I just wanted to post a reply on the "f64"-thread which seemed to raise very basic questions - so I've decided to give it it's own topic.
I hope it's well understandable, and helps to clear up certain missconcepts:
When we speak about "sharpness" we usually don't mean extinct resolution at high contrast from the lens, the film OR the scanner.
What's important is MTF - which basically means contrast at a certain resolution/ frequency. As most of us know, lenses have a MTF, but film and scanners also have MTF - they all "render" detail with a certain contrast.
Of course there is no "absolute" MTF-value for 4x5" or 8x10" - but let's play with a few values that are common:
- 20lp/mm frequency, which results in about 5000 "pixels" resolution @ 4x5", a frequency which is barely visible at 4x magnification (5lp/mm)
- 40lp/mm frequency to see what happens when we become more demanding - that is the Nyquist-limit (0% contrast, we have to scan with higher resolution to render detail at 40lp/mm) at 2000spi scanning resolution.
The lens:
Sironar-S 150mm - pretty much the best you can get for 4x5". It reaches 80% center contrast @20lp/mm at f11 and about 65% at f22 due to diffraction
There are no 40lp/mm-values so we can only guess given other MTFs:
60% @ 40lp/mm at f11 and 30% at f22?
The Film:
Ektar 100 or Portra 400 resolves 20lp/mm nearly perfect with >90% contrast and 40lp/mm with about 70% contrast (varies with the color spectrum)
The scanner:
I have no MTF-value except for the 1M$ 6000ppi ARRISCAN for the movie-industry (it's custom-made Zeiss lens propably costs as much as a drum scanner) - the MTF is worse than you would expect, because a scanner is a MTF-chain for itself (sensor, lens...).
Let's say 80% contrast @ 20lp/mm and 60% contrast @ 40lp/mm.
A very good drum scanner might be a little better, the very best Epsons are a lot worse. Let's say 50% contrast @20lp/mm and 20% contrast @ 40lp/mm
Now starts the fun:
We can calculate the final MTF of the file by multiplying the MTF-values.
Drum-scan from 4x5" at f11: 80% x 90% x 80% = ~57%
Very good rendition of detail at 20lp/mm.
Drum-scan from 4x5" at f22: 65% x 90% x 80% = ~47%
Still good rendition of detail, difference perhaps not visible below 5x magnification
Epson-scan from 4x5" at f11: 80% x 90% x 50% = 36%
The scanner resolves way more than 20lp/mm, the lens has higher contrast than at f22 but still, the file will appear "softer" than the drum-scan from the f22-image!
How far can we push resolution? So let's calculate with 40lp/mm (under ideal conditions):
Drum-scan 4x5" at f11: 60% x 70% x 60% = 25%
Well defined detail at 40lp/mm - that are 10000 x 8000 pixels of actual resolution rendered with 25% contrast - beyond a 80MP-MFDB!
But why are so many MFDB-samples so good in comparison to 4x5"?
Epson-scan 4x5" at f22: 30% x 70% x 20% = 4%
Each component renders detail @ 40lp/mm but the final image barely contains any information at this frequency! The 80MP-MFDB doesn't have contrast there, either but it holds contrast better below this frequency and therefore appears sharper without actually rendering more detail.
Finally a 8x10" negative with a different lens (let's say f22-MTF is the same) with the same file size as the 4x5" (so 20lp/mm at 8x10" equals 40lp/mm at 4x5" in the file).
Drum-scan 8x10" at f22: 65% x 90% x 80% = ~47%
47% vs. 25% - the lens resolves less contrast but due to the large negative, the 8x10" appears sharper, even below extinct resolution.
What have we learned today? ;-)
- Extinct resolution (taking pictures of contrasty resolution charts) doesn't really matter in reality
- Contrast over detail (MTF) is the key to sharpness - not megapixels or file sizes
- Final MTF is the product of the whole image chain - from the lens to the scanner. One weak link can ruin it all while seeming sufficient on it's own
I hope this is helpful to some!?
Bookmarks