How accurate is focus and recompose?
Not sure if this is the right part of the forum for this question but if not maybe a mod will move it?
With all the discussion re the D800 left AF point issue, a number of people have dismissed it as being not very important because they tend to use the centre AF point(s)only.
I take portraits and I like shallow DOF with the eye sharp and the rest going blurred. I've never used the "focus and recompose" technique because I always thought it to be inaccurate.
Now some argue that the difference between focussing a centre AF point on the left eye and then swivelling the camera a bit to the right is neither here nor there, and I realise it will depend on your aperture and just how far you angle the camera away from the original point but I'd still say that in 99% of cases, if your shooting a close up at f1.4 any difference is crucial. I am right am I not?
Sorry if this is a bit daft but it's nearly 2am here in the UK and I'm way past my bedtime
Which of us is right?
#1. "RE: How accurate is focus and recompose?" | In response to Reply # 0JosephK Nikonian since 17th Apr 2006Sun 30-Sep-12 02:58 AM
You are correct in that focus-and-recompose does shift the plane of focus. This can be noticeable up close, getting less noticeable the further way the subject is. Of course, the smaller the aperture, the more more the DOF covers the error.
Seattle, WA, USA
D700, D200, D70S, 24-70mm f/2.8, VR 70-200mm f/2.8 II, 50mm f/1.4 D,
17-55mm f/2.8 DX, 70-300mm f/4.5-5.6 VR, 18-70mm f/3.5-4.5 DX
#2. "RE: How accurate is focus and recompose?" | In response to Reply # 0briantilley Nikonian since 26th Jan 2003Sun 30-Sep-12 06:28 AM
This may be best illustrated with an example
For the sake of argument, let's say you are taking a portrait with an 85mm lens on an FX camera, using an aperture of f/1.4, at a camera-to-subject distance of 6 feet. You decide to focus on the eye with the centre AF point and re-compose. When you have the desired framing, let's say the AF point is over the subject's chin - so it has moved by around 6 inches.
Now, if I remember my Pythagoras correctly, we have a right-angle triangle where the adjacent side (camera to eye) is 72 inches, and the opposite side (eye to chin) is 6 inches. By calculation, the hypotenuse (camera to chin) will be about 72.25 inches - so the distance has changed by 1/4 inch.
With the camera settings and distances described, standard tables tell us that your Depth of Field will be almost 3/4 inch behind and very nearly the same in front of the point of focus.
So in this scenario, the change in focus distance due to re-composing should be easily covered by the available DoF.
However... we should also consider one other factor. Especially if the camera is hand-held, it is very easy (in my experience) for involuntary body movement on the part of the photographer and/or the subject (between acquiring focus and taking the shot) to change the distance to a considerably greater extent than the actual re-framing.
In my opinion, many of the complaints about bad focus due to re-composing are due to this fore-and-aft movement rather than the actual rotation of the lens axis.
Hope this helps
#3. "RE: How accurate is focus and recompose?" | In response to Reply # 2Sun 30-Sep-12 08:52 AM | edited Sun 30-Sep-12 10:58 AM by wooster
Thanks f0lks. Good points. Interesting point you make Brian re the involuntary shift. I hadn't considered that but it makes sense.
I am now led to believe that some of my bad experiences with this technique some years ago is probably due to this rather than the shift on the plane of focus.
This combined with your mathematical example would suggest that if you were using a tripod in a sudio ( as I would do ) then the focus and recompose is a pretty viable way of operating, even with large apertures
Just one thing ( it probably wouldn't affect the outcome much ) but is the triangle actually a right angled one? I would have thought not but I'm mathematically illiterate I'm afraid.
#4. "RE: How accurate is focus and recompose?" | In response to Reply # 3briantilley Nikonian since 26th Jan 2003Sun 30-Sep-12 12:39 PM
In this example, the triangle may not be perfectly right-angled - but since the eye and chin are prety much in the same plane, it won't be far enough out to make a significant difference to the conclusion. Assuming it's right-angled just makes the calculation easier to explain.
If you're moving the camera through a larger angle to re-compose - perhaps when using a wide-angle lens to shoot a full-length subject at one side of the frame with a landmark in the background - then the difference in distance could be significant, but not in the case of genuine portraits (which was the original question)
#5. "RE: How accurate is focus and recompose?" | In response to Reply # 0
Field curvature of the lens will contribute - for better or worse depending on the direction the field curves - though I've no idea if the effect will be significant. It will depend on the lens (and aperture) used.
#6. "RE: How accurate is focus and recompose?" | In response to Reply # 5Sun 30-Sep-12 01:44 PM
Brian. Would the wide angle lens not compensate for the greater swing of the camera by the greater depth of field of the lens? Or would that not be enough to outdo the effect?
AreBee, thanks for the thought. It's a bit too technical for me to understand how it would affect focus for a given situation.
I guess this is a "suck it and see" scenario. I'll stick with using other AF points for now thoug
#7. "RE: How accurate is focus and recompose?" | In response to Reply # 6Sun 30-Sep-12 02:19 PM
>I guess this is a "suck it and see" scenario.<
Why not compose for the shot, but use Live View contrast detect autofocus? You would avoid:
1. Using an off-centre autofocus point, and;
2. Using the central autofocus point prior to focus and recompose.
#8. "RE: How accurate is focus and recompose?" | In response to Reply # 0
I wonder if the nodal point of the lens might have something to do with it.
In doing panos where there is significant detail close up as well as at a distance, we use a tripod attachment with which we can have the camera rotate about the lens' nodal point rather than about the camera or lens tripod socket.
Parallax is readily apparent when this issue is tested with no nodal point correction. And while the big issue is relative movement of far and near objects in the frame, screwing up stitching, it also appears that the focal distance to the subject also rotates away, in an arc, from the original focus plane.
If in the "recompose" process, the photographer rotates his/her head or torso, the camera is essentially being rotated about the camera body, not the nodal point. What would the effect on the resulting focal plane be???
And if it's part of the problem, how in the heck does one solve it handheld?
The issue is more apparent with physically and optically longer lenses. I'd bet an 85mm f1.4 or 1.8, and certainly a 50mm, might not be as susceptible as a 24-70 or 70-200 where the nodal point is far out the lens barrel.
Or maybe the effect is insignificant?
#9. "RE: How accurate is focus and recompose?" | In response to Reply # 8Sun 30-Sep-12 08:20 PM
>I wonder if the nodal point of the lens might have something to do with it.<
Sorry to nitpick, but the term "nodal point" is not correct. Entrance pupil is the correct term.
>If in the "recompose" process, the photographer rotates his/her head or torso, the camera is essentially being rotated about the camera body, not the nodal point. What would the effect on the resulting focal plane be?<
The effect of rotating about an origin non-coincident with the lens entrance pupil will be nil as long as the origin is the same for all rotations.
If the origin is the spine of the photographer then parallax will occur, but will not be relevant because only a single shot is to be captured; if the location is the entrance pupil of the lens then parallax will not occur but is equally irrelevant.
Parallax is an issue for stitched images only because an exact overlap is desired between two or more images.
In addition to the above, the camera still requires to be focussed the same distance from the focal plane of the camera as the final subject will be from it in order to have the focus plane positioned where originally intended.