Some time ago, looking for a way that math-averse photographers can figure out what focus-bracketing step size to use with the GFX 100, I posted this.
Yesterday, I received a question from a reader:
… I’m a bit flummoxed by your asking “how much out of focus is acceptable” re the star target images. My simple-minded answer would be “none.”
It seems there’s another issue to consider. I’ll characterize it as idealism versus pragmatism. Or maybe perfection versus “good enough”.
Maybe because of my engineering background, I view photography as a series of trade-offs that are inherently contradictory. A few examples:
- Portability vs image quality
- Cost vs image quality
- Speed in handling versus image quality
- Depth of field vs diffraction
- Motion blur vs shadow noise
- Available vs artificial light
- Deep depth of field vs shadow noise
That’s far from an exhaustive list, but you get the idea. In photography, if you’re looking for perfection, you’re bound to be disappointed. I’ll go further than that; I think the relentless pursuit of the ideal will harm your photography. The best is the enemy of the good.
Focus is one example. Special effects excepted, we all want our subject in focus, don’t we? Sure. But in general our subjects are three dimensional. With normal single-shot captures, there is only one infinitesimally thin plane of exact focus. With three-dimensional subjects, almost all the subject will not lie in that plane. In fact, with real lenses and their curved fields, even carefully aligned flat subjects won’t lie in that plane. They will be out of focus. Horrors?
Of course not. There are lots of other sources of blur. Diffraction. The sensor pixel aperture. Lens aberrations. Demosaicing. Camera motion. Subject motion. Thermal atmospheric effects. We can’t make all those go away, so what’s the point of picking just one source of blur, focus, to obsess over? Even if we’re trying to obtain an image that is as sharp as possible in the region of the subject, trying to minimize focus blur at all costs will quite likely result in an image that is not as sharp as it could be. If we are trying to minimize focus blurring for a three-dimensional subject, we’ll stop the lens way down. That will introduce diffraction, and lengthen the exposure time, which in turn may cause subject or camera motion to add blur to the image.
When shooting action, minimizing focus error by setting the autofocus priority to “release” may mean the best shots are missed. Manual focus is usually more accurate than autofocus, but, done precisely, takes much longer. Therefore, going for the most accurate focusing may mean missing the best shot.
When trying for the highest sharpness, we should not think only about maximizing focus accuracy. We should think in terms on managing total blur, no matter where it originates. Doing this precisely is not easy. Last year, I spent a lot of time and energy coming up with optimal trade-offs between diffraction and defocus blur. I came up with an algorithm, but wasn’t able to turn my results into something usable in the field.
But I taught myself a lot about the trade-offs and am now better equipped to make them. And that’s made me a better photographer.
Addendum follows:
In another venue, I received the following comment on this post:
Seems to me that a ridiculous question was asked and a ridiculous hyperbolic answer was given. The real answer appears to be trial and error.
So, put a billion monkeys to work at a billion typewriters for a billion years, and see if you get the complete works of Shakespeare? Lotsa luck with that!
But, seriously, this raises the issue of craft versus chance in photography. I am on record as being a fan of chance, but I believe that craft is important as well. In this context, craft means the application of skill and knowledge to achieve a desired end. In combination with chance, craft means the use of skill and knowledge to maximize the probability of the desired end.
If you truly believe in trial and error as a photographic technique, you won’t use exposure meters, zebras, or histograms to set the aperture and shutter speed. Don’t laugh too hard; more than a century ago, it was common to remove the lens cap until it seemed like the film had received enough light. A lot of photographers didn’t believe in sensitometry. But when Edward Weston did that, he wasn’t shooting in the dark; he knew, from long experience, about what to expect.
But let’s turn to the original issue: the selection of f-stop and focus plane in still landscape images. To rely completely on trial and error, you’d want to try every lens aperture and every focus distance within the range of distances spanned by your subject. Say 8 f-stops and 10 focus distances. That’s 80 shots. Now throw in three different exposures, and you’ve got 240 shots. Imagine that you’re shooting moving water and want to get the right amount of motion blur, and you’re up to thousands of shots. All while the light is changing, the clouds are moving, and the universe is headed towards heat death.
Fifty years ago, I used to occasionally do exposure bracketing. I stopped doing that, because I found that, by intentionally making images at other than what I thought was the right exposure, I minimized the probability that the correctly exposed image had all the other elements that I was looking for (It was also expensive, which is no longer an issue).
There’s another element to craft. When employed in service of a vision of the resulting image and its meaning, it not only maximizes the chance that the photograph will succeed, it can be deeply integrated with the creative process. Just don’t forget that the craft must serve the vision, and not the other way around.
Well said.
Mike
the hunting for technical perfection is a mostly amateur thing probably to compensate the lack of creativity. other fields like art photography and I do not have the kitsch in mind which is called fine art in the english speaking world but also commercial advertising photography was and is always focused on content and message first. the use of 35mm did not stop peter lindberg to become one of the greatest fashion photographers of our time and nick knight is shooting absolute impressive flower images with his iphone. the discussion about technical differences and perfection has reached absurd levels maybe a reaction that high quality photography tools have become affordable to more user than ever before. but the need to differentiate is not played out on the creative side which is much harder to accomplish but who owns, uses the best tools or is able to produce technical perfect images.
You highlighted many important facets of the question of focussing; thanks! However, one of them could have been debunked even more:
> “With normal single-shot captures, there is only one infinitesimally thin plane of exact focus.”
This is very far from being correct: inspecting “spot diagrams of the image” (one can google images illustrating this phrase), one can see that there is a whole range of planes where the focus is “at the best” — and there is no way to tell that one of choices in this range is “definitely better” than the others.
In a certain sense, this would give “an objective range” fitting the “idealism/perfection” point of view: choose a dozen of different good objective metrics of PSP: mean spread, median spread, mean quadratic spread, different quantiles, maximum of PSP, etc. They are all would be “reasonable”, but none of them would match the “subjective sharpness” better than the others.
This dozen of metrics would give a dozen of positions of “the best focus” — and the spread of these position would be an “objective measure of depth of focus in the idealistic settings”.
I might put half-a-smiley here — but there might be situations where this “objective measure” may be meaningful…
I’ve done many lens tests that plot MTF50 versus subject distance. The curves are Gaussian-looking “haystacks”. There is no flat spot on top. Whether the differences that I measure are enough to affect normal photography is another issue, that of “good enough”.
Let me repeat: every numerical “objective” measure would have its minimum. Different numerical “objective” measures would have (slightly!) different minima. The question is not 1-dimensional; PSP contains infinitely much information!
The range between positions of minima of DIFFERENT “objective” tests is what I commented on above.
If your point is that different focus metrics yield different results, then I agree with you. But, once you pick a metric, there is only one plane of focus.
Cannot see my post now, but I’m afraid I wrote PSP instead of PSF. Sorry!
One question: I could not understand the last paragraph but one. You write:
> “… I maximized the probability …”
It seems that “maximizing” is good — why then abandon the bracketing? Did some negating slipped through during editing?
I found that, in a bracketed series, that the differences between the images was often more than exposure. If I did a five-shot bracket, the odds that the image that I liked the best was correctly exposed was one in five. If I shot five shots at what I figured to be the right exposure, the odd were much higher than that. This problem doesn’t apply to completely static subject matter.
Thanks, this is clear.
(And, as I thought, it is opposite to what you wrote in the body of the post.)
Thanks. Fixed now.