Please can you expand on this. From my own tests it is striking how the best large-format film lenses approach their diffraction limit at their working apertures. The best medium format lenses for SLR systems are only marginally better at their working apertures and do not approach diffraction limit. Given the substantially larger area of 4"x5" and 8"x10" to even 6x7, let alone the <645 of medium-format digital, this presents a considerable overall optical advantage to larger formats. Of course resolution is not purely a matter of optics and there is significant resolution loss from film/scanning as opposed to sensor capture.
To come to that conclusion I did compare the MTF diagrams of the current Rodenstock lenses. Take the 40 lp/mm graph of the Digaron-S 60mm at f/5.6 for example, it runs very similar to what the Apo-Sironar-S 150mm does at f/11 for 20 lp/mm across twice the image circle. For 8x10" we can look at the Apo-Sironar-S 360mm which is only slightly better at f/22 for the corresponding lines across four times the image circle of the Digaron-S. This pattern even extends to Schneider's Fine Art XXL 1100mm lens which shows only a slightly better curve for 10 lp/mm at f/45.
The Digaron and Digitar lenses for technical cameras seem to be better than most MF SLR lenses. The best MTF for a medium format digital system I've seen so far is the Caldwell 120mm Macro UV-IR Apochromat. The overall optical performance seems to be very close to the Apo-Sironar-S 360mm. Of course the manufacturing tolerances on high resolution lenses for smaller formats have to be much tighter than for the big boys. So there might by a higher percentage of sample variations for digital lenses.
Here is a sample:
This has been my first 8x10" capture. Even though I like the image itself there are a few problems that make it unusable, the bellows shading only being the most obvious. Every step up in format opens new pitfalls to pay attention to.
The lens was a Rodenstock Apo-Sironar-S 240mm at f/16 on Fuji Pro 160S for a little over six minutes exposure time. In fact I scanned this image at 4500 ppi (1.5 GP and 10 GB file size). Back then I was still trying to find the right master file size for my 8x10" images. I even scanned a central chunk of that film at 6000 ppi. At that resolution the full image would be 2.65 gigapixel in a massive 17 GB file. Are there any RIPs that could handle such a file?
I prepared crops at different resolutions to give you an idea on how much pixel density you need to store detail at various contrasts. The crops were sharpened and saved as quality 12 jpegs. From there you can draw your own conclusions.
6000 ppi - 2.65 GP4500 ppi - 1.50 GP (this is a 100% crop from the master file)
2400 ppi - 400 MP (downsized from the 4500 ppi scan)
1440 ppi - 150 MP (downsized from the 4500 ppi scan)
I didn't bother to spot the file. Spotting a multi-GB file is a laborious and time consuming process.
I'm reluctant to pinpoint the resolution of 8x10" to one number. I see it as a range where it takes a very high resolution scan to capture the finest high contrast details while on the other hand the lower contrast textures are getting lost in the film's grain structure. But like I said earlier I see the greatest advantage in smoother tonal transitions in very large prints compared to upscaling by software. The type of film used of course also has an impact. The Fuji Pro 160S has an excellent MTF. It's very fine grained for a color neg but coarser than the new Kodak Ektar 100. Color negs do shine in high contrast situations where the full dynamic range of the film is used. Under normal medium contrast light reversal film has a finer grain appearance.
What is the file resolution of such a 6gb scan?
Typical pixel dimensions for a 3600 ppi file of 8x10" film after cropping the film holder's edges is 34,700 x 27,500.
My own "conclusion" on how many pixels it took to match each medium, assuming the best possible glass being used for all mediums, was ~~50MP to match a perfectly captured and scanned sheet of 4x5 and ~~100MP to match a perfectly captured and scanned sheet of 8x10. Corollary comment here is even when taking extreme care in my set-ups, I got maybe a 30% hit on the "perfectly captured" part for either 4x5 or 8x10. Not saying the other 70% were unusable, far from it -- just not "perfect." (I had a much higher hit rate with the Betterlight because it had electronic focus confirmation and was of course a perfectly flat capture plane.)
Jack, even though I disagree with your MP equivalents for perfect LF exposures on film I do agree that it's almost impossible to make use of the format's full potential for the various reasons you stated. Often it's just some unintentional fall-off in sharpness that doesn't ruin the image but it does compromise the quality. I do mostly long exposures where the wind and the light travelling through the atmosphere can cause all sorts of trouble over time. I even heard of the possibility that the film could buckle inside the holder. Some faults cannot be traced back easily.
May I ask which lenses and what types of film you used for 8x10" back then?
...so, if we want a really large print, and Apo-Digitars do not have enough image circle, shift-and-stitch on 5*7 with a good LF lens is a viable option, and a good alternative to pan-and-stitch?
I was thinking of shift-and-stitch with a digital back, using LF lenses... you would have to invent a stitching back with enough shift.
Don't you think that it would be more effective to use a DSLR and a long lens on a nodal point adapter for super sized stitching? This way you're always using a high quality image circle of a long lens.
-Dominique