Zeiss Touit 32mm f/1.8 Planar: Reader Question on Field Curvature (and field curvature in general)
Nicholas W writes:
I appreciate very much the work you are doing on testing the NEX system, given the increasing lens opportunities and the quality of Sony sensors its future is promising.
Let me just suggest you to include some test on subjects that lay on the same focal plane (could that be also at near infinity, as you did with tile rooftops?). The bike, dolls and cabin have great details to look at, but as your last test on Touit vs Sony showed, provide a very uneven focal plane that makes corners very difficult to judge.
To give you an example, I would like to find out how much quality (if any) increase I would get from a 32mm Touit or Sony compared to my actual setup (SpeedBooster + canon 50 1.8). My guess is little both at 1.8 and 5.6 (if I do not consider very extreme portions of corners) but from your tests is very difficult to tell, mainly due to uneven focal planes of your subjects.
One last question, which software do you use, and which settings, in order to get the sharpening seen on your last tests?
DIGLLOYD: As for sharpening, my approaches involve both Adobe Camera Raw and Topaz InFocus, documented in the Workflow section of DAP.
I hope to make some careful studies at a distance soon.
Of the twenty or so 50mm (equiv) lenses I have tested, all but one has field curvature at distance— even what one would hope to be flat field lenses like the Zeiss 50mm f/2 Makro-Planar (corners)*. The solution: shoot at ƒ/5.6 or ƒ/8 for geometrically planar scenes (such as an infinity scene).
The Leica 50mm f/2 APO-Summicron-M is the best corrected 50mm lens I have yet seen (on all counts), though it might have a trace of field curvature judging by the MTF chart at ƒ/2 and ƒ/2.8. Its sibling Summilux has considerable field curvature.
A flat field lens (no field curvature) is rather the exception, especially at 50mm and wider focal lengths. It is not easy to correct field curvature for apertures faster than ƒ/2, and even at ƒ/2 it requires considerable optical effort. All optical designs trade off some benefit for others.
That said, the correction for field curvature is given far too low a priority in optical design, and is second only to focus shift as a imaging “gotcha”. I prefer lenses that are dead-on predictable in every scenario; field curvature undermines that in spades. Hence I would prefer a highly corrected flat-field ƒ/2.8 lens over an ƒ/1.8 or ƒ/2 lens. This is what Sigma has done with the 30mm (45mm equiv) lens on the Sigma DP2 Merrill, and it is a wonderful thing.
* Field curvature is a curved zone of focus, e.g., the nominal plane of focus is actually curved or wavy. It should not be confused with optical distortion, a warping of the proportions of the subject (straight lines do not remain straight). Nor should optical distortion be confused with perspective, which is purely a camera to subject distance issue not involving optics at all, and obeying the inverse square law. See also the size invariance principle.
Below, an example of a geometrically planar subject with two planes: the foreground railing and the distant background. Few 50mm (equiv) lenses can image this scene with full sharpness across the field on high-res digital until ƒ/5.6 or even ƒ/8. A tiny change in focus can “optimize” the placement of focus for such scenes against the specific field curvature, which also means that comparing lenses fairly on such scene is fraught with risk of erroneous conclusions by very small changes in focus—very hard to do fairly, there are no “quick tests” in spite of the apparent ease of the scene.
For more on this and other topics, see Making Sharp Images.