Reading mode: Light Dark

What will it take to make the Fovean naysayers to S.U.

Go to last post
The Sigma/Foveon samples each color at each pixel. That is VERY important, and it works ... and it works well. More data points ... with only 1/3 of the color data per point ... doesn't work as well.

I have a Canon 1D ... $5K worth of camera ... and the S/F is a match for it, period.
Ken
DMillier said:
The foveon uses 3*3.5 million data samples, true. BUT,they are not
all on separate areas of the image. Every point on the surface has
3 stacked light sensors all occurpying the same spot. These
triplets don't collect 3 pixels worth of data (in the sense that we
conventionally describe a pixel) rather, they are used to generate
a single, more accurate pixel).

The bigger bayer sensors do collect more data points, it's just
that do not form the image directly. Instead the individual
filtered monochrome samples are used as the input to an algorithm
that creates the final full colour pixels.

The foveon method clearly has an advantage pixel for pixel but not
for the reason you suggest.
Jimmy Chen said:
DavidMillier said:
I don't really care how great a 6 megapixel foveon
sensor can be, all I know is we only have a 3 megapixel camera to
deal with.
Click to expand...
What kind of car would you rather buy today? A large lliter engine,
or a smaller engine with a turbo? Almost the same here ... A more
pixel count Bayer that catches 1/3 the data, or a smaller pixel
count X3 that catches 3x the data? Dont forget, 6mp Bayer only have
6mpx12 bits worth of real data. The 3.5mp X3 has 3x3.5mpx12 bits or
10.5mpx12 bits worth of data.

Larger do not always translate to better.
DavidMillier said:
It's priced in the range of the D60/S2/D100, so it must be compared
with those cameras.
Click to expand...
And I believe they will be, as many of us already have in our own
way. We will just have to wait until Oct for Phil's review to see
the real results from a controlled environment.

--
jc
Sony F707
http://www.reefkeepers.org/gallery/f707
http://www.reeftec.com/gallery
Click to expand...
Click to expand...
--

All kinds of old camera and motion picture bodies, lenses, tripods, enlargers, mostly gathering dust, because digital is immediate! NO Canon 1200mm f/5.6.
 
DavidMillier said:
I don't agree with your maths here.

The foveon uses 3*3.5 million data samples, true. BUT,they are not
all on separate areas of the image. Every point on the surface has
3 stacked light sensors all occurpying the same spot. These
triplets don't collect 3 pixels worth of data (in the sense that we
conventionally describe a pixel) rather, they are used to generate
a single, more accurate pixel).
Click to expand...
Quoted directly from Foveon's F7 spec ...
Effective Pixels
2268 x1512 x 3 (in our case we are just saying there are 3.4mp pixels)
Effective Photo Detectors
10.3 million
Total number of measured color data points captured by image sensor

This translate to each pixel does have 3 seperate data collection points, not one.

Then there is this write up from Phil confirming the Foveon X3 F7 chip does have 10.3 mp photosites, not 3.4mp photosites http://www.dpreview.com/news/0203/02030602foveonx3notation.asp

--
jc
Sony F707
http://www.reefkeepers.org/gallery/f707
http://www.reeftec.com/gallery
 
Not quite sure what your "but" is refuting - I never questioned the foveon quality, just the description of the numbers...
Ken Phillips said:
DMillier said:
The foveon uses 3*3.5 million data samples, true. BUT,they are not
all on separate areas of the image. Every point on the surface has
3 stacked light sensors all occurpying the same spot. These
triplets don't collect 3 pixels worth of data (in the sense that we
conventionally describe a pixel) rather, they are used to generate
a single, more accurate pixel).

The bigger bayer sensors do collect more data points, it's just
that do not form the image directly. Instead the individual
filtered monochrome samples are used as the input to an algorithm
that creates the final full colour pixels.

The foveon method clearly has an advantage pixel for pixel but not
for the reason you suggest.
Jimmy Chen said:
Ken Phillips said:
I don't really care how great a 6 megapixel foveon
sensor can be, all I know is we only have a 3 megapixel camera to
deal with.
Click to expand...
What kind of car would you rather buy today? A large lliter engine,
or a smaller engine with a turbo? Almost the same here ... A more
pixel count Bayer that catches 1/3 the data, or a smaller pixel
count X3 that catches 3x the data? Dont forget, 6mp Bayer only have
6mpx12 bits worth of real data. The 3.5mp X3 has 3x3.5mpx12 bits or
10.5mpx12 bits worth of data.

Larger do not always translate to better.
Ken Phillips said:
It's priced in the range of the D60/S2/D100, so it must be compared
with those cameras.
Click to expand...
And I believe they will be, as many of us already have in our own
way. We will just have to wait until Oct for Phil's review to see
the real results from a controlled environment.

--
jc
Sony F707
http://www.reefkeepers.org/gallery/f707
http://www.reeftec.com/gallery
Click to expand...
Click to expand...
--
All kinds of old camera and motion picture bodies, lenses, tripods,
enlargers, mostly gathering dust, because digital is immediate! NO
Canon 1200mm f/5.6.
Click to expand...
 
Ron Parr said:
I have a D30 and a D60 and I've spent a lot of time tweaking my
shots. IMO, no amount of twiddling will give the crisp edges the
SD9 shots have at full size.
Click to expand...
I'm just pointing out the problem with the methodology. Just because the methodology is flawed doesn't mean you won't get the right answer ;-)

Car analogies are fun and popular. Let's try one:

I want to decide what's better: Camero or Mustang. So I'll go down to the the Daytona 500 parking lot and find a dozen Camero drivers. And then I'll get Csaba Csere (editor-in-chief of Car and Driver, my Phil substitute. Not a professional race driver, but a professional reviewer) to drive the Mustang. Now who do you think will produce the best lap times? And is it the car or the driver (or both?)

The SD-9 may be better than the D60 for color and sharpness. However, I think it's too early to make this judgement based on the limited samples seen to to date.

Do you think your D60 equals the non-Phil SD-9 samples?

--
Erik
 
Jimmy Chen said:
Jimmy Chen said:
I think that "the industry" is more surprised by the FillFactory
full-frame CMOS to sell in a camera for $4k than Foveon.
Click to expand...
Dont forget NSC is a much more established and experienced chip
mfg'er then FillFactory.
Click to expand...
I didn't forget. In fact, that's one reason it's surprising. It now seems that you don't have to be a Sony or a Kodak to compete in the sensor biz. (And Foveon may get more price pressure then they wanted in the high-end market from attack of the really big CMOS mosaics.)

--
Erik
 
Erik Magnuson said:
Do you think your D60 equals the non-Phil SD-9 samples?
Click to expand...
It's hard to say without some carefully controlled tests.

I expect that the D60 will excel in some areas like noise at ISO 400, and of course > 400 isn't an option with the SD9. I expect the D60 will have cleaner shadows even at ISO 100.

I'm expecting the resolution tests to be controversial since the SD9 will degrade in different ways from the way D60 degrades as it approaches the resolution limits. If you forced me to bet, I'd predict that the D60 will be able to resolve a little more, but that the SD9 will drop off more gracefully. However, this is really just a guess.

I'm excited because I think it's a neat technology that has demonsrated that it has promise. The fact that it has made it this far and that there is some speculation about how it will compete (in a few weeks, hopefully) with sensors of nominally higher resolution is just great in my book. I really like seeing people try out new ideas and I think that regardless of how the SD9 compares with the D60, consumers will benefit from this in the long run.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
 
I didn't say the foveon doesn't have 10million data points. It's how they are it used that I questioned.

It basically takes 3 colour specific detectors to work out the colour of a single pixel. The foveon does it by have a triplet of sensors arranged in a stack at each pixel position. It measures the triplet to work out the colour of the pixel at that position.

This is quite different from how bayer works. The bayer approach has detectors arranged in little grids covering an area not a point. It measures sample points from an area and uses them to guess the colours of all those points (using software).

This actually means that the final bayer file maps one interpolated colour pixel for every monochrome detector in the sensor whilst the foveon file has one colour pixel for every THREE photodetectors in the sensor.

The reason this profligacy gives better results that the more economical bayer is because the bayer approach requires a little blurring and introduces artifacts. However, a 6MP bayer sensor still measures more separate points on the surface of the sensor than a 3MP foveon.

Kodak makes a black and white version of the 760. I believe it uses the same sensor as the 760 but without the bayer filter and algorithm. That means it is like a black and white foveon but with 6MP not 3MP.

I've no doubt that it will produce a far sharper and more detailed B&W picture than the SD9 could.

However as soon as you want to make a colour image from a monochrome sensor you have to introduce the bayer approach and that's what spoils the final image.
Jimmy Chen said:
Jimmy Chen said:
I don't agree with your maths here.

The foveon uses 3*3.5 million data samples, true. BUT,they are not
all on separate areas of the image. Every point on the surface has
3 stacked light sensors all occurpying the same spot. These
triplets don't collect 3 pixels worth of data (in the sense that we
conventionally describe a pixel) rather, they are used to generate
a single, more accurate pixel).
Click to expand...
Quoted directly from Foveon's F7 spec ...
Effective Pixels
2268 x1512 x 3 (in our case we are just saying there are 3.4mp pixels)
Effective Photo Detectors
10.3 million
Total number of measured color data points captured by image sensor

This translate to each pixel does have 3 seperate data collection
points, not one.

Then there is this write up from Phil confirming the Foveon X3 F7
chip does have 10.3 mp photosites, not 3.4mp photosites
http://www.dpreview.com/news/0203/02030602foveonx3notation.asp

--
jc
Sony F707
http://www.reefkeepers.org/gallery/f707
http://www.reeftec.com/gallery
Click to expand...
 
I'm with you.

The bayer approach appears to have been developed as a way of producing colour images with decent resolution from sensors with a limited number of photosites that couldn't be arranged in layers like a colour film. It has been steadily devloped and now produces very high qaulity results that can rival film in detail and beat it in grainness. There are some problems with the approach (artificacts, moire, softness) but these are usually not noticeable.

I'm only interested in what the pictures look like when printed out at normal print sizes and to me my D100 image seem every bit as good as film.

The foveon approach is very much like a digital colour film and removes some of those problems. It also seems to use an advanced and inexpensive cmos process. But, foveon aren't perfect. The sensor has a smaller pixel count. This is compensated for by the superior accuracy but it seems to have a lower sensitivity (max ISO 400). Does this suggest possible noise problems? Can they make a larger Foveon sensor? At the moment who knows.

Anyway from what I've seem the SD9 images when printed out seem to have about the quality of the D100/D60 etc. Pretty impressive from a 3.5MP sensor, but we are now entering the era of 11 and 14MP bayers. Foveon risk being left behind if they can't increase pixel counts at the same rate and maintain their pixel to pixel quality advantage.

It will be an interesting battle.
Leo Sheikman said:
Matthew Cromer said:
Personally I feel they quite simply destroy D60 and D100 shots in
image quality. The shots look sharp, clean and real, in comparison
the Bayer camera shots simply look artificial and computer
generated.

--
my favorite work: http://www.pbase.com/sdaconsulting/favorite_work
Click to expand...
--
Click to expand...
 
Leo Sheikman said:
Tim,
I do not agree with you. The resolution of a 3Mp sensor is less
than a 4Mp resolution sensor. The sharpness of the image in this
case may be different and may be not.
Click to expand...
As near as I can see, on the issue of "resolution", the thing that
the sensor is trying to "resolve" is color changes. Color changes
are what make an image. It's how our eyes can see that the image
is that of a boat, or that of a lamp, et cetera. If the image was
all white, or all black, or all red, et cetera, it would all be a blank
image of one color. We would see nothing, other than one color.

To detect a color change in a Bayer sensor it takes 6 pixels in a
3 by 2 area.

To detect a color change in a Foveon sensor it takes 2 pixels
in a 1 by 2 area.

This is why the resolution of a 6 million Bayer sensor is equal to
that of a 3.5 million Foveon sensor.
 
To illustrate DMiller's point, let me take things a little farther. Let's say I had a spectrophotometer at every pixel. I capture values at every nm between 400 and 700nm, yielding 300 values per pixel. Wow, that's a lot more data. But for most purposes it does not improve the image -- I would not say that it is 300 times better. I could probably come up with weird examples of lasers and tri-band flourescent lights that would show this solution is so much better than our pitiful artifact-ridden 3 sensor solution.

Another example -- I scan film at 15 bizillion dpi and generate a huge file. Did I really capture an image enormously better than a 8000dpi scan?

Yes, these things do not translate directly over, but there is some truth here. I agree with DMiller's "they are used to generate a single, more accurate pixel." They do get a resolution boost because one doesn't need an anti-alias filter or any smoothing in the interpolation, so they are worth more than one pixel in practice. But not 4x more.

It is, however, exciting to see since the "more accurate pixel" is an unknown quantity right now -- it could be huge! It certainly won't hurt, unless we end up compromising other things (ISO, noise, cost, etc.)
DMillier said:
The foveon uses 3*3.5 million data samples, true. BUT,they are not
all on separate areas of the image. Every point on the surface has
3 stacked light sensors all occurpying the same spot. These
triplets don't collect 3 pixels worth of data (in the sense that we
conventionally describe a pixel) rather, they are used to generate
a single, more accurate pixel).

The bigger bayer sensors do collect more data points, it's just
that do not form the image directly. Instead the individual
filtered monochrome samples are used as the input to an algorithm
that creates the final full colour pixels.

The foveon method clearly has an advantage pixel for pixel but not
for the reason you suggest.
Jimmy Chen said:
DavidMillier said:
I don't really care how great a 6 megapixel foveon
sensor can be, all I know is we only have a 3 megapixel camera to
deal with.
Click to expand...
What kind of car would you rather buy today? A large lliter engine,
or a smaller engine with a turbo? Almost the same here ... A more
pixel count Bayer that catches 1/3 the data, or a smaller pixel
count X3 that catches 3x the data? Dont forget, 6mp Bayer only have
6mpx12 bits worth of real data. The 3.5mp X3 has 3x3.5mpx12 bits or
10.5mpx12 bits worth of data.

Larger do not always translate to better.
DavidMillier said:
It's priced in the range of the D60/S2/D100, so it must be compared
with those cameras.
Click to expand...
And I believe they will be, as many of us already have in our own
way. We will just have to wait until Oct for Phil's review to see
the real results from a controlled environment.

--
jc
Sony F707
http://www.reefkeepers.org/gallery/f707
http://www.reeftec.com/gallery
Click to expand...
Click to expand...
 
DavidMillier said:
It basically takes 3 colour specific detectors to work out the
colour of a single pixel. The foveon does it by have a triplet of
sensors arranged in a stack at each pixel position. It measures the
triplet to work out the colour of the pixel at that position.
Click to expand...
And by doing so, it is one pixel of true RGB.
DavidMillier said:
This is quite different from how bayer works. The bayer approach
has detectors arranged in little grids covering an area not a
point. It measures sample points from an area and uses them to
guess the colours of all those points (using software).
Click to expand...
Exactly, they guess the colors. And those guesses creates many problems because they arent always 100% correct. You may want to read this post from Mike Chaney on these problems ... http://www.dpreview.com/forums/read.asp?forum=1014&message=3455520
DavidMillier said:
This actually means that the final bayer file maps one interpolated
colour pixel for every monochrome detector in the sensor whilst the
foveon file has one colour pixel for every THREE photodetectors in
the sensor.
Click to expand...
But it is one gussed RGB data vs one true RGB data.
DavidMillier said:
The reason this profligacy gives better results that the more
economical bayer is because the bayer approach requires a little
blurring and introduces artifacts. However, a 6MP bayer sensor
still measures more separate points on the surface of the sensor
than a 3MP foveon.
Click to expand...
If you are only comparing the surface, I agree. But the X3 is not just capturing data on the surface.

What it comes down to is this ... Which is more accure and produces more details? 3.4mp X3 worth of true RGB measurement, or 6mp worth of RGRB guesses? We have seen some samples and many believes the 3.4mp worth of true measurement do. But to know the exacts, we will all have to wait for Phil's final review.

--
jc
Sony F707
http://www.reefkeepers.org/gallery/f707
http://www.reeftec.com/gallery
 
Tim,

The Bayer sensor build an images in two steps as I have read: BW and then defines color for each pixel. You are assuming that each pixel defined by interpolation only then 2Mp sensor would never be able to produce an image with 800 line pairs resolution (=1600 divide by 2). The difference is accurate color definition foe each pixel. However, it is only true if the filtering by each silicon layer would be sufficient. Now 5Mp camera is just a standard camera and a high end moves to above 10Mp sensor. All will be decided by designers and us. They should be able to move the price down to make it attractive for us to buy it. If 10 and above Mp camera will be half price of X3 camera then X3 may have a very hard time to stay in still photography business and may have to move to video first to mature.
Leo
tim martin said:
Leo Sheikman said:
Tim,
I do not agree with you. The resolution of a 3Mp sensor is less
than a 4Mp resolution sensor. The sharpness of the image in this
case may be different and may be not.
Click to expand...
As near as I can see, on the issue of "resolution", the thing that
the sensor is trying to "resolve" is color changes. Color changes
are what make an image. It's how our eyes can see that the image
is that of a boat, or that of a lamp, et cetera. If the image was
all white, or all black, or all red, et cetera, it would all be a
blank
image of one color. We would see nothing, other than one color.

To detect a color change in a Bayer sensor it takes 6 pixels in a
3 by 2 area.

To detect a color change in a Foveon sensor it takes 2 pixels
in a 1 by 2 area.

This is why the resolution of a 6 million Bayer sensor is equal to
that of a 3.5 million Foveon sensor.
Click to expand...
 
All this marketing technical jargon is great and all, but again, all it does is dance around the bottom line. The technology is great and all, and one day may serve as the de facto standard for all digital cameras, but IMO, it's just not there yet.

Until Phil does some very controlled tests, I don't think the lot of us have truly made up our minds yet. However, with the small flow of sample images, I must say my initial impression is that they are the best 3 megapixel images I've seen, but just not better then samples from the D60.
Jimmy Chen said:
Jimmy Chen said:
It basically takes 3 colour specific detectors to work out the
colour of a single pixel. The foveon does it by have a triplet of
sensors arranged in a stack at each pixel position. It measures the
triplet to work out the colour of the pixel at that position.
Click to expand...
And by doing so, it is one pixel of true RGB.
Jimmy Chen said:
This is quite different from how bayer works. The bayer approach
has detectors arranged in little grids covering an area not a
point. It measures sample points from an area and uses them to
guess the colours of all those points (using software).
Click to expand...
Exactly, they guess the colors. And those guesses creates many
problems because they arent always 100% correct. You may want to
read this post from Mike Chaney on these problems ...
http://www.dpreview.com/forums/read.asp?forum=1014&message=3455520
Jimmy Chen said:
This actually means that the final bayer file maps one interpolated
colour pixel for every monochrome detector in the sensor whilst the
foveon file has one colour pixel for every THREE photodetectors in
the sensor.
Click to expand...
But it is one gussed RGB data vs one true RGB data.
Jimmy Chen said:
The reason this profligacy gives better results that the more
economical bayer is because the bayer approach requires a little
blurring and introduces artifacts. However, a 6MP bayer sensor
still measures more separate points on the surface of the sensor
than a 3MP foveon.
Click to expand...
If you are only comparing the surface, I agree. But the X3 is not
just capturing data on the surface.

What it comes down to is this ... Which is more accure and produces
more details? 3.4mp X3 worth of true RGB measurement, or 6mp worth
of RGRB guesses? We have seen some samples and many believes the
3.4mp worth of true measurement do. But to know the exacts, we will
all have to wait for Phil's final review.

--
jc
Sony F707
http://www.reefkeepers.org/gallery/f707
http://www.reeftec.com/gallery
Click to expand...
 
Karl,

I wonder that have you ever disigned an analog circiut or mix analog-digital circuit?

Noise is a very difficult issue to address in analog design, especially with mix A-D circuit. And it can not be solved well in first-generation product.

Noise can be happen everywhere, from the AD converter chip, to amplification, and even an ill-designed PCB (printed circuit board) can easily double the overall noise level. Remeber that ISO1000 is just twice the signal level of ISO400, which means noise need to be 1/2 to maintain the current noise level of the ISO400 in the Sigma and this would not directly relate to the noise level of the sensor. If other current digicams go to ISO 1600 or 3200, It would then make sense to suspect the sensor to be the noise source.

My point is that, at this time, you can not assumed that the noise is caused by the sensor, more over that this is the first time Sigma/Foveon is dealing with designing a digital camera (or mix AD circuit problem) while others have at least a decade of experiment. You can not compared the noise (thus ISO level) by the sensor alone. Noise is a system problem and noise CAN be eliminated gradually by more careful designing once you know where it come from.

Hope that you will be more careful in making assumsion.
Just my two cents.
Regards
So.

------------------------------
Karl Guttag said:
We also had to figure that the "stacked" sensors had to be
compromised compared to a simpler unstacked sensor that had a
filter on top of it. This shows up in the ISO. If the stack
sensor was as good as the Bayer sensor, then the X3 should have had
BETTER low noise performance since in theory it did not throw away
1/3rd of the light of the filter. Instead the X3 has worse low
light/high-ISO performance.

There was a lot of concern that the Canon D60 had only ISO1000 when
the D30 when to ISO1600. Image if the D60 only went to ISO400.
Only the still life people would have considered it a clear step
forward over the D30.

While I/we know that the Bayer filtering is a compromise, I/we
think WAY too much has been made over it. If the X3 is going to
clearly beat/obsolete the Bayer sensors, it has to get nearer to
the megapixels of the Bayer sensors AND have the same or less noise
at the same ISO, while at the same time it has to show that it can
ship reliably in high volume (yet to be seen).
Click to expand...
 
I am curious does the X3 sensor REALLY have 12bits per color per sensor-site? Wow if that's true because that would mean 36bit overall per sensor-site right? Am I right? I would like to think that's true but Wow is that true?
Jimmy Chen said:
Jimmy Chen said:
I don't really care how great a 6 megapixel foveon
sensor can be, all I know is we only have a 3 megapixel camera to
deal with.
Click to expand...
What kind of car would you rather buy today? A large lliter engine,
or a smaller engine with a turbo? Almost the same here ... A more
pixel count Bayer that catches 1/3 the data, or a smaller pixel
count X3 that catches 3x the data? Dont forget, 6mp Bayer only have
6mpx12 bits worth of real data. The 3.5mp X3 has 3x3.5mpx12 bits or
10.5mpx12 bits worth of data.

Larger do not always translate to better.
Jimmy Chen said:
It's priced in the range of the D60/S2/D100, so it must be compared
with those cameras.
Click to expand...
And I believe they will be, as many of us already have in our own
way. We will just have to wait until Oct for Phil's review to see
the real results from a controlled environment.

--
jc
Sony F707
http://www.reefkeepers.org/gallery/f707
http://www.reeftec.com/gallery
Click to expand...
 
C.S said:
Karl,
I wonder that have you ever disigned an analog circiut or mix
analog-digital circuit?
Click to expand...
Actually, I have mostly designed digital circuits, but I have been involved with several Analog designs. I am aware of a number of issues invovled with analog designs, but I freely admit my expertice is in digital.
C.S said:
Noise is a very difficult issue to address in analog design,
especially with mix A-D circuit. And it can not be solved well in
first-generation product.
Click to expand...
I fully agree. Which is one of the reasons that the X3 is behind everyone else. But it would be also ignoring reality to think that their don't have more problems than the Bayer, single sensor per site designs have.
C.S said:
Noise can be happen everywhere, from the AD converter chip, to
amplification, and even an ill-designed PCB (printed circuit board)
can easily double the overall noise level. Remeber that ISO1000 is
just twice the signal level of ISO400, which means noise need to be
1/2 to maintain the current noise level of the ISO400 in the Sigma
and this would not directly relate to the noise level of the
sensor. If other current digicams go to ISO 1600 or 3200, It would
then make sense to suspect the sensor to be the noise source.
Click to expand...
The X3 has MORE sources of noise. They have the 9 transitors (versus 3 for the D30/D60). They have the 3 sets of read lines per "pixel" versus 1 for for the Bayer. They have the stacked diodes and whatever effects there are from their interaction.

Then we have the issue of Color separation, being discussed in the Sigma SLR forum. The X3 has a much less than ideal response curve to colors in the various sensors. Where the Bayers can "tune" the filters, the X3 has to "live" with the silicon physics. The color separation curves are far from ideal and this translates into being more sensitive to noise.
C.S said:
My point is that, at this time, you can not assumed that the noise
is caused by the sensor, more over that this is the first time
Sigma/Foveon is dealing with designing a digital camera (or mix AD
circuit problem) while others have at least a decade of
experiment. You can not compared the noise (thus ISO level) by the
sensor alone. Noise is a system problem and noise CAN be eliminated
gradually by more careful designing once you know where it come
from.
Click to expand...
Actually like most things in the real world, it is likely a combination of effects. My guess is that Foveon has hired some very talented people with a lot of experience. But it is also likely that the whole X3 has brought with it a few new challenges.
C.S said:
Hope that you will be more careful in making assumsion.
Click to expand...
I think I was careful in making my assumptions. I know that there are a lot of factors that affect things. I tend to think that the people doing Foveon are smart and that there are some reasons why they are down by over 2X to the obvious competitors in the important area of ISO/noise. I think I know enough about it to realize that the X3 designers have some added challenges that are making it harder for them.
C.S said:
Regards
So.

------------------------------
Karl Guttag said:
We also had to figure that the "stacked" sensors had to be
compromised compared to a simpler unstacked sensor that had a
filter on top of it. This shows up in the ISO. If the stack
sensor was as good as the Bayer sensor, then the X3 should have had
BETTER low noise performance since in theory it did not throw away
1/3rd of the light of the filter. Instead the X3 has worse low
light/high-ISO performance.

There was a lot of concern that the Canon D60 had only ISO1000 when
the D30 when to ISO1600. Image if the D60 only went to ISO400.
Only the still life people would have considered it a clear step
forward over the D30.

While I/we know that the Bayer filtering is a compromise, I/we
think WAY too much has been made over it. If the X3 is going to
clearly beat/obsolete the Bayer sensors, it has to get nearer to
the megapixels of the Bayer sensors AND have the same or less noise
at the same ISO, while at the same time it has to show that it can
ship reliably in high volume (yet to be seen).
Click to expand...
Click to expand...
--
Karl
 
hoang nguyen said:
I am curious does the X3 sensor REALLY have 12bits per color per
sensor-site? Wow if that's true because that would mean 36bit
overall per sensor-site right? Am I right? I would like to think
that's true but Wow is that true?
Click to expand...
There's a 12-bit AD, so can it be very precise? Yes. Will it be very accurate? Don't know yet.
 
I've been away from my computer and never realized my question would create such a debate.

My original post was aimed toward those that claimed the Fovean technology is nothing but vaporware.
Where are those people?
What do they have to say now?

At least now the technology has graduated to a debate comparing it to higher megapixel CMOS and CDD Technology. Sigma has little to do with my original question. What I should have asked is what will this technology offer if Canon or Nikon or some other high quality lens manufacturer employs the Fovean chip?

The pictures posted from Photokina proved that Fovean is a new technology to reckon with and willl find it's way to all types of imaging devices, from video cameras to pro digi-cams to amateur cameras.
 
Well it does not compare at all to Nikon senors since Nikon does not produce any Sensors. Nikon OEMs all of their sensors. In fact the only DSLR maker who produces their own sensor is Canon & Kodak.

I guess we will have to see if others decide to OEM the chip.

I think its funny that everyone thinks this sensor competes against so many companies. The only company I see that would be really threatened if Fovean does make it will be Sony, and this will happen only if Foveon can find a company willing to use their consumer 1.3mp sensor. The 3mp one is too big to fit in a consumer camera.

Ed
Ted Jengle said:
Erik Magnuson said:
Edwaste said:
If this is what's possible with a 3
megapixel sensor, what can a 6 megapixel do?
Click to expand...
About the same as the current state of the art 11 - 14 MP Bayer?
They've only roughly matched 4 - 6 month old cameras at low ISO -
they are still half a generation behind.

--
Erik
Click to expand...
Click to expand...
--
Ed
Canon EOS 3, Sony D700, Canon G1, Canon S330, Minolta D7i
http://www.cbrycelea.com/photos/ Old Pictures
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top