Final output of the Foveon X3 ?

[Jimmy Chen]

Wow, you are claiming the X3 will be perfect? Get real; it's this
kind of overstatement that turns many off of Foveon hype.

No, what I am saying is a X3 pixel will not have the color "bleeding" like a mosaic chip will. Each image pixel will store its own RGB w/o other pixels telling it what is should look like (Bayer).

None of these steps are perfect; for example you
are completely ignoring noise.

I havent, because all sensors have to deal with noise, X3 is no different. But I do not believe better noise was a Foveon claim. As a matter of fact, I still like to know why Sigma is only at ISO 400 with everyone else is already at ISO 1200 or more?

You are also ignoring any issues in
color separation. What happens to a photon that is absorbed exactly
halfway between two of the color layers (say green and blue.) Will
it be counted as green, blue, neither, both?

And how is this issue a non issue with a mosaic sensor? How will the mosaic BGBG row handle this any better then the X3?

It is clear that a 4mp Foveon resolves much better then any 6mp
mosaic chips, and even beats the so call Fuji "12mp".

Sigh. No it shows that a 2048 pixel high X3 has better resolution
than a 2048 pixel high D60/D100. In other words, if the X3 sensor
used had the same 2/3 aspect ratio of the other sensors, it would
also be 6MP. Perhaps you missed the part where Phil says....

And in such case Phil would have taken the res chart with a 3:2 aspect instead of a 1:1 aspect. It would not have changed the outcome. All of the sensors we compared here are of 2048 or similar vertical pixels.

the chip I tested has more vertical resolution and will therefore perform better on this resolution chart.

Phil's note is to warn us not to use that res. chart to say the F7 chip will be as good. But for the 2048 vertical the res chart is still vaild.

Personally, I don't take photos of resolution targets, so I find
this measurement only mildly interesting. Actual samples under
varying conditions is far more interesting. (After all, a 6 MP
single color sensor would whup X3 ass on this one test. However,
that doesn't make an interesting camera for me.)

Agreed.

I've never said X3 = bad. I only comment when people grossly
misstate the likely advantages or disadvantages of either.

As you stated, the final result is yet to be seen.

But Phil himself have already stated many times that in order for a mosaic 6mp sensor to produce the image quality of the F3, you would have to strink the 6mp image down to 1.5mp. Since he first hand experience with the sensor, I tend to accept his POV a little more then other people's.

Btw, I myself didnt believe any of this until two days ago, then Phil explained it to me in one of his replies. Only then did the light bulb went off in my head and said: "aahhhhh ... so thats the difference" ;p

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

 

[Erik Lindahl]

No, what I am saying is a X3 pixel will not have the color
"bleeding" like a mosaic chip will. Each image pixel will store its
own RGB w/o other pixels telling it what is should look like
(Bayer).

Sure, but the multi-level photosites will certainly have other problems not present in the single-layer bayer sensors.

None of these steps are perfect; for example you
are completely ignoring noise.

I havent, because all sensors have to deal with noise, X3 is no
different. But I do not believe better noise was a Foveon claim. As
a matter of fact, I still like to know why Sigma is only at ISO 400
with everyone else is already at ISO 1200 or more?

Well, an obvious problem is that filters aren't perfect, so the lowest layer will see a lower photon count that the medium and top layers, and thus more noise. That's not a problem with a single-layer sensor.

You are also ignoring any issues in
color separation. What happens to a photon that is absorbed exactly
halfway between two of the color layers (say green and blue.) Will
it be counted as green, blue, neither, both?

And how is this issue a non issue with a mosaic sensor? How will
the mosaic BGBG row handle this any better then the X3?

You use an anti-alias filter in front of the sensor that spreads the photons to multiple photosites, so they will see about 50% each of that wavelength simply by statistics.

With the foveon sensor it's probably a difficult job to tune the absorbance as a function of depth so that you get equal signals in two layers in that case. This is one reason you shouldn't trust resolution or color from a single preproduction demo chip. Why do you think it has taken them months to get it into production since they announced the chip?

It is clear that a 4mp Foveon resolves much better then any 6mp
mosaic chips, and even beats the so call Fuji "12mp".

Sigh. No it shows that a 2048 pixel high X3 has better resolution
than a 2048 pixel high D60/D100. In other words, if the X3 sensor
used had the same 2/3 aspect ratio of the other sensors, it would
also be 6MP. Perhaps you missed the part where Phil says....

And in such case Phil would have taken the res chart with a 3:2
aspect instead of a 1:1 aspect. It would not have changed the
outcome. All of the sensors we compared here are of 2048 or similar
vertical pixels.

Yes, but that's exactly the point. The limiting factor is how many transistors you can squeeze onto the CMOS, and the X3 design requires almost three times as many transistors per photosite. In other words, a bayer sensor will probably always have a factor 2-3 more photosites, so that's what it should be compared to.

the chip I tested has more vertical resolution and will therefore perform better on this resolution chart.

Phil's note is to warn us not to use that res. chart to say the F7
chip will be as good. But for the 2048 vertical the res chart is
still vaild.

Yes, but you cannot buy that sensor. I'm sure a 25 Mpixel bayer sensor would be even better, but you can't buy that either. In fact, you can't buy ANY foveon X3 sensor yet.

When/if Foveon introduces a 6Mp sensor the bayer sensors will probably be at 20Mp with the same transitor/area count.

But Phil himself have already stated many times that in order for a
mosaic 6mp sensor to produce the image quality of the F3, you would
have to strink the 6mp image down to 1.5mp. Since he first hand
experience with the sensor, I tend to accept his POV a little more
then other people's.

That's simply not true. The luminance (light) information is much more important to the eye than the chrominance (color), and that's the whole idea with bayer interpolation. If the statement you cite was true, the foveon X3 with 2048 vertical pixels should show twice the resolution in the chart compared to the similar bayer sensors - obviously it doesn't.

Btw, I myself didnt believe any of this until two days ago, then
Phil explained it to me in one of his replies. Only then did the
light bulb went off in my head and said: "aahhhhh ... so thats the
difference" ;p

Phil is very fond of the X3, and for all I know it might be a really nice sensor, but bayer interpolation is a lot more advanced mathematics than just averaging four neighboring pixels.

Cheers,

Erik

 

[JoePhoto]

I am not writing as a tech, but as a layperson. Your post seems at
least to contradict itself.

OK. For the lay person:
1. More MP does not equal more quality

If everything else is equal ... YES it DOES ... (especially if you are dealing in "big" prints) ... albeit in smaller prints it is less noticable.

2. X3 and Bayer sensors are different enough that there is no
simple description of how they relate to each other on a MP x MP
basis.
3. Buy the one that produces the images that look best to you.

What I've been able to get out of listening to techs is that the
Foveon chip gives you true data and the bayer chip does much more
interpolation.

First X3 does not give you "true" data.

It will give you a single "white" pixel ... accurately and without interpolation ... I would call that "true" data.

It only gives you data from
3 different color channels from a single location.

That is GOOD ENOUGH for me .....

Mosaic sensors
use data from a single location plus some additional nearby
locations.

And if your subject is fine enough "detail" that the nearby senors don't see it ... then WHERE is your "data" that you just stated you "needed" ???

However, for comparable price, you can currently get
twice as many mosaic locations.

Yes ... BUT ... you need 3X-4X the number of additional mosaic locations ... (4 X 3.54 = 14mps). (Albeit there is apparently some new cameras that approach that .. so we do indeed need the "next" level of Foveon already.)

Which produces better photos? See 1, 2, & 3 above.

Well here you lose me. My point is that the standard rating
exagerates the existing bayer pattern chip.

Since the standard rating is for mosaic sensors, it better to say
that the X3 technology is "underrated". I'm not going to get into a
discussion as exactly by how much because the only reasonable
answer is "it depends."

How different companies assemble their mosaic is beside the point.

No, it's precisely the point. Theoretical cameras can only take
theoretical pictures. Exactly how much data a camera recovers from
its sensor depends on the implementation. It even depends on the
type of scene that you are photographing.

IOW the ratings are misleading.

That's like saying a car's engine displacement is misleading
because it does not tell you how fast it goes. (X3 = turbo? ;-).
It's only misleading if you don't know what it means.

The top of your post and the bottom of your post are in
contradiction. I realise you disagree, but that's the way it reads.

The point I was trying to clarify is that there is only one
"interpolation" step performed. Your description implied that
both resolution and color were interpolated separately and that was
inaccurate.

--
Erik

--
Thanks for reading .... JoePhoto

( Do You Ever STOP to THINK --- and FORGET to START Again ??? )

 

[Dana Jacobsen]

In my opinion, Erik has presented the information clearly and I think he is right on just about everything he's said. Just my two cents.

I am not writing as a tech, but as a layperson. Your post seems at
least to contradict itself.

OK. For the lay person:
1. More MP does not equal more quality

If everything else is equal ... YES it DOES ... (especially if you
are dealing in "big" prints) ... albeit in smaller prints it is
less noticable.

I think Erik's point was to say that almost always, everything is not equal. There are a lot of other factors. The typical lay person believes more megapixels must mean better quality, as if no other factors came into play at all. I find that even technically savvy people fall into this trap. A 4MP cheapo camera with a plastic lens must take better pictures than a 3MP D30 with 50/1.4 lens because 4 > 3.

2. X3 and Bayer sensors are different enough that there is no
simple description of how they relate to each other on a MP x MP
basis.
3. Buy the one that produces the images that look best to you.

But what would be the fun in that? Don't we have to endlessly speculate with partial information about subjects we don't know well? Working desperately to convince people we don't know of our point because it is crucial that they get it?

What I've been able to get out of listening to techs is that the
Foveon chip gives you true data and the bayer chip does much more
interpolation.

First X3 does not give you "true" data.

It will give you a single "white" pixel ... accurately and without
interpolation ... I would call that "true" data.

It only gives you data from
3 different color channels from a single location.

That is GOOD ENOUGH for me .....

Well, if the colors were dark red, red, and pink, I don't think you'd be very happy. Part of Erik's later posts is that silicon sensors are a little trickier than color filters. But overall I am not sure why Erik is concentrating on this so much -- barring all the fun complications involved in real products, this seems right. The X3 gathers RGB data for a pixel, while the CFA sensor gathers one channel. Later the other channels are produced via interpolation. This interpolation is not as brain-dead as you think it is however.

Mosaic sensors
use data from a single location plus some additional nearby
locations.

And if your subject is fine enough "detail" that the nearby senors
don't see it ... then WHERE is your "data" that you just stated you
"needed" ???

...and if the detail were finer than the X3 resolution, then what? It works both ways. You can't have everything (where would you put it?). This is going on the idea that a CFA sensor will have higher resolution than a Foveon at some point in time. Again barring implementation issues, sure, we'd all like to have the 11MP Canon 1Ds be Foveonesque and gather all three channels at each site. But it doesn't seem likely right now.

However, for comparable price, you can currently get
twice as many mosaic locations.

Yes ... BUT ... you need 3X-4X the number of additional mosaic
locations ... (4 X 3.54 = 14mps). (Albeit there is apparently some
new cameras that approach that .. so we do indeed need the "next"
level of Foveon already.)

You're kidding, right? You seriously think a Foveon sensor is the equal of a CFA sensor with 3 to 4x more pixels? Have you listened to anything Erik has said at all? Comments like this (Foveon = 4x Bayer filter) is why people like me post, which leads to other people claiming I'm "anti-Foveon" and a naysayer.

If you say less than 1.4x, I think you're too low. If you say more than 2.5x, I think you don't understand how CFA filters work. In-between we can have interesting discussions. My informed opinon though -- it's not like this stuff is my job. Oh, and if you say fewer chances for artifacts, you're right -- a key advantage.

 

[Dana Jacobsen]

I thought I'd put my comments in here even though I'm not the person you're replying to.

Thank you, and I think you are saying that theoretically the 3.54Mp
Foveon should produce a print that is equal to the 6Mp Bayer?

Theoretically, yes. The numbers I've seen that seem well thought out say about 1.6x better resolution, so the 3.43MP SD-9 ought to produce images at least as sharp as a Bayer filter camera of 5.5MP. Then you should add some for the lack of artifacts and increased chroma resolution, perhaps leading to 1.8x better. So yes, it should look at least equal to today's 6MP cameras. Note that there are a large number of other things, mainly noise and sensitivity, not to mention the camera itself.

BTW, I think 3.54MP is the number of total pixels, while 3.43MP is the number captured (2268x1512), but I don't see current information about that. The Sigma UK press release actually says 3.3MP. Given these numbers, I would guess the Sigma uses the F7 sensor which the Foveon site says is 3.54MP total and 3.43MP effective. We're quibbling over nothing of course since the "Bayer equivalent" number is so completely fuzzy.

That
is darn good......can't wait for the 6Mp Foveon...without the Sigma
mount.

Oh yes, that would be very nice!

 

[JoePhoto]

In my opinion, Erik has presented the information clearly and I
think he is right on just about everything he's said. Just my two
cents.

Well ... I only have ONE cent ... but I don't agree with any of it.

I am not writing as a tech, but as a layperson. Your post seems at
least to contradict itself.

OK. For the lay person:
1. More MP does not equal more quality

If everything else is equal ... YES it DOES ... (especially if you
are dealing in "big" prints) ... albeit in smaller prints it is
less noticable.

I think Erik's point was to say that almost always, everything is
not equal. There are a lot of other factors. The typical lay
person believes more megapixels must mean better quality, as if no
other factors came into play at all. I find that even technically
savvy people fall into this trap. A 4MP cheapo camera with a
plastic lens must take better pictures than a 3MP D30 with 50/1.4
lens because 4 > 3.

Of course you are correct that often things are "not" equal ... BUT; we are addressing the specific subject of X3 technology vs Bayer/Mosaic technology. So for purposes of comparason ... we can, (and must), assume they are in an "equal" body/lens situation.

Of course you could have an X3 in a cheap/bad lens and produce poor results ... but the same would be said for a 100meg Bayer with a cheap lens. So that is irrelevant to the discussion.

What I've been able to get out of listening to techs is that the
Foveon chip gives you true data and the bayer chip does much more
interpolation.

First X3 does not give you "true" data.

It will give you a single "white" pixel ... accurately and without
interpolation ... I would call that "true" data.

It only gives you data from
3 different color channels from a single location.

That is GOOD ENOUGH for me .....

Well, if the colors were dark red, red, and pink, I don't think
you'd be very happy. Part of Erik's later posts is that silicon
sensors are a little trickier than color filters. But overall I am
not sure why Erik is concentrating on this so much -- barring all
the fun complications involved in real products, this seems right.
The X3 gathers RGB data for a pixel, while the CFA sensor gathers
one channel. Later the other channels are produced via
interpolation. This interpolation is not as brain-dead as you
think it is however.

Once again ... you are making assumptions that X3 may produce poor quality output color. And of COURSE YOU ARE CORRECT. I am not for a minute arguing that X3 will indeed be any good. It may be WORTHLESS. That all remains to be seen. At the moment we can only discuss "theoritical" possibilities. And again ... we are SPECIFICALLY discussing the "resolution" possibilities of the X3 technology - and comparing it to Bayer/Mosiac technology.

Mosaic sensors
use data from a single location plus some additional nearby
locations.

And if your subject is fine enough "detail" that the nearby senors
don't see it ... then WHERE is your "data" that you just stated you
"needed" ???

...and if the detail were finer than the X3 resolution, then what?
It works both ways. You can't have everything (where would you put
it?). This is going on the idea that a CFA sensor will have
higher resolution than a Foveon at some point in time. Again
barring implementation issues, sure, we'd all like to have the 11MP
Canon 1Ds be Foveonesque and gather all three channels at each
site. But it doesn't seem likely right now.

However, for comparable price, you can currently get
twice as many mosaic locations.

Yes ... BUT ... you need 3X-4X the number of additional mosaic
locations ... (4 X 3.54 = 14mps). (Albeit there is apparently some
new cameras that approach that .. so we do indeed need the "next"
level of Foveon already.)

You're kidding, right? You seriously think a Foveon sensor is the
equal of a CFA sensor with 3 to 4x more pixels? Have you listened
to anything Erik has said at all?

Of course I did ... and DID NOT AGREE.

Comments like this (Foveon = 4x
Bayer filter) is why people like me post, which leads to other
people claiming I'm "anti-Foveon" and a naysayer.

If you say less than 1.4x, I think you're too low. If you say more
than 2.5x, I think you don't understand how CFA filters work.
In-between we can have interesting discussions. My informed opinon
though -- it's not like this stuff is my job. Oh, and if you say
fewer chances for artifacts, you're right -- a key advantage.

Look ... a 3.54mps X3 has 3X3.54 (10.62) light sensors which together can provide 3.54m points of "white" light ... without any interpolation; (albeit "mixing" instead of its RGB levels). This 3.54 can be "interpolated" further if it needs/wants to be.

A 6mps Bayer only has 6/4, (1.5mps), "points" of white light. That is ALL. It is already interpolated to provide the advertised 6mps of "white" points of light unless you are using the RAW output.

Therefore .... it remains to be seen if it is true .. but the POTENTIAL is there for the X3 to be DOUBLE the resolution of existing 6mps. (And without color morie/artifacting).

Of course .. and absolutely ... THERE ARE OTHER FACTORS that are UNKNOWN at this time.

I said already that these "other" factors may render the X3 unusable and impractical. But there is no reason to "ass-u-me" that these other factors are already gonna kill it when we simply DON'T KNOW.

--
Thanks for reading .... JoePhoto

( Do You Ever STOP to THINK --- and FORGET to START Again ??? )

 

[Chris Brown]

Look ... a 3.54mps X3 has 3X3.54 (10.62) light sensors

Incorrect. It has 3.54m sensors, and this is vitally important. There are not 3 sensors at each location magically measuring "red", "green" and "blue" light. There is a sensor which is charged according to the light that falls on it. However, this sensor is not a 2-dimensional rectangle, it has depth and the charge generated by light falling on it will be distributed throughout this depth.

It so happens that the charge varies with depth according to the wavelength of the light, and Foven takes advantage of this by sampling the "bucket" at 3 different points.

However, it's nothing like the highly sanitised view you seem to be taking where there are nice distinct layers, marked "red", "green" and "blue", with nice little boundaries between them. In reality, it's much, much more messy than that. The absorbsion is a statistical process, and the "red" light will actually be distributed throughout the "bucket", mingled with other parts of the spectrum. It's just that there's likely to be more of it at a given depth.

When you sample this "red" light, you're also getting plenty of yellow, green and even blue light thrown in for good measure. The clever bit is being able to tidy this mess up, by comparing the "red" sample with the "green" and "blue" samples to try and work out what really is red, and what is just garbage that leaked from elsewhere in the bucket.

If you just took the direct outputs from each of the 3 samples in a Foveon pixel, and made an image based on those with no processing, you'd most likely end up with a sludgy grey/brown mess.

So whilst you may have to do lots of tricky maths with a mosaic sensor to get an estimate of the colour at each location, you laso have to do lots of tricky maths with the Foveon sensor, and what you get is, again, merely an estimate of the colour at any given position. The nature of the maths is different, but they're both still statistical estimates. Which one is better will be revealed by the final images.

 

[Veniamin Kostitsin II]

General consensus among people profecient and experienced in digital photography is, that foveon will be comparable with sensors 1.4 to 1.6 times larger than 3 mp. that means that it will produce the same level of detail as 4.2 - 4.8 megapixel bayer sensor. the colour fidelity of foveon will probably be higher, however it will not be able to capture more detail than physically possible. so far nobody was able to prove this statement wrong. naturally it all depends on the amount of detail in the scene, however how often will you take picture of a plain sheet of plastic which has little detail in it ?

cheers

veniamin kostitsin
http://www.digitalimage.at/

Up to this point in this age of digital technology I found that all
things being equal go with more pixels if you want to print your
pictures larger. That is the reason I have a 6mp DSLR. So my 6mp
file is 3024x2016 if I want to print a 18x12in. (Epson 1280) my
printing resolution is 168 pixel/inch.... barely acceptable. So
here is the question, the Sigma Foveon is 2268x1512 so if I want to
print 18x12in. will be at 125pixels/inch. Will the Foveon X3 CMOS
produce a much better 18X12in. print compared to the present 6mp
DSLR?

Boris

--
Veniamin Kostitsin II
http://www.digitalimage.at/

 

[Veniamin Kostitsin II]

lumina information is captured at every site on the imager. punctum.

 

[Laurence Matson]

That's right.

Go look.

cheers

veniamin kostitsin
http://www.digitalimage.at/

Up to this point in this age of digital technology I found that all
things being equal go with more pixels if you want to print your
pictures larger. That is the reason I have a 6mp DSLR. So my 6mp
file is 3024x2016 if I want to print a 18x12in. (Epson 1280) my
printing resolution is 168 pixel/inch.... barely acceptable. So
here is the question, the Sigma Foveon is 2268x1512 so if I want to
print 18x12in. will be at 125pixels/inch. Will the Foveon X3 CMOS
produce a much better 18X12in. print compared to the present 6mp
DSLR?

Boris

--
Veniamin Kostitsin II
http://www.digitalimage.at/

 

[Erik Magnuson]

And in such case Phil would have taken the res chart with a 3:2
aspect instead of a 1:1 aspect. It would not have changed the
outcome. All of the sensors we compared here are of 2048 or similar
vertical pixels.

Boy, either you have foveonitus or you just don't understand Phil'd resolution charts. Even on a 3:2 camera, his resultion char is SQUARE! He always fills the frame vertically!

Phil's note is to warn us not to use that res. chart to say the F7
chip will be as good. But for the 2048 vertical the res chart is
still vaild.

But it's for a 6MP effective sensor (if it were 3:2)
--
Erik

 

[G. Lentz]

Just thought I'd follow up on my follow-up:

A few things came to mind after I wrote this.

1. Since the Foveon is imaging to (in theory, anyway) a discrete photosite, in my mind, that makes it much more of a 'digital' sensor than the Bayer, at least in terms of its 2D geometry. For this reason, there is the potential for aliasing issues in the Foveon that are perhaps better masked by the 'distributed' nature of the Bayer.

2. Assuming the above has been adequately dealt with, the Foveon should be a lot more resistant to noise during long exposures (e.g. astrophotography) than the Bayer, since infrequent photons would be striking a single (correct) photosite, and not producing spurious off-colored ghosts in the matrixed pixel design.

3. Since the photosites are discrete, there is the ability to 'tune' each and every one for correctness.

4. I wonder what advantages there would be, if any, to Foveon adopting a
pixel layout like the one on the Super-CCD??

Of course, it's all guessing until we start seeing reviews from production cameras, yeah?

-gl

Technically, the Foveon X3 has an 'equivalent' color resolution of
3x3.54Mp=10.62Mp compared to a Bayer sensor. However, you
basically have three pixels (RGB) occupying the same physical
two-dimensional space on the image plane. While this can provide a
'truer' image, it may or may not automatically translate into
increased resolution.

That said, the Bayer must conversely have a lower resultant
resolution, necessitated by the fact that you're using adjacent
pixels of differing colors that are crunched using complex imaging
algorithms in an attempt to derive a final image that more or
less approximates the original source. The success or failure
seems to depend more on the quality of the algorithm than the
sensor resolution, as we see with the Canon G2's apparent better
image quality using a 4Mp sensor than some cameras using 5 or even
6Mp.

As far as final output on paper, one would think one could upsample
the Foveon image using a dithering algorithm and achieve a similar
result to a conventional 'higher resolution' sensor.

-gl

Up to this point in this age of digital technology I found that all
things being equal go with more pixels if you want to print your
pictures larger. That is the reason I have a 6mp DSLR. So my 6mp
file is 3024x2016 if I want to print a 18x12in. (Epson 1280) my
printing resolution is 168 pixel/inch.... barely acceptable. So
here is the question, the Sigma Foveon is 2268x1512 so if I want to
print 18x12in. will be at 125pixels/inch. Will the Foveon X3 CMOS
produce a much better 18X12in. print compared to the present 6mp
DSLR?

Boris

 

[Veniamin Kostitsin II]

Yes, that's right. i did look.

Go look.

cheers

veniamin kostitsin
http://www.digitalimage.at/

Up to this point in this age of digital technology I found that all
things being equal go with more pixels if you want to print your
pictures larger. That is the reason I have a 6mp DSLR. So my 6mp
file is 3024x2016 if I want to print a 18x12in. (Epson 1280) my
printing resolution is 168 pixel/inch.... barely acceptable. So
here is the question, the Sigma Foveon is 2268x1512 so if I want to
print 18x12in. will be at 125pixels/inch. Will the Foveon X3 CMOS
produce a much better 18X12in. print compared to the present 6mp
DSLR?

Boris

--
Veniamin Kostitsin II
http://www.digitalimage.at/

--
Veniamin Kostitsin II
http://www.digitalimage.at/

 

[G. Lentz]

Well, since we all ARE largely pontificating about an as yet non-product.... ;-)

-gl

 

[Boris]

Up to this point in this age of digital technology I found that all
things being equal go with more pixels if you want to print your
pictures larger. That is the reason I have a 6mp DSLR. So my 6mp
file is 3024x2016 if I want to print a 18x12in. (Epson 1280) my
printing resolution is 168 pixel/inch.... barely acceptable. So
here is the question, the Sigma Foveon is 2268x1512 so if I want to
print 18x12in. will be at 125pixels/inch. Will the Foveon X3 CMOS
produce a much better 18X12in. print compared to the present 6mp
DSLR?

Boris

I printed two of the sd9 samples from Phil's site

http://www.dpreview.com/news/0209/02092502sigmasd9samples.aspthe The yellow car w/branches and the apple.

(native size 2268x1512) 226.8 pix.in 10x6.667in. ............13x8.667in. 174.462 pix/in

Then I printed two similar pictures but not exact, almost the same lighting but different lens and apeture from my s2. In place of an apple I had a picture of a tomato

(native size 3024x2016) 302 pix/in 10x6.667 .......13x8.667 232.615

All on a Epson 1280 and high gloss paper. I applied a auto level and sharp 200%-.03-0

to all in Photoshop.

At 10x6.667in. both ( car with tree branches) looked excellent, excellent hard to find fault with either sensor..... I guess that is saying a lot for the Foveon chip..... the other picture of the apple didn't print well at all, the noise in the backgound didn't print even, not exactly blotchy just not smooth. I didnt print that one to 13x8.667in.

At the 13x8.667in. size I would give the nod (large nod) to the S2 . If you scrutinize both 13x8.667in. prints you'll see that you are pushing both chips, but even more so with the sd9. I don't mean this to be a scientific test just a quick observation. And thank you all for explaining the Feveon to me.

Boris
http://public.fotki.com/borysd/

 

[Jimmy Chen]

Sure, but the multi-level photosites will certainly have other
problems not present in the single-layer bayer sensors.

Other then heat, what other additional problems?

Well, an obvious problem is that filters aren't perfect, so the
lowest layer will see a lower photon count that the medium and top
layers, and thus more noise. That's not a problem with a
single-layer sensor.

Same can happen to signle layer sensor. Remember, the given to this problem is the photon is exactly between the green and blue, given that green is green and blue is blue and is no different for an X3 chip as it is to a mosaic chip. If the X3 chip cannot capture that photon, how is the GBGB Bayer is going to capture that photon?

You use an anti-alias filter in front of the sensor that spreads
the photons to multiple photosites, so they will see about 50% each
of that wavelength simply by statistics.

Yeah, and then you have to go guess what the original 50% was.

Yes, but that's exactly the point. The limiting factor is how many
transistors you can squeeze onto the CMOS, and the X3 design
requires almost three times as many transistors per photosite. In
other words, a bayer sensor will probably always have a factor 2-3
more photosites, so that's what it should be compared to.

You seem to foget that larger chips also are harder to make.

Yes, but you cannot buy that sensor. I'm sure a 25 Mpixel bayer
sensor would be even better, but you can't buy that either. In
fact, you can't buy ANY foveon X3 sensor yet.

Nor can you buy any FF CMOS DSLR at this point. Moot point.

When/if Foveon introduces a 6Mp sensor the bayer sensors will
probably be at 20Mp with the same transitor/area count.

And 3 times larger in surface area, which also means many times harder to make and much more expensive. 1Ds is a perfect example. 2x the mp count, but asking for almost 5x the in price.

That's simply not true. The luminance (light) information is much
more important to the eye than the chrominance (color), and that's
the whole idea with bayer interpolation. If the statement you cite
was true, the foveon X3 with 2048 vertical pixels should show twice
the resolution in the chart compared to the similar bayer sensors

• obviously it doesn't.

What is not true, what Phil said, or what you are saying? You can read what Phil said for youself here ...
http://www.dpreview.com/forums/read.asp?forum=1014&message=3432204

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

 

[rick66883]

Up to this point in this age of digital technology I found that all
things being equal go with more pixels if you want to print your
pictures larger. That is the reason I have a 6mp DSLR. So my 6mp
file is 3024x2016 if I want to print a 18x12in. (Epson 1280) my
printing resolution is 168 pixel/inch.... barely acceptable. So
here is the question, the Sigma Foveon is 2268x1512 so if I want to
print 18x12in. will be at 125pixels/inch. Will the Foveon X3 CMOS
produce a much better 18X12in. print compared to the present 6mp
DSLR?

Boris

I printed two of the sd9 samples from Phil's site
http://www.dpreview.com/news/0209/02092502sigmasd9samples.aspthe
The yellow car w/branches and the apple.

(native size 2268x1512) 226.8 pix.in 10x6.667in.
............13x8.667in. 174.462 pix/in

Then I printed two similar pictures but not exact, almost the same
lighting but different lens and apeture from my s2. In place of an
apple I had a picture of a tomato

(native size 3024x2016) 302 pix/in 10x6.667 .......13x8.667 232.615

All on a Epson 1280 and high gloss paper. I applied a auto level
and sharp 200%-.03-0

to all in Photoshop.

At 10x6.667in. both ( car with tree branches) looked excellent,
excellent hard to find fault with either sensor..... I guess that
is saying a lot for the Foveon chip..... the other picture of the
apple didn't print well at all, the noise in the backgound didn't
print even, not exactly blotchy just not smooth. I didnt print that
one to 13x8.667in.

At the 13x8.667in. size I would give the nod (large nod) to the S2
. If you scrutinize both 13x8.667in. prints you'll see that you are
pushing both chips, but even more so with the sd9. I don't mean
this to be a scientific test just a quick observation. And thank
you all for explaining the Feveon to me.

Boris
http://public.fotki.com/borysd/

Boris I trust your opinion but many times I have seen same scene comparisons based on different cameras and you might be surprised how similar they looked and sometimes the pixel and resolution size differed.All I know is when I look at the foreveron, it has something the others don't SOMETIMES.Maybe when all maximized efficiences are in place with monitor and printer, the results might be different.

 

[Ron Parr]

So whilst you may have to do lots of tricky maths with a mosaic
sensor to get an estimate of the colour at each location, you laso
have to do lots of tricky maths with the Foveon sensor, and what
you get is, again, merely an estimate of the colour at any given
position. The nature of the maths is different, but they're both
still statistical estimates. Which one is better will be revealed
by the final images.

The math is not so tricky. It's described in this Kodak patent:

http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=/netahtml/search-bool.html&r=1&f=G&l=50&co1=AND&d=ft00&s1= '5668596'.WKU.&OS=PN/5668596&RS=PN/5668596

It basically involves one matrix multiplication per pixel. Since you need to do at least one such multiplication anyway for white balance, you can just premultiply your white balance matrix by the correction matrix. In actual implementation, it boils down to changing a few numbers in ROM. There is no extra computation done.

For a Bayer pattern sensor, the color filters and substrate also behave stochastically. The difference is that there is somewhat more overlap with the Foveon approach.

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

 

[Ron Parr]

lumina information is captured at every site on the imager. punctum.

Yes, but some pixels such as blue only see a tiny amount of that information since blue's contribution to luminance is quite small. The total luminance information the Bayer pattern sensor sees is about 40% of what's striking the sensor before it goes through the filters.

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

 

[Ron Parr]

No.

The advantages are in better reproduced colour and lack of moiré
patterns in certain situations.

There is no real gain in sharpness or picture resolution.

The goal is to have improved sharpness and picture resolution. Moire is a symptom of the undelrying problem Foveon is trying to fix. It's not the problem.

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