At 12:38 PM +0000 9/23/03, olympus-digest wrote:
>Date: Tue, 23 Sep 2003 09:05:50 +0100
>From: "Julian Davies" <julian_davies@xxxxxxxxxxxxxx>
>Subject: Re: [OM] Nope, It missed it by few hundred miles
>
>I accept the radius of one wavelength is the theoretical limit in this case.
>I don't recall the maths, but it sounds right.
It's actually the diameter, but the edges are not sharp either, so deciding the
diameter is a bit arbitrary.
>The other effects are either reductions of resolution below the theoretical
>limit, or not relevant.
Huh? It's the reduction in resolution that's critical. The point I was making
is that in real lenses, the photons cannot all pass through the same
mathematical point, but instead pass randomly through a disk of about
ten-micron diameter. The reason that all photographic lenses have on-film
resolution not finer than five or ten microns is that this is the resolution
limit of most films of reasonable sensitivity, so there was no reason to make
the lenses better than that.
>The original point is about how wide you have to make the circle of
>probability to ensure that the image is recorded correctly. Widening this
>circle is reducing resolution. Or introducing blur if you prefer.
>Geometric aberrations (pincushion, barrel etc) are irrelevant, as they apply
>equally to all photons destined for the same image point. Blurring
>aberrations (coma, spherical etc) are relevant, but equate directly to a
>reduction of resolution. It's just that in these cases it's not uniform.
We don't "introduce blur", we live with it. The blur can be reduced, but
cannot be eliminated. One can make diffraction-limited lenses, such as those
used for making integrated circuits, but they are very specialized and
expensive. Nor is there an economic rationale to making photo lenses that
good, as film of reasonable sensitivity has far worse resolution than a
wavelength of light. Green light has a wavelength of about 0.5 micron, so if
the focal spot is 5 microns in diameter, it's ten wavelengths across. Film
used to make holograms has resolution of order a wavelength of light, but is
*very* insensitive. (For example, one film rated at 3000 lines (pairs?) per
millimeter requires exposure of 80 to 100 microjoules per square centimeter at
633 nanometers (HeNe red). This is very bright. If I get the energy, I could
figure out the ASA, but I recall it's something like 0.1 ASA.)
>The original point of all this was to say that photographic lenses are
>designed as part of a system, which includes the abilities of the image
>recorder in the parameters.
True.
>We are used to thinking that all limitations (except grain) on our images
>are introduced as failings of the lens design. I have tried to show that the
>film itself introduces a requirement to limit the lens design, and that in
>this requirement is not something which has suddenly appeared with digital
>capture devices replacing film. At the end of the day, if you design a
>perfect lens, but the perfectly resolved edges are lost in the film as an
>"intermediate grey" of lower edge contrast, did the result get better or
>worse?
This isn't quite right. With film, with its random grain structure, there is
no such thing as aliasing, and if one uses a lens too sharp for the film,
nothing bad happens. The overall resolution is limited by the film, but there
are no odd effects or artifacts (like aliasing - moire fringes). If CCDs had
randomly placed pixels, aliasing would not be a problem, but with current
technology, random pixel placement is too difficult, so it isn't done. It's
easier to make the pixel spacing smaller than the resolution of the lens.
>I was also trying to show that the measured level of performance of current
>lenses for film is not universally the result of cost - cutting, but that in
>addition to a Cost - benefit question and law of diminishing returns etc
>(which tends to produce niche products at high cost), there is an "add
>cost - reduce benefit" point beyond which no sensible person would go, and
>explains the absence of those niche products in our market.
>
>This has produced an accepted "envelope of system performance" in which it
>is the film manufacturers who control the right hand edge and the camera
>manufacturers who control the costs within the envelope. Since the envelope
>is already pretty large, there's plenty of scope for brand differentiation
>in there.
>Digital changes that in two ways:
>Film manufacturers will become radically more selling - price sensitive, and
>will stop pushing the right - hand edge of the envelope. After all they are
>now competing with "free - at - source" recording.
Well, we will see action at both ends. Some manufacturers will flee up-market
as well.
>Camera manufacturers will control the system as a whole, and (starting with
>Oly!!!) begin to produce optimised combinations which eventually will
>surpass film in ability. When the E5? arrives with the 11Mp sensor, I expect
>it to be a quantum leap beyond the C*n*n in the appearance of the images.
>Maybe then I'll go digital.
I'd give it a few years.
Joe Gwinn
>Julian
>- ----- Original Message -----
>From: "Joe Gwinn" <joegwinn@xxxxxxxxxxx>
>To: <olympus@xxxxxxxxxxxxxxx>
>Sent: Tuesday, September 23, 2003 3:25 AM
>Subject: Re: [OM] Nope, It missed it by few hundred miles
>
>
> > The ideal lens of geometric optics does that, but no real lens follows
>this ideal, for at least two major reasons. First, all practical lenses
>suffer from abberations, so even in the geometric theory of image formation,
>the light rays don't quite pass through the same mathematical point in
>space. Close, a matter of 5 or 10 microns, but not exact. Second, the wave
>nature of light prevents focal spots from being much smaller than a
>wavelength of light in diameter, about 0.5 micron. Not that any
>photographic lens is diffraction-limited, so the abberations dominate.
< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >
|