[stoppingdown]

So, let's suppose I did this:

1. put the camera on a tripod and make sure everything is stable
   
2. autofocused on the center (a far subject, let's say 500m) and took a shot
   
3. switched to manual focus (and took another shot for a consistency check)
   
4. rotated the camera so the subject is placed at left and right border, and took two more shots
   

The border shots are in general less sharp than the center, as usual. Let's forget defects for now and suppose I'm fine with the results.

 

Now let's suppose I re-took the border shots, this time by reactivating autofocus, and autofocused on them by moving the autofocus cursor. In these shots, the focused border is sharper, but the center is worse than before.

 

Is this a normal behaviour, right? Is it due to field curvature? I have other questions, but first I'd like to understand whether I'm totally off track or not.

[JJ_SO]

Please tell us the axis you rotated the camera around:

A: Was it a turn round the tripod's vertical center column while the camera stayed in landscape orientation?

B: Was it a turn from landscape to portrait?


Answers for

A: no, that's not field curvature. It's simple triangulation, if you draw a line from camera to center of object (given it's a flat object and parallel to the sensor's surface) and another line from camera to the border of the subject, the latter will be longer and therefore the distance must be corrected.

B: yes, that's field curvature. Ideally, the field of sharpness would be only 2-dimensional and a very even wall, if parallel to the sensor, would be very sharp in center and border. But the normal field of sharpness is spherical because a lens without aspherical elements would only show a point or a ring of sharpness. The correction of this spherical area of sharpness into a more or less even area will be in reality a surface similar to the waves a single drop of liquid produces when falling into a volume of liquid.


Maybe that answer helps but if not, pleas clarify the term "rotate the camera" first.

[stoppingdown]

You're right, "rotate" is ambiguous. It's option A.

 

But I disagree with your answer. Neither the subject nor the camera moved, thus the distance is exactly the same.

 

*** edited to add: I understand there's another ambiguity in my post. The subject is relatively small, a bell tower at 500m, so there's not a "center" or a "border" of the subject. The "center" and "border" are those of the framing.

 

PS Apart from the fact that with a subject at 500m I believe there's plenty of DOF to play with: in the worst case, 70mm, at ƒ/8 focused at 500m, the near DOF limit is 29m... but as I said, this shouldn't be ever needed in this case.

[Guest]

You're right, "rotate" is ambiguous. It's option A.

 

But I disagree with your answer. Neither the subject nor the camera moved, thus the distance is exactly the same.

 

*** edited to add: I understand there's another ambiguity in my post. The subject is relatively small, a bell tower at 500m, so there's not a "center" or a "border" of the subject. The "center" and "border" are those of the framing.

 

PS Apart from the fact that with a subject at 500m I believe there's plenty of DOF to play with: in the worst case, 70mm, at ƒ/8 focused at 500m, the near DOF limit is 29m... but as I said, this shouldn't be ever needed in this case.

 

The distance is the same on a spherical curve if you rotate the camera in a circle. However, the lens can have some field curvature, but most likely, it's usually not spherical. Ideally, the focus plane of the lens would be flat.  So, now the question is, how much did you turn the camera. That depends of course on the focal length you used, because that gives you the field of view. A wide angle might give you a 90 degree horizontal angle of view, so you can turn your camera about 45 degrees to get your bell tower from the center to the left or right border of the image.   So, the distance is 500meters, let's assume the lens has a planar plane of focus, no field curvature. Simple trigonometry now  That means, at the border of the lens, which is set to 500 meters, the optimal sharp focus plane is at a distance of square root ( 500 x 500 + 500 x 500 ) = 707 meters.

If you had a tele, say a 500mm lens, then there would be very little difference, because you can't rotate it much to get to edge.

 

Given you don't know the field curvature of the lens, you probably need to make some tests. I.e. using life view set the correct distance at the edge, note the distance setting (say 600meter). Then using life view, set the correct distance for the center (say 500meter).  To have a compromise now, set the distance to 550meter, and choose an aperture that gives enough DOF to have 500 to 600 meters sharp.   Ok, this was just an example, it's pretty impossible to set such distances on a lens, but it illustrates the concept.  

[stoppingdown]

I think I've more or less got it, the idea of tweaking with the manual focus was the thing I was going to do - I needed to understand whether it was meaningful.

 

Thanks to all.

[stoppingdown]

A couple of more questions:

1. To the staff of PZ: when you run the test for MTF, do you only focus once at the center, or re-focus at each point where you measure the sharpness? 
   
2. Concerning the compromise cited by photonius in the previous reply, do you know if any camera body is able to apply this compensation automatically when it autofocuses?
   

Thanks.

[Brightcolours]

PZ uses the optimal focus for each measured area. So, the PZ results do not show field curvature. When a lens does show strong field curvature, usually it is mentioned in the review text, though.

 

No camera can compensate for field curvature in a focus and recompose situation. They focus, then the user recomposes, the camera does nothing at that point. 

[stoppingdown]

For the former point, ok, understood. As far as you know, is it the same procedure that other MFT reviewers follow?

 

For the latter point, I'm not thinking of compensating focus & recompose. I'm thinking of compensating the field curvature and improving the border performance.

[Brightcolours]

For the former point, ok, understood. As far as you know, is it the same procedure that other MFT reviewers follow?

 

For the latter point, I'm not thinking of compensating focus & recompose. I'm thinking of compensating the field curvature and improving the border performance.

Not many reviewers follow the PZ procedure, no. Most just shoot one image and then go through the MTF measuring process.

 

I do not get how you think one would go about "compensating" for field curvature? The camera can't change what is projected on its sensor... Nor can it bend its sensor...  :wacko:

[stoppingdown]

Perhaps between the optimal focus at the center and the optimal focus at the corner there's a point in which the center is just less sharp and the border more acceptably sharp.