["Followup-To:" header set to rec.photo.digital.slr-systems.]
Sosumi <sosumi.TakeThisOut@home.nl> wrote:
> Is everyone asleep here?
No, everyone else knows Sigma's Foveon chip and their
cameras. I remember a guy which was even a bit more rabidly and
uncompromisingly pro-Foveon as Rita (if that can be
imagined). Just mention it and he'd pop up.
In difference to Rita he actually had a brain (even working quite
well) in other areas, like, oh, intricacies of JPEG encoding.
(As in tweaking such things, not just book knowledge.)
> And more lies:
> http://www.sigmaphoto.com/downloads/SD14_SpecSheet_English.pdf
> RAW images are 2640 x 1760 PIXELS. Yet they claim 14 MP because it can
> produce a 14 MP JPEG.
Well, if you would shift the red, green and blue pixels to be
adjacent instead of superpositioned on top of each other, you
get ... 14MP. Which is true enough for marketing (and more
than good enough for any claims Rita would offer regarding
Nikon superiority.)
Pixels, however, do not necessarily count spatial resolution.[1]
Do a Fourier transfomation of some of your images, and look at
the high frequency content. You'll probably find ~70% (ca.
1/SQRT(2)) of the theoretical range actually having content.
That would be basically the distance between 2 green bayer pixels.
Hence, downscaling the image to half the size (70% lenght and 70%
width) will not cause data loss --- or, the other way round,
the 2640x1760 (4.6 Mpix) Foveon should have about the same
*spatial* resolution as a 3734x2489 (9 MPix) Bayer filter camera.
That assumes no unrecoverable blurring on the Foveon, and standard
AA filtering and Bayer restoration on the Bayer filter.
Hence the Sensor should be compared against 8-10MPix Bayer sensors.
When it comes to comparing red and blue colour spatiality,
the Foveon technology is quite ahead of Bayer technology.
The DP1 *has* 2640 x 1760 red sensors, while a conventional Bayer
sensor would need to have 18.6 MPix for the same amount of red
sensors --- and the same is true for blue sensors. Which should
(theoretically!) make Foveon superior when it comes creating B/W
images using mostly red and blue channel data.
OTOH Foveon has severa problems, like colour seperation and
sensor noise at higher gain, which do counteract these inherent
advantages in many cases.
> Sure, then the Nikon D80 should be called a 30 MP camera, because it's raw
> files can easily reach that with software.
No, there simply are no 30 million measured and recorded values.
(not that some ... Ritas ... would be stopped by such small
things.)
There are, however, 14 million measured and recorded values with
the Sigma.
-Wolfgang
[1] Nor are pixels necessarily square, for that matter!
>> Stay informed about: Sigma lies about DP1 
FAQ
Search
Profile
Private Messages
Log in
