Focus & Recompose: Surprise Result

*newarts* · 08-05-2009, 03:06 PM

I wondered how focusing & then shifting the camera to recompose was affected by depth of field, etc. I did some some approximate math* & an experiment.

In the experiment, I put the camera's lens at the center of a rug with circles on it and took photos focused at various distances. To my surprise, even when DOF is small, the rug's circular arcs were in pretty good focus from one edge of the frame to the other!

The lenses were Pentax 50mm F 1:1.7 @ 1:1.7 & Tamron SP XR Di 28-75mm 1:2.8 @ 28mm 1:2.8.

The photos skimming the rug surface (top & bottom) show that the focal plane isn't really rectilinear with these lenses; the DOF is less than curvature of the circular arc centered on the frame.

The center and ends of the arcs all being in focus means one can focus on a subject in the center of the viewfinder then rotate the camera to move the focused object to the viewfinder's edge & shoot without losing the subject's focus.

I guess I'll stop worrying about DOF effects when I Focus & Recompose! The math indicates the result should hold at long focal distance too; it works out to the edge of the rug at least.

Dave

* simple theory implies the ends of a circular arc centered on the camera's optic axis should go out-of-focus when the depth of field is small.

Gooshin · 08-05-2009, 05:10 PM

how far away are YOU from the rug?

*newarts* · 08-05-2009, 05:53 PM

Originally Posted by Gooshin

how far away are YOU from the rug?

I'm sorry that I didn't clearly identify the components of my posted photo composite.

The center photo is looking down at the circular rug so you can get an idea of how much the blue stripe actually curves. (a radius of about 410mm.)

I was lying on the floor behind the camera looking thru the viewfinder for the top & bottom photo strips; the camera's lens was right above the rug's center. The field of view 's center skimmed the rug's surface as indicated by the lines on the top-down photo.

The top strip is the in-focus stripe cropped all the way across the 50mm 1:1.7 lens' photo when focused on the Blue zig-zag stripe on the rug.
The bottom strip is the in-focus stripe cropped all the way across the 28mm 1:2.8 lens' photo when focused on the Blue zig-zag stripe on the rug.

Dave

Gooshin · 08-05-2009, 06:08 PM

ohh, i get it now.

*pause*

suggestion, i would try the re-composition experiment with a target 3-4 meters away, as that is the intended application.

*newarts* · 08-05-2009, 06:27 PM

Originally Posted by Gooshin

ohh, i get it now.

*pause*

suggestion, i would try the re-composition experiment with a target 3-4 meters away, as that is the intended application.

I'll try - I'll use a BB court's circle maybe; or maybe the three point line.

Theory predicts the effect dof effect on recomposing should be less as the distance increases. (Dof increases with the square of distance at a constant f-number, while the depth of the arc across the frame increases linearly with distance.... this implies less of a problem as distance increases.)

Dave

Class A · 08-05-2009, 06:30 PM

Originally Posted by Gooshin

i would try the re-composition experiment with a target 3-4 meters away, as that is the intended application.

The further away the subject, the less critical to get it within the DOF (as the latter becomes larger).

Dave, falconeye and I did some maths in a thread that brought up the "focus and recompose" topic. The main thread topic turned out to be due to a user error but the postings starting from post #69 may be of interest to you.

It turns out that in critical applications "focus & recompose" can introduce significant errors, but we were assuming a flat focusing plane.

I understand that your experiment is meant to demonstrate that typical lenses have sufficient field curvature to make "focus and recompose" a non-issue.

While I see your point, I think that in critical situations one still needs to be cautious as, e.g., one is often unable to perfectly rotate the camera but will involuntarily introduce also some camera to subject distance shift.

Class A · 08-05-2009, 06:50 PM

Originally Posted by newarts

Theory predicts the effect dof effect on recomposing should be less as the distance increases.

I was still writing my posting when you had submitted yours already. Hence the double mentioning.

What could change in practice, is the field curvature of the lenses. Perhaps they may become flatter the further the focus point is away? This is mostly speculation.

Still, even if the focus plane became completely flat, the DOF at 3m with a 50/1.7 is ~25cm while the focusing error (simple flat focal plane theory) is just 8.5cm, if you rotate by the maximum 13.5 degrees. One needs closer distances for the effect to become significant.

The maximum rotation degree results from the 27 degree FOV for a 50mm and the fact that you will use the centre to focus and still will want to see what you focused on at least at the very edge of the frame.

*newarts* · 08-05-2009, 07:31 PM

Originally Posted by Class A

I was still writing my posting when you had submitted yours already. Hence the double mentioning.

What could change in practice, is the field curvature of the lenses. Perhaps they may become flatter the further the focus point is away? This is mostly speculation.

Still, even if the focus plane became completely flat, the DOF at 3m with a 50/1.7 is ~25cm while the focusing error (simple flat focal plane theory) is just 8.5cm, if you rotate by the maximum 13.5 degrees. One needs closer distances for the effect to become significant.

The maximum rotation degree results from the 27 degree FOV for a 50mm and the fact that you will use the centre to focus and still will want to see what you focused on at least at the very edge of the frame.

Thanks for the comments. I'd followed the previous thread closely & set out to solve the analytical problem essentially the way you did, except I used only distances, not angles. I want to describe the situation in terms of field width, focal length,etc.

My results were similar to yours and I thought my best way to check things out experimentally was to use conditions where I thought the effect would be greatest, ie. small dof.

Your numbers are roughly consistent with my observations and the cambridgecolour calculator; although I'm pretty sure my results show better focus tracking across the focus arc than predicted with a flat focal plane.

If you look closely at my 28mm experimental photo you'll see that the red stripe goes in and out of focus across the photo in a manner consistent with a flat focal plane.

Dave

PS you said

I understand that your experiment is meant to demonstrate that typical lenses have sufficient field curvature to make "focus and recompose" a non-issue.

Actually my experiment was intended to demonstrate the refocus problem dramatically by using small distances; I was surprised by my results & called on lens field curvature as a possible rationalization of the observations.

Gooshin · 08-05-2009, 07:43 PM

hmm

in contrast, i would just like to add my recent change in shooting style, i used to center + recompose all the time, and more often than not would get less than desired results

i have since started to put more trust in my cameras (K20D) auto-focus matrix, using the focus-select to focus my non centered shots

overall i am achieving greater focus accuracy.

Class A · 08-05-2009, 09:19 PM

Originally Posted by newarts

Actually my experiment was intended to demonstrate the refocus problem dramatically by using small distances; I was surprised by my results & called on lens field curvature as a possible rationalization of the observations.

Hmh, I guess (I haven't done the calculations for your setup) the magnitude of the focus shift effect is a better/sufficient explanation than field curvarture.

Your experiment is well thought out and well presented but I find the stripes a bit hard to read, i.e., difficult to ascertain where the focal plane really is.

Everyone's notion of acceptable sharpness differs but to me your test indeed (to the extent I can read the results) demonstrates that the circles are not as in constant focus as a "focus & recompose" shooter would wish for. The out-of-focus shifts may be smaller than expected, but they are big enough for me.

Originally Posted by Gooshin

hmm
i have since started to put more trust in my cameras (K20D) auto-focus matrix, using the focus-select to focus my non centered shots

overall i am achieving greater focus accuracy.

That's also my preferred way. While the theoretical focus shift introduced by optimal "focus & recompose" isn't large, in practice there are a number of other factors that are not favourable for the "focus & recompose" technique (higher demands on AF precision, strong likelihood of spatial translation in addition to rotation).

**Marc Sabatella** · 08-05-2009, 09:21 PM

Originally Posted by Class A

While I see your point, I think that in critical situations one still needs to be cautious as, e.g., one is often unable to perfectly rotate the camera but will involuntarily introduce also some camera to subject distance shift.

Right, and in my own experiments, this actually dominates any other effect at short distances. At longer distances, DOF tends to be such that it is much harder to get focus-recompose to fail.

*DanLoc78* · 08-06-2009, 11:57 AM

lately i've been trying to use manual focus along with DOF preview. I never gave a second thought to problems that could arise from focus/recompose...as it was how i naturally taught myself to shoot. oh well, i will definitely pay more attention now! but i've had success using the manual focus technique...depth preview is very helpful.