Focus, recompose, shoot

[Wilt]

This topic started on another thread, and wondering about a statement made, I decided to inquire about 'Why?' I thought I'd open this up on a new thread and invite discussion.

True, true, Mr. Clean.....and I believe most of these "faulty" lenses are not really faulty at all....most shooters don't know how to use AF...they open up the aperture wide, then Lock-Focus-Recompose, seeing that the resulting image is soft in focus....then they shout loud and far that the lens "back focuses"...they fail to manually pick and choose the most appropriate focus points, relying (very badly) just on the center FP.

I have to ask "Why is this any different, than focusing in the microprism area of a manually focused camera, then altering framing and then shooting?" Why suddenly does this practice, which worked for decades, suddenly make for false accusations of 'front focus' with AF lenses?


(keyword: trigonometry) to help others do Searches for this thread

I just happened to pull out the owner's manual for a totally different reason, and stumbled across the following instruction from Canon for focusing an off-center subject:
"Select the desired AF point. Focus the subject. Keep pressing the shutter button halfway and recompose the picture as desired. Take the picture."
In other words Canon's own directions are for Focus, lock, recompose, shoot !

[Mark_Cohran]

"Why is this any different, than focusing in the microprism area of a manually focused camera, then altering framing and then shooting?" Why suddenly does this practice, which worked for decades, suddenly make for false accusations of 'front focus' with AF lenses?

It caused problems back then too - and I've got plenty of prints to prove it. :)

This is one of the better explanations I've run across.

http://visual-vacations.com/Photography/focus-recompose_sucks.htm

Mark

[Wilt]

Thanks, Mark. I read the explanation and at first blush it makes sense. However, upon addition thought, the argument seems to start to fail...

First of all, the assumption is that all lenses are flat field lenses, which is not true. Macro lenses are most often flat field corrected, but other lenses are not necessarily. Some lenses like the Tamron 17-50 exhibit considerable curvature of field (photozone.de test report).

Secondly, the math is not quite right on the web example. If we first assume the eye-to-mid chest distance is 12" (it is on me) a angular shift is just over 5 degrees when you recompose at a shooting distance of 10' and it is 11 degrees when you recompose at a shooting distance of 5'. Now if we plug those numbers into a spreadsheet...

http://i69.photobucket.com/albums/i63/wiltonw/focusshift.jpg

we can see the error is 1/2" at shooting distance of 10' and scarcely an inch when shooting at 5' ! Shooting at 10', you need to have an angular camera angle change of 20 degrees to exceed DOF on a wide open lens. So in theory it matters, but in practice...I'm still skeptical. I could be wrong on the Excel spreadsheet calculations!

[SkipD]

Wilt, I have never done any math to calculate the distance differences, but in 40 years of photographing all sorts of subjects, I have never had a problem with the focus-recompose workflow as long as I wasn't doing anything really radical.

Common sense has always dictated trying to use a focusing point that is roughly the same distance from the camera as the area of interest will be in the final framing of the image.

[Mark_Cohran]

I agree with Skip. Focus-recompose works pretty well when you're not doing anything radical, but I think you'll definitely see issues with shallow depth of field and a signficant recomposition.

Mark

[Wilt]

I agree with Skip. Focus-recompose works pretty well when you're not doing anything radical, but I think you'll definitely see issues with shallow depth of field and a signficant recomposition.

Mark

OK, 'significant recomposition' I buy. But focusing on the eyes, then recomposing so they eyes are placed higher in the frame and the center of the frame is midchest does not seem so be 'significant recomposition' enough, in spite of what the web site and the post both claimed!

I can appreciate that if someone focused at 3'clock (using photographer floor position as center of clock), then recomposed and shot someone at 1 o'clock, the DOF of a wide lens aperture could be exceeded (more than 20 degrees of camera shift), that the object which was in focus originally could be somewhat out of focus even if it had not moved.

And the post made AF sound different than what manual center focus film cameras achieved, as if 'you could get away with it back then', accounting for why 85% did it wrong. Yet the geometry is identical and only the automation is different.

[Tom W]

I have to ask "Why is this any different, than focusing in the microprism area of a manually focused camera, then altering framing and then shooting?" Why suddenly does this practice, which worked for decades, suddenly make for false accusations of 'front focus' with AF lenses?

The difference is that today, many digital shooters inspect and judge the sharpness of their images viewed at 100% with their nose against the monitor. Images weren't generally judged at such stringent levels in the past. Indeed, the "circle of confusion" numbers used to approximate depth-of-field take into account a typical sized print viewed at a "normal" distance, usually close to the length of the diagonal of the print.

In other words, we're often seeing slight softness that would not normally be visible in a print. The practice of focus/recompose hasn't changed, but the method of judgement has.

[Wilt]

The difference is that today, many digital shooters inspect and judge the sharpness of their images viewed at 100% with their nose against the monitor. Images weren't generally judged at such stringent levels in the past. Indeed, the "circle of confusion" numbers used to approximate depth-of-field take into account a typical sized print viewed at a "normal" distance, usually close to the length of the diagonal of the print.

In other words, we're often seeing slight softness that would not normally be visible in a print. The practice of focus/recompose hasn't changed, but the method of judgement has.

Good points. Some observations that support or go counter to the point made...

1. Most of the pixel peeping (until more recent times) has been based on APS-C, which has more DOF than 35mm, which is more forgiving than in the past.

2. The 'normal distance' most often is based upon 8x10 final size prints in DOF tables, not 4x6 miniscule images which mask any DOF reduction. Pixel peeping at 100% would certainly be more stringent if the image were spread across a 17" or 19" monitor rather than the normal 13" diagonal size of the 8x10.

3. The fact that computer monitors use three colors to reproduce what was a single dye cloud on a 35mm print could either mask or accentuate loss of sharpness, depending upon one's perspective.

[Thornfield]

I've tried this myself and it varies a great deal depending on which lens you use and at what aperture you set the lens. Working close in also affects it a lot more than working at a distance to your subject.

[Steiglitz]

I made the comment that sparked this thread, and I stand by it. If you want the best sharpness, don't LFR (lock-focus-recompose). In Canon's own white paper (and I am trying to find it to post) they suggest that users use the multi-focal points provide by Canon DSLR cameras. I often work with other pros that do the LFR dance, and my prints are noticaly sharper, cleaner. LFP did NOT work 40 years ago, and it certainly does not work now. It is lazy for users not to master the technique for manually picking and choosing the most appropriate focal point...it not only throws off focus, it can throw off zone metering too. At first it feels ackward, but over time one can master the changing of FP's very fast and accurately.

Multi-focal points are there for a reason...save the LFR dance for F8 landscapes. In general ,anything at F5.6 and wider could get negatively impacted in regard to sharpness.

A real pro uses the camera as it was intended by the manufacturer. One pro I work with close has a 1D Mark II which offers 45 FP's, yet she relies on the center one for all her shots, then she complains that the camera back focuses...all this from a "pro". I tried to tell her several times to pick and choose the most appropriate FP, and she listens then says "it's to complext, too slow"....then she does more complaining about her 24-70L having "focus issues", yet her copy works perfectly on my 5D....go figure.

[Wilt]

Alfred, I agree in principle, but the analysis (message #3) does not support the principle -- unless the angular movement of the camera is sufficient to cause a significant enough shift of the plane of focus vs. the primary plane of interest.

0.5 to 1.0 " error (resulting from 5-10 degrees of shift) hardly seems a large enough shift, if we use the web site article that had been quoted, as the argument for why the practice of focus-recompose-shoot should not be followed. Care to comment?

[Mark_Cohran]

All I know is from my experience. I haven't done any calculations or any controlled tests. I seldom do - my knowledge is empirical based on my own photography and style. I got away from using FLR in the early nineties (with film cameras) based on advice from a working pro, and I think my photographs have improved for it.

Mark

[Thornfield]

Here's a link to a sample image I put up a while ago while doing a test on this

http://www.photography-on-the.net/forum/showthread.php?t=244240&highlight=focus

[Wilt]

Here's a link to a sample image I put up a while ago while doing a test on this

http://www.photography-on-the.net/forum/showthread.php?t=244240&highlight=focus

I see your test with the 17-40mm lens, on a body unstated.

Assuming it was on a FF body, it has a horizontal FOV of 57 degrees, so a pen at the edge of frame would certainly be at a large enough angular displacement that recomposing to bring it to the center of frame would indeed induce plane of focus shift large enough to result on Out of Focus...if the pen was 3' away, the DOF at f/4 is 4".

Assuming it was on an APS-C body, it has a horizontal FOV of 37 degrees, and the pen at the edge of the frame would still the cause of a large shift of the plane of focus, and if the pen was 3' away, and the DOF at f/4 is 2.4" it would be again exceeded.

So regardless of body used, yes, the shift in plane of focus would exceed the DOF of the lens wide open, exactly as your tests had shown. On that we are in agreement!

But that is not my issue in question. My issue is that a web article depicted a portrait situation, focusing on the eyes then shifting center of frame to midchest, a distance of only 12" and an angular distance of 5-10 degrees, depending upon shooting distance. And THAT situation is the one that I am disputing the validity of the web article!

[SkipD]

A really important test that we could conduct with a few common lenses would be to see how flat the focal plane is. In other words, if you focus on an object at a specific distance from the lens and swing the camera on a tripod does the specific object stay in focus or go out of focus. I guess I have assumed that anything that stays the same distance from the lens will stay in focus. I have never done any controlled experimentation to find out.

[Wilt]

A really important test that we could conduct with a few common lenses would be to see how flat the focal plane is. In other words, if you focus on an object at a specific distance from the lens and swing the camera on a tripod does the specific object stay in focus or go out of focus. I guess I have assumed that anything that stays the same distance from the lens will stay in focus. I have never done any controlled experimentation to find out.

Sounds interesting. A challenge is deciding 'standards' for such testing. For example, if you use a detailed wall as your target for field flatness, how do you deal with the issue that many people have heads without marks for pan angular control, or that the angle marks are too coarse? And how do you ensure that everyone would be absolutely perpendicular -- both horizontally and vertically -- to the wall. And how far away from the way?

I'm all for a test, but standardizing the above (and maybe a few more parameters I hadn't though of) is the issue, as well as the precision of such observations.


For the test of the tolerance for focus-recompose-shoot process, a standard needs to be established there, too, because past tests have at times failed to define distance-to-subject or FL (or Angle of View).

a. on a zoom lens perform the test at shortest FL and also at longest FL.

b. the distance to subject...I assumed 3' on the test Thornfield had done (since I did not see it stated in his test) but maybe 5' is more reasonable distance (or not!) This distance DOES affect the DOF amount and also the angular displacement from edge of frame to the center, so is a critical factor in standardization of a test!

c. angular displacement...we could assume that you start at center of frame and move the rotationally horizontally, until the object is positioned 3/4 of the way between the center and the edge of the frame.

[Steiglitz]

You guys can test all you want, but answer this question: Why does Canon provide 45 focal points, and suggest that shooters manually choose them? Why does Nikon, Olympus, others do same?

[Wilt]

You guys can test all you want, but answer this question: Why does Canon provide 45 focal points, and suggest that shooters manually choose them? Why does Nikon, Olympus, others do same?

For the simple reason that the camera has no idea what object in the frame (or a portion of an object, like eye vs. bridge of nose) YOU WANT to be in best focus. Or for the reason that AI servo can track your item of iterest better.

[Steiglitz]

Correct....so if that thinking is good enough for Canon, then choosing the most appropriate FP manually is good enough for me. ;-)

[Bodog]

Here's another thought. Using the center focus point then shifting away to recompose will put your subject further away from the center of the lens, which normally gives the highest resolution. Since the subject is now in a "softer" portion of the lens, might not someone assume the softness is due to being "out of focus"?

[Mark_Cohran]

Here's another thought. Using the center focus point then shifting away to recompose will put your subject further away from the center of the lens, which normally gives the highest resolution. Since the subject is now in a "softer" portion of the lens, might not someone assume the softness is due to being "out of focus"?

Uhmmm, dude, I don't think it works that way. :)

Mark

[Bodog]

Uhmmm, dude, I don't think it works that way. :)

Mark

Not the first time I've gotten things backwards. What have I missed?:oops:

[Thornfield]

If anything this discussion should make people realise that the center FP is not the only FP on the camera.

Personaly I think the dim viewfinder in the 20D is not helping either.