Higher F-stops
letsc...@hotmail.com
than the 8.0 my G1 goes to. I see F11, F16 etc. What is different
about these higher F stops and if getting to them is important why
does the G1 stop at F8? Is this more important for film cameras
maybe. Seems like I see reference to them in articles that are film
related.
John H. Power
Ken Clause
>does the G1 stop at F8?
An economy move.
Jim
news:20011003074515...@mb-ft.aol.com...
> >why
> >does the G1 stop at F8?
>
> An economy move.
It´s not economy. Digital cameras have sensors smaller than 24x36 film and
with that follows greater depth of field for a given angle of view. With
film you often need those small apertures to get sufficient DOF, but digital
cameras almost always have too much of it.
Apertures smaller than f/8 give less sharpness due to an effect called
diffraction. Since digital cameras don´t need the small apertures, ther is
no point in providing them.
Large apertures are another issue though. Digital cameras woul really need
someting like f/0.2 to be able to make images where small DOF plays a role.
And that´s expensive.
//Jim
Kinon O'Cann
the photographer to use longer shutter speeds so certain effects are
obtainable, like very blurred water, etc. Canon probably did this, like the
other responder said, to save money.
<letsc...@hotmail.com> wrote in message
news:75tlrtsl5daimkm7d...@4ax.com...
Don Stauffer
aperture stop range. Cheaper P&S film cameras also limit the f/# range.
Second, since CCD chips are slower (ISO) than many available films, one
cannot shot with high f/# without a tripod. Manufacturers figure 90+ %
of buyers never use a tripod, so slow exposure with high f/# is not a
big demand. However, more serious photographers would want the higher
f/#.
In particular, any camera offering macro capability should give
availability of high f/#. While depth of field on small format cameras
(including CCD ones) is better than in 35mm, it is not enough better for
macro photography. Macro shots need all the depth they can get. Most
macro photographers do not object to needing tripod.
--
Don Stauffer in Minnesota
stau...@usfamily.net
webpage- http://www.usfamily.net/web/stauffer
Scott Storkel
> It´s not economy. Digital cameras have sensors smaller than 24x36 film
> and with that follows greater depth of field for a given angle of view.
> With film you often need those small apertures to get sufficient DOF,
> but digital cameras almost always have too much of it.
While what you've said may be true, I have to disagree that
digicams don't need aperatures smaller than f8. I've seen
digicams shots made at f8 whihc really could have used a
*lot* more depth of field! The reason digicams have a limited
number of aperatures is, as the previous poster said, money!
> Apertures smaller than f/8 give less sharpness due to an effect called
> diffraction. Since digital cameras don´t need the small apertures,
> ther is no point in providing them.
Bullshit! This is entirely dependent on the design of the lens!
Lenses can be soft at both very large and very small aperatures.
With 35mm gear, you'll often find that f8 and f11 are your
sharpest aperatures, while f2.8 and f32 aren't quite as sharp.
Or, if you've got a really high-quality lens, the differences
may be unnoticable!
-- Scott
--
==========================================================================
Scott Storkel Devious Software & Consulting
Software Engineer ssto...@devious.com
==========================================================================
Dave Martindale
>"Jim" <nos...@nospam.dom> wrote in news:9pf19j$rn0$1...@news.kth.se:
>While what you've said may be true, I have to disagree that
>digicams don't need aperatures smaller than f8. I've seen
>digicams shots made at f8 whihc really could have used a
>*lot* more depth of field! The reason digicams have a limited
>number of aperatures is, as the previous poster said, money!
It's not that simple. Some digicams do have a limited aperture range
because they use the same mechanism for shutter and lens diaphragm, and
it can't close down small enough accurately enough and fast enough.
But even with separate shutter and aperture mechanisms, current digicams
can't really use the small apertures that 35 mm cameras can due to
diffraction (see below).
>> Apertures smaller than f/8 give less sharpness due to an effect called
>> diffraction. Since digital cameras don´t need the small apertures,
>> ther is no point in providing them.
>Bullshit! This is entirely dependent on the design of the lens!
>Lenses can be soft at both very large and very small aperatures.
>With 35mm gear, you'll often find that f8 and f11 are your
>sharpest aperatures, while f2.8 and f32 aren't quite as sharp.
>Or, if you've got a really high-quality lens, the differences
>may be unnoticable!
You're ignoring diffraction. Because of diffraction, *no lens*, no matter
how well made, can be sharper than a certain amount at f/32. And the size
of the blur due to diffraction at f/32 is the same regardless of focal
length. So the blur due to f/32 diffraction is no problem on an 8x10 view
camera, because the negative won't be enlarged very much more when making
a print. On a 35 mm camera, f/32 would produce too much blur for most
purposes, so 35 lenses usually stop at f/16 or f/22.
But the sensors on digital cameras are tiny. Current digicams mostly use
sensors that are about 1/5 the dimensions of 35 mm film, so the image will
be magnified 5 times more than a 35 mm negative for the same size print.
The small sensor size requires the lens focal length to be scaled down
as well (no problem; saves weight and size). But it also requires 5
times as much resolution (in lp/mm) for the same sharpness in the print.
Fortunately, digicam sensors do *not* have the same resolution as film,
so we can get away with lenses that actually only have 2 or 3 times the
resolution of 35 mm camera lenses.
However, for reasonable sharpness, the digicam *also* has to keep the
blur due to diffraction comparable to aberrations. If f/22 is a practical
minimum aperture for 35 mm lenses, then with 1/5 the sensor size and
5 times the magnification in printing, f/4 is the aperture that will give
the same apparent amount of diffraction blur in a digicam. Allowing a
minmum aperture of f/8 actually *doubles* the size of blur compared to
f/22 on a 35 camera. But this is acceptable partly because the digicam
sensor can't resolve as much as the film, and the lens can resolve 5 times
the lp/mm either - so the f/8 diffraction blur is probably comparable to
lens aberration blur, and the expectations are just somewhat lower.
Basically, the physics of diffraction and image scaling mean that the
minumum useful aperture gets larger as the sensor gets smaller.
Compensating for this, depth of field gets greater at the same f/number as
the focal length gets shorter, so the digicam at f/8 actually has more
depth of field than the 35 lens at f/32. But the total aperture span
is reduced. The only way around this is to use a larger CCD sensor and
longer focal length lens.
Dave
Dave Martindale
>In particular, any camera offering macro capability should give
>availability of high f/#. While depth of field on small format cameras
>(including CCD ones) is better than in 35mm, it is not enough better for
>macro photography. Macro shots need all the depth they can get. Most
>macro photographers do not object to needing tripod.
The trouble is that at large f/numbers (small apertures) diffraction
causes blurring, but macro photographers are usually very fussy about
sharpness. At f/8, the typical digicam lens effectively has *more*
diffraction-caused blur than a 35 mm lens at f/22 or f/32 - because
the sensor is 1/5 the size of a 35 mm film frame, and the printing
magnification is 5 times as much.
And diffraction makes *all* of the image fuzzy, while with shallow
depth of field at least something is sharp.
Dave
Leonard Evens
<da...@cs.ubc.ca> wrote:
A good discussion. You will probably be abused because of it by people
who don't understand basic optics but think they do.
--
Leonard Evens l...@math.northwestern.edu 847-491-5537
Dept. of Mathematics, Northwestern Univ., Evanston, IL 60208
r...@nicholson.com
Scott Storkel <ssto...@devious.com> wrote:
>"Jim" <nos...@nospam.dom> wrote in news:9pf19j$rn0$1...@news.kth.se:
>> diffraction. Since digital cameras don´t need the small apertures,
>> ther is no point in providing them.
>
>Bullshit! This is entirely dependent on the design of the lens!
Wrong. Diffraction limits are due to the laws of physics regarding
light of a given wavelength passing through a given size aperture
(and that's size in wavelengths, not in f-stops). Check your old
college physics textbook. Diffraction effects cannot be fixed by
lens design except by making the lens bigger (or worse at all apertures!).
Since most digital cameras use smaller shorter focal length lenses
than film cameras (but use visible light of the same wavelengths),
their aperture will be smaller in linear diameter at the same f-stop,
leading to diffraction limited resolution at much lower f-stops.
Only digital cameras that use film size sensors are immune to this
problem (price those lately?).
IMHO. YMMV.
--
Ron Nicholson r...@nicholson.com http://www.nicholson.com/rhn/
#include <canonical.disclaimer> // only my own opinions, etc.
Leonard Evens
<da...@cs.ubc.ca> wrote:
Keep trying, but it is going to be hard to get through.
Maybe another way to put it would be as follows. Because of the smaller
size of the sensor and the correspondingly smaller focal lengths of the
lenses, one gains somewhere between 4 and 5 stops with respect to depth
of field. So shooting at f/4 with a typical consumer grade digital
camera is equivalent to shooting at f/11 with a 35 mm camera. As Dave
has tried to point out, if you insist on going to f/11 anyway in order to
get even more depth of field, you run into diffraction. THE SIZE OF THE
DIFFRACTION DISK DEPENDS ONLY ON THE F-NUMBER.
(The diffraction or Airy disk is what you get from the image of a point.)
You can tolerate a larger f-number with a camera and format which doesn't
have to be enlarged as much, because you are not going to enlarge that
diffraction disk as much. While you can make an 8 x 10 by enlarging
a 35 mm negative about 8 times, you have to enlarge the image on the
sensor of a digital camera something like 40 times! As he says, you gain
a little because, given the limitations of the sensor array, you are not
going to be so demanding in the first place, but still ultimately
diffraction places significant limits on what you can accomplish.
If you don't believe in diffraction, you should take all your pictures,
or at least those without movement, with a pinhole, and you won't need a
lens at all. Then everything will be in (equally bad) focus.
r...@nicholson.com
Leonard Evens <l...@math.northwestern.edu> wrote:
>Keep trying, but it is going to be hard to get through.
>THE SIZE OF THE DIFFRACTION DISK DEPENDS ONLY ON THE F-NUMBER.
A better way of looking at this is to divide f-number by focal lenth.
Then you get the fact that the angular resolution is dependent ONLY
on the diameter of the aperture (assuming standard visible light
photography and quality lenses & sensors).
More precisely: resolving_angle ~= 1.22 wavelength / diameter
As the lens aperture diameter gets smaller, two objects must be a
greater angle apart in order to not appear as a single blurred disk.
Thus any detail in a photo taken a fixed distance away from some given
subject matter will be diffraction limited by the linear diameter of
the aperture opening actually used for the exposure. Since digital
cameras with small CCD sensors use smaller lenses which thus have a
smaller aperture diameter at the same f-stop, these digital cameras
will require a larger f-stop for the same linear aperture diameter
which will give the same angular resolution of the subject being
photographed.
You can scale two was to avoid this difference: use a larger sensor
(film size CCD); or move the camera closer to the subject matter by
the proportional difference in focal lengths (which would change
composition in most 3D scenes).
IMHO. YMMV. (& someone who actually made & used pinhole cameras)
r...@nicholson.com
>You can scale two ways to avoid this difference: use a larger sensor
>(film size CCD); or move the camera closer to the subject matter by
>the proportional difference in focal lengths (which would change
>composition in most 3D scenes).
Or (I forgot one) use shorter wavelengths of EM radiation. Extreme
Ultraviolet photography anyone?
radiojohn
> >"Jim" <nos...@nospam.dom> wrote in news:9pf19j$rn0$1...@news.kth.se:
> >> Apertures smaller than f/8 give less sharpness due to an effect called
> >> diffraction. Since digital cameras don´t need the small apertures,
> >> ther is no point in providing them.
> >
> >Bullshit! This is entirely dependent on the design of the lens!
>
> Wrong. Diffraction limits are due to the laws of physics regarding
> light of a given wavelength passing through a given size aperture
> (and that's size in wavelengths, not in f-stops). Check your old
> college physics textbook. Diffraction effects cannot be fixed by
> lens design except by making the lens bigger (or worse at all
apertures!).
I agree. Our potty mouth friend apparently hasn't noticed that a lot of
subminiature cameras (whose film size is like a digi sensor) never get down
to F/16. Most stop at 11. That's because the relative aperture would be so
small that diffraction would make the image less sharp.
John
Wtempemail2
Dave--
Thanks for an informative, logical response without descending to the lower
level of those disagreeing with you.
--Wayne
Dave Martindale
>Or (I forgot one) use shorter wavelengths of EM radiation. Extreme
>Ultraviolet photography anyone?
Why stop there? X-rays!
Of course, making the lens becomes a bit of a problem...
Dave
Scott Storkel
> In article <Xns912F51B8B5FBDs...@129.250.35.204>,
> Scott Storkel <ssto...@devious.com> wrote:
>>"Jim" <nos...@nospam.dom> wrote in news:9pf19j$rn0$1...@news.kth.se:
>>> Apertures smaller than f/8 give less sharpness due to an effect
>>> called diffraction. Since digital cameras don´t need the small
>>> apertures, ther is no point in providing them.
>>
>>Bullshit! This is entirely dependent on the design of the lens!
>
> Wrong. Diffraction limits are due to the laws of physics regarding
> light of a given wavelength passing through a given size aperture
> (and that's size in wavelengths, not in f-stops). Check your old
> college physics textbook. Diffraction effects cannot be fixed by
> lens design except by making the lens bigger (or worse at all
> apertures!).
Right. I would claim that current lenses probably aren't
diffraction limited. In fact, I would think that chromatic
abberation and other lens-related distortions are a bigger
problem than diffraction. After all: most complete digicam
systems cost less than a single 35mm SLR lens! The manufacturer
has to be taking short-cuts somewhere and from what I've seen,
it's usually in the design of the lens!
> Since most digital cameras use smaller shorter focal length lenses
> than film cameras (but use visible light of the same wavelengths),
> their aperture will be smaller in linear diameter at the same f-stop,
> leading to diffraction limited resolution at much lower f-stops.
> Only digital cameras that use film size sensors are immune to this
> problem (price those lately?).
Right. But the question is: at what aperature does the lens
become diffraction-limited? Is is f2.0? f11.0? f32.0? So far,
nobody has been able to answer that question convincingly...
Scott Storkel
news:9pfifb$tp$1...@lily.cs.ubc.ca:
> You're ignoring diffraction.
And I think you're ignoring reality... Have you done the actual
calculations to show that diffraction is a problem with the lenses
and CCDs we're talking about?
> Because of diffraction, *no lens*, no
> matter how well made, can be sharper than a certain amount at f/32.
> And the size of the blur due to diffraction at f/32 is the same
> regardless of focal length. So the blur due to f/32 diffraction is no
> problem on an 8x10 view camera, because the negative won't be enlarged
> very much more when making a print. On a 35 mm camera, f/32 would
> produce too much blur for most purposes, so 35 lenses usually stop at
> f/16 or f/22.
Really? So much for those crazy kooks in Group f64, eh? What
were THEY thinking?
I guess you're also using a different sort of "35mm" than I'm
using. The vast majority of my Canon EOS EF lenses are quite
happy to produce f/32 exposures. Some of the close-up shots I've
done at f/32 with my 100mm macro lens look tack sharp (under my
10X magnifying loupe) to me!
Finally: if what you say is true, why does the Canon D30 provide
the option to shoot at f/64, f/72, f/81, and f/91?!?! At least
according to Phil Askey's review:
http://www.dpreview.com/reviews/canond30/page6.asp
Granted, the D30's CMOS sensor is quite a bit larger than the
CCD's typically found in consumer-oriented digicams. Still:
f/91!
> But the sensors on digital cameras are tiny. Current digicams mostly
> use sensors that are about 1/5 the dimensions of 35 mm film, so the
> image will be magnified 5 times more than a 35 mm negative for the same
> size print.
Which is a complete red herring, don't you think? How does the
amount of enlargement needed for a given viewing size have anything
to do with diffraction?
> However, for reasonable sharpness, the digicam *also* has to keep the
> blur due to diffraction comparable to aberrations. If f/22 is a
> practical minimum aperture for 35 mm lenses,
It doesn't appear to be.
> then with 1/5 the sensor
> size and 5 times the magnification in printing, f/4 is the aperture
> that will give the same apparent amount of diffraction blur in a
> digicam.
Show me the math...
> Allowing a minmum aperture of f/8 actually *doubles* the size
> of blur compared to f/22 on a 35 camera. But this is acceptable partly
> because the digicam sensor can't resolve as much as the film, and the
> lens can resolve 5 times the lp/mm either - so the f/8 diffraction blur
> is probably comparable to lens aberration blur, and the expectations
> are just somewhat lower.
This may be true. Unfortunately, you've failed to provide ANY
concrete information which would indicate that there is any
diffraction blurring in these lens systems! Instead, you've
assumed your conclusion was correct from the outset and then
merely repeated that conclusion in a number of unscientific
ways. Why don't you run through the math and show us how much
"blur" we're going to get if the lens on, say, an Olympus C-3040Z
provided an optio to shoot at f/22?
Dave Martindale
But I hope it's wortwhile for others to read - it contains real calculations
of how the maximum resoluton of an image is affected by lens aperture.
------------------------------------------------------------------------
Scott Storkel <ssto...@devious.com> writes:
>> You're ignoring diffraction.
>And I think you're ignoring reality... Have you done the actual
>calculations to show that diffraction is a problem with the lenses
>and CCDs we're talking about?
I hadn't - but I knew how diffraction effects scale with f/number and
sensor size. (And I'm tempted to ask if *you* have even looked at this).
But let's look at actual numbers. For a diffraction-limited lens,
the Rayleigh criteria says that the closest two features can come and
still be resolved as separate features is R = 1.22 * lambda * N,
where "lambda" is the wavelength of light and "N" is just the f-number
of the lens, calculated by dividing the focal length by the pupil size.
The spatial resolution limit, in line pairs per meter, is the reciprocal
of this number.
If we assume a wavelength of 550 nm (green light) and an aperture of
f/16, the Rayleigh resolution limit is 10.7 um. This gives an absolute
upper limit on resolution of 93 lp/mm. At 700 nm (deep red), this changes
to 13.6 um and 73 lp/mm.
Returning to green light, apertures of f/22, f/32, and f/64 give
Rayleigh spacings of 14.8, 21.4, and 42.9 um respectively, and maximum
spatial resolutions of 67.7, 46.6, and 23.3 lp/mm.
As you can see, f/64 is really unacceptable for a 35 mm camera lens.
Normally, a 25 or 30 um blur disc is considered the maximum size
allowed when calculating depth of field, so at f/64 *nothing* would
be sharp enough to be "in focus" by that criterion.
At f/32, the image would be marginally OK, and f/22 and f/16 are definitely
usable.
But that's for 35 mm film. With current digicams (let's use the Canon
G2 as typical), the sensor is about 1/5 the dimensions of 35 mm film,
and the image has to be enlarged 5 times more than a 35 mm image for
the same size print. So for the same standards of sharpness, the
acceptable blur spot has to be 5 times smaller - about 5 or 6 um.
Or look at it in spatial resolution terms: the 4 Mpixel sensor in
the G2 has a pixel spacing of around 3.1 um. It takes two pixels to
resolve a line pair, so the theoretical resolution limit of the sensor
alone is one line pair per 6.2 um, or 160 lp/mm.
Now, the minimum aperture of the G2 lens is f/8. For green light, that
gives a Rayleigh resolution limit of 5.4 nm, and a spatial resolution
limit of 186 lp/mm. And that's the diffraction limit for a perfect lens,
not the real lens in the G2!
Notice how well the maximum possible resolution of the lens at f/8 matches
the resolution of the sensor? At even one stop more (f/11), the maximum
resolution drops to 135 lp/mm, and the image is not of the quality that
you should have with a 4 Mpixel camera - even in perfect focus, even
with no lens aberrations at all.
>Really? So much for those crazy kooks in Group f64, eh? What
>were THEY thinking?
THEY were thinking of using larger format cameras, where f/64 is
a useful aperture.
>I guess you're also using a different sort of "35mm" than I'm
>using. The vast majority of my Canon EOS EF lenses are quite
>happy to produce f/32 exposures. Some of the close-up shots I've
>done at f/32 with my 100mm macro lens look tack sharp (under my
>10X magnifying loupe) to me!
As calculated above, f/32 is really the limit of what's useful on
a 35 camera lens. But because of the scaling, when the sensor is
1/5 the size, the aperture that gives the *same* diffraction-limited
sharpness is 32/5, or about f/6. So a digicam lens at f/8 is LESS
sharp than your 35 lenses at f/32, when the image is printed the same
size.
>Finally: if what you say is true, why does the Canon D30 provide
>the option to shoot at f/64, f/72, f/81, and f/91?!?! At least
>according to Phil Askey's review:
> http://www.dpreview.com/reviews/canond30/page6.asp
>Granted, the D30's CMOS sensor is quite a bit larger than the
>CCD's typically found in consumer-oriented digicams. Still:
>f/91!
The D30 uses a 10.1 um pixel spacing, more than 3 times that of the
sensor in the G2. The sensor's maximum resolution is thus 49.5 lp/mm.
A diffraction-limited lens will drop to that somwhere just short of
f/32.
The dpreview review lists apertures that the camera *recognizes* as
ranging from f/1 to f/91 in 1/3-stop increments. But it says that the
*actual* apertures depend on the lens installed. So, which Canon
lenses stop down further than f/32?
>> But the sensors on digital cameras are tiny. Current digicams mostly
>> use sensors that are about 1/5 the dimensions of 35 mm film, so the
>> image will be magnified 5 times more than a 35 mm negative for the same
>> size print.
>Which is a complete red herring, don't you think? How does the
>amount of enlargement needed for a given viewing size have anything
>to do with diffraction?
Not a red herring at all. Features are sharp or blurry depending on
their angular size to your eye as you look at them. For a normal viewing
distance, you need at least 4 lp/mm on the print. If the print is an 8x10
print from a 35 negative, the magnification is about 8 times, so you need
at least 32 lp/mm on the *negative* to get that on the print. If the original
image comes from a sensor that's only 7 mm wide, the magnification needed
is now about 40 times, and you need 160 lp/mm from the sensor for the same
sharpness on the same-size print. This seems awfully obvious to me - it's
just basic geometry.
And you can get the 32 lp/mm you need from 35 mm film at up to f/32, but
f/8 is as high as you can go when you need 160 lp/mm minimum. That's
why the amount of enlargment directly determines the amount of diffraction
blur that is tolerable.
(It's also worth noting that the 4 lp/mm figure I used is the lower end
of the range of what's considered a sharp print - I've seen figures that
range from 4 to 8 lp/mm. A 4 Mpixel camera can't do better than 4 lp/mm,
but a 35 camera with good lens and film *can*. But to get 8 lp/mm from
35, the smallest aperture possible is f/16, not f/32).
>> then with 1/5 the sensor
>> size and 5 times the magnification in printing, f/4 is the aperture
>> that will give the same apparent amount of diffraction blur in a
>> digicam.
>Show me the math...
My original article had math in it, written in words and based on the
scaling factors between two formats. You complain that it's not good
enough math to convince you - but you present absolutely NO math in your
article, and as far as I can tell you've never even looked up diffraction.
Now there's much more precise math for your edification above. If you
still disagree with my conclusions, show us *your* math, please.
>This may be true. Unfortunately, you've failed to provide ANY
>concrete information which would indicate that there is any
>diffraction blurring in these lens systems!
Well, you could have gone off and actually looked up diffraction,
and found that it *is* a factor in lens design, and *does* limit
the smallest usable aperture. Or you could have looked up diffraction
yourself. Either one of these would have shown that I was basically
right. But instead, you present an even *more* handwaving argument
than I did, with no math at all, while complaining about my lack
of hard numbers. Do I detect a double standard?
>Instead, you've
>assumed your conclusion was correct from the outset and then
>merely repeated that conclusion in a number of unscientific
>ways. Why don't you run through the math and show us how much
>"blur" we're going to get if the lens on, say, an Olympus C-3040Z
>provided an optio to shoot at f/22?
Well, I happened to know that my conclusion was basically correct, even
without detailed calculations. While you apparently attacked it
without benefit of any knowledge or research into diffraction at all.
My argument itself may have been unscientific, but it was reporting
knowledge obtained by science. Where did your assertions come from?
Anyway, the result you ask for is above: at f/22, with green light, the
Rayleigh resolution limit is 14.8 um, and the spatial resolution limit
is about 68 lp/mm. Assuming that the C-3040 has about the same size
sensor as the G2, that's about 484 line pairs or 967 pixels across the
width of the sensor.
In other words, no matter what resolution you set the camera to, the
actual image quality you'd get would be equivalent to about 967 x 725
pixels, or a 0.7 Mpixel image. If users of the C-3040 got images of
this quality from the camera, they'd go back to Olympus as defective,
despite that fact that this is the best image a *perfect* lens can
produce at that f/number and scale. Olympus isn't stupid, so they
don't let you shoot at apertures that would give such poor images.
Dave
PS: The formulas used in this article come from "Applied Photographic
Optics", Sidney F. Ray, ISBN 0-240-1350-9. In particular, the first
couple of pages of chapter 16, "Resolving power of lenses and imaging
systems" has the formulas and some interesting diagrams.
Dave Martindale
>Right. I would claim that current lenses probably aren't
>diffraction limited. In fact, I would think that chromatic
>abberation and other lens-related distortions are a bigger
>problem than diffraction.
It depends on aperture! Do you know that most camera lenses are
at their sharpest about 2 or 3 stops down from maximum aperture?
At that aperture, aberrations and diffraction are roughly balanced
in the amount of effect they have on the image.
At all larger apertures, diffraction decreases but aberrations
increase, so the image is less sharp. At all *smaller* apertures,
aberrations decrease but diffraction increases, so the image is less
sharp.
So, most normal camera lenses are diffraction-limited for about the
smaller half of their aperture range. Some special-purpose lenses
like macro and graphics arts lenses may be diffraction-limited at *all*
apertures.
Dave
letsc...@hotmail.com
straightforward fashion. I think the intial answer related to cost
considerations which was not really what I was confused about.
John H. Power
Jim Heikkinen
<letsc...@hotmail.com> wrote in message
news:d0gort46q75furjc9...@4ax.com...
> I am not sure my original question was ever answered in a simple and
> straightforward fashion. I think the intial answer related to cost
> considerations which was not really what I was confused about.
Your question *is* answered! Since digicams can´t produce good images at
small apertures, manufacturers don´t give them small apertures.
I guess cost *could* be a completing issue, but I doubt it.
//Jim
Leonard Evens
<da...@cs.ubc.ca> wrote:
> Warning: this is a long article. Scott asked for math, so he gets it.
> But I hope it's wortwhile for others to read - it contains real
> calculations of how the maximum resoluton of an image is affected by
> lens aperture.
Let me applaud you for a really excellent discussion. It should
convince anyone who is willing to do a little work and is really
interested in the optics behind digital photography.
But I suspect you aren't going to convince your debating opponent.
There is an interesting phenomenon going on here. "Practical men"
distrust theory and like to rely on their experience. While experience
in its totality should always trump theory, such people don't realize
that one's preconceptions can significantly affect what one "sees". It
isn't that we don't see what we think we see, but we haven't
looked at sufficeintly many situations, and we can be misled by limited
observations. It is easy to make up some conceptual model to explain in
our minds what occurs in common situations we are very familiar with and
then overgeneralize to extend them to ranges where they don't apply.
One only finally realize that the conceptual model is faulty when one is
presented with very different situations. Theory of course must
explain everything in the applicable range of observations, not just
those of interest to "practical men".
The paradox here is that those of us who do have solid scientific
training are usually much more aware that we could be wrong, even about
things we thought we knew very well, than the "practical men" we find
ourselves debating. I'm sure that statement will enrage some readers
who will treat it as being elitist, but science and mathematics are
basically not elitist in the sense that anyone who is willing to put in
the work has a chance to learn the subject. But there are no short cuts
and making it up yourself on the basis of limited knowledge is not
allowed.
> PS: The formulas used in this article come from "Applied Photographic
> Optics", Sidney F. Ray, ISBN 0-240-1350-9. In particular, the first
> couple of pages of chapter 16, "Resolving power of lenses and imaging
> systems" has the formulas and some interesting diagrams.
Thanks for the reference. I've been using my old copy of Photographic
Optics by Arthur Cox, but it is a bit out of date and not very clear on
some points. Of course, the basic optics goes back to the 19th century,
but there are always additional facts, so a more up to date reference
will be useful. Do you know the publisher of that reference? It is
not in the Northwestern University Library, which is rather poor in the
area of photography.
Don Stauffer
'shape' than defocus blur. That is, the spot profile or point spread
function of diffraction is not the same shape as defocus, and gives a
somewhat different effect. To many people, the effect of diffraction is
not the same as defocus.
In fact, the once-popular soft focus screens made use of this effect.
Some portrait photographers used to shoot at very high apertures to use
the effect to do an artistic softening that softened skin pores, but did
not really make the photo 'fuzzy'.
Dave Martindale wrote:
>
>
> The trouble is that at large f/numbers (small apertures) diffraction
> causes blurring, but macro photographers are usually very fussy about
> sharpness. At f/8, the typical digicam lens effectively has *more*
> diffraction-caused blur than a 35 mm lens at f/22 or f/32 - because
> the sensor is 1/5 the size of a 35 mm film frame, and the printing
> magnification is 5 times as much.
>
> And diffraction makes *all* of the image fuzzy, while with shallow
> depth of field at least something is sharp.
>
> Dave
--
Charlie Ih
Thank you for giving an excellent explanation. There are so much information
and mis-information on the NG. Many readers who is not knowledgeable
in this field don't normally know which is right and which is wrong.
Cameras are normally designed to match the lens resolution to that of
the sensor (film or CCD). There is absolutely nothing wrong using a
larger f# than the optimum, because it still can produce a more than
acceptible image except that you don't get the most out of the sensor
which, BTW, often provides more than we normally need.
In DC, there may also be a physical limitation. For instance, the actual
size of the aperture of a 8 mm lens is 0.25 mm at f/32 and 0.36 mm at f/22.
In article <9ph82s$31s$1...@lily.cs.ubc.ca>,
Dave Martindale <da...@cs.ubc.ca> wrote:
--
Charles S. Ih
302-831-8173, FAX 302-831-4316
e-mail, i...@mail.eecis.udel.edu
Dave Martindale
>> PS: The formulas used in this article come from "Applied Photographic
>> Optics", Sidney F. Ray, ISBN 0-240-1350-9. In particular, the first
>> couple of pages of chapter 16, "Resolving power of lenses and imaging
>> systems" has the formulas and some interesting diagrams.
>Thanks for the reference. I've been using my old copy of Photographic
>Optics by Arthur Cox, but it is a bit out of date and not very clear on
>some points. Of course, the basic optics goes back to the 19th century,
>but there are always additional facts, so a more up to date reference
>will be useful. Do you know the publisher of that reference? It is
>not in the Northwestern University Library, which is rather poor in the
>area of photography.
The Ray book is published by Focal Press, the same publisher as Cox.
I actually just got a used copy of Cox and haven't had a chance to look
at it yet. Cox was originally published in 1943 (and seems to use
Imperial units) while Ray was first published in 1988.
Warren J. Smith also has a couple of good books on real-world lens
design that may well be better yet; Ray is simply the first book I
took off the shelf and it had enough info to answer the questions.
Dave
Dave Martindale
>Thank you for giving an excellent explanation. There are so much information
>and mis-information on the NG. Many readers who is not knowledgeable
>in this field don't normally know which is right and which is wrong.
I'd like to point out that I'm *not* a physicist or optical designer
or even an engineer. I'm a computer scientist by training, and everything
that I know about optics came from buying a few books and reading them.
There are a number of books on photographic optics that are written for
the practical *user* of optics, and don't require a university education
in *any* field to read.
>In DC, there may also be a physical limitation. For instance, the actual
>size of the aperture of a 8 mm lens is 0.25 mm at f/32 and 0.36 mm at f/22.
That's a good point, which I missed. At these apertures and focal lengths,
the lens really *is* approaching a pinhole!
Dave
Dave Martindale
>In addition to fuzzing the whole field, diffraction blur is a different
>'shape' than defocus blur. That is, the spot profile or point spread
>function of diffraction is not the same shape as defocus, and gives a
>somewhat different effect. To many people, the effect of diffraction is
>not the same as defocus.
Yes, when they are of similar size. Defocus blur is theoretically disc-
shaped, while diffraction blur is the shape often described as the Airy disc -
closer to a Gaussian in intensity profile than a disc. At the same diameter,
the Airy disc will look less unsharp.
But stop down enough, and the diffraction blur will completely swamp
everything else.
In a view camera, you can afford to stop down well beyond the optimum
aperture of the lens, because even relatively large diffraction blur
won't get enlarged very much. In a typical digicam, stopping down very
much beyond optimum makes the diffraction blur visible because the
image is enlarged so much in printing. So the small image format giveth
extra depth of field at the same aperture, but taketh away some of the
aperture range.
Dave
Steve West
>
>>
>
> Right. I would claim that current lenses probably aren't
> diffraction limited. In fact, I would think that chromatic
> abberation and other lens-related distortions are a bigger
> problem than diffraction. After all: most complete digicam
> systems cost less than a single 35mm SLR lens! The manufacturer
> has to be taking short-cuts somewhere and from what I've seen,
> it's usually in the design of the lens!
>
>
That's what I would have thought too. However, quite some time ago, a
couple of us got our cp 950s and 990s together. We shot resolution
targets. I was expecting that lower f/ratios would actually have worse
resolution than higher f ratios. My reasoning was that aberrations
would grow faster than diffraction size shrunk. I was wrong. The
faster the beam, the sharper the images. So I'd say that diffraction is
still the dominate effect, but that doesn't necessarily mean they are
diffraction limited... However, they appear to be pretty darned good.
swest
Arjen van Andel
any
further? ;o)
I've made a series of testshots with my Fuji Finepix 6900 Zoom. Manual
focus, aperture and WB. F2.8-11. The results are not exactly what your
explanation led me to expect. The 2.8 end was overexposed because the camera
has 1/2000 sec, but not in manual mode (duh!?), so that might have caused
some extra blurring. Perhaps the camera was not focussed well enough. I had
to do a number of series before I got some amount sharpness at 2.8.
Autofocus or manual, eitherway I could not make out if the focus was spot on
through the viewfinder or LCD. But at a mile's distance, that really
shouldn't have mattered; it should not be so difficult to focus such a far
removed object?
Except from the overexposed images, all images (from F4.5-F11) have more or
less the same amount sharpness. I've uploaded an example to my gallery at
http://www.freehomepages.com/avaphoto/ under the heading Miscellaneous. Keep
in mind that the camera was set to soft and apart from some darkening for
the overexposed shots no editing was done.
Dave, can you explain why I can't see any apparent trace of diffraction in
these shots? I did another example on a closer range (1 foot) and couldn't
realy see all that much difference between F2.8 and F11. The F11 was
somewhat fuzzier than F2.8, and the middle values were a tad sharper but
only at an onscreen magnification large enough to cover four 21" monitors.
Arjen.
Dave Martindale
>All this theory is ok, but how about a real-live test to trouble the waters
>any
>further? ;o)
Excellent! To paraphrase someone else, "One well-performed experiment is
worth a thousand expert opinions".
>I've made a series of testshots with my Fuji Finepix 6900 Zoom. Manual
>focus, aperture and WB. F2.8-11. The results are not exactly what your
>explanation led me to expect.
Well, first let's see what we should expect. The 6900 is somewhat
weird because of its sensor with the pixel rows tilted at 45 degrees.
It's a 3.3 megapixel sensor that produces 6 megapixel images. In
theory, the maximum horizontal and vertical resolution could be the
same as a 6 Mpixel conventionally-oriented sensor, while the diagonal
resolution is only equivalent to a 3 Mp sensor. But, in these images,
most of the more visible fine detail is either horizontal or vertical
(the buildings) so we should be able to treat it as basically a 6 Mp
camera for these tests.
Now, it's described as a "1/1.7 inch" CCD, meaning the horizontal width
of the sensor's active area is about 7.5 mm. There are 2832
(interpolated) pixels across that width, so the (virtual) pixel pitch
is 2.66 um. (The actual pixel pitch will be something like 3.75 um).
So the highest resolution we can theoretically get from the *sensor* is
one line pair per 5.32 um, or 188 lp/mm. From one of my previous
articles, an aperture of f/8 has a limiting resolution of 186 lp/mm for
550 nm green light - this is almost the same.
In other words, for apertures smaller than f/8, we should expect the
image to be visibly less sharp than the sensor can resolve, with the
loss of spatial resolution being proportional to the aperture. So at
f/11 we should have 1/sqrt(2) as much resolution, and at f/16 we should
have half the resolution - easily visible. At apertures *larger* than
f/8, we wouldn't expect to see the image get any better, because now
the CCD is the limiting factor.
Now, those are theoretical calculations. In practice, of course,
diffraction isn't the only cause of image unsharpness - lenses have
aberrations. With typical lens designs, the lens is not terribly sharp
wide open due to aberrations. These improve rapidly as the lens is
stopped down through 2 or 3 stops. At that point, aberrations get
smaller than diffraction blur, and as you stop the lens down further
the image gets less sharp again. of the range the lens is unsharp
because of aberrations
There's another factor too: The image processing done in the camera to
rotate the image, from 45 degree diagonal pixel rows to
horizontal/vertical pixel rows, will likely lose a bit of information.
So the CCD-resolution-limited images at apertures wider than f/8 are
probably not as sharp as theory says, which means that the loss in
sharpness at f/8 and smaller may not be obvious until you get quite
a bit smaller than f/8.
So, not having looked at your pictures at all, and without any specific
information about your camera's lens, I would predict that:
- Wide open (f/2.8 in this case) it will be noticeably less sharp. As
you stop it down, it will get sharper.
- At some point, the lens will be sharper than the CCD can resolve, so for
some range of f/stops the digital image sharpness will remain the same.
- At around f/8, the image will become less sharp, but you may not be
able to see the difference until f/11. At f/16 you should definitely
see the loss of sharpness.
>Autofocus or manual, eitherway I could not make out if the focus was spot on
>through the viewfinder or LCD. But at a mile's distance, that really
>shouldn't have mattered; it should not be so difficult to focus such a far
>removed object?
Well, on an old manual SLR you'd just set the lens to the infinity stop
and trust that the stop was accurately set at the factory. But with
autofocus lenses, the manufacturer probably puts the stop somewhere *beyond*
infinity and depends on the autofocus to focus the lens. And yes, the
adjusment is pretty critcal if you want decent sharpness at f/2.8, even for
objects at infinity.
>Except from the overexposed images, all images (from F4.5-F11) have more or
>less the same amount sharpness. I've uploaded an example to my gallery at
>http://www.freehomepages.com/avaphoto/ under the heading Miscellaneous. Keep
>in mind that the camera was set to soft and apart from some darkening for
>the overexposed shots no editing was done.
>Dave, can you explain why I can't see any apparent trace of diffraction in
>these shots? I did another example on a closer range (1 foot) and couldn't
>realy see all that much difference between F2.8 and F11. The F11 was
>somewhat fuzzier than F2.8, and the middle values were a tad sharper but
>only at an onscreen magnification large enough to cover four 21" monitors.
Well, see how your results compare to the predictions above.
Unfortunately, the lens doesn't go to f/16, where I predicted clearly
visible loss of sharpness. But f/11 should be at least a bit less sharp
than f/8, and I think it is. (It's easier to compar f/11 and f/7 since
they're adjacent). In fact, I think f/11 is clearly less sharp than f/7.
Opening up, f/5.6 is sharper than f/7, and in fact f/5.6 seems to be
the best of these images. Above that, f/4 is clearly worse than f/11,
and f/2.8 much worse yet. Now, some of that may be due to the overexposure
as you mention - but f/4.5 is worse than f/5.6, and f/2.8 is worse than f/4,
so aberrations are pretty clearly involved in the worsening image too.
So, I'd say you *can* see the effects of diffration at f/11, but they're
still mild, and Fuji prevented you from trying f/16 where they would be
considerably more visible. But when looking for diffraction, you should
be comparing a mid-aperture (say f/5.6) where the CCD is likely the limit
to f/11. At f/2.8, the lens is unsharp because of aberrations, though
diffraction is minimized.
Now, as you point out, you really don't need all this sharpness for on-screen
display. But if you took these images and made good prints from them, you
should be able to see the difference between f/11 and f/5.6, and between
those and f/8. The 188 lp/mm resolution limit of the best of these images
(due to the CCD) drops to 5.6 lp/mm on an 8x10 print, which is about the
middle of the range of what's considered sharp. So you don't really have
excess resolution for printing 8x10 or larger. (By good print, I mean
a continuous-tone printer, not an inkjet or laser printer).
The executive summary: If you care about maximum sharpness, shoot at
f/5.6. Diffraction isn't terribly significant until near f/11, and the
manufacturer doesn't let you go beyond that. But if the camera *did*
have smaller apertures, you would see the image get noticeably worse -
that's one reason why the camera doesn't have f/16 or f/32. And avoid
wide open whenever possible - true for almost any lens of any focal length
for any film format, except some special copy and macro lenses.
Dave
radiojohn
> Except from the overexposed images, all images (from F4.5-F11) have more
or
> less the same amount sharpness.
I don't think you get it yet. There is no f/16 or f/22 on your camera
because THOSE smaller apertures would show diffraction softness.
Testing the larger apertures proves nothing.
John
Don Stauffer
for common easy stuff. However, if I want anything special, like lots
of depth of field, I go to the film camera and scanner.
In particular, I do almost all macro photography on the SLR film
camera. I can see focus easier, and no parallax in viewfinder. Yeah, I
can use the LCD screen as a viewfinder on my Oly, but it is so much
easier with SLR camara, especially since it has depth-of-field preview.
--
Arjen van Andel
> weird because of its sensor with the pixel rows tilted at 45 degrees.
> It's a 3.3 megapixel sensor that produces 6 megapixel images.
Actually, it's the software that creates those extra pixels, not the CCD.
> In theory, the maximum horizontal and vertical resolution could be the
> same as a 6 Mpixel conventionally-oriented sensor, while the diagonal
> resolution is only equivalent to a 3 Mp sensor. But, in these images,
> most of the more visible fine detail is either horizontal or vertical
> (the buildings) so we should be able to treat it as basically a 6 Mp
> camera for these tests.
I don't feel that moving the pixels a few um (thats basically what they did.
Although the rows are at an 45 degree angle, they form also horizontal
rows with a small offset every other row) would suddenly create a CCD with
much higher
resolution than another chip with the same amount of pixels.
> Now, it's described as a "1/1.7 inch" CCD, meaning the horizontal width
> of the sensor's active area is about 7.5 mm. There are 2832
> (interpolated) pixels across that width, so the (virtual) pixel pitch
> is 2.66 um. (The actual pixel pitch will be something like 3.75 um).
I forgot to mention that the pics have 3 Mpixels for 2048x1636 resolution;
so no upsampling has taken place. That might invalidate some of your
calculations. I also forgot to tell that the 'test array' on my gallery is a
printscreen from a 21" monitor dumped into jpeg; which may not interfere
with your explanation at all. ;o)
> So the highest resolution we can theoretically get from the *sensor* is
> one line pair per 5.32 um, or 188 lp/mm. From one of my previous
> articles, an aperture of f/8 has a limiting resolution of 186 lp/mm for
> 550 nm green light - this is almost the same.
>
> In other words, for apertures smaller than f/8, we should expect the
> image to be visibly less sharp than the sensor can resolve, with the
> loss of spatial resolution being proportional to the aperture. So at
> f/11 we should have 1/sqrt(2) as much resolution, and at f/16 we should
> have half the resolution - easily visible. At apertures *larger* than
> f/8, we wouldn't expect to see the image get any better, because now
> the CCD is the limiting factor.
>
> Now, those are theoretical calculations. In practice, of course,
> diffraction isn't the only cause of image unsharpness - lenses have
> aberrations. With typical lens designs, the lens is not terribly sharp
> wide open due to aberrations. These improve rapidly as the lens is
> stopped down through 2 or 3 stops. At that point, aberrations get
> smaller than diffraction blur, and as you stop the lens down further
> the image gets less sharp again.
something missing, something missing
>of the range the lens is unsharp
> because of aberrations
>
> There's another factor too: The image processing done in the camera to
> rotate the image, from 45 degree diagonal pixel rows to
> horizontal/vertical pixel rows, will likely lose a bit of information.
> So the CCD-resolution-limited images at apertures wider than f/8 are
> probably not as sharp as theory says, which means that the loss in
> sharpness at f/8 and smaller may not be obvious until you get quite
> a bit smaller than f/8.
I don't think that the exact positions or layout will influence the amount
of information, there are just somany pixels per square mm. And don't forget
that all CCD's are interpolated to get the color information for each
defined pixel.
Yes, F7 is clearly sharper. I've made some prints (A4 and A6) and that
doesn't change the outcome. I've also made some shots on the 6 Mpixel
resolution on F2.8, 5 and 11. Same difference. F2.8 is the least sharp, F5
is sharpest and F11 is sharper than F2.8.
I always try to stay away from both 'ends' of the f-scale, but with only ISO
100 (200 and 400 become grainy) there's not much oa a choice, really. This
test seems to explain why I had so many unsharp pictures. I must try harder
to use at least F4. And I would, but since Fuji forgot to put in a
stabilizer...
Thanks Dave,
Arjen.
Dave Martindale
>Actually, it's the software that creates those extra pixels, not the CCD.
Of course. But the tilted sensor *is* capable of higher resolution in
the horizontal and vertical direction than a plain 3.3 Mpixel sensor,
and a 6 Mpixel output file is necessary to show all that resolution.
>> In theory, the maximum horizontal and vertical resolution could be the
>> same as a 6 Mpixel conventionally-oriented sensor, while the diagonal
>> resolution is only equivalent to a 3 Mp sensor. But, in these images,
>> most of the more visible fine detail is either horizontal or vertical
>> (the buildings) so we should be able to treat it as basically a 6 Mp
>> camera for these tests.
>I don't feel that moving the pixels a few um (thats basically what they did.
>Although the rows are at an 45 degree angle, they form also horizontal
>rows with a small offset every other row) would suddenly create a CCD with
>much higher
>resolution than another chip with the same amount of pixels.
What it does is "redistribute" the resolution.
With a normal CCD, with the sensors arranged in simple rows and
columns, the resolution for diagonal detail is higher than for
horizontal and vertical detail. For example, if the CCD had 5 um pixel
pitch, the greatest possible horizontal or vertical resolution is one
line pair per 10 um, or 100 lp/mm. But if the image is composed of 45
degree diagonal lines, the sensor has higher resolution. If you look
at the sensor pattern rotated 45 degrees, you'll see that there are
lines of sensors at that angle too, and the lines are spaced only 3.5
um apart. The sensors *within* each diagonal line are spaced 7.1 um
apart, but the *lines* are spaced 3.5 um apart. And that means that
the CCD can actually resolve up to 140 lp/mm in the diagonal direction
- a factor of sqrt(2) higher than for horizontal or vertical detail.
If an image is reproduced at a size where the horizontal and vertical
resolution of the image matches the human eye's resolution limit, the
extra diagonal resolution is wasted. The human eye is *less* sensitive
to diagonal information than horzontal or vertical. So, one way to
make better use of the pixels you have is to rotate the pixel grid 45
degrees in both camera and display, aligning the direction of highest
resolution of the imaging system with the direction of highest
resolution of the eye.
Unfortunately, the image output devices we use all have
horizontal/vertical rasters, for a number of good reasons. And in
something like a CRT display, the "extra" resolution isnt' really
wasted, because the CRT resolution (about 4 pixel per mm, or 2 lp/mm)
is well below the limit of what the eye can see. But when printing at
400 pixels per inch, which can resolve 8 lp/mm, it would be useful to
have the highest resolution in the horizontal/vertical orientation.
So what Fuji does is use a rectangular-grid sensor oriented at 45
degrees. The pixel pitch should be something like 3.75 um. *If* this
sensor was used in the normal orientation, you'd expect the resolution
limit of the sensor to be about 133 lp/mm horizontal/vertical and about
188 lp/mm for diagonal features. By rotating the sensor, it can now
resolve 188 lp/mm horizontal/vertical and only 133 lp/mm diagonal.
Effectively, the horizontal and vertical pixel row spacing has become
about 2.66 um, while the diagonal pixel spacing is 3.75 um. (Within
one row or column, the pixel spacing is 5.3 um, but the rows or columns
are only 2.66 um apart). Effectively, they've obtained the
horizontal/vertical resolution of a 6 Mpixel sensor oriented
conventionally, but using only the pixel density of a 3 Mpixel sensor,
and with only half as much data to read from the sensor.
Now, there's no free lunch. The image you get from the rotated sensor
has a resolution limit of 188 lp/mm horizontal and 133 lp/mm diagonal.
This *is* better than a conventional 3.3 Mpixel camera, which gives you
133 lp/mm horizontal and 188 lp/mm diagonal, because the eye is more
sensitive to H/V detail. But it's *not* as good as a true 6 Mpixel
sensor, which would give you 188 lp/mm horizontal/vertical and 266
lp/mm diagonal. The Fuji gives you equal H/V resolution, but only
*half* the diagonal resolution. It is not a 6 Mpixel camera.
But when the Fuji image is written to a *file*, the file implicitly has
the normal row/column raster, not a 45 degree rotated raster. In order
to preserve the 188 lp/mm horizontal resolution, the effective pixel
pitch needs to be 2.66 um, so the output file must use interpolated
pixels in order to actually contain the resolution that the sensor
captured. And that means a 6 Mpixel output file.
>I forgot to mention that the pics have 3 Mpixels for 2048x1636 resolution;
>so no upsampling has taken place. That might invalidate some of your
>calculations. I also forgot to tell that the 'test array' on my gallery is a
>printscreen from a 21" monitor dumped into jpeg; which may not interfere
>with your explanation at all. ;o)
Oops. That's bad. The trouble is that, at 2048x1636, you'd expect the
Fuji to look *worse* than a normal 3.3 Mpixel camera. You don't
actually get the "extra" output resolution that the rotated CCD gives
you, because the output file doesn't allow it. You don't get the
higher diagonal resolution that a normally-oriented 3 Mpixel image has,
because the Fuji CCD doesn't capture it. So you're not seeing the best
that the camera can do. In addition, at 2048x1636, the output pixel
grid is *not* aligned with the grid of sensors on the CCD, so there may
be additional interpolation losses due to that.
If you redid the test at the camera's maximum resolution, you'e probably
see more difference between f/11 and f/5.6
>something missing, something missing
>>of the range the lens is unsharp
>> because of aberrations
Actually, nothing missing. Those two lines were just accidentally left
in from a previous version of that paragraph - they should have been
deleted.
>I don't think that the exact positions or layout will influence the amount
>of information, there are just somany pixels per square mm. And don't forget
>that all CCD's are interpolated to get the color information for each
>defined pixel.
It doesn't influence the amount of information, but the rotated sensor
redistributes the information. The 6 Mpixel output file is necessary
to preserve the horizontal/vertical resolution that the sensor captures,
even though the *file* has unnecessary extra resolution in the diagonal
direction. It's an unavoidable consequence of switching from a 45 degree
raster in the sensor to a H/V raster in the file.
>Yes, F7 is clearly sharper. I've made some prints (A4 and A6) and that
>doesn't change the outcome. I've also made some shots on the 6 Mpixel
>resolution on F2.8, 5 and 11. Same difference. F2.8 is the least sharp, F5
>is sharpest and F11 is sharper than F2.8.
It would be interesting to see them. The difference between f/11 and f/5
is probably all diffraction. The difference between f/5 and f/2.8 is
probably all lens aberration.
>I always try to stay away from both 'ends' of the f-scale, but with only ISO
>100 (200 and 400 become grainy) there's not much oa a choice, really. This
>test seems to explain why I had so many unsharp pictures. I must try harder
>to use at least F4. And I would, but since Fuji forgot to put in a
>stabilizer...
Yeah, with a sensor this small, there's not much aperture range of high
sharpness. Diffraction becomes important about 4 stops earlier than with
a 35 mm camera lens, while the wide end isn't any faster. It's due to
unavoidable physics of light. The only way around it is to make the
sensors bigger, and make the spacing between pixels larger. Someday that
will happen, I hope.
Dave
radiojohn
Digicam optics are a lot like 8mm movie cameras. You can practically be in
focus from 3 feet to infinity with many of the wide angle models.
John
Dave Martindale
>In particular, I do almost all macro photography on the SLR film
>camera. I can see focus easier, and no parallax in viewfinder. Yeah, I
>can use the LCD screen as a viewfinder on my Oly, but it is so much
>easier with SLR camara, especially since it has depth-of-field preview.
Yeah. One of the nice things about an SLR is that the viewfinder
contains all the information that your eye can make use of - your eye
is the resolution limiting part of the system. Digicams with direct-view
LCDs and LCD viewfinders are operating well below the resolution of the
CCD sensor itself, and just don't give you any idea of the amount of
detail in the final image - nor even a very good idea if it's in focus.
It seems you have to take the image, then zoom in and pan around the
image in review mode before you know what you're going to get.
A real SLR digicam like the E-10 ought to get around this problem,
but I've never looked through one.
Someday maybe will get a 4 Mpixel viewfinder on a 4 Mpixel camera.
But I'm not holding my breath.
Dave
Arjen van Andel
Your explanation is the first one I have seen about Fuji's tilted design.
Even Fuji doesn't seem to be telling why (they think) their design is better
than the conventional H/V layout. Their explanation (what I have seen from
them) was that this design would allow for bigger cells and shorter
pathways. The latter would make it faster to scan. And indeed their
Super-CCD is fast; bursts of four to five 6 Mpixel images per second is
unseen in prosumer digital camera's. Pity that flash media is as slow as it
is; after the burst it takes a good 20 seconds to save the five shots.
I never gave this specific design much thought, and as I said, Fuji doesn't
seem to share this insight with their consumers. Even Phil Askey <SP?>
apparently didn't have this information and stated there would be no more
than 3.3 Mp detail in the 6 Mp images. I saw one review somewhere else that
stated they had actually measured more detail than an ordinary 3.3 Mp CCD
should give, but offered no explanation why. Even Phil stated something like
"funny, but a 6 MP file downsampled in PS to 3 Mp looks better than the
original 3.3 file from the camera". When I read that I knew there must be
something 'fishy' going on, or else the pics wouldn't have looked better.
Your story explanes that, and also why the 6MP mode works good for
achitecture as Phil mentioned; lots of horizontal and vertical data there.
After reading your enlighting explanation I see that there must be some
resolution gain in the tilted design, but only you seem to know about it (or
took the time to delve into the matter). ;o)
So I should do more or less what I was doing when I just got the camera, and
use the 6 Mp mode again. Even with compression set to Normal instead of Fine
(to limit file size; no microdrive here you know) I might gain some
resolution over the 3.3 Mp file. I stopped using the 6 Mp mode more or less
because 'everyone' was telling it basically wasn't any good. But the 3.3 MP
mode does not result in very sharp images, no matter what I do. One of my
brothers has a Nikon 990 and I allways felt his images were sharper. I have
older 6 Mp macro shots that are sharper than what I'm getting now. If you're
correct that the 3.3 mode loses some definition than that's the answer to
that.
Mental note: Use 6Mp and F5.6 for best results.
And also use TIFF format. But with the 17.7 MByte filesize only 3 will fit
on a 64 MB card...
> It would be interesting to see them. The difference between f/11 and f/5
> is probably all diffraction. The difference between f/5 and f/2.8 is
> probably all lens aberration.
I will try (weather permitting) to make the same series in 6Mp and upload
them. I'll notify you when I did; the weather forecast looks promissing for
sunday. But now it rains and I hear thunder. And somebody is taking
nightshots outside; I can see the flash. ;o)
Arjen.
Dave Martindale
>Your explanation is the first one I have seen about Fuji's tilted design.
>Even Fuji doesn't seem to be telling why (they think) their design is better
>than the conventional H/V layout. Their explanation (what I have seen from
>them) was that this design would allow for bigger cells and shorter
>pathways. The latter would make it faster to scan. And indeed their
>Super-CCD is fast; bursts of four to five 6 Mpixel images per second is
>unseen in prosumer digital camera's.
Well, it allows larger cells (and fewer of them) than a conventional
6 Mpixel sensor - because it's really 3.3 Mpixels.
>I never gave this specific design much thought, and as I said, Fuji doesn't
>seem to share this insight with their consumers. Even Phil Askey <SP?>
>apparently didn't have this information and stated there would be no more
>than 3.3 Mp detail in the 6 Mp images. I saw one review somewhere else that
>stated they had actually measured more detail than an ordinary 3.3 Mp CCD
>should give, but offered no explanation why. Even Phil stated something like
>"funny, but a 6 MP file downsampled in PS to 3 Mp looks better than the
>original 3.3 file from the camera".
It's true that there is "no more than 3.3 Mpixel detail" in the
images. What Fuji has done is to simply reorient the 3.3 Mp worth of
detail so the horizontal/vertical resolution is better, and the
diagonal resolution is worse, than a conventional 3.3 Mp sensor. If
(and this is an important qualifier) you are looking at the image at a
scale where your eye *can* see the extra horizontal detail provided by
the Fuji sensor, but *cannot* see the extra diagonal detail provided by
a conventional sensor, then the Fuji should look sharper than a
conventional 3.3 Mp sensor.
My own initial reaction when the camera was released was that it must
be all smoke and mirrors. But then I looked at the Fourier transform
of a square-grid sensor, and saw that the resolution *was* higher in
the diagonal direction. I already knew from somewhere that the eye is
more sensitive to horizontal/vertical detail, so it became apparent why
the Fuji design *ought* to be useful in theory. And I have no doubt
that the Fuji engineers know all this, but the marketing people don't
understand it.
>When I read that I knew there must be
>something 'fishy' going on, or else the pics wouldn't have looked better.
>Your story explanes that, and also why the 6MP mode works good for
>achitecture as Phil mentioned; lots of horizontal and vertical data there.
For those subjects it makes especially good sense to provide the highest
resolution in the H/V directions.
>So I should do more or less what I was doing when I just got the camera, and
>use the 6 Mp mode again. Even with compression set to Normal instead of Fine
>(to limit file size; no microdrive here you know) I might gain some
>resolution over the 3.3 Mp file. I stopped using the 6 Mp mode more or less
>because 'everyone' was telling it basically wasn't any good. But the 3.3 MP
>mode does not result in very sharp images, no matter what I do. One of my
>brothers has a Nikon 990 and I allways felt his images were sharper.
Yeah, a 3 Mpixel image from the Fuji is in some sense the worst mode to
use. You don't get the extra horizontal/vertical resolution of the
rotated sensor, because the output file doesn't have the pixel density
to support it. You *also* don't get the increased diagonal resolution
of a normal 3.3 Mp sensor, because the rotated Fuji sensor doesn't
provide it - diagonal is its poorest direction. And there's more image
processing necessary to get the image you do get.
For smaller images (2 Mp and below), all 3 Mp cameras should have about
the same resolution. At 3 Mp image size, the conventional array has an
advantage over the 3 Mp rotated array. It's *only* at the 6 Mp image
size that the Fuji arrangement has any advantage over a conventional
3Mp camera.
And, of course, a 6 Mp non-rotated sensor would beat the Fuji at all
orientations - but that's a much more expensive camera. It would be
interesting comparing the 3.3 Mp Fuji with a 4 Mp conventional camera
like the G2. In theory, the Fuji could still be better than the G2 for
horizontal/vertical resolution, though its diagonal resolution would be
much less.
>And also use TIFF format. But with the 17.7 MByte filesize only 3 will fit
>on a 64 MB card...
Yeah, that's the problem with TIFF. The colour interpolation is done
in the camera, giving a file that's 3 times the size of the actual
data. The Canon RAW format makes much more sense for conveying data
from camera to computer. (Biased opinion warning: I have a G2, and
this is one of the reasons why).
Dave
Bernard Hill
<da...@cs.ubc.ca> writes
>And I have no doubt
>that the Fuji engineers know all this, but the marketing people don't
>understand it.
Ha! A very familiar story in my experience!
Bernard Hill
Selkirk, Scotland
Arjen van Andel
> is probably all diffraction. The difference between f/5 and f/2.8 is
> probably all lens aberration.
Dave,
I've uploaded a replacement taken with 6 Mpixel filesize at
http://www.freehomepages.com/avaphoto/.
I could not get a perfect focus on F2.8, just as before. So this time I
tried another approach and focussed on an object nearby at roughly 20 meters
and used that setting for the silo a mile away. That seems to have a
dramatic improvement on the sharpness at F2.8.
I couldn't test if the increased sharpnes at F2.8 was caused by the absence
of direct sunlight (no overexposure by several stops on the white silo as in
the previous test). Note that I overexposed these pics by one stop to add a
little brightness to the scene.
Since this (mine only or this model) camera clearly cannot focus accurately
on objects far away, sharpness of ALL the fragments in the earlier test are
compromised, and therefore no final conclusions can be taken on them.
It's hard to determine which setting produces the sharpest image (but your
eyes may be better than mine). Above F8/9
sharpness drops visibly. F10/11 are clearly suffering from diffraction.
Arjen.