Hello,
A new member dabbling in. This is my first technical post here - hope I don't immediately embarrass myself by getting into a well discussed subject.
I play with a Zetopan and a couple of experimental rigs (DIY fluoro etc., more on those perhaps later). For visual observation I use Reichert Plan achros and for photos I've gotten a few apos, mostly Zeiss. I stack almost always, midsize, 10-50 layers. Which brings about my question.
Is the planarity of the objective relevant at all when shooting stacks?
The stacking programs presumably assume that the sharp areas of the layer images lie in parallel planes, not on patches of spherical surfaces (as in the case of a non-plan objective). But does it really show anywhere in the final stack that the layer data is slightly curved? Is there some subtle artifact, a telltale sign, in the final product that reveals the non-planarity of the used layers.
My apos are not plans so I can't compare but perhaps more experienced people could comment. My apo stacks beat the plan achro stacks most of the time but that's apples to oranges.
Of course say plan apos often have e.g. better coatings than apos, since they are usually more modern lenses (and they better have them given the multitude of elements), but that is not my point here. So this is really a geometry question.
Cheers,
-Karl
Plans & stacks
Moderators: rjlittlefield, ChrisR, Chris S., Pau
-
rjlittlefield
- Site Admin
- Posts: 23624
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
Re: Plans & stacks
ModelZ, welcome aboard!
The normal 2D image produced by a focus stacking program will not be affected by curvature of field. This is because a stacking program cannot tell the difference between a curved field with a flat subject, and a flat field with a curved subject. Either way, various parts of the image will be in focus and others will not. But of course this happens all the time, in all sorts of patterns, when photographing 3D subjects. Focus stacking programs just preserve the sharp bits, wherever they are in the images, and discard the un-sharp bits.
If you use the stacking program to generate some sort of 3D model based on focus, for example by using the depth map that can be produced by most programs, then curvature of field in the optics will produce a corresponding reverse curvature in the model. For example if the optical focal surface is lower in the center than on the edges (convex downward), then a flat subject will be captured as higher in the center than on the edges (convex upward). This matters if you are making say 3D models of coins and want to be sure that the model ends up flat instead of curved. With more work the curve can be characterized and then removed from the model, but this extra step is not needed if you use a plan objective in the first place.
--Rik
The normal 2D image produced by a focus stacking program will not be affected by curvature of field. This is because a stacking program cannot tell the difference between a curved field with a flat subject, and a flat field with a curved subject. Either way, various parts of the image will be in focus and others will not. But of course this happens all the time, in all sorts of patterns, when photographing 3D subjects. Focus stacking programs just preserve the sharp bits, wherever they are in the images, and discard the un-sharp bits.
If you use the stacking program to generate some sort of 3D model based on focus, for example by using the depth map that can be produced by most programs, then curvature of field in the optics will produce a corresponding reverse curvature in the model. For example if the optical focal surface is lower in the center than on the edges (convex downward), then a flat subject will be captured as higher in the center than on the edges (convex upward). This matters if you are making say 3D models of coins and want to be sure that the model ends up flat instead of curved. With more work the curve can be characterized and then removed from the model, but this extra step is not needed if you use a plan objective in the first place.
--Rik
Re: Plans & stacks
Thanks Rik for a thoughtful explanation.
The coin is indeed a useful example. Integrated circuits or such starkly geometric objects would probably also suffer under non-plans. I mostly shoot biological stuff but this illuminates there, too. If I understand it right in a well made preparate the specimen is supposed to be against the coverglass i.e. kind of "floating" on the mountant. If shot with a non-plan, then the top in the stack will slightly curve downwards (convex upward as you point out). But given that this is often some kind of malleable tissue stuff, this shouldn't necessarily look unnatural at all. On the contrary, since the sample is likely to be slightly flattened against the cover glass when the prep. was made, the stack may show it closer to it original form!
The mundane motivation for this question is whether one should get into updating the apos to plan-apos for photo work. Money probably better spent elsewhere. Especially since plan achros are more user friendly in visual work with their better WD & DOF.
The coin is indeed a useful example. Integrated circuits or such starkly geometric objects would probably also suffer under non-plans. I mostly shoot biological stuff but this illuminates there, too. If I understand it right in a well made preparate the specimen is supposed to be against the coverglass i.e. kind of "floating" on the mountant. If shot with a non-plan, then the top in the stack will slightly curve downwards (convex upward as you point out). But given that this is often some kind of malleable tissue stuff, this shouldn't necessarily look unnatural at all. On the contrary, since the sample is likely to be slightly flattened against the cover glass when the prep. was made, the stack may show it closer to it original form!
The mundane motivation for this question is whether one should get into updating the apos to plan-apos for photo work. Money probably better spent elsewhere. Especially since plan achros are more user friendly in visual work with their better WD & DOF.
Re: Plans & stacks
I wonder if there has been a misunderstanding of Rik's answer? Are you making 3-D models? If not, Rik said the field curvature does not matter when stacking.Integrated circuits or such starkly geometric objects would probably also suffer under non-plans.
Re: Plans & stacks
No, I wasn't thinking of 3D models, just plain stacks. (Stack is just the vertical projection of the 3D model, right?)
But maybe I'm still not understanding something. I do get it when Rik says "a stacking program cannot tell the difference between a curved field with a flat subject, and a flat field with a curved subject".
But isn't it so that the curvature of the object and the field (as the lens has it) do add up in the stack? The final stack is a 2D image but has the curvature represented in it, otherwise we wouldn't see the image as natural. Does this adding of a non-trivial thing, the curvature of the field, result in some kind of perceptible defect (not unsharpness) in the final 2D stack image was my original question. I suspect that for most subjects it probably doesn't, but for a strongly geometric objects (e.g. IC with lotsa straight lines and right angles) it should show up.
Oh well, I guess I have to think this a bit more. Thanks for feedback Lou.
-Karl
Re: Plans & stacks
No, Rik's answer was that if you are only making a stacked photograph, the curvature does not matter. I am sure he can elaborate on that.
In fact, for stacking, sometimes a non-plan objective is better than a plan objective. Optical design always requires compromises, and if a designer has fewer constraints, he or she can do a better job at acheiving the desired result. It takes extra corrections to make a plan objective. Perhaps a designer can make better color corrections if he doesn't have to worry also about correcting curvature of field. For fluorescence or UV objectives, the number of elements matters a lot, and some of the non-plan versions are quite likely to be better for stacking than the plan versions because they have fewer elements. An example is the UAPO-340 series of Olympus objectives for 340nm UV light.
In fact, for stacking, sometimes a non-plan objective is better than a plan objective. Optical design always requires compromises, and if a designer has fewer constraints, he or she can do a better job at acheiving the desired result. It takes extra corrections to make a plan objective. Perhaps a designer can make better color corrections if he doesn't have to worry also about correcting curvature of field. For fluorescence or UV objectives, the number of elements matters a lot, and some of the non-plan versions are quite likely to be better for stacking than the plan versions because they have fewer elements. An example is the UAPO-340 series of Olympus objectives for 340nm UV light.
-
rjlittlefield
- Site Admin
- Posts: 23624
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
Re: Plans & stacks
Lou knows me quite well...
I think what's in play here is the difference between "what you can actually see" and "what you know must be in there".
In principle, curvature of field could introduce lateral distortion. For example, if the surface of focus is convex downward, and you focus the center of field on a planar subject, then you have to move the objective slightly closer to focus the edges. Unless the lens is perfectly telecentric, that moving closer will cause a slight change of scale, which the stacking software may or may not propagate to the final output. If the scale change does propagate to final output, then you have a change of scale that depends on distance from center, and that's the essence of lateral distortion.
But then two questions arise: (1) can you see the distortion at all, and (2) if you can see it, can you distinguish the focus-induced distortion from ordinary common aberrations like barrel and pincushion?
If you actually run the experiment, and look only at the usual 2D stacked output image (the projection of the stack), then I'm quite confident that at least the second answer will be "no". The focus-induced distortion will be small, if any, and will be indistinguishable from ordinary barrel/pincushion.
This is very different from looking at a 3D model or stereo pair, where the curvature can be obvious and unambiguous.
Another example, with quite mundane lenses, is discussed at viewtopic.php?p=57815#p57815 and viewtopic.php?p=57849#p57849.In fact, for stacking, sometimes a non-plan objective is better than a plan objective. Optical design always requires compromises, and if a designer has fewer constraints, he or she can do a better job at acheiving the desired result. It takes extra corrections to make a plan objective. Perhaps a designer can make better color corrections if he doesn't have to worry also about correcting curvature of field. For fluorescence or UV objectives, the number of elements matters a lot, and some of the non-plan versions are quite likely to be better for stacking than the plan versions because they have fewer elements. An example is the UAPO-340 series of Olympus objectives for 340nm UV light.
Briefly summarized, two 10X finite objectives were pushed down to 6.9X by using shorter than normal extension. The CF N Plan Achromat 10X NA 0.30 developed severe streaking in the corners of APS-C. But the non-plan 10X NA 0.25 retained good sharpness even in the corners, despite developing a ridiculous amount of field curvature, about 125 microns difference in focus from center to edge, over a distance of less than 2000 microns on subject.
--Rik
Re: Plans & stacks
Interesting reading, I appreciate Rik spelling these things out. In links, too. Good use of sandpaper! I'd imagine the etched end of a preparate, obliquely lit, would be a good, flat target, too.
I'm not a pixel peeper, in fact I'm a rather lazy tester after the first few "rolls". If later on something in real pics starts sticking to my eye I may investigate what the catch is. But it makes sense to try to understand what is relevant improvement in a new product and to what end. In matters optical there is a fair amount of manufacturers' hyperbola and gadget geeks' drooling distracting from essentials.
These things are rather interesting to follow. Somewhat off along a photographic tangent... superwides are notoriously difficult designs. A few years ago, when sensors started pushing film to margin it became clear that the optimal symmetric design was doomed. Light hitting emulsion ok was too oblique to a sensor array. Goodbye to Super-Angulon, Hologon and such beauties. The first generations of retrofocus designs were not great. Then later on manufacturers came up with more advanced formulas, marketed with ablomp and hideous price tags. Most buyers were perhaps happy but soon clear that not architectural photogs. The subtle "moustache" distortions on corporate headquarters. Vibrating lines on their glass towers. Best paying customers annoyed! Flip side of higher order aspherical corrections exposed. Fixable in PP, but an unnecessary pain.
Later on read that some reputable manufacturers skipped a few optical corrections in these lenses altogether. The computing power in the body made it possible to fix them on the fly, in the firmware. Perhaps the problem above etc. hidden in bytes. Haven't followed this closely, don't know how much of this goes on today. Of course software in PP is great, at least if it works and user knows what is going on.
Always interesting to know more about these thing. What is hoopla, what is niche, what is essential and to whom.
Cheers,
-Karl
I'm not a pixel peeper, in fact I'm a rather lazy tester after the first few "rolls". If later on something in real pics starts sticking to my eye I may investigate what the catch is. But it makes sense to try to understand what is relevant improvement in a new product and to what end. In matters optical there is a fair amount of manufacturers' hyperbola and gadget geeks' drooling distracting from essentials.
These things are rather interesting to follow. Somewhat off along a photographic tangent... superwides are notoriously difficult designs. A few years ago, when sensors started pushing film to margin it became clear that the optimal symmetric design was doomed. Light hitting emulsion ok was too oblique to a sensor array. Goodbye to Super-Angulon, Hologon and such beauties. The first generations of retrofocus designs were not great. Then later on manufacturers came up with more advanced formulas, marketed with ablomp and hideous price tags. Most buyers were perhaps happy but soon clear that not architectural photogs. The subtle "moustache" distortions on corporate headquarters. Vibrating lines on their glass towers. Best paying customers annoyed! Flip side of higher order aspherical corrections exposed. Fixable in PP, but an unnecessary pain.
Later on read that some reputable manufacturers skipped a few optical corrections in these lenses altogether. The computing power in the body made it possible to fix them on the fly, in the firmware. Perhaps the problem above etc. hidden in bytes. Haven't followed this closely, don't know how much of this goes on today. Of course software in PP is great, at least if it works and user knows what is going on.
Always interesting to know more about these thing. What is hoopla, what is niche, what is essential and to whom.
Cheers,
-Karl