Should AF focus point affect sharpness?

Merlin

Merlin

Messages
47
Edit My Images
Yes
Playing with Sigma 24-70 2.8 last night I noticed something a bit odd.

At F8 with suitable shutter speed and just the centre AF point selected I took a shot of a newspaper square on. Newspaper filled the frame.

Centre of the image was sharp but gets soft towards the edges.

Taking the same shot with all AF points selected the whole image is sharp.

Can anyone explain why this is?
 
Sounds to me as if the same shot has in fact been taken with slightly different focus as the camera has detected a slightly different focus point on the subject. When you're working quite close it doesn't take too much to throw it out. I assume you are quite convinced everything else in the shot was the same, tripod used? Aperture definately the same etc etc. Could you post the pics c/w exif data?
 
I'll repeat the 'test' and post the results.

But yes, everything was the same apart from AF points selected.
 
:thinking: I'll watch this one with interest..
 
Off the top of my head and possibly wrong... my guess would be it is because the distance from the lens to what is in focus is constant. So the sharp zone is on a spherical, curved plane and the newspaper is flat. When you use an off centre focus point more of the curved focal plane covers the newspaper with some oof in front and some behind - but a better average.
I'm sure I read somewhere macro lenses have a flat focal plane so are the best type of lens to use for copy stand type work.
 
This is how the AF would work....it's normal.

If you're sighting perpendicular to the target the the centre point is the nearest part of the target and all other points are covering an area slightly further away....in the rear portion of the DOF.
If you choose multiple AF points then it will be the average focus distance and the focus length will be somewhere on the target between the centre and the portion covered by the furthest out AF points. This puts the centre in the front DOF region and the edges in the rear DOF region.
If you really crop it down then you'll find that the centre of the multiple target shot is slightly softer than the centre of the single point AF shot.
The further away the target is, the less obvious the difference is when using the centre AF point....that's provided you don't change the focal length if using a zoom.

Bob

As usual, Robert 's beaten me to it....so it's "what he said"
 
makes a change for me to get in first :)
 
Canon Bob said:
If you choose multiple AF points then it will be the average focus distance and the focus length will be somewhere on the target between the centre and the portion covered by the furthest out AF points.
Click to expand...

interesting, I thought the AF picked out the nearest of the focus points and used that :thinking: :shrug:
 
Wookie said:
interesting, I thought the AF picked out the nearest of the focus points and used that :thinking: :shrug:
Click to expand...

My understanding is that it puts as many AF point in focus as possible and then averages. Most subjects have a larger depth of field and would not respond in the same way. A 2d sheet of paper would allow all points to be valid...except on extremely wide closeups.

Bob
 
RobertP and CanonBob, you've both got it horribly wrong I'm afraid.

Cameras focus on a plane, not on a specific distance. If you focus on a point that's 3m away, say, it does notfocus on all points which are 3m away. (This is easy to confirm experimentally.) Instead, it focuses on all points which are in the plane which includes the focus point and lies parallel to the sensor or film plane in the camera.

EDIT: Some of this (below) is rubbish. It just goes to show how confusing this subject is! I've made the bits that are wrong small and grey. I'll try to correct it in a later post.

Here's what should happen here, if there's no experimental error - i.e. if the camera is oriented so that the axis of the lens is perpendicular to the newspaper, the newspaper is flat, etc. When the centre AF point is used, the sensor plane of the camera is parallel to the plane of the newspaper, so the entire newspaper will be in focus. But when an off-centre AF point is used, the line joining the focus point to the camera is not perpendicular to the plane of the newspaper, so the entire newspaper will not be in focus.

What Merlin is seeing is, in fact, the opposite to this. This strongly suggests that the experimental setup is faulty. The camera is not oriented correctly, so the centre AF point is not pointed perpendicular to the newspaper, but one of the other AF points is (or at least more nearly so). Hence the observed results.

Incidentally, this sort of test is incredibly difficult to do properly. If you ever see someone claiming they have focus issues as the result of a newspaper test, there's a high probability that the test setup is faulty. (Sorry Merlin.)

When I can get to my PC I'll draw a picture, which will be a lot easier to understand.
 
tdodd said:
I'm not sure if this is the situation being described in the original post (my brain hurts this morning) but here's some examples of the danger in misunderstanding how lens focal distance works....

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

http://www.mhohner.de/recompose.php

http://www.bobatkins.com/photography/technical/focus_recompose.html

Might save reinventing the diagrams.
Click to expand...
Thanks tdodd. I'd seen the first of these before, but not the other two.

I hadn't wanted to introduce the focus-recompose issue, because that's not what Merlin is doing and I was concerned that it would be a red herring. However, (re-)reading the focuus-recompose articles has shown me that my logic was faulty. I've edited my previous post to show which bits were wrong.

Let's try again.

If the test setup is sound, then the plane of the newspaper will be parallel to the plane of the sensor. Whichever AF point is used, the entire newspaper should be in focus.

If the entire newspaper is not in focus when the centre AF point is used, it strongly suggests that the plane of the newspaper is not parallel to the plane of the sensor.

I can't work out how to interpret the results using off-centre AF points at the mpment. Like tdodd, my brain hurts! I'll try to come back to it later.
 
The discussion here may have some relevance, and could explain why a less than perfect lens design can show soft edges/corners when using the centre focus point vs the averaged focus distance obtained from all focus points used at once. I didn't read through the whole discussion but it seems all lenses are not created equal...

http://photo.net/bboard/q-and-a-fetch-msg?msg_id=004Mvi

Can you imagine if, when using a wide angle lens (say 10mm on a cropper, or 16mm on a 35mm camera), you had to arrange everything in you shot to be on the boundary of a sphere? Imagine trying to do any architectural sots with a lens that didn't focus on a flat plane. You wouldn't be able to shoot anything at all that had a flat surface, like a wall/building or a flat sheet of newspaper. Apart from fish-eyes, a lens should focus on a flat plane.

It sounds like maybe the OP's lens is not focusing on a true flat plane. If the focus accuracy with a single focus point is on the edge of acceptable DOF limits, rather than being completely accurate and bang in the middle of the DOF zone, then any defect in the focal plane flatness could easily show up. If the lens is front focusing and the lens has a very slight curve to its focal plane then the corners/edges could well fall outside the safe zone for DOF.

EDIT : bloody hell - I just noticed the problem is at f/8 :eek:. Wide open I can see the potential for a problem, but at f/8 :thinking:, I have no idea what the problem could be.
 
tdodd said:
EDIT : bloody hell - I just noticed the problem is at f/8 :eek:. Wide open I can see the potential for a problem, but at f/8 :thinking:, I have no idea what the problem could be.
Click to expand...
Crikey. Good point. Let's hope it was a very dodgy test set-up, because the alternative would appear to be a very dodgy lens...
 
I's be very wary of these results. Setting up an accurate test shot of newspaper so that the whole of it is parallel to the sensor is extremely difficult, verging on impossible. A good copy stand would make matters somewhat easier.

Whilst not pretending to understand the intricacies of the multi point AF system, it's fairly obvious from my use of it against birds in flight etc, that some sort of averaging between the points must go on. Given that many lenses now transmit distance information to the camera, I'd be surprised if some sort of intelligence isn't at work to determine the likely most relevant AF point, e.g. closest to the camera in many cases.
 
I just plugged some numbers into the DOF calculator here - http://www.dofmaster.com/dofjs.html. Now the only thing we know is the f/stop, so I assumed 55mm focal length (sort of between 24-70) and a 10' subject distance. That should give a DOF of 3' with a 1.6X cropper or 5' on a full frame body. There's no way, with those figures, that even a humungous error in setup could account for the problem. If you had the newspaper turned edge on (and focused on the centre) you should everything in focus from the front edge right through to the back.

But with a 1' subject distance (assuming the lens can focus that close) you get a DOF of just 0.3" on a 1.6X crop body. Perhaps the OP could supply some figures for the DOF calculator from his test setup.
 
Hey Merlin,

In short the issue here is user interface error. ;)

If you post up a little more info on how you took these shots then folks will no doubt be able to tell which bit of weirdy camera physics you have invoked here but regardless, the AF points cannot affect a the sharpness of the lens. :)
 
StewartR said:
RobertP and CanonBob, you've both got it horribly wrong I'm afraid.

Cameras focus on a plane, not on a specific distance. If you focus on a point that's 3m away, say, it does notfocus on all points which are 3m away....
Click to expand...

Sorry for the slow response....been travelling.

I'm trawling back many years here and will throw this into the pot for discussion.

I believe that your statement would be true for an uncorrected lens....ie, optically true. Does this still hold true for a lens that has been corrected for distortion...ie, to eliminate pincushion or barrel distortion?
These corrections are neccessary to project an image on a flat sensor or film but they change the properties of the lens.

I may be talking dangly bits here as it's 35 years since I studied the physics of these things.

Opinions from more learned folks very welcome and apologies if I've got it wrong (as usual :crying:). Senility is a wonderful excuse when you haven't got it.

Bob
 
Sorry or not replying sooner.

I will replicate the test at the weekend doing my best to ensure everything is lined up correctly, think it was originally but will take extra care. I appreciate Stewart's comments on the setup and agree there may well be an element of user error but I thought F8 would be enough to ensure focal plane issues were ruled out, I'll go smaller next time.

Just to make it clear - I've been more than happy with the lens since I got it and never had reason to question it's performance before this test. For the price it offers bloody good value for money over the Canon equivalent.
 
You must log in or register to reply here.
Back
Top