[philbaum]

In the Sept/Oct 2009 magazine "Photo Techniques" (Photo Techniques Magazine - The Magazine dedicated to professional photographers) there's this article on field curvature's effect on lenses. It makes me wonder which WEB lens reviews are written by folks that really know some of the intricacies of lens performance.

===========================
Some quotes from this magazine article:

1. A lens might exhibit a flat field at close range, yet show strong field curvature near infinity. The Canon EF 14/2.8L II is one such lens, and even f/11 won't overcome the field curvature at infinity focus. Yet another lens might exhibit a mostly flat field at infinity, but show substantial field curvature at close range- the Zeiss ZF 25/2.8 Distagon is one such lens. Such variation usually leads to erroneous conclusions about the merits of any given lens, especially in lab settings and "quick tests"

2. The photos in Fig 2 show modest field curvature using the Nikon 20mm f3.5 AI-S. Optimizing focus for a crisp center yields blurred corners, and optimising focus for the corners yields a blurred center.

3. There is no choice or workaround with field curvature-the optical design of a lens is fixed. All the photographer can do is to understand the behavior and ideally exploit it to advantage while avoiding compositions that are in conflict with the shape of the curvature.

4. Some Leica designs for wide angle have very wavy MTF curves showing both strong astigmatism and field curvature; the Leica Super-Elmarit-R 15mm f2.8 ASPH is one example. The best focus overall for across the frame sharpness is a compromise: adjusting focus for a neutral middle ground, then stopping down, can yield the best results (keep this in mind next time you read simplistic rules about depth of field).

(Phil: I once tried to get the center and edges of a sea wall in focus with a Sigma 10-20 f4. No matter how i stopped down, i could not do it. The above technique might have helped.

5. Detecting Field Curvature: The quickest way to see field curvature without even taking a picture is to use the Live View feature of most cameras (at maximum aperture). A test chart or newspaper taped to a wall can work well. Using Live View, focus at center, then move towards the edges, observing image sharpness. Refocus off-center and determine if the image becomes sharper-if so, you're seeing field curvature (assuming you've aligned the camera squarely to the target)

6. The optical designs of most 50mm lenses are very similar, and nearly all show field curvature that is easily misinterpreted as a lens being "soft". I tested eight different 50mm lenses from Nikon, Canon, Zeiss, Olympus, and Sigma, and found that all had field curvature that became obvious with the appropriate subject matter. I did so because in my testing of the Zeiss ZE 50mm f1.4 Planar, I detected strange variations in image sharpness across the frame, even at f5.6. At first i though this was a bad lens, but subsequent testing of the eight 50mm lenses showed this to be common.

7. The good news is that with most lenses and subject matter you can safely ignore field curvature. But don't be surprised to find odd variations in sharpness at the same desired plane of focus....Especially when comparing different lenses, field curvature is hugely important, because very small changes in focus can shift the sharpness/contrast dramatically. In fact, comparing the same lens to itself can "prove" that it is both better and worse than itself, depnding on very subtle shifts in focus - one reason that making real images the only realiable way to assess lens performance and a compelling reason to be skeptical of casual lens tests you might find on the WEB.

8. An MTF curve that is flat or slopes off gradually usually indicates little or no field curvature...However, an MTF curve might be measured at a distance at which the lens has a relatively flat field, and neither Nikon nor Canon is clear on their measurements, so MTF is not always a reliable indicator...With Zeiss and Leica, MTF is a very good indicator of field curvature, and they indicate the test distance...
============

---

Last edited by philbaum; 11-10-2009 at 02:13 AM.

[Banjo]

Well, very interesting: thanks for the summary!

[georgweb]

Phil thanks for sharing (I can see moiré in your avatar :-),

another good read about this subject I found here Astigmatism and Field Curvature including that same Planar 1.4/50 (design unchanged since 1973).

I guess there is a sweet spot focal length for good flat field performance around 85ish - 105ish in small format lenses. As you point to Zeiss indicating correct data, here's an example for this, Contax Planar 2/100 data sheet. This is the old non-macro version but you see the rare case of a lens exhibiting sort of flat horizontal MTF lines at any aperture - indicating at least at longer distances a good 'corner sharpness' in any f-stop imo.

I always wondered why digital sensors could not be built spherically and the picture being 'flattened' in postprocess in the camera - like we humans do, too.

Best, Georg (the other)

[philbaum]

Phil thanks for sharing (I can see moiré in your avatar :-),

another good read about this subject I found here Astigmatism and Field Curvature including that same Planar 1.4/50 (design unchanged since 1973).

I guess there is a sweet spot focal length for good flat field performance around 85ish - 105ish in small format lenses. As you point to Zeiss indicating correct data, here's an example for this, Contax Planar 2/100 data sheet. This is the old non-macro version but you see the rare case of a lens exhibiting sort of flat horizontal MTF lines at any aperture - indicating at least at longer distances a good 'corner sharpness' in any f-stop imo.

I always wondered why digital sensors could not be built spherically and the picture being 'flattened' in postprocess in the camera - like we humans do, too.

Best, Georg (the other)

Georg and Banjo, thanks.
(Excuse my moire :-)) They have some interesting technical articles in that magazine, wish they had more of their published stuff on the WEB, one can search for and buy individual articles, but it'd be nice if they had more "samples" up.

This field curvature is perhaps one reason why using center focus and reframing (which i often do now:-)) is not theoretically correct. Although in most cases, folks will never see the effect.

Your comment about a curved sensor is interesting. Although i didn't quote it, the article made mention that film had some thickness to it that made focusing less critical than current sensors that have no tolerance of OOF.

After i got my 50-135, i'm embarassed to say that i taped a newspaper to the wall, etc. Incredibly, i found no areas of weakness. But reading this article makes me wonder how many fine lenses are being returned because we may not understand the field curvature. I'm thinking of using the newspaper test for my Sigma 10-20 so that i can better understand its personality. Also, i'm wondering how field curvature on it varies between the wide and tele ends.

Phil

[MattGunn]

I always wondered why digital sensors could not be built spherically and the picture being 'flattened' in postprocess in the camera - like we humans do, too.

Best, Georg (the other)

The field curvature varies from lens to lens and can also varie with object distance for any given lens. For a curved sensor to work it would have to be flexible and adapt to the conditions. This is theoretically possible as flexible image sensors have been developed (see here Characterization of flexible image sensor arrays with bulk heterojunction organi). However, having spent in excess of two years designing and building a camera which can automatically alter its geometry to acheive the Scheimpflug condition, I can assure you that if a camera capable of altering the sensor curtvature to adapt to the lens and imaging conditions were built at present, pretty much no one would be able to aford it.
If field curvature were that much of a problem then it would be cheeper to use special glasses and aspheric surfaces to correct it than to make the camera correct it.

[Ben_Edict]

I always wondered why digital sensors could not be built spherically and the picture being 'flattened' in postprocess in the camera - like we humans do, too.

Best, Georg (the other)

Indeed you could - but in reality only with very large sensor arrays, which are built of many single sensors on a curved surface. You do that for astronomical high res sensor arrays. For a single small sensor. like in our cameras, the production of spherical sensor surfaces would be prohibitively expensive, because it would afford a completely re-designed microlithographic process, which is not feasable, considering the small numbers of sensors (compared to standard microchips, that is).

Ben

[Ben_Edict]

The field curvature varies from lens to lens and can also varie with object distance for any given lens. For a curved sensor to work it would have to be flexible and adapt to the conditions. This is theoretically possible as flexible image sensors have been developed (see here Characterization of flexible image sensor arrays with bulk heterojunction organi). However, having spent in excess of two years designing and building a camera which can automatically alter its geometry to acheive the Scheimpflug condition, I can assure you that if a camera capable of altering the sensor curtvature to adapt to the lens and imaging conditions were built at present, pretty much no one would be able to aford it.
If field curvature were that much of a problem then it would be cheeper to use special glasses and aspheric surfaces to correct it than to make the camera correct it.

One way to achieve that, would be to transfer the principle of adaptive optics onto the smaller sensors - for which the flexible sensor would be a prerequisite. Adaptive optics is well-proven, but not on such a small scale - so, this might be a solution for the future.

EDIT: Curved sensor planes were more common in film days for astronomical Schmidt-Cameras. The focal plane of a Schmidt-Camera (or telescope) is strongly spherical and you would simply bend the film or glass plate to the required curvatore to achieve edge to egde sharpness. If curved correctly, the sharpness of a Schmidt optic is phenomenal.

Ben

[Jonson PL]

Thanks, Chambers has a very fine website too. Though some of it is pay site :

diglloyd.com blog: November 2009

[Ben_Edict]

For anybody interested in the basics, I can recommend a couple of books:

The classic is Conrady's "Applied Optics and Optical Design", which is re-published by Dover very cheaply.
If you are more interested in photographic optics the bible would be Ray's "Applied Photographic Optics" by Focal Press.
An excerpt of this fundamental book is available as "The Photographic Lens". This books does not delve too much into the physics and mathematics, and gives a good and very comprehensible overview.

Ben

[georgweb]

Thanks all for the insights!

I can comprehend Matt's notes that a concave sensor would have to be flexible in order to adjust to different lenses and also object distances.

What about this then: A fixed lens (ultra-)wide angle digicam. I'd think this would be possible with a constant IQ throughout the focus distances.
Why not build it with a fixed aperture like f/8.
Why not bring the sensor very close to the lens.
All this has been done already in the Contax G 16/8 Hologon (plus a special anti-vignette-filter).


Contarex Hologon 15/8
It does look intriguingly similar to a human eye, moreso its predecessors, Sutton's Patent Panoramic Water Lens, 1859 (scroll down for the Hologon).
If there will be small curved sensors one day you'd have a very fine palm-sized camera - I'd bet you will be making phone calls with it, too.

Thanks again for all the valuable insights,
Georg (the cellphone cameraman .-)

[MattGunn]

Curved sensor planes were more common in film days for astronomical Schmidt-Cameras. The focal plane of a Schmidt-Camera (or telescope) is strongly spherical and you would simply bend the film or glass plate to the required curvatore to achieve edge to egde sharpness. If curved correctly, the sharpness of a Schmidt optic is phenomenal.
Ben

But these were designed for fixed aperture and fixed conjugates (infinity focus) so the curvature was fixed.

For anybody interested in the basics, I can recommend a couple of books...
Ben

"Camera Lenses: From Box Camera to Digital" and "Practical Computer-Aided Lens Design" both by Gregory Hallock Smith are also well worth a read if you are interested in lens design.

All this has been done already in the Contax G 16/8 Hologon (plus a special anti-vignette-filter).


Contarex Hologon 15/8
It does look intriguingly similar to a human eye, moreso its predecessors, Sutton's Patent Panoramic Water Lens, 1859 (scroll down for the Hologon).

The Hologon is a particularly simple and elegant lens design, unfortunately its not suitable for DSLR's due to the very short back focus distance. However it is a well traveled lens design having made it to Mars on the the NASA mars rovers cameras (see a chapter from "Camera Lenses: From Box Camera to Digital" here: http://www.mwoa.org/Ch31.pdf). If you want a really wide angle, non retrofocus prime then have a look at the Goerz Hypergon which had a fenominal coverage angle with a simple design (although the ultra thin meniscus lenses must be a real nuisance to make).
Incidentally if you are interested in looking seriously at lens designs then there is good free ray tracing package available here:OpticalSoftware.NET - How To do lens design with WinLens [optical design software], Tolerancer, Glass Manager & PreDesigner, MachVis [lens calculators]