Re: Sony announce 25Mp 35mm sensor
- On Friday, February 01, 2008 at 10:41, Keith Martin wrote:
> But then, a full-frame sensor camera does mean working with differentThis is what I would suspect people to do with a full frame sensor,
> lenses to get the equivalent effect. So a rough equivalent of the
> 10.5mm would be 16mm, wouldn't it? Something like the old 16mm
> fisheye that I used briefly on my old Canon A1.
yes. But apparently most of them use a 10.5mm or even 8mm lens in
order to need less shots. With a 25MP full frame sensor they would
get roughly the same output resolution as with the same lens on a
10MP crop factor camera.
> Assuming the manufacturing and glass quality was similar, that wouldYes, of course. But you always can go for even higher quality using
> give approximately the same view but reduced chromatic abberation.
longer lenses and more shots...
> (Slightly reduced depth of field too, but that's physics for ya!)There is a frequent misunderstanding about DOF and spherical
panoramas mostly because people use DOF values intended for single
printed images as a comparison. For spherical you have to calculate
> I don't think it is really a matter of being beyond the resolution ofThe corners of a crop factor 1.5 image from a 10.5mm lens are very
> a fisheye, as that's just analog-world optics. The Sigma 8mm and
> Nikon 10.5mm fisheyes are designed to produce acceptable images on a
> cropped-area sensor, and trying to capture an image using a broader
> part of the image means going beyond the design intentions.
close to the outer image circle. Hence if you talk about fisheyes you
can't simply say "designed for...". Anyone should know that close to
the image circle there is a lower resolution, not only due to lens
design flaws but due to the fisheye mapping.
> So... isn't the important thing simply using a lens that is actuallyThe pre-digital Sigma 8mm lenses where meant to cover a full-frame
> meant to cover a full-frame sensor?
sensor. Nevertheless the image quality was bad near the image circle.
You could use it on a full-frame sensor for a 3-around workflow,
where each image contributes about 120° - which is pretty inside the
image circle and (coincidentally!) a crop by 1.5
Same applies if you use a 10.5mm lens on a full-frame. In a 3-around
workflow you more or less use only the parts visible on a 1.5 crop
sensor anyway. The ecxess parts are only necessary because you need
some overlap to find control points.
- Sometime around 1/2/08 (at 13:35 +0100) Erik Krause said:
>close to the image circle there is a lower resolution, notGot it. Although really it is is lower *quality* that we're talking
>only due to lens design flaws but due to the fisheye mapping.
about. Resolution, although related in a sense, means something
slightly different. At least, with digital images it is used to refer
to the sensors and the final pixels.
Thanks for the further info and the DoF link! I was thinking in terms
of individual shots, but that's interesting data on that wiki. Stuff
for me to ponder. :-)
I'm not a rocket scientist but did learn that the delay for the introduction of teh Betterlight new 10k scan back was all about that no lenses resolved what it would do properly.. Some how they did get to a compromise and did release the back.. In my converstaions with them at that time they spoke of the issues related to that level of resolving using optics.. Further we must realize that today optics are multi part.. So it's probably not just a matter of saying make another one that does it.. In the smaller sensors they have been packing more and more pixels.. but as spoken in these posts that has been at a price.. So it makes sense to make larger sensors..so the info that is captured isn't shrunk as much. The optics are actually doing a conversion...making a big image fit on a small sensor.. It appears that what we are experiencing is degradiation when we get to a certain threhshold at out current optical technology. The present answer is larger cameras it seems.. So tommorrows Nikon may look like my Pentax 6x7 with a face lift and a Nikon logo on it ( hopefully a little lighter too)....or a new version of a Sinar 8x10 with a fixed sensor in the back of it.with large pixel sizes... Whooo Hooo.
----- Original Message -----
From: Keith Martin
Sent: Friday, February 01, 2008 3:41 AM
Subject: RE: [PanoToolsNG] Re: Sony announce 25Mp 35mm sensor
Sometime around 1/2/08 (at 01:42 -0800) Paul D. DeRocco said:
> > From: Erik Krause
>> ...then 10MP on a crop factor 1.6 sensor is beyond the resolution of
>> any fisheye, too. 25MP on a full frame sensor has the same absolute
>> resolution (pixel density) like a 10MP sensor at crop faktor 1.6
>Probably. Even on my 6Mp 10D, which is a 1.6x crop sensor, I can see a lot
>of CA near the edges, so it's obvious that I wouldn't be getting any more
>sharpness if I stuck it on my 10Mp 40D. And a 25MP FF sensor would probably
>be even worse, because it reaches into the worst part of the lens.
But then, a full-frame sensor camera does mean working with different
lenses to get the equivalent effect. So a rough equivalent of the
10.5mm would be 16mm, wouldn't it? Something like the old 16mm
fisheye that I used briefly on my old Canon A1. Assuming the
manufacturing and glass quality was similar, that would give
approximately the same view but reduced chromatic abberation.
(Slightly reduced depth of field too, but that's physics for ya!)
I don't think it is really a matter of being beyond the resolution of
a fisheye, as that's just analog-world optics. The Sigma 8mm and
Nikon 10.5mm fisheyes are designed to produce acceptable images on a
cropped-area sensor, and trying to capture an image using a broader
part of the image means going beyond the design intentions. So...
isn't the important thing simply using a lens that is actually meant
to cover a full-frame sensor?
(I think that's what you meant in your first post, but I wasn't sure...)
[Non-text portions of this message have been removed]
- Hey Carel,
So in theory it would be possible to develop a ~4mp image from a 12.8mp
RAW file, is it right? Have you ever tried this? How different in
practice would that be from reducing a 12.8mp picture to a 4mp size with
a good interpolator?
I can hardly imagine a 4mp image that surpasses it's 12.8mp version in
Is this really used in star photography?
> One could use all these pixels in combination with DCRaw in the Super-pixel
> mode to circumvent the Bayer matrix induced artifacts. The main disadvantage
> of the super-pixel method is that you end up with an image that is only
> 1/4th size of the original, so with this sensor you would end up with a
> 6Mpixel image, but more detailed image.
- Fabio Bustamante-2 wrote:
>No, it would not surpass the original sized quality, but might be an
> Hey Carel,
> So in theory it would be possible to develop a ~4mp image from a 12.8mp
> RAW file, is it right? Have you ever tried this? How different in
> practice would that be from reducing a 12.8mp picture to a 4mp size with
> a good interpolator?
> I can hardly imagine a 4mp image that surpasses it's 12.8mp version in
> any way...
> Is this really used in star photography?
interesting way to use this overkill of 24Mpixels for web purposes. I only
have 5D images, but a test is on the way. I will use the super-pixel method
with DCRaw versus ACR. My reasoning was along the lines of Bernhard Vogl,
his complaints about the shortcomings of the Bayer array and his observation
that one retains more detail when downsizing an image that was taken with a
longer lens to the size of the same image taken with a shorter lens. But
maybe that is not a good analogy.
>Is this really used in star photography?The inclusion of this method in DeepSkyStacker would indicate so. When I
asked about this on my recent visit to the Mt Wilson Observatory, it did not
seem to ring a bell, while the recently discussed method of getting a
sharper image by using a slightly misalligned stack of images ("dribbling")
View this message in context: http://www.nabble.com/Sony-announce-25Mp-35mm-sensor-tp15207204p15249800.html
Sent from the PanoToolsNG mailing list archive at Nabble.com.