To infinity and beyond!
I was working on a section about focus for a new book when I got diverted on a search and ended up on someone's page touting the "best hyperfocal tables yet." This caught my attention and I got caught deep in Googledom: following one link after another just out of curiosity.
But before I get to those "best yet" tables, I should mention mine: I include hyperfocal information in most of my books, and I used to provide a calculator that would allow you to create your own custom table. Do I use these tables? Not usually. Should you use these tables? Maybe. Keep reading.
"Hyperfocal focus distance" is basically the focus distance for any given camera, lens, and settings that maximizes the apparent focus distance while including infinity at the far end. Normally you carry a table of hyperfocal values with you and just set what it says as focus distance for the lens and aperture settings you're using. Do so and everything from about half the hyperfocal distance to infinity is "in focus." Since a lens can only focus one distance at a time, hyperfocal focus depends upon the blur circle that is created by things in front or behind the focus distance to be small enough that we still detect it as a "point."
The first problem with some of those tables is that the circle of confusion used is often larger than the photosite size of the camera. Do you think it might be possible that the photosite size actually dictates the smallest detail that can be resolved? Some of us think so. Digital is different than film in that we don't have overlapping recordings of light: we have distinct and separate recordings of where adjacent photons hit the sensor. Digital thus records the blur circle slightly differently than film. On film, the blur circle is pretty much always recorded as a circle (of blur). On digital, the blur circle is recorded as a point until it is larger than a photosite, then gets messed up a bit by the Bayer pattern as it starts to cover two photosites, then begins to get fully resolved once it occupies several adjacent photosites. The corollary is that setting a circle of confusion smaller than the photosite size is nonsensical: the camera can't distinguish any finer detail.
So, your first thought in trying to figure out how much blur is too much blur is to consider how it is being recorded, and when differences will begin to show. Based upon my experience, for discerning users printing big the circle of confusion should probably be between the photosite size and about 1.5x the photosite size. Smaller than that doesn't get recorded. Larger than that starts to put the blur deeply into adjacent luminance values.
Of course you might be saying at this point: "so what if some blur makes it into adjacent luminance values, I'm going to be printing at 300 dpi and I can't even see 150 dpi without a loupe." The contention here would be that low levels of blur are going to get buried in the printed pixels without making it up to the visible level we resolve. True. If we're talking about a 12mp camera and a 12" print, then yes, we have 300 dpi to give the printer and two adjacent blurred pixels aren't really going to be a problem. But when we get to a 24" print from that camera, we're now giving the printer 150 dpi and the blur is now at 75 dpi—we're well into the realm where you'll perceive a difference.
So, first rule: make sure that you're using hyperfocal values calculated with a circle of confusion that isn't too lax for your use.
What makes me laugh when I look at hyperfocal charts—including my own—is how precise they are. Yes, we've got floating point math in our calculation engines these days and we can often get precise calculations out to many decimal points, but, ahem, what's the point? Consider this value I found in one chart: 190.13 meters. Say what?
If we're going to focus at 190 meters (more on that problem in a bit), that's about 600 feet away. It's more than a football field (or soccer pitch). Can we actually get the lens to focus the other .13m precisely? Not a chance in...well, you know what I was going to write. (Bonus points for 200-400mm users: who knows what the lens is "focused" on at 190m. ;~) It seems absurd to publish numbers that you have no hope of actually achieving with any precision using your equipment.
But this same chart has an entry for an 600mm f/1 lens! (My tables only go up to 135mm f/2, which is a real lens. Still, I feel so inadequate now. ;~) I suppose there's a curiousity value in knowing what the number is ("Alex, I'll take Hyperfocal Values for 500." "The answer is 18,947.37 meters." "Alex, 'what is 600mm f/1?'").
Next in the sins of hyperfocal is the usual statement "everything from half the hyperfocal distance to infinity will be sharp." Sharp? No. Theoretically within an acceptable blur that we can't distinguish from an actual point at normal viewing distances? Maybe. But sharp? No. There's also an implication of "equal sharpness." Also not true. If you're using an 18mm DX lens at f/5.6 the calculated hyperfocal distance is about 11.9 feet and while 11.9 feet will be perfectly sharp the further you get from that distance the more blur you'll get, right up until the point where the blur circle gets obvious. In other words, 11.9 feet is sharp, 20 feet is less sharp, 60 feet is even less sharp, and infinity is the least sharp. What you're counting on with hyperfocal focus is that "least sharp" is still not within the bounds of visibility when you print. Print small and it is, print large and it might not be.
18mm at f/5.6 and 11.9 feet was a trick on my part, though. 11.9 is what the tables I include in my books say is the hyperfocal for 18mm and f/5.6, but if you plug 11.9 feet into the distance you'll find that only 5.95 to 7144.91 feet is in focus. Why? Because I use rounding in my tables. You can't actually focus the lens at 11.91 feet (if you can you're doing better than me, and even 11.9 feet is questionable—I should probably round up to 12 feet [which would be about six feet to infinity in focus]).
So all that precision (and all those long focal lengths and focus distances) in most hyperfocal tables are a snipe hunt. Try to use the tables as is and you'll be hunting the illusive snipe, not "focus."
At this point you're probably wondering whether hyperfocal distance is useful at all. Good. You should wonder that. My answer is that it is "sort of good for a limited number of conditions."
- First, hyperfocal is really only usable for wide angle lenses. By the time you get to a 50mm lens, hyperfocal at f/11 is pretty much out beyond all the focus markings on the lens. I doubt that you'll be able to focus precisely enough to use even a rough hyperfocal number with that much lens. That said, DX users in the 10-24mm range, and FX users in the 10-36mm range will indeed find times when hyperfocal is useful. 600mm f/1? Well, I've already dealt with that.
- Second, you can't focus as precise as you can calculate. Moreover, you can't rely upon modern lens markings to get you to a particular distance (instead estimate the distance visually and focus on something at that distance). Thus, round up. If the table says 3.7 feet (14mm DX 12mp f/11), round up to 4 feet. Focus on something you believe to be at 4 feet and everything from about 2 feet to infinity should be within the acceptable blur circle.
- Third, if you're printing large, make sure that the circle of confusion equals your photosite size and round up the resulting values liberally. You don't want the extremes (near and far distances) to have blurs that are large enough to actually get into visible range.
But that's not all folks. I wrote earlier that I don't really use hyperfocal distances. Why did I state that? Hyperfocal focus done right looks wrong. It's an artificial construct promulgated by us photo writers who ran out of things to write about and started inventing things that seem useful but really aren't.
Our brains don't do "hyperfocal." Indeed, a critical depth cue our brain uses is that detail = near, lack of detail = far. See the individual whiskers on a face? The lion is too near. Can't tell if that bump on the horizon is a lion? The lion is an acceptable distance away (though he might be able to recognize you).
So when you use hyperfocal focusing techniques you're actually trying to violate the natural methods by which our eye/brain determines distance. Yes, your big print now is razor sharp. But it looks artificial. You've taken out a depth cue. Just as the latest craze of using tilt shift lenses to restrict focus depth makes images look artificial by adding an artificial depth cue.
That's not to say that there aren't times when infinity being at focus is useful. I often point to Ansel Adams' Denali and Wonder Lake shot: Denali, though it is about 28 miles away from where he set up his tripod, is critically sharp, as is the edge of Wonder Lake, which is only a 100 yards or so away. By having Denali sharp, Adams has thrown away depth cues for something else. Indeed, in his later prints, he uses more burning and contrast within Wonder Lake itself to make it seem more out of focus than it is, further forcing our eyes to run up to the Denali massif. That's pretty much what most people do at that spot. Denali is big. You can't take your eyes off it (assuming weather isn't shrouding it, as happens a majority of the time). Still, you get no sense of how far away Denali might be because Adams has robbed us of the depth cue. Considering that Denali is that big in your eyesight but is 28 miles away, that depth cue is actually important for understanding the true nature of the landscape, in my opinion.
Personally, I like having distance recorded in my shot, so I usually prefer having a slight bit of blur circle on objects that are at infinity. Thus, I rarely use hyperfocal distances unless I have a reason, like Adams did, to do otherwise.
So, if you're using hyperfocal focus charts religiously and your prints all are sharp but dull (sharp in acuity, dull in impact), think harder about how you're using "focus." Consider putting infinity slightly off focus, especially if the far feature is instantly recognizable.