Circle of Confusion - applied values

Im aiming to get pinsharp landscapes with max depth of field so use
Hyperfocal Length, using Circle of Confusion values applied as
follows:

A4 prints - 0.0189mm
A3 prints - 0.0137mm
Larger prints - 0.0094mm

but Im wondering...

My EOS40D has a sensor size of 22.2 x 14.8mm and max print size of
3888 x 2592 pixels. Applying simple logic and maths, this would
suggest an individual 'cell' on the sensor is 0.0057mm.... which in
turn suggests that even shooting for the larger prints above, a single
point of light falling on the sensor is spanning 2 cells.

Is my logic sound, or am I missing some variables ...?

Interestingly, this also begs questions
- what's the circle of confusion tolerance applied by the EOS
Autofocus system?
- if I set hyperfocal distance by autofocusing on an object at a known
distance, is the Circle of confusion tolerance on the AF system going
to introduce significant error (distance to object)?

Thanks

"M" <mspencer_home@hotmail.com> schreef in bericht
news:bd822e30-9cae-42a8-8907-5a70076abf2f@t54g2000hsg.googlegroups.com...
> Im aiming to get pinsharp landscapes with max depth of field so use
> Hyperfocal Length, using Circle of Confusion values applied as
> follows:
>
> A4 prints - 0.0189mm
> A3 prints - 0.0137mm
> Larger prints - 0.0094mm
>
> but Im wondering...
>
> My EOS40D has a sensor size of 22.2 x 14.8mm and max print size of
> 3888 x 2592 pixels. Applying simple logic and maths, this would
> suggest an individual 'cell' on the sensor is 0.0057mm.... which in
> turn suggests that even shooting for the larger prints above, a single
> point of light falling on the sensor is spanning 2 cells.

A single point of light if it is not infinite small can fall on four cells
if it just hits the corner of the four cells. This is why the resolving is
less than the pixel size. Roughly the resolving 'power' of the CCD is number
of available pixels/(sqrt(2)) This is in horizontal and in vertical
direction. So to get resolved the line or point should allready be the area
of 2 pixels.
(A line to get resolved should be sqrt(2) pixels wide, a point to get
resolved should be 2 pixels in area).

Or two lines should be seperated by a line of sqrt(2) wide.

Adding to this a circle of confusion does reduce the resolving power even
more. Say that the circle of confusion again is the area of 2 pixels, maybe
this does reduce the resolving power to half of what is was. (Or does is
only reduce the resolving power with 50 % or 25 %, I am not sure).

So the circle of confusion does reduce the resolving in addition to the
limited resolving of the CCD.


http://en.wikipedia.org/wiki/Zeiss_formula
Dof calculations should be base on c = d/1730
(This is a fairly critical number, often for 35 mm format 1440 was used as
the/a standard).

For 22.2mm x 14.8 mm this works out to 0.015423 mm.
(this works out to a dof of 2.702 pixels length's).
This formula is 'size' independend.
So there is still almost 3 pixels, which is above more than the resolving
limit.
Having a dof or 2 pixels, comes close to the resolving limit. And although
the effect is an added effect, I guess that a dof of less than 2 pixels is
not functional any more.

If pictures get larger, probably more dof is needed because we also take a
close look at the picture, but I do not agree that a picture twice the size
should have a dof halved on the CCD. I think that the dof should be a
fraction of the diagonal, and maybe the fraction get's a bit smaller for
larger sizes but not liniear. Say for A4 1/1730 can be used for A3 we make
it a bit more critical say 1/2000, for A2 1/2400 (but that's about the
limit).
If more dof is needed, you are to close to the picture, or you have
extremely good eyes.
(The 1/2400 is still 2 pixels in size so you are also reaching the limit of
the CCD).

But it all adds up offcourse:
The resolving power of the lens.
The resolving power of the CCD. (About 1/SQRT(2) of the number of pixels).
Errors in focusing. (Shifts where sharp is, does not realy alter DOF).
Distance from focus which causes a COC.



>
> Is my logic sound, or am I missing some variables ...?
>
> Interestingly, this also begs questions
> - what's the circle of confusion tolerance applied by the EOS
> Autofocus system?
The focus error does depend on the lens and the focusing system, a focus
error will shift the DOF towards or from the camera. But the depth will
remain almost the same.

> - if I set hyperfocal distance by autofocusing on an object at a known
> distance, is the Circle of confusion tolerance on the AF system going
> to introduce significant error (distance to object)?

The AF does not have a Circle of confusion tolerance. But a focusing
tolerance.
Some people speak of back or front focus and say this does cause a
significance in shifting the Depht of Field. Some say this is cause by some
camera types, some say this is only in some camera's of the type. On the
internet you will be able to find some examples of the phenomem.
Good Example:
http://www.focustestchart.com/chart.html

Does this 'resolve' your question ?
ben brugman

>
> Thanks

M wrote:
> Im aiming to get pinsharp landscapes with max depth of field so use
> Hyperfocal Length, using Circle of Confusion values applied as
> follows:

You might want to check diffraction issues too.

BugBear

On Jun 17, 8:42 am, M <mspencer_h...@hotmail.com> wrote:
> Im aiming to get pinsharp landscapes with max depth of field so use
> Hyperfocal Length, using Circle of Confusion values applied as
> follows:
>
> A4 prints - 0.0189mm
> A3 prints - 0.0137mm
> Larger prints - 0.0094mm
>
> but Im wondering...
>
> My EOS40D has a sensor size of 22.2 x 14.8mm and max print size of
> 3888 x 2592 pixels. Applying simple logic and maths, this would
> suggest an individual 'cell' on the sensor is 0.0057mm.... which in
> turn suggests that even shooting for the larger prints above, a single
> point of light falling on the sensor is spanning 2 cells.
>
> Is my logic sound, or am I missing some variables ...?
>
> Interestingly, this also begs questions
> - what's the circle of confusion tolerance applied by the EOS
> Autofocus system?
> - if I set hyperfocal distance by autofocusing on an object at a known
> distance, is the Circle of confusion tolerance on the AF system going
> to introduce significant error (distance to object)?
>
> Thanks

A couple of schools of thought. One is just what you have done- set
blur circle approximately equal to pixel size. Another is one third
pixel size. The later results in a much smaller circle of confusion,
so that optics is NOT the limiting factor. I personally believe the
right value is somewhere in between, say about 5/8 pixel size so that
circle area is about 1/3 pixel area.