View Single Post
  #32  
Old February 13th 14, 10:38 PM posted to sci.engr.color,sci.image.processing,rec.photo.darkroom,rec.photo.digital,comp.soft-sys.matlab
nospam
external usenet poster
 
Posts: 24,165
Default my take on Kodak downfall

In article , PeterN
wrote:

bayer is the best solution that exists today and will be for the
foreseeable future.

foveon's layered approach has been a disaster.


facts?


oh, where to start.

with a market share of under 1% and shrinking, sigma's cameras have
been a complete disaster. not only do they not sell particularly well,
but they are a money loser for the company.

foveon originally was backed by two venture capital firms, who soon
realized they'd been had and what a ****up it was, so they bailed.

that left foveon, whose only customer was sigma, without any money and
about to file for bankruptcy.

had that happened, sigma would have been completely screwed, so sigma
bought foveon at firesale prices and has been sinking money into it
ever since.

the cameras are utter crap (and yes, i've used them).

they're anything but consistent. two photos in a row might look totally
different, despite the settings being identical. the sd14 was pig slow
(around 6 seconds write time) and you actually had to wait until it
finished, or the camera could lock up and you'd lose photos. the sd14
was *really* buggy and sometimes locked up even if you weren't taking
photos one after another

the sd14 originally was $2000 msrp ($1600 street), which quickly
dropped due to slow sales, and around a year later, it was being sold
off for $300-400, new, and even at that price, people still weren't
buying all that many.

the sd1 originally was pitched at a ridiculous $9700 msrp (~$6000
street) which was complete insanity for a 15 megapixel camera. even the
fanbois were stunned. not surprisingly, sigma did not sell very many
cameras at that price.

as best as i can tell, they sold about 10, total, based on posts in the
sigma forum and serial number analysis (which is encoded in every
photo, btw, so it's easy to track).

after tens of thousands of unsold cameras sat in warehouses, sigma
slashed the price by roughly $4000 in one day, and the price has
dropped even *further* since then.

the dp1/2/3 series have been riddled with lens motor failures, where
the lens just gets stuck extending out or back in, along with all the
usual problems with the sensor.

sigma tries to claim more accurate colour, but the delta-e is much
higher than bayer, which means *less* accurate colour.

sigma claims 'no guessing' of colour data, yet there's more 'guessing'
than bayer because they don't actually capture rgb at every location
(another lie of theirs). the spectra of the layers overlap by quite a
bit and there's a ****load of very complex math to extract rgb from it,
which is one reason why the software is slow and the results are noisy
and with weird colour casts.

the first foveon camera, the 3.4 megapixel sd9, did not sell well, so
they decided to lie about the number of pixels in the sd10 because
'bigger numbers are better', despite the camera having the same sensor.

normally that's called fraud, but somehow, they managed to get away
with it.

sigma's software is slow and buggy and there aren't any viable options
from third parties. even adobe has given up supporting it. part of the
'sigma look' is heavy sharpening. if you set the sigma software to 0
sharpening, you're actually getting a wallop of sharpening. you have to
set it to -1 to -2 to get 'none', depending on version.

foveon sensors are theoretically interesting, but they are riddled with
problems and actually don't offer anything the eye can see anyway. it's
a solution in search of a problem.

foveon sensors have substantially higher noise, lower colour accuracy,
lower resolution, worse high iso performance and higher manufacturing
cost.

that's a huge price to pay for 'full colour', something humans can't
even see anyway.

three layer sensors sound like a good idea on paper, and one day they
might be feasible without significant tradeoffs, but they sure as hell
are not now.

if that day comes, the technology won't need lies to market it. it will
sell itself because it's actually better.

on the other hand, bayer is a very clever design based on how the human
eye works. it's cost-effective to manufacture and works exceptionally
well for creating photos that humans look at.