on the 56×56 = the lense will have to be used at higher spatial frequencies – at lower contrast. 10um would correspond to 50 lp/mm versus 18.3um to 27 lp/mm. A lens transmitted contrast a 27lp/mm is better than at 50lp/mm (especially at wide apertures, see page 8 graph of Carl Zeiss’ paper on reading MTFs)!
Summary: The larger the format, the better the image – produced by the lens – quality.
And seeing (at Paris Photo 2011) a Richard Learoyd 8h exposure, giant direct-positive image, taken with his room sized camera obscura is an amazing experience c.f. Uncomfortably Close: Richard Learoyd’s, Presences, that wont contradict this observation!
An other way to look at it: a large format will “capture” more analogical optical detail at a certain transmitted micro-contrast (High spatial frequencies with >= 50% of transmission [MTF50 curves] are important for the apparent “pep” of an image, maybe even 90% which is very low-resolution).
It’s cost (and weight), not quality that keeps sensors small!
n.b. Obviously to have the same field of view going from e.g. 24×36 to 56×56 one would need to use a lens with an 1.83x longer focal (with a shallower depth of field at the same distance & apperture which is nice btw) and a larger image circle.
Contrarily to a common conception, medium format lenses as they need to produce a larger image circle are more difficult to produce and therefore might have worst MTF than the best 24×36 lenses, but the size advantage still holds…
n.b. Obviously having a shallow depth of field even with semi wide angles + a square format (in the case of 6×6 medium format) helps in creating the medium format “magic”.
n.b. Obviously having a high number of megapixel is nice as lenses can produce (barely) visible/usable “details” up to MTF 5-2% !? at very high spatial frequencies. To sample those a high resolution is needed. High megapixel is good, but could not replace using the lens at its best! As long as the sensor can sample at least the lens details at its MTF50 threshold, a 56×56 mm image will always be better than a 24×36 image, no matter the pixel count (as long as the lenses are good).
n.b. I’m wondering if semi / cropped – medium format digital camera with sensors as (too) small as 33×44 mm (Hasseblad h4D-40), 30×45mm (Leica S2), up to (better) 56×36mm (Leaf Aptus-II 10) 53.7×40.4 (Leaf Aptus-II 12) are worth the price! Compared to 24×36 they will, respectively, get a reduction in lens used spatial frequencies (= uncrop factor) of 1.27, 1.24, 1.54, 1.55 (a detail corresponding to a spatial frequency of 50lp/mm would be reduced to, respectively 39.4, 40.3, 32.47, 32.25 lp/mm)...
Introduction to resolution and MTF curves