noirist
Newbie
All rangefinder cameras have a single focus zone in the middle of the frame. We all know that focus-and-recompose introduces focusing errors. How then do you accurately focus on an off-center subject with a rangefinder camera? Thanks.
ferider
Veteran
It takes a very fast moderate wide angle, and a very off-center subject to make a difference at all. If you stay within the center "1/3rd" of the picture, focus error will always be covered by DOF, also for the fastest lenses.
bukaj
Established
unless you're doing close up shots, in which case you just move the camera rather than pan/pivot to keep it in a parallel frame of your shot. no need to be too accurate here either as you still have some leeway.
-jakub
-jakub
mile
Member
peterm1
Veteran
Thats correct. If you take a pace or two to one side - keeping the same distance from the subject you can recompose with no change of distance to the intended subject. If you have ever fired a gun you will know this phenomenon well. If you move a gun an inch to one side keeping it parallel to its previous position, the bullet will end up an inch to one side of where it would have . But if you angle the gun so the front of the barrel is an inch out of alignement (keeping the rear in the same position as before) the bullet can land meters away from where it previously would have done. Something more or less like this happens when you just twist the angle of a camera to reframe - it effectively changes the distance to the "target" and it is no longer exactly in the plane of focus.
In reality though I have not found it to be a problem unless shooting a very fast lens wice open.
Last edited:
tlitody
Well-known
Are all lenses designed for curved(circular) or flat field subject focus or somewhere(ish) in between ?
Last edited:
Lss
Well-known
You move sideways instead of turning the camera. If the subject is only slightly off center or you are not using a very fast lens wide open, framing off center by turning the camera is not a real problem either for the kind of photography rangefinder cameras are generally used.
ferider
Veteran
Alright, let's do an example:
Full frame, Noctilux (f1), wide open, 1m focus distance, 39 degrees FOV. Focus in the center first, then deflect into a "1/3rd" composition, meaning rotate by 39/6 = 6.5 degrees. Assuming the Noctilux has a flat focus plane (which it doesn't, it's curved in your favor), the focus error is 100cm * (1 - cos(6.5)) = 0.64cm. The Noctilux has 2cm DOF in this scenario which well covers the error.
With a longer lens, the angle goes down, with a wider lens, the DOF increases, either reducing the impact of recomposing on focus accuracy.
Roland.
Full frame, Noctilux (f1), wide open, 1m focus distance, 39 degrees FOV. Focus in the center first, then deflect into a "1/3rd" composition, meaning rotate by 39/6 = 6.5 degrees. Assuming the Noctilux has a flat focus plane (which it doesn't, it's curved in your favor), the focus error is 100cm * (1 - cos(6.5)) = 0.64cm. The Noctilux has 2cm DOF in this scenario which well covers the error.
With a longer lens, the angle goes down, with a wider lens, the DOF increases, either reducing the impact of recomposing on focus accuracy.
Roland.
Last edited:
tlitody
Well-known
So why have hasselblad designed a focus system which compensates for it. not saying your maths is wrong but are they doing this just because they can ?
uhligfd
Well-known
All answers so far (moving parallel and then shooting) work only in landscape mode, right. [In portrait orientation, how do I get up there where my subject is? Question of flight into the air ... ?]
The simplest trick is to realize that the plane of focus swings along with the central axis of the lens, so if you rotate this axis, what was in focus at the center once will be in front of the plane of focus. Hence in both camera orientations, I simply focus on the nose or chin instead of the eyes (for example) and get sharp pics when the face is at the edge of the image.
The simplest trick is to realize that the plane of focus swings along with the central axis of the lens, so if you rotate this axis, what was in focus at the center once will be in front of the plane of focus. Hence in both camera orientations, I simply focus on the nose or chin instead of the eyes (for example) and get sharp pics when the face is at the edge of the image.
PatrickONeill
Well-known
furcafe
Veteran
rxmd
May contain traces of nut
Hi Roland,
Of course the main reason that this seems so harmless is that you are limiting yourself to the center third. As soon as your subject is outside the center third, the effect becomes a lot more prominent, due to the mathematical properties of the cosine..
As a counterpoint to your case, let's assume the opposite extreme of an object at the edge of the frame. To make up for this, instead of the Noctilux let's consider a rather conservative wideangle, a 24/f2.8, wide open at f/2.8, 1m focus distance, 75 degrees FOV. Focus in the center first, then deflect to the edge of the frame. Rotate by 37,5 degrees. Assuming your lens has a flat focus plane, the focus error is 100cm * (1 - cos(37,5°)) = 21 cm. The 24/2.8 lens has 24cm DOF in this scenario, 8cm of which are in front of the focus point and 16cm to the back, so you're well outside your DOF zone and your object will be out of focus. And this is not a particularly fast wideangle.
Take a 21/f4 in the same situation. The focus error is 24cm now, the lens has 31cm DOF (EDIT: no, actually it has 46), so wide open we're still out of focus (EDIT: no, actually barely in focus, but with no rear DOF tolerance left) at the edge of the frame, in spite of the rather slow lens.
Also, the effect is completely independent of distance. This is rather obvious mathematically, too. In the 24/f2.8 example, if instead of 1m focus distance you take 100m, you're still outside focus; the focus error is 21m now, the lens has 24m DOF, 8m to the front and 16m to the back. In fact this is a much more difficult case, because you can't just shift to the side for focusing instead of tilting your head.
It's really only a function of the focal length, the aperture, and most importantly the deflection between the optical axis and the object. If you shoot wideangles, as soon as your object is near the edge of the frame, you have a problem, even if your lens is slow.
Of course the main reason that this seems so harmless is that you are limiting yourself to the center third. As soon as your subject is outside the center third, the effect becomes a lot more prominent, due to the mathematical properties of the cosine..
As a counterpoint to your case, let's assume the opposite extreme of an object at the edge of the frame. To make up for this, instead of the Noctilux let's consider a rather conservative wideangle, a 24/f2.8, wide open at f/2.8, 1m focus distance, 75 degrees FOV. Focus in the center first, then deflect to the edge of the frame. Rotate by 37,5 degrees. Assuming your lens has a flat focus plane, the focus error is 100cm * (1 - cos(37,5°)) = 21 cm. The 24/2.8 lens has 24cm DOF in this scenario, 8cm of which are in front of the focus point and 16cm to the back, so you're well outside your DOF zone and your object will be out of focus. And this is not a particularly fast wideangle.
Take a 21/f4 in the same situation. The focus error is 24cm now, the lens has 31cm DOF (EDIT: no, actually it has 46), so wide open we're still out of focus (EDIT: no, actually barely in focus, but with no rear DOF tolerance left) at the edge of the frame, in spite of the rather slow lens.
Also, the effect is completely independent of distance. This is rather obvious mathematically, too. In the 24/f2.8 example, if instead of 1m focus distance you take 100m, you're still outside focus; the focus error is 21m now, the lens has 24m DOF, 8m to the front and 16m to the back. In fact this is a much more difficult case, because you can't just shift to the side for focusing instead of tilting your head.
It's really only a function of the focal length, the aperture, and most importantly the deflection between the optical axis and the object. If you shoot wideangles, as soon as your object is near the edge of the frame, you have a problem, even if your lens is slow.
Last edited:
tlitody
Well-known
that'll be why hasselblad did it then
Tim Gray
Well-known
In practice, I've found that the error introduced by just pivoting is covered by the DOF in most situations. If you are really worried, stop down or sidestep.
In fact, when DOF is really thin with a 50/1.4 at f/1.4 and minimum focusing distance, I'm likely to sway back and forth by a cm, screwing up my focus. And it's rare that I am taking a picture of someone in said scenario, focus on their eye, and recompose to put their eye at the edge of the frame. Not saying that the focal point never ends up at the far edge, just that it's very rare for me - it's often closer to the center than that.
To further complicate matters, most lenses are designed for flat fields. But they often fall down on that design parameter somewhere near the edges.
In fact, when DOF is really thin with a 50/1.4 at f/1.4 and minimum focusing distance, I'm likely to sway back and forth by a cm, screwing up my focus.
And it's rare that I am taking a picture of someone in said scenario, focus on their eye, and recompose to put their eye at the edge of the frame. Not saying that the focal point never ends up at the far edge, just that it's very rare for me - it's often closer to the center than that.To further complicate matters, most lenses are designed for flat fields. But they often fall down on that design parameter somewhere near the edges.
ferider
Veteran
dofmaster tells me DOF is 57cm for the 21 f4, at 1m distance and on full frame.
But you are right, when you focus outside of the center 1/3 x 1/3 of the picture, the situation degrades rather quickly. Personally, however, I rarely do. And I don't use anything wider than 28. Your mileage may vary, of course.
Roland.
wpb
Well-known
When I focus and recompose I am basically pivoting the camera on a point. The only distance change is the slight change in distance when the camera is rotated. If you focus on an object 1 meter away and pivot the camera to recompose the camera has moved only slightly. No matter what focal length the object is still, for all intents and purposes, 1 meter away.
Last edited:
tlitody
Well-known
But the focus plane has moved. If the lens is perpendicular to a wall then the wall should be sharp from edge to edge of the frame. But the bit in the middle is nearer than the bits at the edge of the frame. Because the subject focus plane is flat the whole wall is sharp. But rotate the camera from perpendicular and the plane rotates with it so that only the bit on the axis in the middle will be sharp(and anything else on the focus plane). The wall at either edge will no longer be sharp because its not all on the subject focus plane.
rxmd
May contain traces of nut
Thanks for pointing this out; I had an error in my calculation (fixed above). The calculator I use (http://www.erik-krause.de/schaerfe.htm) gives me 46cm for the 21/f4. The CoC it uses for 35mm is a bit more conservative than DOF Master's, hence the divergence.
(These are still rather generous values, though. If you have an M9 and use its pixel pitch of about 7 µm for the maximum allowable circle of confusion, you get a DOF of about 6.5 cm for the 24/f2.8 at 1m; i.e. if you have an M9 and are pixel peeping, you'll start to see focusing errors even near the center third.)
Well, I shoot both 21mm and 24mm lenses a lot, so it does matter for me
Last edited: