Focus and recompose is an extremely common method for getting around AF sensors with limited AF point spread, but it's an imperfect technique that could be causing you more issues than it's worth.
The Problem
A common complaint (particularly with DSLRs as opposed to mirrorless cameras) is that poor AF spread point across the frame makes it difficult to create off-center compositions. This frequently becomes an issue in portraits, where images that do not place the subject dead center are often desired. It's particularly compounded by the fact that portrait photographers often use medium-length telephoto lenses at short subject distances with wide apertures, meaning very narrow depth of field. Here's why that's a problem.
The Math
Remember trigonometry? SOH-CAH-TOA? It's back. Sorry, but we're about to get triggy with it. Let's say I have a right triangle, and I label the bottom angle theta.
Then, I have three sides, the hypotenuse, the side opposite theta, and the side adjacent to theta. Having fun yet? Me too. The trig function I'm interested in is cosine, which tells us the ratio of the length of the adjacent side to the hypoteneuse for a given angle θ. Why am I interested in that? Let me show you.
Meet Ryan. He recently turned 30, loves his work as a writer and teacher, and though he wishes he weren't still single, he's holding out hope and has managed to avoid the existential crises that normally hit when changing decades. As part of his thirty, flirty, and thriving phase, Ryan is having his portrait taken for his new dating profile. To really make Ryan pop, his photographer is shooting him on an 85mm lens at f/1.8 from 5 feet away (this makes side "a" 5 feet long). This gives us a depth of field of 0.11 feet, or about 1.3 inches. Now, let's say Ryan's photographer uses the focus and recompose method, rotating the camera to the left after using the center point to get focus. Here's the key thing to note: the side o represents the plane of focus. Notice how it directly intersects his eyes.
The plane of focus is in red.
The angle has been exaggerated for the purposes of illustration.
Notice that the plane of focus is now behind Ryan. That's the curse of having flat image sensors. But just how much has it shifted? Well, that's just the difference between the length of side h of the triangle and side a. Here's the math:
In other words, the focus point where Ryan is located has shifted 5.077-5= .077 feet or 0.92 inches behind him. Not a big deal, right? It wouldn't be, except the near DOF limit is this case is 0.63 inches in front of him, meaning the entire window of acceptable focus has shifted behind him, and his eyes are no longer within the DOF limits. Whoops.
The Solution(s)
There are a number of ways around this, each with benefits and drawbacks.
1. Tell Camera Manufacturers to Spread Their AF Points Across the Frame
Benefit: This problem wouldn't exist.
Drawback: They're likely not going to listen to us.
2. Shoot With a Normal Focus Point, Then Crop In
Benefit: Your plane of focus will not shift.
Drawback: You'll lose resolution. If you're shooting on a high-resolution camera such as the Nikon D810 or Canon 5DS, you probably won't care. Even on my 30-megapixel 5D Mark IV, it really doesn't bother me.
3. Get Clever With AFMA
Since your focal plane is shifting in a predictable manner, you can take advantage of your camera's AFMA adjustment to dial in a bit of front compensation, pulling the focal plane back toward the subject. However, I don't recommend this for the reasons below.
Benefit: It will alleviate the issue.
Drawback: This method should only be used if you constantly compose portraits the same way and focus and recompose at about the same angle. Otherwise, if you try to take a center-composed shot, for example, it will be front-focused. So, unless you make your living by taking the exact same portraits via the exact same method all the time, skip this.
4. Take More Centered Portraits
Benefit: You'll be using the AF points of your camera as intended and can expect a higher keeper rate.
Drawback: It limits you compositionally.
I like centered portraits. Call me crazy.
5. Stop Down
Benefit: By increasing the depth of field, you can keep the shift within acceptable limits.
Drawback: If you're someone who likes melty wide-aperture portraits, this kills that. Also, you're not solving the problem; you're just reducing its severity.
6. Shoot in Live View
Benefit: You'll get a wide focus range across the frame.
Drawback: It can be difficult to shoot in live view, battery life suffers, contrast-detect AF is typically slower, and it may be tricky to precisely select focus on a smaller feature such as the eyes.
7. Lean Back
So, since we're talking about corrective distances on the order of a few inches, you could conceivably focus and recompose, then lean back ever so slightly to pull the plane of focus back where it belongs.
Benefit: It allows you to compose as you please, but still keeps the image sharp.
Drawback: You're estimating distances. If you typically shoot with the same lens at about the same distance, you can probably develop decent instinct and muscle memory for how much to lean back. I've tried this, and while it's not perfect, I do notice an increase in my keeper rate.
8. Compromise
Shoot with the AF point nearest your subject in the desired composition, then focus and recompose.
Benefit: You get your desired composition and reduce the error.
Drawback: You don't completely eliminate the error, and depending on the spread of AF points on your particular camera, the correction may not be significant.
Conclusion
Focusing and recomposing is an imperfect solution to a peculiar problem. It's important to be aware of the consequences of it, however, particularly if you often notice your portraits being slightly out of focus. I typically use the solution of leaning back or simply cropping in; both work sufficiently well for my shooting. Use whatever works best for your purposes and style.
Leica/Rangefinder users have been dealing with this forever and "lean back" is just what you've got to do. Unfortunately it's a bit unreliable, even as you become more experienced, and especially with super fast f/0.95 to f/1.4 lenses. The newest ones with focus peaking really change the game.
I also believe I read somewhere that Canon actually told 5D Mk II users to focus/recompose with the middle point because the AF was so lacking as a whole. I had to laugh.
This has been a problem for ages, and interesting how many people have not noticed it and notice their images are not as sharp as they can be. The only camera system I know that tries to correct for it is Hasselblad's True Focus and True Focus II system used on their cameras since the H4D. While it is not perfect and still requires some technique to perfect usage, it basically uses motion sensors in the camera to analyze and correct for the horizontal and vertical shift from recomposing the image (and True Focus II now even compensates for the curvature of the lens). It would be nice if this was more exploited by other manufacturers, but until then, choosing the focus point you use with care, repositioning yourself for framing (when possible) to avoid recomposing, or "lean back" would be go-to short-term solutions we can use for crisp images, especially at shallow apertures and when close to your subject.
There is a sense in which it's even worse than this.
Traditional DOF calculations (the basis of http://www.dofmaster.com/dofjs.html) for the near and far limits are based on circle of confusions sizes for a image format in relation to an 8x10 print enlargement. But if you're simply looking at an image on your screen (say 27" 2560x1440) at 100% magnification on a 36MP FF sensor, then you're talking about the equivalent of a ~55" x 36" print. At that size, your near and far limits give you just fractions of an inch.
Of course, if you then make that 8x10 print you're back to the traditional calculations and the numbers are right again.
It's even larger than that. A 27" 2560x1440 monitor would have a pixel pitch in the neighborhood of 110 ppi. A 7360x4912 pixel (36.3MP Nikon D800) image viewed at 100% would be the equivalent of a 67x45 print. But then most of us would view a 27" monitor at a distance greater than the 10"/25cm viewing distance upon which the 8x10 DoF chart is based.
Ah, you're right. When I was punching the numbers into https://www.sven.de/dpi/ I never changed it from 22" to 27". 22" is 133ppi = 55"x36" print.
This is completely new information to me. I was using the focus and recompose method a lot and becoming more and more frustrated with the sharpness of the subjects. Now I understand why. Thanks for sharing.
Dumb question (hey, I'm a photographer, not a mathematician!): Instead of panning the camera left or right, couldn't the photographer simply track left or right by taking a step in the desired direction (or even better, using a tracking mount on a tripod) to maintain the same focus plane while repositioning the subject within the frame? Again, it would be imperfect, but it seems more precise than leaning back, no?
Technically, you could absolutely do this. The two problems I see are the extra time involved to move, and the variance of distance that gets introduced as you move over. You might as well just lean back after recomposing.
Not a dumb question at all! Mathematically, that's a perfect solution. The problem is that the distance required to move the subject across the frame via horizontal translation is much bigger than that needed when leaning back, and I suspect one has much more precision control of their body when keeping their feet planted and simply shifting their upper body slightly.
I thought this was common knowledge as its very basic math/physics. Your solutions are however pretty good to work around it. Personally I would probably use the a point as close as possible and then just very slightly recompose.
Also worth mentioning, there is a lot easier to calculate the difference in distance by just using A^2+B^2=C^2. Didn't really see the point of complicating it with trigogeometry.
It's common knowledge to experienced photographers, yes. There are a lot of people who are relatively new to the craft. There are also a lot of people who are good creatively, but not technically-minded.
Hi David,
I used trigonometry for two reasons. First, it directly connects the angle of rotation to the difference, which I think better represents the physical act. Second, the length of side o is not explicit, as the initial presentation is an angle-side-angle (ASA) problem, so we need some trig anyway. What you can say heuristically is that the hypotenuse is always longer than the two other sides, so we know automatically that the plane of focus will shift backward, but the trig gives us exact results.
At apertures F/2.8 and smaller, depth of field will suffice, as it did in the decades before when the only focusing aid was in the center.
The reason (at least as given by Canon's Chuck Westfall) that focus points are not spread farther around the 24x36mm format is because the focusing system becomes increasingly inaccurate as the light ray angle moves away from perpendicular. If you measure it out, the actual physical spread is the same for APS-C and 24x36mm...but it covers more of the area of the smaller format. Of course, on-sensor focusing doesn't face that limitation.
Lean back is a very useful technique. Stop down too. I shoot a lot with old manual focus film cameras or old autofocus film cameras with single point or very limited multi-point spreads and have adapted my technique consequently. Live view or immediate chimping not being an option, you tend to get better at distance judgement by eye alone.
I have never been a fan of autofocus, the focus point is rarely convenient and shooting nearly wide open allows very little margin for error. I tried back in the Nikon F4 and 5 era but chose to focus manually. Now that I use Canon, I love my L lenses and still use MF. Imagine explaining the law of physics to your client and being too lazy to take control of the shoot.
or go mirrorless
+ it will fix potencial focusing misalignment
Another option... upgrade your d*** camera! That's what I did! lol I went from the 6D to the 5D Mark III solely for the spread of the AF points. I later moved to the 5D Mark IV and enjoy the even further spread (although minimally so) AF points.
Here is my two cents worth.......because of the curvature of the lens, many moons ago doing group shots, we would run a string from the tripod to the group and arrange the subjects along the arch the string would make. Perfect focus except for lens distortion. If you want to add another piece of equipment to your load, purchase a panoramic head and find you nodal point for the lens of choice. Problem solved. Ya a bit cumbersome but it works like a charm! P.S. Tell your subject not to move a muscle!
Hmm, I remember something like this as well for shooting flat lay posters (curving them to match the curved focal plane). If the focal plane was curved as you're suggesting then the entire premise of this article would be incorrect and there wouldn't be any problem with recomposing as your focal shift would be along a curved line.
Alex do you wanna weigh in here with science?
You're describing field curvature, which is undesirable in a lens and is mostly designed out of it these days (Zeiss Planar is a great example).
Uncorrected field curvature is why lenses such as the EF 85mm f/1.2 are so popular for portraiture. They don't do well shooting flat test charts because they weren't designed to shoot flat test charts. They're misunderstood to be *soft* on the edges when in reality they can be very sharp on and near the edges if that is where the lens is focused when aimed at a flat test chart.
The current obsession with corner to corner flat field sharpness is great if you are a landscape photographer or do document reproduction, but actually exacerbates the problem of focus and recompose in portraiture.
There are other advantages to a lens with field curvature for portraiture. Just look at an example shot with a flat field lens that is razor sharp at wide aperture of the classic "shoulder forward" pose when the subject is wearing a textured fabric and the spot on the leading shoulder that is the same distance from the camera as the subject's eyes sticks out like a sore thumb.
Even the most well-corrected lenses have a field of focus that is shaped more like a lasagna noodle than a perfectly flat plane. So the trig is still as much an art as it is a science.
Or you could shoot large format, use a loupe, and tell your model if they move, they die. Maybe that's just a good one for me, though. Great article, Alex. I knew I'd use trig someday!
you should get one of those neck clamps they used in the olden days to hold your models in place :-)
I should! I havena friend who does tin types. Ill borrow his.
Solution: Learn to manual focus.
That wouldn't fix the problem with focusing before composing :-)
Modern DSLRs are not designed for manual focusing, and it can be very difficult to do at wide apertures through the viewfinder.
Agreed. I have the Canon 5D III and it doesn't have interchangeable focusing screens. I own two film cameras, Canon A-1 and New F-1; both have interchangeable focusing screens. I've been using the split-image microprism since 1980 and that is one thing lacking on my 5D. On occasion, I have had to switch from auto focus to manual focus, either due to low light or because the subject of interest was in the background instead of the foreground.
I exclusively shoot in isosceles triangles. Problem solved.
Very well written article, Alex! I use the focus and recompose method a lot and am able to get tack sharp shots shooting at wider apertures. However, for every sharp shot I get, I can count on getting one that's not sharp almost exclusively because I use this method. Even though trig would not be my strong suite (that's a huge understatement), your article made perfect sense. I also loved how you presented multiple solutions as well as the pros and cons to each one. Very informative and well thought out.
Thank you, Korey! I very much appreciate the kind words. :)
Surprised the hasselblad tech isn't mentioned
Prob because we are trying to solve the prob without having to fork more than 20K on a camera body but you are correct. True Focus on a Hasselblad deals with this and it does it very accurately i must say if you follow some basic principles
Come on... This is as a problem as dead pixel...
I used to shoot with 30D and 5DmkII and used that recomposing technique all the time. True that I was selecting focusing point closest to my subject (face, eye, etc)... You can adopt different techniques depends on the style of the image you are going for, but nevertheless I never saw that as a problem. Just stop watching your images blown 100% on the screen.
Finally!
Alex, have you seen the AF point spread on cameras like A7Rii, A6500? Problem solved. Also, for the MF freaks, the instant optional zooming is a wondrous feature.
In my book, focus and recompose = chimping, soooo 2010! Time to upgrade to cameras with modern AF and live view ;)
I couldn't agree more.. That technique has never made an ounce of sense to me. I see so many photographers talking about the hit rate of their 1.2 or 1.4 lenses wide open and they're the same ones who you see videos of focusing on the face then dropping the lens where the center focus is now on the chest.. so now they're back focused. If your desired focus point is somewhere different than your center point, change your focus point to where the eye is, focus and shoot. Even if you have limited phase detection focus on your lens, it's still better than focusing on an axis point and then dropping the lens.
Hasselblad has solved this problem with their smart focus and only uses one focus point.
I shoot on a 5d mkii and focus/recompose is the norm.I'm a big "lean back" kind of guy. My back is so sore after a long shoot!
I'm sure the math is correct, but I just don't have an issue with focus and recompose. I shoot 99% of my work with an 85mm at f/1.4. I love the results.
I'm no expert in mathematics, but I see a common sense problem here. When you 'rotate' your camera to recompose, how does the plane of focus remain on side 'o'? it should also 'rotate'. in other words all the points of focus should lie, not on a straight line as proposed, but on a circle with the same radius as side 'a'. Shouldn't they?
An optically perfect lens exhibits a true focal plane parallel to the sensor and normal to the lens. In practice, there is a small amount of curvature, but that's due to imperfect optics, not the physics.
It is actually due to the physics involved in the apochromat and superapochromat designs used to correct for chromatic aberration.
Why not simply change the position of the camera to align a focus point with the model's eyes, or move the focus point in the viewfinder? No trig required.
Because the spread of points is often not sufficient to achieve the desired composition.
I'm not familiar with Canon or Nikon focusing systems, but I can put a focus point where I want with my Olympus. I've never done any testing to determine if moving the focus point to the edge of the frame results in sharp focus, but I seem to get in focus images.
A typical focus and recomposed shot would be to like a rule of thirds composition. I think the 10° is a bit overstated as rule of thirds is only 4° from center. 10° would really pin the subject at the very edge of the frame or even partially off it with an 85mm at 5 ft. This mathematical example really took it to the extreme.
And this is where Nikon's new AF shines! (For DSLRs, especially the D500).
I personally would use a lcd view finder... good contrast / larger screen / you can zoom in on one particular point and manual focus or let the damn thing focus slowly on the point you want sharpness. Ok - it's slower to focus - but then you are spot on.
Hold on a second and let me pull out my HP-35 calculator.