Julian Loke wrote:
> There is a worrying misunderstanding that DEP and A-DEP mode are calculated
> using Circle of Confusion and Depth-of-field data.
> The modes are apparently implemented using defocus information at the sensor
> plane, and use a "guestimate" that the best DOF occurs at 7/17 between the
> near and the far points. Circle of Confusion does not enter directly in the
> picture, literally.
But two different things must be determined in DEP mode (and perhaps A-DEP
mode)
1. The point of focus. Even an analytical solution does not involve CoC:
U = 2 Uf Un / (Un + Uf)
Where
U = Focused [object] distance
Un = Near limit of DoF
Uf = Far limit of DoF
The EOS DEP algorithm hopefully effects a reasonable approximation
of the above; if the 7/17 is correct, the EOS algorithm agrees with
the analytical solution only when Un = 0.7 Uf.
2. The f-number. This necessarily involves the CoC or something
equivalent. From the FAQ:
"At the same time the focusing position is established, the EOS
camera's CPU calculates the aperture value required to bring
DEP 1 and DEP 2 within a 0.035mm circle of confusion. This
calculation is also extremely accurate because it is based on
the total defocus amount at the film plane as previously
described."
I don't pretend to know exactly how it works ...
Incidentally, I suspect that on most manual-focus lenses, the algorithm
actually is implemented in image space:
V = 2 Vf Vn / (Vn + Vf)
At small lens extensions, V ~ Vn ~ Vf, so that
V ~ 2 VV^2 / (Vn + Vf)
or
V ~ (Vn + Vf) / 2
For the f-number:
N = (f/c)(Vn - Vf)/(Vn + Vf)
but V ~ Vn ~ Vf ~ f, so that
N ~ (f/V)(Vn - Vf)/(2c) ~ (Vn - Vf)/(2c)
Where
c = Circle of confusion
V = Focused [image] distance
Vn = Image distance at which object at near limit of DoF is sharply focused
Vf = Image distance at which object at far limit of DoF is sharply focused
Both final algorithms are easy to implement with a mechanical DoF scale;
they shouldn't be too surprising given that most lens DoF scales are
symmetrical about the focus mark. Some of the usual caveats apply: the
lens is assumed to be unit focusing, and accuracy decreases as lens
extension increases.
So even with the time-tested lens DoF scales, some compromise often is
involved. Yet another reason to close down a step ...
Jeff Conrad
*
****
*******
***********************************************************
* For list instructions, including unsubscribe, see:
* http://www.a1.nl/phomepag/markerink/eos_list.htm
***********************************************************
