MFlenses - A new life in a digital age
MFLenses 2026

Manual Focus Lenses Forum

Explore the world of vintage lenses through repair notes, real test shots, and detailed archive threads packed with images.

Explore the world of vintage lenses through repair notes, real test shots, and detailed archive threads packed with images.
Spotlight

Featured partner placement for the forum front page.

Sigma buys Foveon
View previous topic :: View next topic  


Reply

http://www.dpreview.com/news/0811/08111101sigmafoveon.asp

I guess this rules out seeing this sensor in any other mounts.


Reply

Probably (although Sony sensors show up in other brand cameras for example).

I suspect that Sigma thought this was the best way to bump up the rate of development. A 4MP Foveon sensor is still only equivalent in resolving power to a 6 to 8 MP Bayer sensor (i.e. 1.5x, not 3x (4Mp -> 12) as they claim) so to compete with existing 12 to 24 MP bayer sensors they need to be making 8 to 16 MP ones ...


Reply

Yeah, Chris I think you are correct.
This is a ramp up for Sigma.
They could be a real if they developed a full frame 35mm as well as larger sensors like the new Leica S system (not necessarily for one of there own cameras).
It seems that Foveon has stalled in it's evolution where the rest of the industry refines their sensors.
I don't know enough about it to know whether it has hit it's ceiling or not.
I guess we wait and see. Sigma moves terribly slow.


Reply

I suspect this is the early stages of a long slow death for foveon. They clearly had to sell because of shortage of funds and Sigma does not have particularly deep pockets. This kind of research takes a lot of money. It is a shame that the big players did not adopt it - I for one have always loved the results of this sensor - it is the closest to film.


patrickh


Reply

I wouldn't understimate Sigma financial situation. I don't know its balances but the real money in reflex photography comes from the lenses, not from the bodies, and Sigma lenses sell pretty well.


Reply

F16SUNSHINE wrote:
http://www.dpreview.com/news/0811/08111101sigmafoveon.asp

I guess this rules out seeing this sensor in any other mounts.


Actually, if I was the ceo of Sigma, I would buy Foveon not to use it in monopole but to make a lot of money by licensing it to all the other camera makers.

It's not different for them than using their own lens technology to complement cameras by other brands.

If you think about it, now that the megapixels race is approaching the end (how many more MP can be in a small sensor?), companies will need something different to trigger the desire of purchase. So what better for this than a new different type of sensor?

(obviously this means that Sigma thinks to have either the money or the brainpower (or both) to accelerate development of Foveons)
-


Reply

Quote:
Actually, if I was the ceo of Sigma, I would buy Foveon not to use it in monopole but to make a lot of money by licensing it to all the other camera makers.

It's not different for them than using their own lens technology to complement cameras by other brands.

If you think about it, now that the megapixels race is approaching the end (how many more MP can be in a small sensor?), companies will need something different to trigger the desire of purchase. So what better for this than a new different type of sensor?

(obviously this means that Sigma thinks to have either the money or the brainpower (or both) to accelerate development of Foveons)


This would be ideal. Sigma although successful at selling consumer grade products does not have a culture of high quality standards.
It would be great if they did license out this technology and one of the big boys took the ball over the goal line.



Hummmmm........ A Full frame Foveon M mount rangefinder Razz


Reply

Orio wrote:

obviously this means that Sigma thinks to have either the money or the brainpower (or both) to accelerate development of Foveons)

I would certainly agree here, Orio. There would almost HAVE to be a thought similar to this by the CEO(s). It would be really nice to see a boost in Foveon development, whether or not Sigma does it within its own company, or farms it out.

I've always been impressed with the color quality of images from the Foveon sensors.


Reply

ChrisLilley wrote:
I suspect that Sigma thought this was the best way to bump up the rate of development. A 4MP Foveon sensor is still only equivalent in resolving power to a 6 to 8 MP Bayer sensor (i.e. 1.5x, not 3x (4Mp -> 12) as they claim)

I can't agree. This can be true only if you compare resolution of luminance channel. But for resolution in colors, 4.7MP Foveon can outperform even 14MP cameras, because their resolution of red and blue channel is pretty poor and additionaly downgraded by low-pass optical filtering (AA filter). Luminance resolution is good for architecture and low-light capures (lower luminance noise), but for nature, where color is more important, Foveon can do a better job.

Here are some samples taken from SD14 and DP1 reviews and upscaled to competitors resolution using Genuine Fractals software:

vs. 12MB Fuji S5 PRO:

http://i.mavrev.com/i/08327344810.jpg

vs. 12MP Nikon D300:

Nikon:


Foveon:


Just look at the red... I've often heard, that the Nikon sample is not focused properly, but that can't be true, because black and gray fabric is sharp, only colored parts are washed out...


Reply

In my experience Red is depend from lens too, most of the lens can't provide good red on my Olympus E-1 , some lenses like Olympus OM 50mm f3.5 produce superb red.


Reply

no-X wrote:
ChrisLilley wrote:
I suspect that Sigma thought this was the best way to bump up the rate of development. A 4MP Foveon sensor is still only equivalent in resolving power to a 6 to 8 MP Bayer sensor (i.e. 1.5x, not 3x (4Mp -> 12) as they claim)

I can't agree. This can be true only if you compare resolution of luminance channel.


Yes, thats what I was referring to.

no-X wrote:
But for resolution in colors, 4.7MP Foveon can outperform even 14MP cameras, because their resolution of red and blue channel is pretty poor and additionaly downgraded by low-pass optical filtering (AA filter).


Agreed that the red and blue channels have lower resolution (only one quarter of the photosites capture red; same for blue). The influence of the AA depends on the camera (one reason i like the D40 is that it was the last camera Nikon made with that 6MP sensor, they knew wel how to use it by then, and the AA filter is very mild).

no-X wrote:
Here are some samples taken from SD14 and DP1 reviews and upscaled to competitors resolution using Genuine Fractals software: ...
Just look at the red... I've often heard, that the Nikon sample is not focused properly, but that can't be true, because black and gray fabric is sharp, only colored parts are washed out...


Interesting. However, comparing using JPEG is problematic because its stored as YCC and most software performs a 4:1 downsampling of the Cr and Cb channels. Which (if you are trying to determine the influence of a 4:1 reduction in red and blue) is an issue.

Do you have TIFF or PNG of these images?


Reply

Hi Guys.

I am wondering why DXOMark is not testing Sigma/Foveon sensors.
http://www.dxomark.com/

It is a mystery to me how come the Foveon technology, if superior, has not managed to bubble up to the top. On paper, it seems to make sense, but I cannot get a grasp of an apples to apples comparison.

I can't even figure out how to understand the pixel counting with it compared to anything else. 4Mpxls times 3 colors = 12Mpxls? How does that compare to 12 Mpxls times one color each?

Just curious. Rolling Eyes


Reply

Henry Fisher wrote:
Hi Guys.

I am wondering why DXOMark is not testing Sigma/Foveon sensors.
http://www.dxomark.com/



I guess they don't consider Sigma a serious player. They tested the Fuji S3/S5, though, so that might not be it. A more conspiratorial guess is that they produce data that are serious outliers. Though again, Fuji is a counter-argument if you look at the dynamic range graph.

But I don't put a lot of weight in DXOmark. DXO in general has seemed to be a hype-heavy organization, and lots of the DXOmark results don't match real world experience or other test reports from sources I trust.

Quote:
It is a mystery to me how come the Foveon technology, if superior, has not managed to bubble up to the top. On paper, it seems to make sense, but I cannot get a grasp of an apples to apples comparison.


Well, Foveon only ever got Sigma on board (I'm ignoring the ill-fated Polaroid x530) And Sigma as a camera maker has always had challenges, stemming largely from a very odd mount design decision. It's fine for hackers and modders like us, but you aren't going to get a lot of mainstream adoption unless you either have an established mount like Pentax, or you come to the table with a Canon or Nikon mount. But the politics among camera makers are strange indeed. I think if Sigma had embraced its membership in FourThirds and released a Foveon sensor 4/3 camera, it would be a different story. (for everyone involved, actually)

Quote:
I can't even figure out how to understand the pixel counting with it compared to anything else. 4Mpxls times 3 colors = 12Mpxls? How does that compare to 12 Mpxls times one color each?


That I can help you with. When we're talking about a 12MP camera, be it a Foveon, Fuji's "SR" sensors in the S3 and S5, or a conventional Bayer array (everyon else), there are 12 million photosensor elements, each with a color filter over them. I'll call these "sensels" to distinguish them from output pixels.

In a conventional Bayer sensor, those 12 million sensels are in a regular square grid, with half the sensels having green filters over them, and one quarter each of red and blue. To get a 12MP image out, you have to interpolate two of the color channels for each pixel from the corresponding sensel's neighbors. This interpolation, or 'demosaic' is the reason Bayer array cameras have an anti-aliasing filter, to prevent moire and color fringing artifacts. If you don't want to interpolate, you end up with an image with 1/4 the number of pixels (and less noise in the green channel because you're averaging two green sensels). You can get this result using `dcraw -h`.

With Foveon, those 12 million sensels are arranged in a regular grid of 4 million stacks of three. There isn't an explict color filter, instead the design relies on the fact that silicon absorbs more blue light than green, and more green light than red. So the blue sensels are on top, then the green, then red buried in the sensor. If you use the 4MP image from the Sigma RAW converter, you don't get any interpolation. You can upscale to 12MP for comparison if you want, but that's as much subject to the scaling algorithm as anything else.*

For completeness, a 12MP Fuji "SR" sensor looks a lot like a Bayer array sensor except that the grid is a staggered grid of octagonal sensels. Half of the sensels are large, and half are small. The smaller sensels extend the dynamic range in the highlights, which is the SR's claim to fame. In the SR (the S3) each octagon (under a single color filter) contains two sensels, one large and one small. In the SR II (the S5), the small sensels are in the square gaps at the corners of the octagonal grid, and I haven't found a reference that shows the color layout of the mosaic filter. With the SR, it's clear that this should be considered a 6MP sensor, as each pair of sensels shares a microlens. With the SR II, well, it's not so clear. Smile

From an information theory standpoint, there's 12MP of data in all of these. Translating that data into image information is where the trick (and all the debate) lies.

I hope I've enlightened more than I've confused. Very Happy

*The samples that no-X posted are interesting, but difficult to evaluate. In the Nikon sample, I'd like to see a third image converted with dcraw -h and then upconverted with GF. I honestly don't know offhand if it would be better or worse than the demosaiced image. With the Fuji, well, see "SR II" above. Wink


Reply

Greetings, n6mod.

Thank you for the articulate explanation above, and for the education about the various schemes used to capture color information in the various arrangements of "sensels."

It's remarkable to me how far this technology has come in the past few years. All I ever wanted from digital was a digital back for my old F mount glass. Now it appears that a rough approximation of that has emerged in the form of Nikon's full frame bodies. I'm pretty happy about that. With ten old Nikkors and one otherbrand zoom already in the toy box, I am pretty much determined to have an F mount FX digital body sometime in the not-too-distant future.

I have been somewhat aware of Foveon for several years. And I have wondered how that it might stack up against a comparable sized sensor from Canon or Nikon. And now I wonder how an FX Foveon sensor (if it emerged) would stack up against the D3 or D700.

Probably does not matter. If the D3/D700 sensor is good enough, then it is good enough. And no matter what comes along, it will continue to do what it does as well as it ever did for its natural lifetime.

I noticed the dominance of the red in the example images upthread. I'm left wondering if this "over-emphasis" is not something that can be managed well enough in post processing, or with filtering.

BTW, I see you that you are an electronics enthusiast and a radio operator. Me too (kc4i).

Thanks again.

Happy day.


Reply

Henry Fisher wrote:
Greetings, n6mod.
Probably does not matter. If the D3/D700 sensor is good enough, then it is good enough. And no matter what comes along, it will continue to do what it does as well as it ever did for its natural lifetime.


As Orio is figuring out, the latest generation of sensors are better than a lot of lenses. This doesn't necessarily mean you shouldn't still use them, but if you go pixel peeping you'll figure out that you've run out of lens before you run out of sensor.


Reply

Hello again n6mod.

Quote:
... As Orio is figuring out, the latest generation of sensors are better than a lot of lenses. This doesn't necessarily mean you shouldn't still use them, but if you go pixel peeping you'll figure out that you've run out of lens before you run out of sensor.

I guess I am not much of a pixel peeper, actually. Or a brick counter either. I cannot see a lot of practical difference between a 50f2 Jupiter 8M and a 50f2 Leitz Summicron, both of which I use.

To the idea that the modern sensor is better than my old glass, I say GREAT! If the new sensors are that good, then quibbling about Foveon versus Bayer and such is probably beyond the scope of my old tackle's abilities anyway, and purely academic. One thought I have with respect to FX is that it will help me get the most out of my old lenses by using the full frame, and thereby requiring less enlargement per sensor plane unit area.

These are my "now virtually worthless" old lenses: 24f2.8, 28f3.5, 35f2.8, 35~105f3.5, 43~86f3.5, 50f1.8, 50f1.4, 105f2.5, 135f3.5, 200f4, R500f8.

If they perform as well with "digital film" as they have with Kodak and Fuji products, I will be okay with that. Since I began in this hobby back in the mid 60s, I have found that my biggest problems with photography have had very little to do with technological limitations of the gear. "Seeing" and execution seem to have been the essence of "the chase" all along.

Thanks for your insights.