Camera phone technology article (Photonics-Spectra Nov iss..

Rich wrote:
>
> Now Foveon is going after that market. Nice to see they've found a
> place, but I figured their technology was too costly to implement in
> such a vulgar, cheap imaging environment?

Actually their technology is very cheap to implement in that market.

They cannot compete in P&S or D-SLR cameras because of the noise
problems, color separation problems, and the difficulty of scaling their
technology to high resolution. However for low-resolution cameras such
as the ones used in cell phones, their sensors can be manufactured very
cheaply, using older semiconductor fabs that aren't usable for chips
such as processors. There are lots of those fabs still around, some of
which are being crated up and sent to third-world countries where the
labor is very inexpensive.
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

In article , Mr. Strat
says...

> I'd like to know how they stay in business selling their weird-o
> technology.

What's wrong in having the full colour information in each pixel?
--

Alfred Molon
------------------------------
Olympus 50X0, 7070, 8080, E3X0, E4X0 and E5X0 forum at
http://tech.groups.yahoo.com/group/MyOlympus/
http://myolympus.org/ photo sharing site
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

In article , Alfred
Molon wrote:

> In article , Mr. Strat
> says...
>
> > I'd like to know how they stay in business selling their weird-o
> > technology.
>
> What's wrong in having the full colour information in each pixel?

nothing is wrong with having full colour information at every pixel,
all things being equal. unfortunately, there are a lot of shortcomings
with sigma/foveon's current implementation, such as:

- fewer pixels than with a comparably sized bayer chip (4.6 versus
10-12 megapixels). even the low end nikon d40 and pentax k100d have
more pixels, and at about 1/3rd the price! the additional luminance
resolution of a bayer chip is *much* more important to human vision
than having three samples per pixel.

- noisier pixels due to each layer having a lower well capacity (as
much as 1/8th for the top layer). as a result, high iso (and even
medium iso) is nowhere near as good as with other cameras.

- poor colour separation between the layers (and it isn't rgb, unlike
what the ads want you to believe), requiring a compelx transform to
actually get rgb, and that adds more noise.

- metamerism and strange colour casts that can't easily be removed.

- no anti-alias filter, resulting in alias artifacts and false detail.
this isn't really a flaw of the chip itself, but since the artifacts
don't show up as colour fringing, they can get away without an
anti-alias filter, hoping that people perceive the result as having
more resolution than it actually does.

- deceptive and misleading advertising (it is *not* 14 megapixels).

- more expensive to manufacture.

- currently only available in a single sigma body which has a very
limited feature set, very poor performance and a significantly limited
selection of lenses and accessories compared to any mainstream slr.
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

On Mon, 19 Nov 2007 13:50:32 +0100, Alfred Molon wrote:

>In article , Mr. Strat
>says...
>
>> I'd like to know how they stay in business selling their weird-o
>> technology.
>
>What's wrong in having the full colour information in each pixel?

Nothing, Nikon is trying to do just that with their new sensor design, but in a
different and less efficient way.

Just the sign of another ignorant resident-troll.
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

On Sun, 18 Nov 2007 23:20:27 -0800, nospam wrote:

>In article , Alfred
>Molon wrote:
>
>> In article , Mr. Strat
>> says...
>>
>> > I'd like to know how they stay in business selling their weird-o
>> > technology.
>>
>> What's wrong in having the full colour information in each pixel?
>
>nothing is wrong with having full colour information at every pixel,
>all things being equal. unfortunately, there are a lot of shortcomings
>with sigma/foveon's current implementation, such as:
>
>- fewer pixels than with a comparably sized bayer chip (4.6 versus
>10-12 megapixels). even the low end nikon d40 and pentax k100d have
>more pixels, and at about 1/3rd the price! the additional luminance
>resolution of a bayer chip is *much* more important to human vision
>than having three samples per pixel.

Wow. Let's let idiocy reign supreme, shall we?

First off, 10-12 megapixels on a Bayer sensor equates to 1/4th that number to
define the actual color image resolution. You need 1 red, 1 green, 1 blue, and 1
green pixel to define each of those final RBG display pixels of which you are so
proud. Okay, so now we're actually down to 4 megapixel color resolution.

Whereas 4.6M tri-color pixels equates to 4 times that amount if duplicated in an
RGBG array. or the same amount of data as would be acquired by an 18.4 megapixel
Bayer sensor.

No sense even reading the rest of your brain-barfage after witnessing what you
already wrote.


Really now, if you're going to choose a new newsgroup in which to set up house
to troll in, shouldn't it be one where you can display some meager sign of
knowing the subject? This could be why there's such a large turn-over with you
resident-trolls. No matter what newsgroup you become a resident of your
stupidity and/or ignorance is all too quickly revealed.

Maybe you should stick with some current-events newsgroup, then you need nothing
more than your cable-tv for the extent of all your experience in life.


























>
>- noisier pixels due to each layer having a lower well capacity (as
>much as 1/8th for the top layer). as a result, high iso (and even
>medium iso) is nowhere near as good as with other cameras.
>
>- poor colour separation between the layers (and it isn't rgb, unlike
>what the ads want you to believe), requiring a compelx transform to
>actually get rgb, and that adds more noise.
>
>- metamerism and strange colour casts that can't easily be removed.
>
>- no anti-alias filter, resulting in alias artifacts and false detail.
>this isn't really a flaw of the chip itself, but since the artifacts
>don't show up as colour fringing, they can get away without an
>anti-alias filter, hoping that people perceive the result as having
>more resolution than it actually does.
>
>- deceptive and misleading advertising (it is *not* 14 megapixels).
>
>- more expensive to manufacture.
>
>- currently only available in a single sigma body which has a very
>limited feature set, very poor performance and a significantly limited
>selection of lenses and accessories compared to any mainstream slr.
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

In article , JollyOllie
wrote:

> >- fewer pixels than with a comparably sized bayer chip (4.6 versus
> >10-12 megapixels). even the low end nikon d40 and pentax k100d have
> >more pixels, and at about 1/3rd the price! the additional luminance
> >resolution of a bayer chip is *much* more important to human vision
> >than having three samples per pixel.
>
> Wow. Let's let idiocy reign supreme, shall we?

apparently, based on what you write:

> First off, 10-12 megapixels on a Bayer sensor equates to 1/4th that number to
> define the actual color image resolution. You need 1 red, 1 green, 1 blue,
> and 1
> green pixel to define each of those final RBG display pixels of which you are
> so
> proud. Okay, so now we're actually down to 4 megapixel color resolution.

that's not remotely close to how bayer works.

> Whereas 4.6M tri-color pixels equates to 4 times that amount if duplicated in
> an
> RGBG array. or the same amount of data as would be acquired by an 18.4
> megapixel
> Bayer sensor.

you must be a graduate of the george preddy institute of mathematics.
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

JollyOllie writes:

>>- fewer pixels than with a comparably sized bayer chip (4.6 versus
>>10-12 megapixels). even the low end nikon d40 and pentax k100d have
>>more pixels, and at about 1/3rd the price! the additional luminance
>>resolution of a bayer chip is *much* more important to human vision
>>than having three samples per pixel.

>First off, 10-12 megapixels on a Bayer sensor equates to 1/4th that number to
>define the actual color image resolution. You need 1 red, 1 green, 1 blue, and 1
>green pixel to define each of those final RBG display pixels of which you are so
>proud. Okay, so now we're actually down to 4 megapixel color resolution.

Typical Sigma propaganda. But, in fact, a 12 megapixel Bayer sensor
gives you "12 megapixels worth" of luminance resolution, the same as
you'd get from a 12 megapixel monochrome sensor, and far more real
resolution than a 4 megapixel Foveon sensor (that Foveon would like to
call 12 megapixels, but is not). That's because the Bayer sensor and
the 12 megapixel monochrome sensor measure light intensity at 12 million
different locations in the image plane, while a Foveon sensor with 12
million numbers output measures at only 4 million locations.

Which can easily be verified by looking at actual resolution charts shot
by various Bayer and Foveon-sensor cameras.

Now, it's true that a 12 MP Bayer sensor is equivalent to about 3 MP
when resolving detail that differs in chroma *only* with no luminance
difference, but it's hard to find such detail in a real image unless
it's a test chart specifically designed to measure this. And the lower
resolution doesn't matter for general photography, because the human
eye's colour-only resolution is about 1/10 as good as its luminance
resolution, so a camera that gives chroma that is 1/2 as good as
luminance is actually considerably better than necessary.

>No sense even reading the rest of your brain-barfage after witnessing what you
>already wrote.

If you'd look at actual resolution tests, or check out the history of
Foveon discussions in this group, you wouldn't be rehashing Foveon's
incorrect propaganda.

There *are* applications where you want colour resolution to be as good
as luminance resolution (e.g. blue screen matting), and a Bayer sensor
may cause problems, but Foveon's noise would equally be a problem. In
those applications, some sort of full tricolour sensing is used (3 CCDs,
or a trilinear CCD, or colour-sequential).

Dave
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

In article , Dave Martindale says...

> Typical Sigma propaganda. But, in fact, a 12 megapixel Bayer sensor
> gives you "12 megapixels worth" of luminance resolution,

How could that be the case, when such a Bayer sensor does not measure
luminance at 12 million locations? The green channel may be a good
approximation of luminance, but not the red or blue channels.
--

Alfred Molon
------------------------------
Olympus 50X0, 7070, 8080, E3X0, E4X0 and E5X0 forum at
http://tech.groups.yahoo.com/group/MyOlympus/
http://myolympus.org/ photo sharing site
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

On Tue, 20 Nov 2007 07:02:22 +0100, Alfred Molon wrote:

>In article , Dave Martindale says...
>
>> Typical Sigma propaganda. But, in fact, a 12 megapixel Bayer sensor
>> gives you "12 megapixels worth" of luminance resolution,
>
>How could that be the case, when such a Bayer sensor does not measure
>luminance at 12 million locations? The green channel may be a good
>approximation of luminance, but not the red or blue channels.

This is why there's so much discussion and experiments over the various RAW
interpolation algorithms. That 12 megapixels worth of luminance resolution is
interpolated from the lower-resolution color data, it's not true luminance
resolution. Some RAW to JPG/TIF algorithms when inventing that luminance
resolution do it better than others.

The aforementioned resolution differences that they are arguing about still
stands. A same sized Foveon sensor will always produce 4x's the true color
resolution of any Bayer sensor. The luminance resolution will be interpreted
more accurately from those individual Foveon pixels instead of trying to invent
them out of 4 Bayer pixels by doing guesswork from data spread over a larger
area.

I still hope they get the bugs worked out of that method someday. It's very
promising. Maybe they just need to tackle it from a whole new angle. It'll be
the next great advancement in digital cameras when it happens the right way.
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

In article , Alfred
Molon wrote:

> In article , Dave Martindale says...
>
> > Typical Sigma propaganda. But, in fact, a 12 megapixel Bayer sensor
> > gives you "12 megapixels worth" of luminance resolution,
>
> How could that be the case, when such a Bayer sensor does not measure
> luminance at 12 million locations? The green channel may be a good
> approximation of luminance, but not the red or blue channels.

because it calculates the two missing components at every spatial
location. the red pixels measure red and then calculate green and blue
from as many as 24 neighboring pixels (sometimes even more). the
result is a very accurate representation of the subject except in some
edge cases that don't generally occur in the real world. in other
words it is a very good engineering tradeoff.
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

In article , Alfred
Molon wrote:

> How could that be the case, when such a Bayer sensor does not measure
> luminance at 12 million locations? The green channel may be a good
> approximation of luminance, but not the red or blue channels.

Please don't start defending Foveon technology. It's been proven to not
work - unless you like Homer Simpson skin tones and artificial
sharpness.
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

On Nov 21, 2:41 am, Ted Marks wrote:

Hello, navasgroup!

>
> This is why I think it's going to have to come from some other advancements in
> research. I sometimes suspect it might have to come from nano-tube technology
> where the tubes are grown and oriented perpendicular to the substrate (the
> reading circuitry), duplicating the structure of the rods and cones in the human
> eye. Doing away with any need for intermediate color filtering (as Nikon is
> trying).

Fascinating. How are tubes of radius of the order of nanometres going
to remove the need for colour filtering?

> Then again, something might come out of a lab as another ingenious
> "accidental" discovery, as all great discoveries have happened. It's the same
> way that I discovered the real reason that moths are drawn to lights long ago
> and it is now rewriting a century of textbooks (as well as giving insight into
> the disappearance of amphibians and other species world-wide). A new discovery
> in a lab could just as easily make the Bayer method disappear.

So why are they drawn to light, then? I've always wondered (of course
I have my theories but they probably pale in comparison to yours)..
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss.. 

In article , nospam says...

> > How could that be the case, when such a Bayer sensor does not measure
> > luminance at 12 million locations? The green channel may be a good
> > approximation of luminance, but not the red or blue channels.
>
> because it calculates the two missing components at every spatial
> location. the red pixels measure red and then calculate green and blue
> from as many as 24 neighboring pixels (sometimes even more). the
> result is a very accurate representation of the subject except in some
> edge cases that don't generally occur in the real world. in other
> words it is a very good engineering tradeoff.

The result is something, but not something that has 12 million pixels of
spatial resolution. The effective real resolution is lower than the
pixel count for Bayer sensors.
--

Alfred Molon
------------------------------
Olympus 50X0, 7070, 8080, E3X0, E4X0 and E5X0 forum at
http://tech.groups.yahoo.com/group/MyOlympus/
http://myolympus.org/ photo sharing site
 >> Stay informed about: Camera phone technology article (Photonics-Spectra Nov iss..