The PetaPixel View of the "Failure" of the Foveon

[xpatUSA]

Link posted for what it's worth!

PetaPixel is known for occasional (!) plagiarism but an interesting read nonetheless, at least for me:

Foveon: The Clever Image Sensor That Has Failed to Catch On

How Sigma's Foveon sensor design works and why the clever idea has failed to catch on so far in the digital camera industry.

petapixel.com

 

[D Cox]

It's quite well written -- I didn't see any obvious mistakes. They might have mentioned that the Quattro cameras can save DNG raw files, which makes life much easier.

Nice to see a good number of Sigma camera users commenting.

Don

 

[Tom Schum]

Well-written and generally factual, so far as I can tell.

 

[Jeffry7]

I have several problems with this article.

The first, and easiest, is the dismissal of Sigma as a camera company. To do this, you have to ignore the fp line. These are cameras. People are buying them. Sigma is a camera company.

The second is resolution BS. Canon's most recent camera aimed at Pros is the R3. It has a sensor pitch of 6 microns for a 24 megapixels FF sensor. The R5, the high rez sibling to the R6, has a pitch of 4.4. The big dog here is actually the R7, with a pitch of 3.2, but an APS-C sensor size. It is aimed squarly at enthusiasts.

The sd Quattro cameras have a top layer pitch of 4.3.

If you really want hi rez, get a Nikon 1 camera. The V3 has a pitch of 2.8. It is also a discontinued line.

If Sigma just makes a full frame camera with the same rez as the top layer of the sd Quattro, then they will have higher resolution than the pro cameras and be quite competitive with the consumer and enthusiast cameras. And that is before the advantage the 3 layer sensor brings.

Finally there is the idea that Sigma is trying to compete with Nikon, Sony and Canon. They aren't. The Sigma FP should have made that clear. All the other camera manufacturers make 'weird' cameras. They differ in some way from the idealized camera Canon, Nikon, and Sony are pursuing. We buy them partly despite that, and partly because of it.

 

[Tom Schum]

I have several problems with this article.

The first, and easiest, is the dismissal of Sigma as a camera company. To do this, you have to ignore the fp line. These are cameras. People are buying them. Sigma is a camera company.

If Sigma just makes a full frame camera with the same rez as the top layer of the sd Quattro, then they will have higher resolution than the pro cameras and be quite competitive with the consumer and enthusiast cameras. And that is before the advantage the 3 layer sensor brings.

Sigma fp L Standard Full frame
36x24mm = 864 sq mm
9520x6328 = 60,242,560 pixels
36mm / 9520 = 3.7815126 micron pixel pitch
24mm / 6328 = 3.7926675 micron pixel pitch
Average = 3.787 micron pixel pitch

Sigma sd Quattro
23.5x15.6mm = 366.6 sq mm
5424x3616 = 19,613,184 pixels
23.5mm / 5424 = 4.332596 micron pixel pitch
15.6mm / 3616 = 4.314593 micron pixel pitch
Average = 4.3236 micron pixel pitch

However, if we assume 3-layer images to be equivalent to a CFA of twice the number of pixels, the pixel pitch of a competitive Bayer would be 4.32u / sqrt 2 = 3.0572 micron and this would be a full frame Bayer sensor of 11,775 x 7850 = 92.4mp.

And, these densities are on the horizon. As it is, the 60mp sensor of the fp L has more dynamic range than the sd Quattro sensor, as one might expect.

Meanwhile, a FFF with sd Quattro pixel pitch would be 8326 x 5551 = 46.2mp.

The existing Sigma plan, if I understand it correctly, is for a 20mp full frame 1:1:1 Foveon. Is this correct?

If so, pixel layout will be something like 5477 x 3652 = 20.002mp

Pixel pitch will be 0.036 / 5477 = 6.573 microns.

This compares favorably with the SD15 pixel pitch of 7.84 microns, and it is larger than the SD1M pixel pitch of 5.00 microns. Average of these is 6.42 microns.

It's interesting that the pixel pitch for the new sensor is just slightly larger than midway between SD15 and SD1M!

If Sigma can make a product out of this, image quality ought to be pretty darn good.

 

[Doppler9000]

I have several problems with this article.

The first, and easiest, is the dismissal of Sigma as a camera company. To do this, you have to ignore the fp line. These are cameras. People are buying them. Sigma is a camera company.

The second is resolution BS. Canon's most recent camera aimed at Pros is the R3. It has a sensor pitch of 6 microns for a 24 megapixels FF sensor.

Some market segments require speed. The full frame Foveon, if it is ever made, will not be even in the same galaxy as the R3 speed-wise, which is the key reason the resolution is 24 MP.

The R5, the high rez sibling to the R6, has a pitch of 4.4. The big dog here is actually the R7, with a pitch of 3.2, but an APS-C sensor size. It is aimed squarly at enthusiasts.

The chip in the fp L and Sony A7R4 has a 3.76 micron pitch.

The sd Quattro cameras have a top layer pitch of 4.3.

If you really want hi rez, get a Nikon 1 camera. The V3 has a pitch of 2.8. It is also a discontinued line.

Hopefully the Nikon 1 isn’t the business model Sigma is emulating here. Interestingly, the Nikon 1 forum has roughly the same amount of traffic as this one.

If Sigma just makes a full frame camera with the same rez as the top layer of the sd Quattro, then they will have higher resolution than the pro cameras and be quite competitive with the consumer and enthusiast cameras.

The recently-announced Fuji X-H2 is 40 MP on APS-C. The Foveon will likely be competing against circa 100 MP resolution full frame sensors. Not sure if the Sigma will be “quite competitive” on resolution. It would seem that it will be well behind in noise/low light performance and autofocus speed.

And that is before the advantage the 3 layer sensor brings.

Finally there is the idea that Sigma is trying to compete with Nikon, Sony and Canon. They aren't. The Sigma FP should have made that clear.

All the other camera manufacturers make 'weird' cameras.

”Weird” compared to what?

They differ in some way from the idealized camera Canon, Nikon, and Sony are pursuing. We buy them partly despite that, and partly because of it.

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

 

[Ceistinne]

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

Doppler9000,

I have no doubt that they will have that figured out. They have 60 odd years of experience behind them & getting better and better as they go along.

S

 

[D Cox]

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

Doppler9000,

I have no doubt that they will have that figured out. They have 60 odd years of experience behind them & getting better and better as they go along.

S

Yes. Experience tells.

I'm not sure that any of the Sigma Foveon cameras have been failures. They are almost all sold out. If you make, say, 4000 cameras and people buy them all, that's a success in my book.

Don

 

[Iain G Foulds]

D Cox wrote

I'm not sure that any of the Sigma Foveon cameras have been failures. They are almost all sold out. If you make, say, 4000 cameras and people buy them all, that's a success in my book.

Don

… Right. If I make 4000 cameras and people buy them all, it is a success.

… However, if a major international company like Sigma, after massive design, engineering, and manufacturing costs, sells only 4000 cameras and remains unknown in the photographic world, it is a massive failure.

 

[rick decker]

Well certainly a full-size 20mp 3-layer camera would hold more appeal than the Quattro. That said, my feeling is that if they can't do (really do) high iso and not just ISO400, they won't get great sales. There is are good reasons why the FP line is so successful and if it didn't have high ISO (12500) it wouldn't have the numbers it has now. That is the main reason I got an FP and that is the reason I got one.

 

[D Cox]

Well certainly a full-size 20mp 3-layer camera would hold more appeal than the Quattro. That said, my feeling is that if they can't do (really do) high iso and not just ISO400, they won't get great sales. There is are good reasons why the FP line is so successful and if it didn't have high ISO (12500) it wouldn't have the numbers it has now. That is the main reason I got an FP and that is the reason I got one.

Yes, any future camera must give Foveon quality at ISO 800, which is a good setting for general walkabout photography. That means two stops more than the Merrill. It won't be easy, unless they have found out why the existing X3 sensors are so insensitive (which is a mystery to me).

I'm sure the folk working on this project know all this as well as we do.

I bought the fp mainly because the L mount allows me to use all my classic lenses from film SLRs (except a few that I foolishly sold), and because Sigma's controls and menus are easy to use.

Don

 

[D Cox]

D Cox wrote

I'm not sure that any of the Sigma Foveon cameras have been failures. They are almost all sold out. If you make, say, 4000 cameras and people buy them all, that's a success in my book.

Don

… Right. If I make 4000 cameras and people buy them all, it is a success.

… However, if a major international company like Sigma, after massive design, engineering, and manufacturing costs, sells only 4000 cameras and remains unknown in the photographic world, it is a massive failure.

I don't think the development costs of the fp will have been any more than those of an Art lens. Probably less, as it has no moving parts. Lenses today are extremely complicated, while the fp is rather simple, and so are the various DP models.

However, development of a new sensor is an expensive business.

Don

 

[Doppler9000]

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

Doppler9000,

I have no doubt that they will have that figured out. They have 60 odd years of experience behind them & getting better and better as they go along.

S

I would agree that Sigma has a great track record designing and manufacturing lenses.

Your optimism with regard to their track record when applied to sensor design seems ill-founded, however. Designing sensors has essentially nothing to do with designing lenses. After years of trying and who-knows-how-much money invested, Sigma has had to go back to the drawing board, starting over and prototyping a small version of the full size sensor. If this seems to work, they will begin to prototype a full size version, and develop the associated electronics. How much money has been sunk so far? So it doesn’t seem that they are really getting “better and better” at designing larger, updated sensors, does it?

In the meantime, the camera market continues to contract, with most of the new releases targeted largely toward the video market. Bayer sensors continue to improve, as Sigma stumbles along. I am a Sigma fan and own four of their cameras. I worry they are going to do themselves significant financial harm by chasing a white whale.

And keep in mind that successful companies, with long track records of success, make mistakes and go out of business all the time.

 

[Doppler9000]

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

Doppler9000,

I have no doubt that they will have that figured out. They have 60 odd years of experience behind them & getting better and better as they go along.

S

Yes. Experience tells.

I'm not sure that any of the Sigma Foveon cameras have been failures. They are almost all sold out. If you make, say, 4000 cameras and people buy them all, that's a success in my book.

Don

Did they make a profit? That would be a more realistic measure of success, from a business perspective.

 

[D Cox]

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

Doppler9000,

I have no doubt that they will have that figured out. They have 60 odd years of experience behind them & getting better and better as they go along.

S

Yes. Experience tells.

I'm not sure that any of the Sigma Foveon cameras have been failures. They are almost all sold out. If you make, say, 4000 cameras and people buy them all, that's a success in my book.

Don

Did they make a profit? That would be a more realistic measure of success, from a business perspective.

It would for the DP cameras. As Sigma made a whole series of them, they must have been profitable.

If a camera takes interchangeable lenses, you can sell it at cost, or even at a loss, and make the profits from lenses. What you lose on the swings, you make up on the roundabouts.

Don

 

[Jeffry7]

I have several problems with this article.

The first, and easiest, is the dismissal of Sigma as a camera company. To do this, you have to ignore the fp line. These are cameras. People are buying them. Sigma is a camera company.

The second is resolution BS. Canon's most recent camera aimed at Pros is the R3. It has a sensor pitch of 6 microns for a 24 megapixels FF sensor.

Some market segments require speed. The full frame Foveon, if it is ever made, will not be even in the same galaxy as the R3 speed-wise, which is the key reason the resolution is 24 MP.

Fair enough. Still there are high end cameras with ~20mp sensors and not everyone who wants one is a sports or BiF photographer.

The R5, the high rez sibling to the R6, has a pitch of 4.4. The big dog here is actually the R7, with a pitch of 3.2, but an APS-C sensor size. It is aimed squarly at enthusiasts.

The chip in the fp L and Sony A7R4 has a 3.76 micron pitch.

The sd Quattro cameras have a top layer pitch of 4.3.

If you really want hi rez, get a Nikon 1 camera. The V3 has a pitch of 2.8. It is also a discontinued line.

Hopefully the Nikon 1 isn’t the business model Sigma is emulating here. Interestingly, the Nikon 1 forum has roughly the same amount of traffic as this one.

I have a V3. It is a fun if imperfect camera.

If Sigma just makes a full frame camera with the same rez as the top layer of the sd Quattro, then they will have higher resolution than the pro cameras and be quite competitive with the consumer and enthusiast cameras.

The recently-announced Fuji X-H2 is 40 MP on APS-C. The Foveon will likely be competing against circa 100 MP resolution full frame sensors. Not sure if the Sigma will be “quite competitive” on resolution. It would seem that it will be well behind in noise/low light performance and autofocus speed.

I doubt that will happen. Canon, Nikon, and Sony can make these sensors now if they want, and haven't. Meanwhile, sensor pitch has remained very stable. The Canon 550d, released in 2010, has a 4.3 pitch.

There is a lower limit to a lens systems ability to resolve detail. Sometimes this is referred to as diffraction limited. One measure of this is the Rayleigh Criteria. This gives the angular resolution possible with an idea lens given the diameter of the lens and the wavelength of light. To convert this to spatial resolution, you include the flange distance. Well none of these factors change for a given camera system. Throwing a smaller pitch at a given system eventually gets to diminishing returns.

Smart phones have ridiculously small flange distances and can benefit from smaller pitch. I would imagine mirrorless systems will have smaller pitches because that have smaller flange distances. But not to the degree of smart phones.

Angular resolution - Wikipedia

en.m.wikipedia.org

And that is before the advantage the 3 layer sensor brings.

Finally there is the idea that Sigma is trying to compete with Nikon, Sony and Canon. They aren't. The Sigma FP should have made that clear.

All the other camera manufacturers make 'weird' cameras.

”Weird” compared to what?

Compared to Canon, Nikon, or Sony.

They differ in some way from the idealized camera Canon, Nikon, and Sony are pursuing. We buy them partly despite that, and partly because of it.

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

I have no idea. Since they have done so previously, I would assume yes.

 

[Tom Schum]

I bought the fp mainly because the L mount allows me to use all my classic lenses from film SLRs (except a few that I foolishly sold), and because Sigma's controls and menus are easy to use.

Don

If Sigma offers an L-mount three-layer product it will be a lot easier for existing L-mount users to buy one, since their investment in L-mount lenses and adapters is not going to be lost.

I think in the past the SA mount was a dis-incentive to potential users, since it invalidated practically any previous lens investments they might have made.

L-mount is such a great step forward that even a Sigma APS-C body ought to sell well. Numerous Leica users are lamenting the end of the Leica L-mount APS-C bodies, so the market might be ripe for one. I wouldn't be surprised to see Sigma jump into this.

Meanwhile, can a 20mp full frame Foveon be any better than a fp L image downsized to 1/3 of original pixels? My guess is no, it can't. Colors might be better (in terms of art, not quantitative accuracy) from the Foveon though.

 

[Doppler9000]

I have several problems with this article.

The first, and easiest, is the dismissal of Sigma as a camera company. To do this, you have to ignore the fp line. These are cameras. People are buying them. Sigma is a camera company.

The second is resolution BS. Canon's most recent camera aimed at Pros is the R3. It has a sensor pitch of 6 microns for a 24 megapixels FF sensor.

Some market segments require speed. The full frame Foveon, if it is ever made, will not be even in the same galaxy as the R3 speed-wise, which is the key reason the resolution is 24 MP.

Fair enough. Still there are high end cameras with ~20mp sensors and not everyone who wants one is a sports or BiF photographer.

The R5, the high rez sibling to the R6, has a pitch of 4.4. The big dog here is actually the R7, with a pitch of 3.2, but an APS-C sensor size. It is aimed squarly at enthusiasts.

The chip in the fp L and Sony A7R4 has a 3.76 micron pitch.

The sd Quattro cameras have a top layer pitch of 4.3.

If you really want hi rez, get a Nikon 1 camera. The V3 has a pitch of 2.8. It is also a discontinued line.

Hopefully the Nikon 1 isn’t the business model Sigma is emulating here. Interestingly, the Nikon 1 forum has roughly the same amount of traffic as this one.

I have a V3. It is a fun if imperfect camera.

If Sigma just makes a full frame camera with the same rez as the top layer of the sd Quattro, then they will have higher resolution than the pro cameras and be quite competitive with the consumer and enthusiast cameras.

The recently-announced Fuji X-H2 is 40 MP on APS-C. The Foveon will likely be competing against circa 100 MP resolution full frame sensors. Not sure if the Sigma will be “quite competitive” on resolution. It would seem that it will be well behind in noise/low light performance and autofocus speed.

I doubt that will happen. Canon, Nikon, and Sony can make these sensors now if they want, and haven't. Meanwhile, sensor pitch has remained very stable. The Canon 550d, released in 2010, has a 4.3 pitch.

Do you think the parent sensor tech used in the Fuji won’t be be used in a full frame camera? This is not how the sensor business works, outside a few corner cases. The same wafer is the source sensors for a range of cameras. For example, the 26 MP sensor tech used in the Fuji X-T3 is also found in the 61 MP Sony A7RIV and fp L, the 100 MP Fuji GFX 100x, and the 150 MP Phase IQ4.

The full frame variant of the X-H2 sensor will be ~100 MP. I would bet, barring further Covid supply issues, you will see announcements in the next several months of larger sensors from the same source wafers.

There is a lower limit to a lens systems ability to resolve detail. Sometimes this is referred to as diffraction limited. One measure of this is the Rayleigh Criteria. This gives the angular resolution possible with an idea lens given the diameter of the lens and the wavelength of light. To convert this to spatial resolution, you include the flange distance. Well none of these factors change for a given camera system. Throwing a smaller pitch at a given system eventually gets to diminishing returns.

Using the Sparrow criterion, how about 1.4 gigapixels at f/1.4? Or 357 MP at f/2.8?

Smart phones have ridiculously small flange distances and can benefit from smaller pitch. I would imagine mirrorless systems will have smaller pitches because that have smaller flange distances. But not to the degree of smart phones.

https://en.m.wikipedia.org/wiki/Angular_resolution#The_Rayleigh_criterion

And that is before the advantage the 3 layer sensor brings.

Finally there is the idea that Sigma is trying to compete with Nikon, Sony and Canon. They aren't. The Sigma FP should have made that clear.

All the other camera manufacturers make 'weird' cameras.

”Weird” compared to what?

Compared to Canon, Nikon, or Sony.

They differ in some way from the idealized camera Canon, Nikon, and Sony are pursuing. We buy them partly despite that, and partly because of it.

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

I have no idea.

Indeed.

Since they have done so previously, I would assume yes.

This really isn’t a very compelling argument. The new sensor is clearly presenting significant new challenges.

 

[Doppler9000]

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

Doppler9000,

I have no doubt that they will have that figured out. They have 60 odd years of experience behind them & getting better and better as they go along.

S

Yes. Experience tells.

I'm not sure that any of the Sigma Foveon cameras have been failures. They are almost all sold out. If you make, say, 4000 cameras and people buy them all, that's a success in my book.

Don

Did they make a profit? That would be a more realistic measure of success, from a business perspective.

It would for the DP cameras. As Sigma made a whole series of them, they must have been profitable.

This is a new sensor that is proving challenging - literally back to the drawing board.

If a camera takes interchangeable lenses, you can sell it at cost, or even at a loss, and make the profits from lenses. What you lose on the swings, you make up on the roundabouts.

Doesn’t overall profitability depend, in part, on the cost of bringing the sensor to market? Sigma is putting more money in and admits nothing may come of it.

Don

 

[Ceistinne]

Can Sigma sell enough, profitably, to not lose their shirts on the FFF R&D?

Doppler9000,

I have no doubt that they will have that figured out. They have 60 odd years of experience behind them & getting better and better as they go along.

S

I would agree that Sigma has a great track record designing and manufacturing lenses.

That is true.

Your optimism with regard to their track record when applied to sensor design seems ill-founded, however. Designing sensors has essentially nothing to do with designing lenses. After years of trying and who-knows-how-much money invested, Sigma has had to go back to the drawing board, starting over and prototyping a small version of the full size sensor. If this seems to work, they will begin to prototype a full size version, and develop the associated electronics. How much money has been sunk so far? So it doesn’t seem that they are really getting “better and better” at designing larger, updated sensors, does it?

I did not say they are getting better & better at designing sensors. I believe the sensor they are working on at the moment is the first designed in house by Sigma. I think all the rest were designed by Foveon and used by Sigma.

The new one will only be used if Kazuto Yamaki is satisfied that it will be an absolute major improvement over any X3 sensor to date. I think that is a certainty from various things he has indicated about it. If it turns out to be good enough for him we will see it, if not they will go down a different road.

Having said that I have used almost all versions of the Foveons since the SD9 and still do with the SD 15 occasionally, the SD1M, sd Q & sd Q H along with the dp2 Q with as much success in the right circumstances as with any other cameras such as Nikons & Sonys. The major difference for me is the Nikons & Sonys can be used successfully in low light at higher ISOs, that being the only reason I have always had either a Nikon, previously and currently a Sony, alongside my Sigmas, and I can use all my Sigma lenses on the Sony with the MC-11. So I suppose that is a kind of win win situation.

s

In the meantime, the camera market continues to contract, with most of the new releases targeted largely toward the video market. Bayer sensors continue to improve, as Sigma stumbles along. I am a Sigma fan and own four of their cameras. I worry they are going to do themselves significant financial harm by chasing a white whale.

And keep in mind that successful companies, with long track records of success, make mistakes and go out of business all the time.

m