Sorry folks this is a long posting, but I hope it helps the discussion.
Garth Wood wrote:
"Well, perhaps Oly wasn't so wrong after all. A friend of mine who is a
professional commercial photographer recently
invested in a Fuji FinePix S1 body (the one that takes Nikon F-mount
lenses) to augment his film work, and to preserve
his investment in Nikons (although, like all good addicts, he used the
purchase to also justify acquiring several
extra Nikon lenses... 8^> ). While the digital camera has fast become
a favourite of his for ultra-high-speed
turnaround for clients, he has noted an interesting problem with the
resulting image files which has been noted -- and
complained about! -- on several Usenet news groups, but which to date
has received no official confirmation from Fuji.
The problem, in a nutshell, is non-perpendicular light rays striking the
CCD sensor!
The symptom is that, especially in photos where there is a large area of
smooth, even gradations of colour (like, say,
when one does a product shoot against a specified background), the
resulting images show a decided colour shift from
the centre of the image out to the edges."
This has come up several times regarding the use of 35mm lenses on
digital cameras.
As an OM user and a physicist I've been interested to look further into
the problem. I additionally have the advantage of working for a major UK
government research centre, in which I'm surrounded by many people with
deep expertise in these sorts of areas.
First let me say to Garth that I don't doubt his report for a moment or
that his colleague is having some problem with his Fuji Finepix.
But I don't think non perpendicular rays are the problem.
First I was able to talk to a colleague who is developing CCD devices,
and is one of several experts in this field, at our lab.
I suggested that such a color shift effect would be most likely to occur
if the pixel elements used an interference effect to differentiate the
three primary colours into individual pixels. But I doubted that CCDs
used interference filters, more likely a set of dye filters. She was
able to confirm that at least one device being applied to digital
photography made by Matsushita (we don't know if this is the one used by
Fuji), has dye filters, not interference filters. It seems unlikely that
this is what is causing the color shift. We don't know of any other
structure which could be causing colour shifting in a CCD (she designs
and make such devices so she is a reliable source). She suggested that
the reviews of the Fuji Finepix do not mention such effects. It could be
that only this persons sample has this problem.
Could Garth obtain some images showing this problem. The direction of
shift (red to bue or blue to red might help pin it down.
The other possibility is that the Fuji chip has non-rectangular pixels,
and a special algorithms to construct the image. Could this be some
effect of aliasing? Again a sample image would be useful.
The other side of this goes back to some recent mail reffering to a
statement by Olympus that OM system lenses could not be used because of
the non-perpendicular ary effect.
As I understood the statement from olympus, it was that 35mm format
lenses could not be used with CCD chips as the angle of the rays was to
oblique for the imaging elements to detect.
I disagree with this because:
1 I have successfully used a Nikkor 105mm Macro lens with a 600x400
element CCD camera.
2 Video cameras use much shorter lens to film distances with their zoom
lenses which would give more oblique angles than the equivalent 35mm
lenses.
3 For the angle to be limited, the CCD sensor ?well? would need to be
significantly deeper than its in plane dimensions. As the latter is of
the order of 10micron, a similar depth is implied. I do not know of any
reason why the layered structure of the imaging elements would be as
deep as several microns. Deep dry etching techniques are available and
are used for micromachining, but they would not be needed for the
structure of a CCD.
4 The explanation implied that the problem is greater for wide and super
wide angle lenses. It is indeed true that a lens with a focal length
shorter than standard, and therefore sitting closer to the film plane,
would indeed generate rays with shallower angles at the film plane then
the equivalent standard lens. However, SLR wide angle lenses are not of
shorter focal length. They are what is known as a ?retro-focus? design.
This means that they produce an image with a field of view equivalent to
the stated focal length. In terms of lens design, their rear nodal point
N2 lies between the rearmost lens element and the film plane. If the
rearmost element of a wide angle retrofocus design is the same diameter
as the standard lens for that format, then the minimum angle of rays
meeting the film plane would be the same as for the standard lens
design. Looking at my own standard and wide angle lenses, and at the
illustrations in my Zuiko lens manual (IEMZLG 0578 30MT [ca 1978]), I
note that wide angle Zuikos have smaller diameter rear elements than
standard lenses. My 50mm F/1.8 has a rear element of 22.5mm dia., and my
24mm F/2.8 has a rear element of 17mm dia. I estimate that the 18mm
F/3.5 has a rear diameter of 16mm.
Calculating the angle between a ray meeting the film plane at the
furthest corner of the 35mm frame, having emerged from the outermost
edge of the element at a point diametrically opposite. I get the
following angles (with respect to the CCD plane):
Focal length Rear Element Dia. Angle
[mm] [mm] [deg]
50 23 26
24 17 58
18 16 63
Therefore the reverse of the Olympus statement is true; the wide angle
Zuikos would actually work better with a ?deep? CCD element than would
the longer focal lengths, because of the diameter of their rearmost
elements.
That said I would personally welcome Olympus producing a digital version
of the multilens SLR system.
I would expect Olympus to start from scratch with the optimum size of
CCD. It wouldn't make sense to base the system around a 35x24mm CCD.
This would be more expensive than needed. For an equivalent resolution
to a 35mm film system we'd be loooking at something with about 2000x1500
pixels (assuming resolution of 50line pairs/mm for a 35mm lens).
This means a 20mm x 15mm CCD. The equivalent standard lens focal length
for this would be 28mm.
So, we'd be expecting an image plane to lens distance of about 25mm, and
a standard lens length about half that of the 50mm Zuiko. That is a
camera about 60-70mm deep with its standard lens. The standard lens,
being a 28mm focal length could have an aperture of 28mm, and an overall
diameter of say 35-40mm. Even better, its F number would f/1.0. All in a
lens of say 1/4 the weight of the current 50mm Zuiko.
Similar ratios would apply at other focal lengths. Wide angle (retro
focus) lenses of say 14mm focal length could eisily have f/2.0 maximum
apertures for the same size and weight as the standard. Our present neat
little 100mm f/2.8 would be equivalent to a 180mm f/2.8, and so on.
In other words Olympus have a great opportunity to produce a new
interchangable lens camera system for the 21st century. One that
continues the OM tradition of being small, light weight and excellent
quality. I'd be surprised if Mr. Maitani hasn't thought of it already.
Sorry its a bit more than £0.02 worth, but I thought it needed to be
said.
Chris Barrett
Malvern
< 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 >
|