How to think about the effect of film flatness?
Leonard Evens
noticed any major degradation of images due to lack of film flatness. At
one point when I was being somewhat fanatic about such things, I did get
a special back for my Rolleiflex TLR which used a glass plate to insure
flatness, but I've seldom used it.
Still since the issue has been raised in various posts, I've tried to
think about it. I started with the standard formula(s) for depth of
focus, but I've since decided this is the wrong way to approach it. The
problem is that the same simple geometric considerations---i.e., similar
triangles---used to derive formulas for depth of focus are also used to
derive formulas for depth of field. so one can't use these formulas as
though they were independent of one another.
I've come up with another way of thinking about it, and I wonder if
others think it makes any sense.
If the film is not flat, then some parts of it are not in the expected
film plane. That is equivalent to focusing the lens at a different
distance for that part of the film. That could either bring that part
of the image into better focus or it could put it further out of focus.
This obviously will have some effect on depth of field, as it appears in
different parts of the image. But getting a quantitative idea of its
effect requires posing the problem correctly. I have some initial ideas
about that, but I'm not sure how useful they are. Anyway, here they
are.
Assume one is doing normal photography with typical lenses and one
focuses on the hyperfocal distance. Then some back of the envelope
calculations seem to show that the distance of the image plane from the
focal plane is about C*N where C is the diameter of the circle of
confusion and N is the f-number. (Let's ignore diffraction for the
purposes of these calculations.) So it seems to me the natural unit for
thinking about film flatness is C. If the deviation from flatness is
just a small multiple of C, it would seem the effect on depth of field
would be minimal. But clearly if it were high, e.g., N/3 times C, it
would be pretty dramatic.
Does this make any sense? If so, what does it say about film flatness
as a problem for different formats? For example, one might conclude
that in some sense, the larger the format the less the potential problem.
The argument would be that the larger the format, the larger the
acceptable C, so the larger the acceptable absolute measure of
displacement from flatness. Also, with larger formats, one tends to use
higher f-number, both to achieve adequate depth of field and because
such f-numbers don't create diffraction problems as they would with
smaller formats (and smaller Cs).
I stand ready to be corrected for any faults in my reasoning. Also,
what are typical numbers for departures from flatness for various
formats. Clearly, the smaller the format, the easier it should be to
control flatness, at least up to a point. But how much easier is it?
--
Leonard Evens l...@math.northwestern.edu 847-491-5537
Dept. of Mathematics, Northwestern Univ., Evanston, IL 60208
Andrew Koenig
Leonard> expected film plane. That is equivalent to focusing the lens
Leonard> at a different distance for that part of the film. That
Leonard> could either bring that part of the image into better focus
Leonard> or it could put it further out of focus. This obviously will
Leonard> have some effect on depth of field, as it appears in
Leonard> different parts of the image. But getting a quantitative
Leonard> idea of its effect requires posing the problem correctly.
I wonder if the following way of looking at it will help.
Imagine that the film is in the right place, and look at one point
on the sharply focused image. Light from all parts of the lens must
be reaching that same point; otherwise it wouldn't be a point.
Now displace the film slightly. The light that had been reaching a
single point will now fill a circle, called the circle of confusion,
the perimeter of which can be determined by drawing straight lines
from the original point to all the points on the edge of the lens
diaphragm as seen from the film.
In other words, suppose the correct lens-to-film distance is F,
the exit pupil--that is, the diameter of the lens as seen from the
film--is E, and you displace the film by a distance D. Then the
diameter of the resulting circle of confusion will be D*E/F.
Note that (D*E)/F is equivalent to D*(E/F), so the parentheses don't
matter. Note further that E/F is nothing more than the reciprocal of
the aperture, at least for symmetric lenses (I'm not quite sure how to
account for asymmetric lenses, such as retrofocus lenses, but I would
not be surprised to find that it doesn't matter).
So we have a simple rule: Divide the deviation from film flatness by
the aperture and you have the circle of confusion. For example, a
displacement of 0.1mm at f/4 will yield a circle of confusion of 0.025mm.
--
Andrew Koenig, a...@research.att.com, http://www.research.att.com/info/ark
Struan Gray
>Leonard> If the film is not flat, then some parts of it are not in the
>Leonard> expected film plane. That is equivalent to focusing the lens
>Leonard> at a different distance for that part of the film. That
>Leonard> could either bring that part of the image into better focus
>Leonard> or it could put it further out of focus. This obviously will
>Leonard> have some effect on depth of field, as it appears in
>Leonard> different parts of the image. But getting a quantitative
>Leonard> idea of its effect requires posing the problem correctly.
>
> I wonder if the following way of looking at it will help....
Readers might want to look at this diagram:
http://www.sljus.lu.se/People/Struan/pics/dofconfuse.jpg
The top diagram illustrates the conventional way of calculating
DOF, as described in Andrew's post. The extension to a buckled film
should be obvious.
The lower diagram might amuse DoF afficionados. It occurred to me
that the conventional derivation, while great for considering buckled
film, was actually misleading for the usual case of a photographer who
has focussed on one thing and wants to know what else will be 'in
focus'. If you look at the light from the far DoF limit plane (green)
in the top diagram, you see that on-film it is actually spread out by
*more* than the circle of confusion. Similarly, the light from the
near DoF limit is spread out on the film by *less* than the circle of
confusion.
The lower diagram adjusts the position of the near and far DoF
planes so that on-axis points create a spread on the film equal to the
circle of confusion. If you churn through the trig, you find that the
new near DoF is almost exactly the same as the old one, but that the
far DoF doesn't recede to infinity quite as fast as the old one did
when the lens is stopped down. I have convinced myself that some of
my hyperfocal-distance landscapes with surprisingly blurred distant
mountains are due to this mismatch.
Mind you, it's completely cured by stopping down one more stop for
safety.
Struan
Bob Gurfinkel
ligtht up a pipe, close your eyes.
Bob G.
ArtKramr
You shed light on the obvious.
Arthur Kramer
Visit my WW II B-26 website at:
http://www.coastcomp.com/artkramer
ArtKramr
>ligtht up a pipe, close your eyes.
>
>Bob G.
Well said,
And when complete your nap saunter leisurely to a camera dealer and buy a
camera where film flatness is not a problem. But don't buy film. If you shoot
pictures that will interfere with your naps.
McLeod
and started making some images.
"Leonard Evens" <l...@math.northwestern.edu> wrote in message
news:20011204.085012...@math.northwestern.edu...
Robert Monaghan
previously noted the same problem with getting infinity sharp and my own
workarounds too ;-) better to give up a bit of foreground when the subject
is mostly at infinity ;-)
I've been pretty interested in film flatness, and have documented many
related notes at http://people.smu.edu/rmonagha/mf/flat.html which answer
some of Leonard's and other questions on magnitude of the effect etc.
what may be more interesting in regard to the question of how much small
errors in focusing (or displacements of film plane by bulging film by
analogy) hugely impact the loss of resolution with thin emulsion films.
see http://people.smu.edu/rmonagha/mf/critical.html critical focusing
pages with tables by photographer/physicist Dr. Eugen J. Skudrzyk
as an example, a 2mm shift or error in focusing a normal 35mm SLR lens at
the circumferance of the lens focusing scale equates to circa a 100 micron
shift in the positioning of the plane of focus, with high resolution (thin
emulsion) film this error can reduce resolution from an observed 100-120
lpmm to circa 40 lpmm at f/2. Wow!
So obviously a 0.1mm or 100 micron height film bulge away from the plane
of focus would have a deleterious effect with such a high resolution film.
similarly, Norm Goldberg noted (Norman Goldberg, Shoptalk, Pop. Photogr.
May 1986, p.82) a 48% drop in contrast of a 35mm SLR lens due to a 0.08mm
film bulge. Again, this is stunning, that less than the equiv. of a 2mm
focusing error can also reduce potential contrast (by causing a 0.1mm film
bulge) by costing roughly half or more of the potential fast lens (eg f/2)
contrast (wide open) with high resolution films. Wow!
With thicker and lower resolution color film emulsions, as Skudrzyk's
tables show, the losses from focusing errors are masked by the thick film
emulsion, as are the effects of diffraction (the film is that BAD ;-). So
from these studies, I would say that the issue of film choice is a hidden
but critical element in how critical your film bulge/position/DoFocus is.
on the other hand, for thin emulsions (quote:
a very important conclusion: they show that focusing is very critical for
high-resolution, thin-emulsion films. Since the resolution of such a film
decreases greatly if the F-stop is increased above F/4, the loss in
resolution because of a focusing error can never be compensated by
stopping the lens down... [Ibid., p. 255].
end-quote
so the issue of film emulsion type and thickness is a major concern in
real-world photography when addressing issues like film flatness, focusing
errors, Depth of focus, and their impact on results. Again, losses of 66%
of potential resolution of wide open f/2 lenses and 48%+ of their contrast
with thinner high resolution films is a major surprise, given the small
2mm focusing error or film bulge magnitude (100 microns) needed to bring
such losses about...
the good news is we don't have many fast lenses in medium format, I guess
;-) Usually such errors are covered up by our slower lenses, use of
mid-range f/stops or "sweet spots", and hopefully more accurate focusing
mechanisms (e.g., 3X and 5X prisms and chimney viewfinders etc) in med
fmt. But I find it interesting that the major losses are from issues like
film choice, focusing errors, film positioning errors, and so on rather
than relatively modest but more talked about differences in pro lens
resolutions which seem to be limiting our (or at least my ;-) results...
grins bobm
--
* Robert Monaghan POB752182 Southern Methodist University, Dallas Tx 75275 *
* Third Party 35mm Lenses: http://people.smu.edu/rmonagha/third/index.html *
* Medium Format Cameras: http://people.smu.edu/rmonagha/mf/index.html *
Leonard Evens
<a...@research.att.com> wrote:
Perhaps I misunderstand what you mean, but you seem to be saying that E/F
is the reciprocal of the f-number. That is of course true for object
points at infinity. But for points at a finite distance, what you call
F is greater than the focal length and F/E is not the f-number. How
close it is to the f-number depends of course on how far away the object
point is.
However, the similar triangles you used to calculate still work fine at
any film distance. To relate this to the f-number, however, you have
to use the usual rule 1/obj_distance + 1/im_distance = 1/focal_length,
and do some algebra. If you assume the object is at the hyperfocal
distance, the usual approximation for the object distance is
focal_length^2/C*N
where C is the maximal acceptable diameter of the circle of confusion and
N is the f-number. Putting this in the previous formula yields the
approximation I derived for the distance between the focal plane and the
film plane when focused on at the hyperfocal distance.
But maybe we are talking at cross purposes.
Leonard Evens
<strua...@sljus.lu.se> wrote:
> Andrew Koenig, a...@research.att.com writes:
>
>>Leonard> If the film is not flat, then some parts of it are not in the
>>Leonard> expected film plane. That is equivalent to focusing the lens
>>Leonard> at a different distance for that part of the film. That
>>Leonard> could either bring that part of the image into better focus
>>Leonard> or it could put it further out of focus. This obviously will
>>Leonard> have some effect on depth of field, as it appears in Leonard>
>>different parts of the image. But getting a quantitative Leonard> idea
>>of its effect requires posing the problem correctly.
>>
>> I wonder if the following way of looking at it will help....
>
> Readers might want to look at this diagram:
>
> http://www.sljus.lu.se/People/Struan/pics/dofconfuse.jpg
>
> The top diagram illustrates the conventional way of calculating
> DOF, as described in Andrew's post. The extension to a buckled film
> should be obvious.
I'm afraid I don't recognize that diagram. The bottom diagram is the
one I'm familar with.
>
> The lower diagram might amuse DoF afficionados. It occurred to me
> that the conventional derivation, while great for considering buckled
> film, was actually misleading for the usual case of a photographer who
> has focussed on one thing and wants to know what else will be 'in
> focus'. If you look at the light from the far DoF limit plane (green)
> in the top diagram, you see that on-film it is actually spread out by
> *more* than the circle of confusion. Similarly, the light from the near
> DoF limit is spread out on the film by *less* than the circle of
> confusion.
>
> The lower diagram adjusts the position of the near and far DoF
> planes so that on-axis points create a spread on the film equal to the
> circle of confusion. If you churn through the trig, you find that the
> new near DoF is almost exactly the same as the old one, but that the far
> DoF doesn't recede to infinity quite as fast as the old one did when the
> lens is stopped down. I have convinced myself that some of my
> hyperfocal-distance landscapes with surprisingly blurred distant
> mountains are due to this mismatch.
I think I'm missing something here since I don't know where your first
diagram comes from. The second diagram is what I use to derive the
formulas for depth of field. However, the exact formulas one gets this
way are a trifle complicated and one usually simplifies them by assuming
the object distance is large compared to the focal length.
>
> Mind you, it's completely cured by stopping down one more stop for
> safety.
From experience, I have to agree with you completely about that. But I
would like to see why it is true from a theoretical point of view. That
is why I raised the issue in the first place. Not only would the theory
be enlightening, but it would give a way to understand
quantitatively how large a
departure from film flatness is acceptable as it relates to format size.
>
> Struan
Struan Gray
>> I wrote:
>>
>> Readers might want to look at this diagram:
>>
>> http://www.sljus.lu.se/People/Struan/pics/dofconfuse.jpg
>>
>> The top diagram illustrates the conventional way of calculating
>> DOF, as described in Andrew's post. The extension to a buckled film
>> should be obvious.
>
> I'm afraid I don't recognize that diagram. The bottom diagram is the
> one I'm familar with.
Lucky for you. The books I have all use the top diagram, as do
all the internet sources I have looked at closely. They confused the
hell out of me when I started to trying to figure out what these DoF
discussions were all about.
The idea is that the light from an on-axis point source forms a
cone behind the lens, with its base formed by the exit pupil and its
point at the focal plane. It then spreads out in a second cone behind
the focal plane. If the film moves away from the focal plane, the
light will form a circle on it rather than a point. Once you have
defined an 'acceptable' circle of confusion, that then translates into
an allowable movement of the film. The end points of that allowable
movement define new image planes, which you translate into positions
in front of the lens using the Gaussian formula.
That all makes sense in terms of film movement, but the same
argument and diagram are often given to derive DoF. In my readings,
'often' means 'always', and although I am happy to accept that might
say as much about my readings as it does about the world of
photography, I have seen the argument presented many, many times.
The formulea given by the two methods are very close for all
likely photographic situations. I plugged them into an analysis
package I use and plotted DoF limits for various combinations of focal
length, aperture and object distance. The only practical difference I
could find was the one I mentioned: if you take a landscape and focus
on the hyperfocal distance given by the first diagram, your horizon
will be blurred by more than you expected.
> The second diagram is what I use to derive the
> formulas for depth of field. However, the exact
> formulas one gets this way are a trifle complicated
> and one usually simplifies them by assuming
> the object distance is large compared to the
> focal length.
If you assume that you know the actual f-number the formulea are
not too bad. They certainly allow you to play with parameters and see
what is important and what is not. For any given lens, you should
work out what the f-number really is (often called the 'effective
aperture' in photography books), but outside the macro range it
doesn't change enough to worry about.
> I would like to see why it is true from a
> theoretical point of view.
Me too, so bugger the armchair beer swillers who think an interest
in mathematics precludes an interest in art. They are simply misled
by their own inability to engage in more than one mental function at a
time into thinking that it is a universal affliction.
> Not only would the theory be enlightening, but it
> would give a way to understand quantitatively how
> large a departure from film flatness is acceptable
> as it relates to format size.
For a true quantitative understanding you need to add diffraction
and aberrations to the mix. Diffraction can be treated analytically,
especially if you don't mind jumping into Fourier space, but
aberrations quickly lead you into a morass of special cases and
empiricisim. Worse, the effects of aberrations are such that choosing
'best focus' becomes a subjective issue, and 'acceptable' depends on
whether you like sharpened edges or softly-glowing highlights.
There are freeware or demo ray-tracing packages if you want to get
into this in detail, but for practical photography outside of the
technical and scientific realm, remembering the relationships between
quantities is more important than knowing exact values for the
prefactors in the equations.
In the case of film bulge, Bob M's page gives numbers and facts
for a variety of cameras and photographic situations. It's no
accident that those doing photogrammetry or film-based remote sensing
take great care to keep the film flat. It is also no accident that in
almost all scientific imaging outside wide-area survey problems the
field has been taken over by nice, flat CCD chips.
I personally regard film bulge in the same way I regard
reciprocity failure. It's one of those things that is irrelevant for
almost all my photography, but worth konwing about for those (very)
few occasions where it can ruin the shot.
Struan
John Stafford
on a camera. Focus on an object. Then turn your focusing mount out
.006". The image will be thrown out of focus to some degree. That's
happening in your camera when the film isn't flat. Stopping down is
covering the effect, so shoot a dozen or so rolls wide open at closer
ranges. If you can't see the consequences of non-flat film, then
there's nothing to worry about.
:) I chose .006" above because that's the best tolerance I can made
when planing wood, for example when I hacked out my super-wide,
bellowless 5x4. Rather arbitrary, maybe too big a number for
film-flatness, but it illustrates the point.
I do appreciate your math, Professor. Maybe you could be our Martin
Gardiner of Photography. Did you know that every mechanical 3-number
dial combination lock has two different combinations? Oh. I think I
learned that from you. :) -5, +2.5
Robert Monaghan
unfortunately, lenses don't throw totally flat images, nearly all have
some degree of field curvature, with many designs (telephotos..) having a
lot of field curvature in many examples ;-) Most lenses are setup to
compromise, so the center loses resolution and the edges are sharper than
they would otherwise be.
I suspect field curvature has as much practical effect on the issues of
f/stops needed to ensure depth of focus is adequate to cover this issue as
well, esp. in the corners of film, as focusing errors and film bulges.
somewhat related is the issue with illumination, esp. overcoming cos^4
losses in simpler lens designs the exit pupil as seen from the corners is
reportedly not necessarily a circle, to provide more illumination etc.
So I suspect the math is a whole lot more complicated than the simple
cones and circular exit pupils seemingly being assumed. Oh well, this will
make Leonard happy, at least ;-)
Y'all carry on... ;-) bobm
Clint O'Connor
your lens wide. That minimizes your depth of field and film flatness may
become a consideration.
You could toss out your lens and use a pinhole. Nearly infinite depth of
field and film flatness is NOT an issue.
Clint
Austin, TX
"Leonard Evens" <l...@math.northwestern.edu> wrote in message
news:20011204.085012...@math.northwestern.edu...
FLEXARET2
At a Photo Expo I met one of the experts at the Kyocera booth to discuss the
Contax 645. While realizing he had an axe to grind to promote his product, I
feel he did speak truthfully to me when he said-
"The medium format lack of film flatness is the best kept secret in the
photographic industry."
He told me that his company's Contax 645 had partially solved that problem, but
only in their vacuum back - which only took 220 film, as it would not work with
the backing paper on 120 film.
If I ran a business where film plane flatness was a critical issue
I would buy and use a Contax 645 with that back and 220 film.
I have been examining a variety of cameras old and new to see
how that problem is dealt with.
You can read my articles on Bob Monaghan's Medium Format/Bronica site-
http://people.smu.edu/rmonagha/bronica.html
My short conclusions - the current Rollei 6x6cm SLR cameras -very flat film
plane. On the Pentax 67 - very flat film plane.
On less expensive equipment - Bronica S2A with 12/24 back and
improved one top roller film insert (sometimes marked "A") - very flat film
plane - so that I can use my adapted 180MM f2.8 Sonnar wide open
and get sharp portraits with the background blurred out.
New Kiev NT backs - available in Kiev 88 and Hasselblad versions
for under $99 - very flat film plane.
Other manufacturers have dealt with this in various ways, better or worse. If
the user wants to shoot with wider apertures and still get sharp pictures, he
should not be forced to stop down and lose a creative effect to make up for the
lack of a flat film plane.
- Sam Sherman