Best Non-Digital LF Lens with Digital Backs

[Greg Miller]

I'd be interested to know more about this. I assumed sharpness was sharpness, and that a lens would act the same regardless of what it was projecting onto. How is a lens designed differently for digital?

One big difference is altering how light gets to the wells nearing the perimeter of the sensor. The wells collect photons best when light enters them straight on, rather than at an angle.

Think of shooting ping pong balls into a tube. You will be more successful when shooting the balls at the tube straight, rather than at an angle.

For wells in the middle of the senor, light tends to enter straight on. But as you move away from the center of the sensor, light enters at more and more acute angles. The design of digital lenses attempts to minimize the angle of entry as much as possible.

[rdenney]

One big difference is altering how light gets to the wells nearing the perimeter of the sensor. The wells collect photons best when light enters them straight on, rather than at an angle.

Think of shooting ping pong balls into a tube. You will be more successful when shooting the balls at the tube straight, rather than at an angle.

For wells in the middle of the senor, light tends to enter straight on. But as you move away from the center of the sensor, light enters at more and more acute angles. The design of digital lenses attempts to minimize the angle of entry as much as possible.

Yes, this is one of the reasons why the digital lenses have more of a retrofocus design. But I wonder if this effect is overstated. There were many who stated that only the digital design lenses would avoid obscene falloff on digital sensors, but I have used a variety of pre-digital lenses on my Canon 5D without seeing any worse falloff than I would have seen on film. Those lenses run the gamut from a Russian 16mm fisheye to a 14mm rectilinear Sigma that definitely predates the move to digital. Sure, these are strongly retrofocus because of avoiding the mirror box, but they certainly were not made specifically to deal with this issue.

It might be an issue with a very short view camera lens used at the edge of its coverage, like, say, a 47mm Super Angulon (pre-XL), which was designed decades before digital sensors. It is only slightly retrofocus, and even on film there is a 3-4-stop falloff in the corners at the edge of its coverage at f/22. The light rays are definitely approaching that image surface at an extremely shallow angles. But that lens has a focal length about 40% of its image circle, which is far beyond the coverage of any of the digital lenses.

Rick "suspecting much of this has been talk without test" Denney

[engl]

There is a very real problem when using digital sensors together with non-retrofocus wide lenses, and that is color shift. Instead of going though one color filter then hitting a sensel, the light rays go though two color filters before hitting the sensel. This results in odd color casts outside the center of the image.

Schneider made a Super Angulon, more or less no retrofocus at all, for Leica. Ken Rockwell has a demonstration what happens when that lens is used with a digital Leica M9:
http://www.kenrockwell.com/leica/21mm-f4.htm

There were many who stated that only the digital design lenses would avoid obscene falloff on digital sensors, but I have used a variety of pre-digital lenses on my Canon 5D without seeing any worse falloff than I would have seen on film. Those lenses run the gamut from a Russian 16mm fisheye to a 14mm rectilinear Sigma that definitely predates the move to digital. Sure, these are strongly retrofocus because of avoiding the mirror box, but they certainly were not made specifically to deal with this issue.

Anyone claiming that a lens has to be "designed for digital" to avoid vignetting is of course wrong. But, digital sensors work better with light coming straight on, so any "designed for digital" wide is going to be retrofocus to avoid color shifts.

[Adamphotoman]

As the pixel size decreases in size the lens needs to resolve better. Fine grained film requires a lens to resolve to what the film is capable of. The newer smaller chips have need of finer resolving power as they are essentially, finer grained with smaller pixel sizes.
Even with the Betterlight scanning backs when you go from a Super 6K to an 8K and then to a 10K the lenses need to be better and better to resolve the smaller pixel sizes; which go from 12 microns to 9 microns and to 6 microns in size. The Super 6 K may not see much difference between an Apo Symmar and an Apo Sironar S but a Super 10K sure does. The 180mm Digital Apo Sironar and the 180 Apo Sironar S that I tested were similar in performance so the extra $ for the digital lens was not merited. Add that the digital had less coverage for movements and the case is closed.
Smaller 645 or 3X3 inch chip sizes need better lenses. You might see the difference on film too but it would depend on the film's capability.
Most colour emulsions are stacked in layers. Theoretically the focus or the magnification of each of the RGB layers would want to be stacked as well. With digital scan backs and Bayer sensors the RGB magnifications are on the same flat plane. Better coatings on the back elements help stop internal ghosting and reflections.

[rdenney]

As the pixel size decreases in size the lens needs to resolve better. Fine grained film requires a lens to resolve to what the film is capable of. The newer smaller chips have need of finer resolving power as they are essentially, finer grained with smaller pixel sizes.

This is commonly stated but it is only true if 1.) resolution is the goal and 2.) one desires the system to be limited by the sensor. If the lens in question provided sufficient resolution for film of the same format size, then it will provide that resolution with a digital back, even if the digital back is able to see the flaws in the lens (I'm excepting the more narrow considerations of very short lenses described in previous posts). Again, one only notices the flaws a 6-micron sensel sees if one is enlarging to the point where those flaws become a problem. Those limits would exist for that lens for film as much as for a digital sensor.

The issue is not that digital sensors have smaller sensels, but rather that they are smaller overall, and require greater enlargement for a given print size.

I have frequently used film that resolved better than the lens in use and still been perfectly satisfied with the results. It is no different with digital sensors.

Rick "who measures need in terms of desired prints" Denney

[Steve M Hostetter]

I used a 115mm Rodenstock and the image got muddy (out of focus) as you shifted on distant objects but did better on closer subjects..
I used a 165mm Schneider and it gave much better results even when shifted.

[Ed Kelsey]

I think a Super Symmar HM 120mm would be good enough. They problem will be with wider angles.

[Noah A]

Yes, this is one of the reasons why the digital lenses have more of a retrofocus design. But I wonder if this effect is overstated. There were many who stated that only the digital design lenses would avoid obscene falloff on digital sensors, but I have used a variety of pre-digital lenses on my Canon 5D without seeing any worse falloff than I would have seen on film. Those lenses run the gamut from a Russian 16mm fisheye to a 14mm rectilinear Sigma that definitely predates the move to digital. Sure, these are strongly retrofocus because of avoiding the mirror box, but they certainly were not made specifically to deal with this issue....

I've never used LF lenses on digital, but I did try out a Leica M9 for a while, and it was a nightmare in terms of edge color shifts with wide lenses. With the 28mm and wider lenses, the edges would have a strong color shift. It was (somewhat) fixed with firmware trickery, but it was a real problem.

The M9 has microlenses on the CCD, as do some digital backs. Those particular backs, if I'm not mistaken, aren't recommended for shift lenses or technical cameras because of asymmetric color shifts and possible sharpness issues when light hits the microlenses at extreme angles. Digital backs with CCDs without microlenses are recommended if you're using lens shifts. Even in that case, software fixes may be required to get rid of color casts with wide lenses or large lens movements.

I would suspect that current longer lenses could work decently well on digital. Lenses like the Apo Sironar S for example. I know the 135mm, 150mm and 210mm versions are extremely sharp and because of the longer flange distance, I doubt the edges will suffer even if you use some movements. You'll have to focus very carefully and avoid stopping down too far to avoid losing too much sharpness to diffraction.

[engl]

This is commonly stated but it is only true if 1.) resolution is the goal and 2.) one desires the system to be limited by the sensor. If the lens in question provided sufficient resolution for film of the same format size, then it will provide that resolution with a digital back, even if the digital back is able to see the flaws in the lens (I'm excepting the more narrow considerations of very short lenses described in previous posts). Again, one only notices the flaws a 6-micron sensel sees if one is enlarging to the point where those flaws become a problem. Those limits would exist for that lens for film as much as for a digital sensor.

The issue is not that digital sensors have smaller sensels, but rather that they are smaller overall, and require greater enlargement for a given print size.

I have frequently used film that resolved better than the lens in use and still been perfectly satisfied with the results. It is no different with digital sensors.

Rick "who measures need in terms of desired prints" Denney

For people buying 40k$ digital backs and 10k$ lenses, I think it is pretty safe to assume that resolution and massive print size potential is the goal. There is no "need" of a sharper lens in the sense that a soft lens will result in a blank photo, but a top of the line lens is needed to see the resolution potential that you paid 40k$ for.

It is no stranger than all the people here who will tell you that you need a sturdy tripod for LF.

[engl]

It might be an issue with a very short view camera lens used at the edge of its coverage, like, say, a 47mm Super Angulon (pre-XL), which was designed decades before digital sensors. It is only slightly retrofocus, and even on film there is a 3-4-stop falloff in the corners at the edge of its coverage at f/22. The light rays are definitely approaching that image surface at an extremely shallow angles. But that lens has a focal length about 40% of its image circle, which is far beyond the coverage of any of the digital lenses.

The HR Digaron-S 23/5.6 has a 70mm image circle, in other words, a focal length of about 33% of its image circle. It covers 112 degrees, while the SA 47/5.6 covers 105 degrees.

Edit: and none of them are close to the Canon TS-E 17mm, with an 67mm image circle (focal length is 25% of its image circle), and 125 degrees coverage.