So, I finally managed to upload the test files to our server. Good job, Dallas.
Now, to the field experiment. I'm shooting a horse pen surrounded by an electric fence. My focus is on the birch trunk in the middle. Precise focus was achieved using LiveView on the D3, at maximum enlargement of the LCD. I double-checked by taking a picture and examining it again. After the critical focus was achieved, the focus collars were taped down, and focus rechecked once more to ensure nothing had moved.
I shot the 300/2.8 AFS on the D3, at f/2.8, ISO 200, and the 200/2VR on D2X, also ISO 200, both cameras using mirror lock-up and a cable release. NEFs processed i Capture NX. No sharpening applied. After the initial evaluation of the images, I returned to the site and using a surveyor's rod, measured the relative psoition of key elements I had found on the shots. Although I aimed to shoot the fence at 45 degrees, closer examination showed it to run in a slight zig-zag manner so I felt it necessary to do actual field measurements.
Shooting distance was 54 m (zero point: the tree trunk).
First, the overviews
D3:
D2x:
Note: the control points are marked with relative distance to the zero point, + indicating a greater (rear) distance, - the corresponding frontal position.
One will appreciate that there is servere foreshortening of the apparent perspective at this distance.
I did not aim to achieve perfect colour match between these images, so they are slighlt different despite being shot with only a minute apart.
Now, to the details. The positions refer to the poles of the fence unless otherwise noted.
D2X/200 to the left and D3/300 to the right on all A/B comparisons.This is the comparison at the +7.5 m point, 100% crop no sharpening.
It can be seen that both systems are quite similar, with the D2X being ever so slightly softer. The background blur is different too, one should expect the D2X /200 to provide the smaller blur, but the opposite is seen. Note that DOF concerns what's in focus, not what's out of focus. But the different blur sizes do indicate which lens will increase DOF the most by stopping further down.
After having poured over the 100% crops for a while, I'd say this point is close to the rear DOF extension. Possibly one could add 0.5 m to it, but no much more.
Moving closer to the reference point, here is the +3.6m comparison. Again 100% crop and no sharpening.
Obviously, both systems are well within the zone of high sharpness here.
Now, here is the -1.2m comparison. It includes the birch tree that was the zero point (in the background) and it is apparent the focus there is positioned correctly. Both DX and FX shows a severe loss of sharpness and the pole at -1.2 no longer is in critical sharpness. Details of grass at the ground level confirms that we now have reached the front of the DOF.
The final point is -4.2m in front of the birch tree. Here, we very evidently are way outside the DOF region.
Running across in a slanted direction in the background is the transition zone between sharp and unsharp. The transition is much better defined here in the front than we saw for the rear section. This can be seen to be ever so slightly more forward in the FX shot.
**************** so much for the actual field data **************
The test shots show that DOF is very, but not quite, similar between DX/200 @ f/2 and FX:300 at f/2.8. The difference there is may be more of an academic value. So if we adjust the hypothesis of "one stop advantage" to read "more than 0, but less than 1 stop advantage", that should cover the observed facts. The demarkation lines between sharp/unsharp are not prominent enough when the distance is this great to allow us to assign an accurate DOF estimate. So the best guess would be around 9 m in the setup I used. Scrutiny of the originals show also that there is curvature in the DOF intersection with the ground, thus one can see that the details of the grass becomes blurred at the same distance from the camera that the poles are well defined (the poles were better aligned to the optical axis).
We can, however, say that DOF is very asymmetrical in its distribution, in this case, at most 1.5 m in front and at least 7.5 m to the rear.
So, how does this compare to the DOF models? This is where the CoC criteria enter the fray. I've run several DOF calculators and they all agree of a value around 5.5 m, a little more for the DX format (contrary to observation). We observe nearly twice that value. What is more striking is that the models all insist on DOF being quite distributed quite symmetric around the target, which very obviously is not the conclusion one draws on the basis of these test shots. I've tried to use the full set of parameters including pupillary magnification, but the model estimates are pretty much the same. So at least this discrepancy illustrates that the modelled DOF (based upon geometry) need not have a direct counterpart in the perceived DOF.
