hyperfocal distance

Big Bill writes:

wrote:
I'd say the DSLR focal plane. When you are figuring the acceptable
circle of confusion, you must tailor it to to the resolution capability
of the focal plane device being used. We have made up general rules of
thumb for the average film resolution, but that doesn't apply even for
film if it isn't normal res film, let alone a completely different
imaging device.

But Dave wrote:
The one that's actually there, because the actual sensor size determines
the magnification needed to produce a same-size print, and CoC depends
on printing magnification.

Interesting...
Two exactly opposite answers.
I guess it really is relative. :-)

No, they're not opposite at all. Don says you should use the DSLR focal
plane [size], not the 35mm film frame size. I said you should use the
actual sensor size, not the 35 mm film frame size. So we both agree on
that.

Don *also* says that you may have to tailor it to the device being used.
I agree; I just didn't go into that level of detail. Basically, if your
sensor has *lots* of resolution, the CoC size should be set entirely by
some threshold of perception of blur in the final print under some
assumed viewing conditions. In other words, CoC size is set by human
visual acuity and not sensor resolution - as long as resolution is good
enough. But when resolution is poor (e.g. 1 megapixel camera), even the
"in best focus" subject material is visibly somewhat unsharp to the
viewer, and you can have things further out of focus before they are
visibly "worse".

What this means is that with low-resolution sensors, there's little
point in using a CoC smaller than about twice the pixel pitch. So,
here's a revised rule which I suspect Don would agree with: For CoC,
use twice the pixel pitch, or the sensor diagonal divided by about 1740,
*whichever is larger*.

Finally, note that this is still just a rule of thumb; feel free to
modify it if necessary. I knew a cinematographer who calculated his own
DOF tables, using a smaller CoC than the usual one for the format,
because he simply had higher standards for what is "in focus".

Dave

Big Bill wrote in
:

It seems that you're saying that "it's all relative".
In which case, the formulae presented here are nice, but don't mean
much except to 'pixel peepers'.

I kinda agree; it's what you see that counts.


The formulas is a try to formalise "what you see".
They are based upon how we look at and appreciate
pictures - what kind of unsharpness we can accept
in out of focus areas. The formulas are actully
the oposite to pixel peepers. They were invented long
before digital photography saw the light (sic).

This is not precise science though, and it can certainly
vary with both time, person, fashion and technical
circumstances. But I assume that at normal viewing
distances and rather high demands, the formulas
are better than nothing. They are practical guidelines that
makes it possible to focus your camera without first
looking at the result. When you set a program focus to
"mountains", the formula will be used, totally without
using any meassuring auto focus mechanism.


/Roland

(Dave Martindale) wrote in news:cc48hi$muh$1
@mughi.cs.ubc.ca:

Typical Tony. In a situation with two legitimate interpretations, he
picks a third irrelevant one, comes to the wrong conclusion, and then
defends it agressively.


In this case, it is a perfect compromise though

(const/x) * x = const.


/Roland

Big Bill wrote in :

I'd say the DSLR focal plane. When you are figuring the acceptable
circle of confusion, you must tailor it to to the resolution capability
of the focal plane device being used. We have made up general rules of
thumb for the average film resolution, but that doesn't apply even for
film if it isn't normal res film, let alone a completely different
imaging device.

But Dave wrote:

The one that's actually there, because the actual sensor size determines
the magnification needed to produce a same-size print, and CoC depends
on printing magnification.

Dave

Interesting...
Two exactly opposite answers.

I guess it really is relative. :-)


Nope. The second answer is the correct one. The first is wrong.
At least according to the definition of hyperfocal distance.
The first answer describes something else.


/Roland

On Fri, 2 Jul 2004 18:22:45 +0000 (UTC), (Dave
Martindale) wrote:

Big Bill writes:

wrote:
I'd say the DSLR focal plane. When you are figuring the acceptable
circle of confusion, you must tailor it to to the resolution capability
of the focal plane device being used. We have made up general rules of
thumb for the average film resolution, but that doesn't apply even for
film if it isn't normal res film, let alone a completely different
imaging device.

But Dave wrote:
The one that's actually there, because the actual sensor size determines
the magnification needed to produce a same-size print, and CoC depends
on printing magnification.

Interesting...
Two exactly opposite answers.
I guess it really is relative. :-)

No, they're not opposite at all. Don says you should use the DSLR focal
plane [size], not the 35mm film frame size. I said you should use the
actual sensor size, not the 35 mm film frame size. So we both agree on
that.

The focal plane is the plane the image formed by the lens falls on;
whether it's a film SLR or a DSLR, the image is the same size.
So, the DSLR has the same image on it a film SLR has.
That's what I saw Don say.

Don *also* says that you may have to tailor it to the device being used.
I agree; I just didn't go into that level of detail. Basically, if your
sensor has *lots* of resolution, the CoC size should be set entirely by
some threshold of perception of blur in the final print under some
assumed viewing conditions. In other words, CoC size is set by human
visual acuity and not sensor resolution - as long as resolution is good
enough. But when resolution is poor (e.g. 1 megapixel camera), even the
"in best focus" subject material is visibly somewhat unsharp to the
viewer, and you can have things further out of focus before they are
visibly "worse".

What this means is that with low-resolution sensors, there's little
point in using a CoC smaller than about twice the pixel pitch. So,
here's a revised rule which I suspect Don would agree with: For CoC,
use twice the pixel pitch, or the sensor diagonal divided by about 1740,
*whichever is larger*.

Finally, note that this is still just a rule of thumb; feel free to
modify it if necessary. I knew a cinematographer who calculated his own
DOF tables, using a smaller CoC than the usual one for the format,
because he simply had higher standards for what is "in focus".

Dave


Bill Funk
Change "g" to "a"

Roland Karlsson wrote:
lid wrote in news:10eabjg8nir0247
@news.supernews.com:

I don't think that the CoC size it is all that subjective on a digital
camera.

We know exactly what the spatial frequency response of the sensor is
-- in most digital cameras you lose contrast dramatically at about 70%
of the Nyquist frequency. So, a CoC smaller than that is pointless,
but a CoC much larger will reduce contrast. A CoC of about 2 x the
sensor pitch seems about right.

Happily, that's close to the 1/1730 of the "Zeiss formula", at least
on a 6 Mpixel camera.

This is no coincidence. A 4 Mpixel camera (or more) is about the
minimum to get perceptive sharp large pictures. So - as long as
we have 4-8 Mpixels, the CoC in the hyper focal distance is
approx one or two pixels.

But his has really nothing to do with digital cameras in general.
If we have a 250 Mpixel camera, the CoC will become 1/1730 again
and not 2 pixels.

Well, I was arguing against that conclusion, and I explained why. I'm
assuming that very high resolution sensors implies large prints, and
prints that people will want to get close to -- you don't always want
to look at the whole image. And if you have a 250 Mpixel camera, a
CoC of 1/1730 is a nasty blur.

Andrew.

lid wrote in news:10edvtr6kf0in57
@news.supernews.com:

Well, I was arguing against that conclusion, and I explained why. I'm
assuming that very high resolution sensors implies large prints, and
prints that people will want to get close to -- you don't always want
to look at the whole image. And if you have a 250 Mpixel camera, a
CoC of 1/1730 is a nasty blur.


Yepp - you have a point here. The CoC based formula does only
work when you examine a picture at "normal" viewing distance.
For extra ordinary situations, like very small pictures or
very large very sharp pictures, it is not valid.

Now - most pictures are viewed at "normal" viewing distances
and therefore the CoC based meassure is useful. But - sometimes
(like all us pixel nerd here we want mor, and then the
hyper focal and DOF scales and calculations are not valid any more.


/Roland

Roland Karlsson writes:

Yepp - you have a point here. The CoC based formula does only
work when you examine a picture at "normal" viewing distance.
For extra ordinary situations, like very small pictures or
very large very sharp pictures, it is not valid.

You mean the standard CoC formula based on 1/1500 or 1/1700 of
the image diagonal. However, you can also calculate an acceptable CoC
from first principle, picking an acceptable angular size at the eye of
the viewer and taking into account the print size and viewing distance.
This works for any print size and any viewing distance.

Now - most pictures are viewed at "normal" viewing distances
and therefore the CoC based meassure is useful. But - sometimes
(like all us pixel nerd here we want mor, and then the
hyper focal and DOF scales and calculations are not valid any more.

The preprinted tables are not valid. But calculated DOF tables that
allow you to change the CoC size cover all possible situations. You can
even build your own tables using a spreadsheet.

I knew a cinematographer that calculated his own DOF all the time,
because he was very fussy about sharpness and demanded a smaller CoC
than the standard tables were based on. I think he did this via a small
program on the HP-41 calculator he always carried, but these days one
could easily calculate and present whole tables on a Palm or PocketPC.

Dave

DOF questions/topics come up quite frequently and naturally lead
to hyperfocal distance and then to CoC. We all know that the CoC is not
hard number which depends only many factors including "enlargement".
For digital cameras and using "digital" printers, we may be able to
define CoC more scientifically. The following is a post I made almost
exactly one year ago. It is slightly edited and repeated below. I
hope this will clear some of the CoC issues.

For digital cameras in general we may be able to "determine" the CoC more
"scientifically and systematically". Before we start, let's assume the
final printing quality is defined in the following manner. (You may
define them differently)

Best 250 - 300 ppi
better 200 - 250 ppi
Good 150 - 200 ppi
Acceptable 100 - 150 ppi
Unacceptable 100 ppi

Clearly CoC must be in the unacceptable range. We may choose
CoC to be the equivalent of 50 ppi (you may use other numbers).
Then if your final prints are 300 ppi, then the CoC is 6 pixels.
The following is the list of CoC size for different final print quality.

Print quality Size of CoC

Best (300 ppi) 6 pixels
Better (250 ppi) 5 pixels
Good (200 ppi) 4 pixels
Acceptable (150 ppi) 3 pixels

These definitions are universal and indepandent of cameras and/or
total pixel number. When we use the final print quality as a criteria,
we don't have to take into account of the 1.5 (or 1.6) factor neither.
This also takes care of the "enlargement" factor.

If your final print quality is better than the best, the
CoC is proportionally increased. For instance, if an image
has 12 MP (4000 x 3000) (Canon EOS-1Ds or Kodak Pro 14n/c) and is printed
on a 10" x 8" paper, the resolution is 400 ppi. Then the
CoC is 8 pixels. I hope this will clarify some of the CoC issues.



In article ,
Roland Karlsson wrote:
wrote in news:10edvtr6kf0in57
:

Well, I was arguing against that conclusion, and I explained why. I'm
assuming that very high resolution sensors implies large prints, and
prints that people will want to get close to -- you don't always want
to look at the whole image. And if you have a 250 Mpixel camera, a
CoC of 1/1730 is a nasty blur.


Yepp - you have a point here. The CoC based formula does only
work when you examine a picture at "normal" viewing distance.
For extra ordinary situations, like very small pictures or
very large very sharp pictures, it is not valid.

Now - most pictures are viewed at "normal" viewing distances
and therefore the CoC based meassure is useful. But - sometimes
(like all us pixel nerd here we want mor, and then the
hyper focal and DOF scales and calculations are not valid any more.


/Roland


--
Charles S. Ih
302-831-8173, FAX 302-831-4316

(Charlie Ih) wrote in
:

DOF questions/topics come up quite frequently and naturally lead
to hyperfocal distance and then to CoC. We all know that the CoC is
not hard number which depends only many factors including
"enlargement". For digital cameras and using "digital" printers, we
may be able to define CoC more scientifically. The following is a post
I made almost exactly one year ago. It is slightly edited and repeated
below. I hope this will clear some of the CoC issues.

...

Hi ... nice analysis!

Yes - you can choose CoC according to a fixed resolution, e.g. 50 ppi.
Then you can use that to calculate DOF and hyperfocal distance if you want.

But - this is not how it is done. CoC is not based upon acceptable
sharpness in general - it is based upon percepted sharpness for out
of focus details. Out of focus details may be unsharper than what
is unacceptable for in focus detail; and the basis for CoC is that
the CoC is really independent of how sharp the sharper parts are;
it is approx. proportional to the picture size.

Of course, this assumption may be wrong. If the tendency is towards
super sharp pictures in general, the the CoC migh be what you propose.
If the tendency is to use DOF artistically, then it is rather the old
CoC method that is valid IMHO.


/Roland