Foveon news article is complete nonsense!

[Gary Eickmeier]

Correct, as you point out the "purity of the line" is important and
has a direct effect on output resolution. Did you see the
resolution chart crop I posted in my X3 preview? The Foveon X3 is
capable of resolving more monochrome resolution than an identical
MP bayer chip. And just as importantly the resolution chart could
use red, blue, orange, purple or any coloured line and you'd get
the same resolution from the Foveon.. Try that with a mosaic chip.

But Phil, was that chart image taken with the 2048 x 2048 square chip or the 1512 x 2268 Sigma chip? Not fair comparison if the former, unless you compare with a 6mp Bayer. You zoomed in more!

Gary Eickmeier

 

[Gary Eickmeier]

That's right! Before the Foveon X3 existed, we accepted it, as it
was the only logical solution. If we had X3 first and Bayer pattern
CCD later, we would probably called those X1/3, although I can live
with X1 for (1 layer) Bayer pattern and X3 for 3 layer Foveon.

Bullpucky! The Sony f707 CONSUMER camera got 1850 lines extinction resolution on the chart that Phil uses. How do you account for this kind of performance if it should have only half the resolution of the Foveon?

Gary Eickmeier

 

[Phil Askey]

I didn't post it for comparison, go back and read the article.

http://www.dpreview.com/news/0202/02021103foveonx3preview.asp

Correct, as you point out the "purity of the line" is important and
has a direct effect on output resolution. Did you see the
resolution chart crop I posted in my X3 preview? The Foveon X3 is
capable of resolving more monochrome resolution than an identical
MP bayer chip. And just as importantly the resolution chart could
use red, blue, orange, purple or any coloured line and you'd get
the same resolution from the Foveon.. Try that with a mosaic chip.

But Phil, was that chart image taken with the 2048 x 2048 square
chip or the 1512 x 2268 Sigma chip? Not fair comparison if the
former, unless you compare with a 6mp Bayer. You zoomed in more!

Gary Eickmeier

 

[Jack40001]

Huh? Did I talk about resolution?
Jack.

That's right! Before the Foveon X3 existed, we accepted it, as it
was the only logical solution. If we had X3 first and Bayer pattern
CCD later, we would probably called those X1/3, although I can live
with X1 for (1 layer) Bayer pattern and X3 for 3 layer Foveon.

Bullpucky! The Sony f707 CONSUMER camera got 1850 lines extinction
resolution on the chart that Phil uses. How do you account for this
kind of performance if it should have only half the resolution of
the Foveon?

Gary Eickmeier

 

[Tom Meeks]

Absolutely!

It was a real eye-opener to me when I first learned how Bayer Pattern camera's worked. The number of articles on the web discussing carious interpolation techniques leads one to realize that we're at the mercy of the 'Solution of Day' that happens for be packed in our camera's firmware of RAW image processing software.

This wasn't too critical when digital cameras were low resolution 'toys' but if we want to make money with a camera it starts to be a very serious issue. The clue to how critical this is can be found in the fact that 'banding' fixes can be made at the firmware level.

I, for one, welcome the fact that Foveon has brought this issue to the fore.

It's hard to compare apples and oranges, so in the end it's going to come down to image fidelity and effective print sizes. But, in the meantime, Foveon's claim loks pretty good to me and if I have any displeasure it's aimed at the noise my CCD and firmware deliver... no matter how small.

The absurdity is in saying that Bayer sensors have the full
advertised resolution. We've grown to accept this little lie and
now the inconsistency of it all bothers us when we apply the same
logic to the Foveon sensor.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Ron Parr]

You can think of it as the square part from a 6MP Foveon sensor.

Compare this with the D60's test results. Same lens, same sensor resolution - the foveon sensor does better.

(The one thing we don't know is the physical size of the Foveon sensor. It's possible that the D60 sensor ran into the some lens limitations if the D60 sensor was physically smaller.)

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

 

[Ulysses]

I'm really not sure of what the bigger achievement is by Foveon here:
1) the actual high quality of the images produced
2) the way this sensor forces us to reexamine how images are produced
3) the potential impact that they will have on other chip makers

Methinks a retraining of sorts will be in order over the next year or so.

I'm with Ron, this is where the whole thing gets sticky. It's our
acceptance of existing single-colour-pixel sensors as providing per
photosite per pixel resolution that causes us to believe the X3
isn't what it says it is.

-- Ulysses

 

[Gary Eickmeier]

You can think of it as the square part from a 6MP Foveon sensor.

Compare this with the D60's test results. Same lens, same sensor
resolution - the foveon sensor does better.

(The one thing we don't know is the physical size of the Foveon
sensor. It's possible that the D60 sensor ran into the some lens
limitations if the D60 sensor was physically smaller.)

Yes, that's the idea. These resolution charts are a little misleading because of the way "resolution" is defined, as lines per picture height. As I understand it, he shoots the test target so as to fill the frame from top to bottom, no matter what the size or aspect ratio of the sensor. For larger sized sensors, this is a distinct advantage. And this is as it should be, if we are comparing sensors with similar aspect ratios. However, in the case of the Foveon test camera that Phil was using, the sensor was a square 2048 x 2048 pixel sensor, versus the Sigma (or Canon or anything else) 1512 x 2268 rectangular sensor. This means when he shoots the test target with the Canon, it fills a smaller surface area and the image is smaller on the chip.

I think it will be fair to compare resolutions between the Sigma and the Canon, but not between the square Foveon chip and the Canon.

Gary Eickmeier

 

[Gary Eickmeier]

I didn't post it for comparison, go back and read the article.

No, you didn't post it for comparison in your article, but you are using it for comparison in the text below:

Correct, as you point out the "purity of the line" is important and
has a direct effect on output resolution. Did you see the
resolution chart crop I posted in my X3 preview? The Foveon X3 is
capable of resolving more monochrome resolution than an identical
MP bayer chip. And just as importantly the resolution chart could
use red, blue, orange, purple or any coloured line and you'd get
the same resolution from the Foveon.. Try that with a mosaic chip.

Gary Eickmeier

 

[misha]

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!

I think that any claims pro and con are not really valid until we see actual images from working cameras, like the new Sigma.
--Misha

 

[Ulysses]

I did a little simulation recently, not for real comparison of resolution, but just to demonstrate the effects within these two different styles of sensor. The mosai chip is obvious.

(It's Foveon not Faveon..)

capable of resolving more monochrome resolution than an identical
MP bayer chip. And just as importantly the resolution chart could
use red, blue, orange, purple or any coloured line and you'd get
the same resolution from the Foveon.. Try that with a mosaic chip.

-- Ulysses

 

[Ron Parr]

Yes, that's the idea. These resolution charts are a little
misleading because of the way "resolution" is defined, as lines per
picture height. As I understand it, he shoots the test target so as
to fill the frame from top to bottom, no matter what the size or

It's measuring the maximum resolution you could hope to get using your camera.

aspect ratio of the sensor. For larger sized sensors, this is a
distinct advantage. And this is as it should be, if we are
comparing sensors with similar aspect ratios. However, in the case
of the Foveon test camera that Phil was using, the sensor was a
square 2048 x 2048 pixel sensor, versus the Sigma (or Canon or
anything else) 1512 x 2268 rectangular sensor. This means when he
shoots the test target with the Canon, it fills a smaller surface
area and the image is smaller on the chip.

I think it will be fair to compare resolutions between the Sigma
and the Canon, but not between the square Foveon chip and the Canon.

If the square part of the 6MP Canon sensor is the same physical size as the Foveon sensor, then it's fair to compare them as 6MP vs. 6MP.

Even if they aren't the same physical size, it's fair to compare them if the lens resolution isn't the limiting factor.

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

 

[Gary Eickmeier]

I did a little simulation recently, not for real comparison of
resolution, but just to demonstrate the effects within these two
different styles of sensor. The mosai chip is obvious.

Sorry, U, but I can't just buy your comparison as is. You've got to tell us a little about how you did it. I'm betting it is straight off of (from) Phil's charts.

Gary Eickmeier

 

[glowluzid]

The biggest problem with the Foveon CMOS is....

IT AIN'T IN A CANON or NIKON body!!

As for technical details about the 10.3 photoSITES (sites, not pixels, folks). Here's my take.

Foveon/Sigma claim 10.3 photo sites compared to 3.1 actual pixels. True, but photo sites don't necessarily give you equivalent resolution of pixels, especially if they're stacked--each of the three stacked sites can't resolve any additional luminance info from the others in the same pixel.

The bigger point here I think is that a traditional Mosaic CCD is undersampled by a factor of two in the green and three in the red and blue channels. So the equivalent sampling of a 3.1 Mos CCD is actually at best only 1.55 MP of real info (or 1.2 average RGB). The Sigma/Foveon would have 3.1 of real sampling. (All of this ignores obscuration which further degrades the sampling).

3.1 MP of a 2D array compared with 1.55 MP of a 2D array tells me that you'll get SQRT(3.1)=1.76x the nominal resolution of a Mosaic 3.1 CCD. (SQRT because its 2D).

In other words, resolution wise, the Fov/Sig 3.1 CMOS has at least as much resolution as a 5.5MP Mosaic CCD. Probably more because the colors are stacked and registered, making the color accuracy at least twice to three times better, and removing interpolation artifacts. So the claims that the Fov 3.1 CMOS will match 6MP mosaic CCDs is about right.

Now, being CMOS, it should be slightly cheaper to mass produce. But they have significant R&D to make up right now.

--Photos, tips and tests at: http://www.geocities.com/glowluzid

 

[Kok Chen]

Now we realize
we were duped FAR worse, that there was only 1/4 of ANY actual
colour !!! and each "site" was a guess of TRUE COLOUR.

Not only is it a guess of the color, it is a guess of the luminance as well .
You can't even get a clean high resolution black and white image.

Actually, I won't go as far as saying we were duped. I knew that Bayer
technology was used as far back as the Apple QuickTake cameras. Nobody
really hid the data (Phil actually calls them out as "R-G-B-G," or some such).

It is just that most people did not go through the reconstruction mathematics
to realize how bad the problem is. I became a much happier person when I

finally viewed my Coolpix 990 images -only- at (reprocessed) half resolution .

And if you think you are being duped by Bayer technology, you might want
to visit the anti-aliasing filters ahead of the sensors. Remember that you will
need different filter resolutions for different color sensors, since the Bayer
does not have equi-resolution sensors. You might want to find out which
manufacturer does that. This is assuming that they even filtered down to
Nyquist in the first place.

• kc

 

[Kok Chen]

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!

I agree it is absurd. To someone in digital image processing, the Foveon
provides 3.5 million pixels.

But, what if I create a color film where the color layers have holes in them,
would you still call that a 24x36 mm film with full resolution? Afterall,
you can interpolate the colors, right?

I think that Phil, at least, is trying to find a way to describe the increased
quality from a sensor which provides vector data at each sampling
point, now that the Bayer camera manufacturers have it ingrained in our

minds that the density of (scalar) sensors equates directly to (color) resolution.
The other writers will also eventually have to wrestle with that.

• kc

 

[Johnnycb]

After reading all these rants and raves.....WHO CARES!!

As long as it delivers who gives a RAT'S A* !

Not me........

After all, the buyer's of digital cameras have been lied
to since day one about mp effictiveness. By not speaking
out with a loud voice we have accepted the lies.
You buy a 4mp camera only to find out it is really only
3.8mp effectively....did you rant then. Did you return the camera?
NO, you lived with it because the final print looked just fine
to your eye.

jb

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!

I agree it is absurd. To someone in digital image processing, the
Foveon
provides 3.5 million pixels.

But, what if I create a color film where the color layers have
holes in them,
would you still call that a 24x36 mm film with full resolution?
Afterall,
you can interpolate the colors, right?

I think that Phil, at least, is trying to find a way to describe
the increased
quality from a sensor which provides vector data at each sampling
point, now that the Bayer camera manufacturers have it ingrained in
our
minds that the density of (scalar) sensors equates directly to
(color) resolution.
The other writers will also eventually have to wrestle with that.

• kc

 

[Iain West]

I think the cells on the Foveion chip are arranged in a square grid. Maybe that is why it is good a picking up vertical lines on the chart.

I don't believe it will outperform a Bayer chip with the same pixel count for a monohrome image of a normal subject.

It's unbeleivable that this debate has gone on for so long, its
SIMPLE MATHEMATICS folks !!! Fer chrissakes -

Issue_1 : A 3.4 MP Faveon and a 3.4 MP Bayer Pattern have the SAME
RESOLUTION. ie If you are taking a shot of a very fine grid-pattern
(Phils V and H Res charts), they both are "capable" of resolving
the same number of "Lines". The "purity of the Line" would be far
superior with the Faveon though, because it can "resolve" perfect
colour at eack "pixel" or "line element". In the bayer, you only
have a red then blue then green / green pixel, so if it was a "red
line" you'd miss resolution every 4 !! You can't MAKE UP what it
was supposed to be !!!

Correct, as you point out the "purity of the line" is important and
has a direct effect on output resolution. Did you see the
resolution chart crop I posted in my X3 preview? The Foveon X3 is
capable of resolving more monochrome resolution than an identical
MP bayer chip. And just as importantly the resolution chart could
use red, blue, orange, purple or any coloured line and you'd get
the same resolution from the Foveon.. Try that with a mosaic chip.

 

[Ron Parr]

I think the cells on the Foveion chip are arranged in a square
grid. Maybe that is why it is good a picking up vertical lines on
the chart.

Foveon sensor was equivalent to the square part of a 6MP sensor.

I don't believe it will outperform a Bayer chip with the same pixel
count for a monohrome image of a normal subject.

Well, I guess I should ask why you mean by "normal", but it's clear the the Foveon sensor gets more accurate luminance information, so why would you not expect it to perform better?

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

 

[Kok Chen]

I don't believe it will outperform a Bayer chip with the same pixel
count for a monohrome image of a normal subject.

Would you consider a monochome image that has all sorts
of colour fringes still a monochrome image?

If not, then I am willing to wager that the 3.5 MP Foveon
beats a 6 MP Bayer. I.e., the position on a monochrome resolution
chart which exibits, say, 1% (7 bits) hue error?

• kc