Foveon news article is complete nonsense!

To make a fair comparison between these sensors, other factors must also be considered, take operating speed for example: The 3.4 Mp Foveon chip can run at 2 fps max while the 6 Mp chip in the D60 will do 8 fps. A 6 Mp Foveon will be impractically slow.

What bothers me about Foveon is the inconsistency in their operation. 2 years ago they announced a 16 Mp chip with much fanfare, then they quietly shelved it. Even with these new sensors, if you look at the pdf spec sheets, the specs of the F7 chip can be found in the F10 spec sheet (and vice versa). This inconsistency is very unprofessional.

Richard C said:

Only the issue of what constitutes a "fair" comparison - I don't
think a 6 MP mosaic has a snowball's chance in hell of competing
with a 6 MP Foveon. The fair comparison will be between, for
example, a soon-to-be-available 3.5MP SD9 and a
soon-to-be-available 6MP D60. I honestly think the 3.5MP Foveon
will have a decisive edge, but, time will tell.

Click to expand...

And I realized that the resolution chart shot s are vertically matched sizewise not horizontally, so that makes the foveon sensor test shot a crop of a 6mp sensor (with 3:2 aspect ratio) and not the sigma sensor a crop of the foveon.

Richard C said:

Only the issue of what constitutes a "fair" comparison - I don't
think a 6 MP mosaic has a snowball's chance in hell of competing
with a 6 MP Foveon. The fair comparison will be between, for
example, a soon-to-be-available 3.5MP SD9 and a
soon-to-be-available 6MP D60. I honestly think the 3.5MP Foveon
will have a decisive edge, but, time will tell.

Click to expand...

I see.

The point I was making was just that based upon what we've seen, we can conclude that if Foveon were to build a 6MP sensor by extending the type in Phil's prototype to make it rectangular, it would probably do better than the 6MP D60.

Now, I agree that we can also infer that there's a point less than 6MP where a Foveon sensor will do as well or better than the D60's 6MP Bayer sensor. Is this point 3.5MP? I don't know.

--Ron ParrFAQ: http://www.cs.duke.edu/~parr/photography/faq.htmlGallery: http://www.pbase.com/parr/

Yes. What length anti-aliasing filters would be required, is a great way to think about it.

This lets us calculate the 'N' in the equation 1 3-color pixel, is worth N bayer pixels.

If you have a bayer sensor, your anti-aliasing filer for red and blue would have to prevent any spacial detail, with a wavelength shorter than 4 pixels, (twice the red and/or blue sampling periods).
So, the linear resolution is 1/2 as compared to a 3-color pixel.

So 1/2 linearly, means 1/4 area wise, so a bayer sensor is worth a 3-color sensor approximately 1/4 the number of pixels.

We shall see...

Don

K.Chen said:

Don Erway wrote:
I think your observation is correct. If a Foveon camera did not use
a proper anti-aliasing filter ahead of the sensor, you can still
get aliasing
errors.

But, all else being equal, Foveon should have an easier time
getting it
right.

With a Bayer mosaic, to meet Nyquist requirements, you need a
lower resolution anti-aliasing filter for the red and green components
and a sharper anti-aliasing filter for the green component.

With a Foveon type sensor, the anti-aliasing filter that is optimal
for one
color component is also optimal for the others.

But, honestly, I have no idea how the manufacturers implement their
anti-aliasing filters. Perhaps both are equally easy/hard.

• kc

Click to expand...

--Don Erway http://www.pbase.com/derway/kona_underwater_g2

I agree that many other factors need to be taken into account, but the readout speed is not a particular problem of the Foveon CMOS chip. The D60 has a max of 3 fps (for 8 frames, that spec has caused a bit of confusion). Given that the Foveon 3.5 MP is generating 50% more RAW data, 2 FPS seems reasonable. The Sony CCD as used in the D100 is probably inherently faster, but I doubt if many implementations of it will exceed 3 fps, the D100 doesn't.

It's only fair to see all sensors CCD, Mosaic CMOS and 3-layer CMOS as works-in-progress so we can undoubtedly expect improvements in all of them over the next few years, I would expect Canon & Foveon to be improving CMOS readout speeds and Sony & Fuji to be improving CCD production so as to get costs down.

As far as Foveon's record of production is concerned, well they have been producing 4MP x 3 CMOS studio cameras for a couple of years now. From all accounts those 3 chip cameras are capable of spectacular results. The 16MP monochrome CMOS hasn't seen production yet, it still may but a 16MP x 3 CMOS camera was never going to be a big volume seller - who could blame Foveon for putting their efforts into an idea which could make them big players in the market.

I don't know why (apart from natural caution, of course) people are loath to accept that Foveon have a good idea and are serious about competing with the big boys. As consumers, this can all be nothing but good news for us.

Berg Na said:

Richard C said:

Only the issue of what constitutes a "fair" comparison - I don't
think a 6 MP mosaic has a snowball's chance in hell of competing
with a 6 MP Foveon. The fair comparison will be between, for
example, a soon-to-be-available 3.5MP SD9 and a
soon-to-be-available 6MP D60. I honestly think the 3.5MP Foveon
will have a decisive edge, but, time will tell.

Click to expand...

Click to expand...

--Richard C. South Australia

Richard C said:

I agree that many other factors need to be taken into account, but
the readout speed is not a particular problem of the Foveon CMOS
chip. The D60 has a max of 3 fps (for 8 frames, that spec has
caused a bit of confusion). Given that the Foveon 3.5 MP is
generating 50% more RAW data, 2 FPS seems reasonable. The Sony CCD
as used in the D100 is probably inherently faster, but I doubt if
many implementations of it will exceed 3 fps, the D100 doesn't.

Click to expand...

The Sony CCD that is assumed to be used in the D100 maxes out a 3FPS anyway, so don't worry about this.

Those guys who are desperately searching for reasons for the Foveon approach to fail are forgetting that even with a Bayer sensor, you need to construct 3 dimensionsal color values before passing the image in to the jpeg compression algorithm, so the amount of internal bandwidth required will be the same. The only difference will be the amount of data that you need to pull of the sensor.

--Ron ParrFAQ: http://www.cs.duke.edu/~parr/photography/faq.htmlGallery: http://www.pbase.com/parr/

D100 is based on F80, not F100, so there is no chance it would be sealed like F100. Just some clever marketing from Nikon.

Bill Pitcher said:

I'm trading the F80 in for a F100 next week - it's certainly a more
rugged camera. I also plan on buying the D100, but only if it's
sealed like the F100, not the F80. $2000+ is a way too much to pay
for a "sunny" day camera. I'll keep the Sigma though, it's
definately a good value. Here's a shot I took with the F80 and the
Sigma 400mm last fall:
http://www.goldenbcphotography.com/0_most_recent/dogtooth.htm

Bill

Click to expand...

Well actually 4 points:

1. Highly rated body (at least at F80 level according to most reviews).

2. Dust protection (You talk about weather protection, but try changing your lenses on 1D in a desert when wind blows. I think sigma will behave better...)

3. Always on software (Phil said you can start shooting regardless of the mode you are currently in. Even from playback!!!)
4. Foveon sensor !!!

As for lenses..., most of us lowly amateurs would buy sigma or sigma-class lenses even if we had canon or nikon body. Price/performance issues.
How much for that Nikkor 50-500?

Rgrds,
Moshe

Ger Bee said:

SIGMA.

Turn over and go back to sleep!!

Click to expand...

Berg Na said:

Since the Foveon image sensor only contains 3.4 M spatially unique
sampling pixels, it is complete nonsense and a mockery of the
published guidelines to claim that the chip has 10.3 M pixels. To
do so is akin to saying that the imaging area of a 35mm frame is
actually 24mm x 36mm x 3 (3 layers of emulsion), or equivalent to
51 mm x 51 mm, a size which is comparable to the dimensions of
medium format. That's just absurd!

Click to expand...

Well actually, a COLOR 35mm film area DOES have the equivalent of 3 times the area! One 24 by 36mm area for EACH of the film's 3 color sensitive layers!

What the X3 refers to on the X3 sensor is that the effective 'color resolution' is three times what an equivalent bayer pattern sensor has.

We have become so used to accepting that the bayer pattern sensor actually HAS such and such number of pixels, when in reality it only has 1/4 of that number as far as color is concerned.

--Sincerely,Bob the Printer

My understanding the bayer sensors is that they have 4 photosites (2 green, 1 red, 1 blue), which are 'merged' into a pixel, then, the cameras, extrapolate 4 pixels out of that one.

So, if we had a mythical 2x2 bayer sensor, it has 4 photo sites, which gives one pixel (merged between the 2 greens, 1 red and 1 blue), which is up-sampled to 4 pixels.

The Foveon (and correct me if I am wrong), of a mythical 2x2 sensor would have 4 photo sites (three levels deep, red, green and blue), give 4 pixels of 'in quotes' 'correct' color with no interpolation.

The whole argument appears to have been driven by marketing selling the 'x megapixel photosites' sensor, rather than the non-extrapolated pixel output of the camera.

Sounds a lot better when you have a 1mp bayer sensor and you can advertise your camera as a 1mp camera (in tiny print, extrapolated) than being honest and saying you are selling a 1mp sensor with 256kp (in tiny print, non-extrapolated) output, camera, doesn't it?

--...Lyall

Since the detail information is typically mostly in the green, the bayer sensor is worth a 3 color sensor with approximately 1/2 the number of pixels. Of course, this is somewhat image content dependent. If the image was red detail on a black background, then the number would be 1/4.

Don Erway said:

This lets us calculate the 'N' in the equation 1 3-color pixel, is
worth N bayer pixels.

If you have a bayer sensor, your anti-aliasing filer for red and
blue would have to prevent any spacial detail, with a wavelength
shorter than 4 pixels, (twice the red and/or blue sampling periods).
So, the linear resolution is 1/2 as compared to a 3-color pixel.

So 1/2 linearly, means 1/4 area wise, so a bayer sensor is worth a
3-color sensor approximately 1/4 the number of pixels.

We shall see...

Don

K.Chen said:

Don Erway wrote:
I think your observation is correct. If a Foveon camera did not use
a proper anti-aliasing filter ahead of the sensor, you can still
get aliasing
errors.

But, all else being equal, Foveon should have an easier time
getting it
right.

With a Bayer mosaic, to meet Nyquist requirements, you need a
lower resolution anti-aliasing filter for the red and green components
and a sharper anti-aliasing filter for the green component.

With a Foveon type sensor, the anti-aliasing filter that is optimal
for one
color component is also optimal for the others.

But, honestly, I have no idea how the manufacturers implement their
anti-aliasing filters. Perhaps both are equally easy/hard.

• kc

Click to expand...

--
Don Erway
http://www.pbase.com/derway/kona_underwater_g2

Click to expand...