Huge New Foveon Sensor Debuts in Sigma Flagship SD1

[Voice of Jerry Seinfeld saying "Newman!"]: Sigma!

Sigma, one of the major independent lensmakers, is also one of the most quixotic, frustrating, and yet interesting cameramakers on the planet. It was the original licenser of the Foveon sensor invented by Carver Mead, an early pioneer of microelectronics and the man who coined the term "Moore's Law," in the year 2000, and it bought Foveon Inc. lock, stock and barrel in 2008.

Although Sigma cameras with the Foveon X3 sensor have had modest success, when I think of Foveon I can't help but think of the innovative cone loudspeaker driver of Lincoln Walsh (1903–1971). The Walsh driver has some real advantages over conventional loudspeaker drivers, but the technology has been locked away in the hands of a tiny speaker manufacturer called Ohm for lo these many years. The intent was to capitalize on the design; the effect has been to marginalize it. Ohm is not a bad company—a small band of dedicated guys toiling away in Brooklyn—but there's no sense pretending it's ever going to take the world by storm. Meanwhile, mainstream speaker design and manufacture has gone around the Walsh driver like water in a river going around a rock.

I can't help but think that if Foveon technology were available freely to all cameramakers, the Bayer array sensor would be a quaint remnant of digital photography's history by now. Consider a parallel Universe where all the intensive R&D by Sony, Canon, Kodak and Fuji had gone into Foveon-type 3-layer sensors—it boggles the mind as much as science fiction. Instead, Foveon technology is the property of Sigma, a large lensmaker and small cameramaker. Sigma has done a decent job of marketing it—but let's not pretend it's taken the world by storm. For eight years now, Sigma has ambled along making a small number of adequate cameras with this interesting sensor in them...cameras that are typically not on time to market and a half skip to a skip-and-a-half behind the moment, and undeniably outside of the main current of digital imaging technology. Sigma has always insisted in giving the resolution figure of its rather modest 13.8 x 20.7mm*, 2640 x 1760-pixel** basic sensor a "times three" value, which, while sort of correct, works against its own interests, because in fact it doesn't match the resolution of a 14-MP Bayer array (for whatever reason), and it's generally seen as a marketing exaggeration by the camera-buying public. Meanwhile, you get the sense that if Panasonic or Samsung got their mitts on this technology, they would use it annhiliate the rest of the camera market and stand astride our world like a conquering colossus. In Sigma's hands, by contrast, it's sort of like the weird kid in class off doing something strange in the corner.

And yet. There is something undeniably very special about the Foveon X3 sensor. Purely aesthetically, Foveon pictures have a quality that can often be easily seen. I once likened the Foveon's relationship to the Bayer array as being like that of Kodachrome to a tri-color camera (a tri-color camera was a complicated, klunky Rube-Goldberg device that essentially used beam splitters to record the red, green, and blue components of a color image on three pieces of black-and-white film simultaneously). The Bayer array is essentially inelegant, the Foveon sensor essentially elegant. A Bayer array essentially recreates a color image, from a patchwork of information that has to be interpolated; a Foveon sensor records it directly. The Foveon sensor doesn't suffer as much from moiré effects, so it doesn't need a sharpness-destroying anti-aliasing filter. Sigma says it is sharper and purer "pixel for pixel" than any Bayer-array sensor. True...except, what do you mean by "pixel"? It certainly is better if you compare it to a conventional 4.6-MP sensor. The advantages aren't quite as clear if you use the times-three numbers and put it up against everything the biggies have been able to accomplish with the Bayer array.

Big day...sort of
In any event, Sigma has just announced, at Photokina, an all-new flagship camera called the SD1. Although that's big news, it's not the big news. The big news is that after using versions of similar-spec'd Foveon sensors approximately since digital crawled out of the emulsified swamp with four legs and a tail, Sigma has debuted a much larger Foveon sensor. The new sensor is 24×16mm in size, or true APS-C, and it has 4,800×3,200 photosite locations. That would be 15.4 megapixels if measured only by the standards of conventional Bayer arrays, which of course isn't accurate or fair, and it's a whopping 46 megapixels by Sigma/Foveon's habitual X3 computation—which is also probably not exactly accurate, subjectively speaking and all.

And yet. Ctein's back-of-the-napkin calculation is that the SD1 sensor will grab the mantle from the Sony A900 as the highest resolution sensor in a conventional DSLR body.

And image quality? That of course remains to be seen. And therein lies a whole 'nuther tale. Sigma has a well-deserved reputation for lagging behind its promises when it comes to product introductions, sometimes very significantly. (Remember how the weird kid never turned anything in on time, but when he eventually did, half the time it was brilliant?) The SD1 is slated to ship in February, and I'm tempted to say that if it does I'll eat it. But Sigma the cameramaker and especially the Foveon sensor maker is a cut-'em-some-slack case in the camera firmament if there ever was one. In the parallel Universe of What Might Have Been, this would have been by far the most significant announcement not only of Photokina, but of 2010. Even in this world, it more than bears watching. More news as the months tweedle by.

Mike

* That's in between APS-C and 4/3rds size.
** That would be considered a 4.6-MP sensor if it were a Bayer array. These figures are from the SD14, by the way.

Send this post to a friend

Note: Links in this post may be to our affiliates; sales through affiliate links may benefit this site. More...
Original contents copyright 2010 by Michael C. Johnston and/or the bylined author. All Rights Reserved.

Comments

[Nico Burns]

some interesting news in the world of foveon at last!

I really had given up hope on there ever being a new foveon sensor, this could actually be a reasonably good seller for sigma if they can deliver it quickly and don't have any other performance issues.

[Joshua]

This sounds very cool, can't wait to see a sample image.

[max]

"...you get the sense that if Panasonic or Samsung got their mitts on this technology, they would use it annhiliate the rest of the camera market and stand astride our world like a conquering colossus. In Sigma's hands, by contrast, it's sort of like the weird kid in class off doing something strange in the corner."
Hahahahahaha, spot on description, Mike!!!

[KeithB]

One technical problem is that since the red sensor is so deep in the silicon, it is sensitive to the angle of incidence of the light. This means that the red signal will fall off at the edges of the chip. This can only get worse with a larger chip.

[Stuart]

I want one! I want one! a foveon with more than 4MP, yay! But which of Sigma's lenses would I need to buy to make for a sharp and wide kit?

[GKFroehlich]

"The Foveon sensor can't suffer from moiré effects"

Are the photosites discreet, and placed on a rectangular array? If so, the sensor can exhibit moiré effects, as they are a by-product of sampling.

If they're not discreet, then reading the info off of a continuum of them is the real breakthrough with this sensor.

[Aaron]

I'd be really excited about this, if I didn't have to use Sigma lenses. Man, if they came out with a version of this with a Nikon mount, I'd definitely check it out.

Being limited to Sigma lenses is a deal breaker though. I've tried two on my Nikons and ended up returning both of them.

[Adam Lanigan]

One idea that's been bandied about, that I've raised my eyebrow at a few times, is that Sigma's SLRs should, like their lenses, come in a variety of mounts (or 4/3rds, at least). The thinking being that people would be a lot more likely to test the Foveon waters if they didn't have to invest wholesale in a new system of Sigma-only lenses. Imagine using your cherished Canon, Nikon, Olympus, etc. glass with a Foveon sensor. Like you've mentioned above Mike, at this point Sigma's had to have realized the water rushing around them, and this might be an interesting way to jump in and get swimming.

I still bring out my little turtle of a DP1 when I purposely want to do some 'slow' photography. It's hesitancy can sometimes be a virtue in that usage and I'm constantly surprised at the qualities of the Foveon sensor.

[Antony Shepherd]

It's good to see that Sigma haven't been idle since they bought Foveon, and given what I've seen from my DP-1 I can only imagine how sharp and excellent the pictures from this must look.

But I'm saving my pennies for the Fuji...

[Bahi]

I love the look of Foveon pictures online (I've never seen prints) but am now seeing much of what I like about them from shots taken with high-resolution cameras like the 5D Mk II when they're reduced to screen size. This will be interesting to watch.

[Dennis]

Sigma already makes lenses in Canon, Nikon, Sony, Pentax mounts. How hard would it be to make this new camera in those same mounts ? Not easy, I'm sure, but sales in each of those mounts would probably be ten times what they'll be with the Sigma mount. A 16MPxRGB APS-C sensor is truly intriguing ... but a 16MPxRGB sensor in a camera limited to Sigma lenses loses its appeal.

[Nicholas Condon]

Aaron beat me to it, but the Sigma lenses are a real problem here. The sensor may be able to theoretically out-resolve an A900, but will that ever happen in practice?

When does the Foveon patent expire? I'm not sure it will make any difference when it does, but it might be nice to think of a world where Sony et. al. are making multilayer sensors.

[beuler]

Nail hit squarely on the head Mike.

In this universe, cameras are not judged on image quality, but on how little noise they have at ISO 6400 (it used to be ISO 1600 and ISO 400 before that). Only medium format cameras escape this yardstick, and being an APS camera, the SD1 is doomed by this factor alone.

[Ken Tanaka]

"And yet. There is something undeniably very special about the Foveon X3 sensor. "

I can deny it. The Foveon spec promise is what drew attention years ago. But having spent time with that crappy DP2 last year...yuck. Yes, perhaps like a talented but uncultivated athlete the Foveon could have had a better future on a better team. But I've no interest in "could have". I'm a photographer who has placed Sigma on my ignore list. Why waste the time or money on junk?

[Paul Van]

Thanks for reminding me of the Walsh driver. I can still remember the first time I heard a pair of Ohm F's.

[Mike Johnston]

"it might be nice to think of a world where Sony et. al. are making multilayer sensors."

Nicholas,
You mean when they're *marketing* them. They're already making them, make no mistake about that.

Mike

[deejjjaaaa]

> Man, if they came out with a version of this with a Nikon mount, I'd definitely check it out.

< sigmacumlaude.com sells kits to have F mount on Sigma bodies... no autofocus though

[Jim Kofron]

@Keith,

An 'oft-quoted' issue that's not an issue. The distance from sensor to back of lens is reasonably far, so the light going in is pretty perpendicular to the sensor. On top of that, the index of refraction for SiO2 is around 4.5, so any light coming in at an angle is strongly bent to the perpendicular.

What does cause problems is reflection of light (especially wavelength dependent reflection) off of sensors. But this is much more of a problem with compacts that have a short registration distance (like the DP1). In this case, offset microlenses help with the situation.

@GKFroehlich
Go to the Foveon website for a cartoon of the sensor and the 'sensels' there are three discrete layers that sense the light, stacked on top of one another. Color moire is not an issue. Aliasing (luminance channels) can be an issue, but is usually not too much of a problem.

It's very cool to see Sigma come out with this less than 2 years after the acquisition of Foveon. That's pretty rapid progress. Hopefully the release date won't slip too much.

And for those who want a different mount---either buy one of the compacts and try the foveon stuff out for yourself, or check out Sigmacumlaude for a mount replacement (at least on the SD14/15). You can even get a Leica mount... But I'm sure that this is a razor/razorblade situation, and Sigma will make a nice profit off of lenses. With this pixel pitch it's going to be hard to hold the camera steady enough to not see motion blur (I had that issue when I moved from the 3.4 MP SD10 to the 4.6 SD14---imagine a 14 MP SD1). So you'll need that nice new EX glass with OS... And I've used many excellent Sigma lenses (8-16, 10-20, 50, 70 Macro, 70-200), so there will be a nice selection. The 70 macro is among their sharpest, and that would be one of the first I'd try.

[Daniel]

I think there are some caveats to the "sampling directly" the colour information at each site. The top layer samples most of the blue, some of the green, and little of the red. The middle layer samples a little bit of the blue, a fair bit of the green, and some of the red. The lowest layer sample a little bit of green, and mostly red. This sampling of the spectral information still has to be deconvolved to arrive at a "colour", something which represents the tri-stimulus values of human colour vision. Issues could arise in this step due to the transformation required to change the recorded data into a colour value and its sensitivity to noise.

If we consider a patch of colour, a well designed Bayer colour filter array can actually give better colour than a Foveon sensor if the filters are chosen correctly and the the de-mosaicing algorithm is up to snuff. When it gets to colour varying at high spatial frequency, then the situation gets a bit more complicated.

Ideally, in designing a stacked pixel array, you want to tailor the absorption profile of each layer, rather than use the naturally occurring silicon absorption, this why Fuji has been developing (amongst others) dye-coupled photo-diodes hoping to leverage their experience with colour films.

Foveon sensors can still suffer from moire, this will always occur if there are frequency components above the Nyquist limit reaching the sensor. What it does not suffer from is colour moire, the cross-talk between the spectral and spatial sampling domain. I would like to bring to wider attention work done on combined spatial-spectral sampling and reconstruction work, e.g. http://www.accidentalmark.com/research/papers/Hirakawa08CFA_TIP.pdf Fig. 8. and http://www.greyc.ensicaen.fr/~lcondat/publis/condat_icip09_newCFA.pdf for similar work. These references also explain why colour moire occurs in Bayer sampling reconstruction and how to reduce it by careful CFA patterns, see Fig. 1 of the second reference above.

I do think the Foveon sensor is elegant but Bayer still has a lot of life left. My proposal is a bi-layer sensor. A green layer with a small pixel pitch, and a second layer which is checkerboard red/blue but at twice the pixel pitch. Or perhaps use a more sophisticated colour array in the spirit of the references above to optimise spatial-spectral sampling characteristics. Spatial resolution would be close to a full RGB stacked pixel design, but the data overhead would be slightly larger than for a single layer Bayer. Drawbacks include complex pixel architecture, relies on being able to tailor the absorption profile of each layer, increased complexity of de-mosaicing, potential for extra pixel crosstalk, and I'm sure lots of other things. But stacked pixel have been shown to be viable by Foveon. Fuji has demonstrated a prototype stacked organic CMOS sensor http://image-sensors-world.blogspot.com/2009/10/fujifilm-organic-cmos-sensor.html. So my idea seems at least plausible?

[Michael Korak]

Foveon/Walsh. What a great comparison. I do not use Sigma/Foveon digital SLRs, however, I WAS the very proud owner of a pair of Infinity Monitor IIA column speakers. They had the Walsh conical super tweeters. Purchased new in 1971 when I was single and a vinyl audiophile. Linked to my Marantz component system, they were the cat's a-- back in the day.

If Foveon chips are near as industry revolutionary and deliver the same level of performance as the Walsh designed tweeters, many more digital camera manufacturers should persue a relationship with Sigma for the rights.

[Thom Hogan]

A. Announce is not ship. Sigma has a tendency to announce way earlier than they're capable of delivering.
B. The Foveon sensor can most certainly exhibit moire. It just won't show color artifacts in moire.
C. "Highest resolution camera" means little. Resolution is a chain of things, and if you have anything in the chain that is low in resolution, it impacts the overall resolution number. As you go up in "sensor resolution," other factors start to be the restrictive factor, like lens resolution. But alignment becomes a big issue, too.
D. I'm not convinced that three-layer is "the solution" nor that Bayer is not the solution. Adding additional color resolution is actually not as good as adding additional monochromatic resolution. Most of the researchers that I talk to believe that when we get to very large megapixel sizes (say 100mp in FX), that you actually should go the opposite direction and make color information (blue/red) more sparse in the array. This better conforms to the way our eyes work.
E. As someone else noted, the Foveon technique is angle dependent. They need better microlenses and probably even BSI and offset microlenses to do large sensors with the technology.
F. What no one is talking about is whether Sigma has improved the Foveon noise tendencies. Since I believe we're talking about smaller photosites, this is something that seems important ;~).
G. One real issue with the Sigma's has been the internal JPEG processing engine. I haven't tried the very latest (SD15), but previous incarnations have been a bit of an issue. Coupled with this, of course, is that the sensor designs and internal ASICs of all their competitors now support HD video, and this too seems to be an area where Sigma is lagging.

Overall, I love Sigma's idiosyncrasy and their entrepreneurial nature, which has them making interesting and different decisions. But one real problem I see is that they just haven't solved the "camera problem." In almost every way, the Sigma cameras I've used to date would have been competitive with, oh, late 1980's and early 1990's cameras. They could have gone a different way and broken the mold, but they're designing 20 years back (and as everyone knows, I think that ALL Japanese camera makers are still not getting camera designs right.)

I hope they do well with this new camera. But I see nothing that indicates that they've really made the amount of progress they need to to be competitive. So I'm with Michael on this one: interesting technology being marginalized by a small player.

[Paul]

"Being limited to Sigma lenses is a deal breaker though"
I was inclined to think the same as Aaron until last year I tried out and later purchased Sigma´s 50mm F1.4 EX DG HSM which turned out to be a very capable lens. It´s only handicap is it´s a bulky lens.
Paul

[psu]

Color me skeptical. What the Foveon sensor does, arguably, is increase color resolution in a particular way. But most of the time you don't need that much color resolution. if you did, then JPEG compression would not be so effective.

It remains to be seen whether the Foveon sensor can match the other perfomance characteristics of the more developed sensors from S, C and N, especially the astounding sensitivity of some of the newer large sensor cameras.

My bet is it can't.

Also, any sampling device can suffer from moire effects. You can't escape from the information theory jail.

[David L.]

The problem with many of the earlier Foveon chips was an inability to capture the color red properly. KeithB explains why above. I reviewed the DP1 a couple of years ago and could not get a red Porsche to render in a color aside from magenta with the Foveon sensor. I found a third-party application that ran an algorithm to compensate when shooting RAW images, but JPEGs were stuck with poorly captured red tones. Hopefully the new SD1 has developed a workaround, if not, it is another camera doomed to obscurity.

[Nicholas Condon]

@Mike Johnston
I'm aware of some (non-Foveon) multilayer sensors for specialty applications; is that what you're referring to, or does one of the big semiconductor companies actually make the Foveon chips?

[RobJ]

I love Sigma's Foveon Sensors, but who buys their DSLRs? I don't understand how the camera side of their business is even remotely profitable. They should focus on the DP line of compacts and get this sensor into a DP3 as quickly as possible (I know, I'm dreaming-- this won't happen until 2013). Also, as others have mentioned, a Foveon sensor in a Nikon or Canon mount could be a big hit-- though I'm assuming they would need Canikon's approval to do this?

[amcananey]

I have to admit that I was surprised that this didn't get more publicity on the other camera/photography websites. If Sigma really meets the announced date, they will have a camera that is competitive in terms of megapixel count even if you compare it directly to Bayer-array models from competing companies (and of course I would expect it to be much sharper in practice).

As to the claim that if Nikon/Canon/Panasonic, etc. had gotten their hands on Foveon's technology, they would rule the camera world...well, I have trouble believing that. If any of them thought there was a reasonable chance of that occurring, they presumably could have bought Foveon themselves. Instead, I suspect that the mainstream camera manufacturers believed one or more of the following:

1. The Foveon technology had/has some fundamental flaw that made/makes it unattractive.

2. Manufacturing Foveon sensors is so expensive that it is never likely to be cost-effective compared to mass-produced, higher "nominal" pixel count Bayer-array sensors.

3. They would develop their own non-Bayer sensors.

Of course, it is also true that the mainstream manufacturers may have been banking on #3 only to find out that developing a non-Bayer sensor isn't as easy as they thought...

Best,
Adam

[GKFroehlich]

@Jim Kofron
Indeed, the distinction between "color moiré" and "luminance moiré" is a good point. The latter will always exist if the photosites are discreet and regular (i.e., periodic). But the spatial frequency of the latter will often (usually?) be high enough that the moiré patterns won't be excessive or objectionable. "Color moiré" is more visible than "luminance moiré" with the lower spatial frequency of the Bayer color samples. By the way, there is an excellent discussion of moiré patterns at Wikipedia <http://en.wikipedia.org/wiki/Moir%C3%A9_pattern>, with nice illustrations that explain the phenomenon. I use their examples when teaching folks about signal processing.
-gkf-

[Tim]

It seems to me that Foveon sensors have always given too-small images in need of some kind of up-scaling, while theoretically the bayer/anti-alias filter means APS-C should be scaled *down*, in order to get a version of "the image" in the middle for comparison.

The metric to consider is the number of bytes of image-data per pixel when stored as JPEG (starting from same lens in different mounts, using the same processing, both converted to the same size, compressed at the same quality).

I *had* a SD9. Would not buy another one.

[James]

OK, fantastic news - who hasn't been hoping they'd bring the resolution up?

Two thoughts:

- so we now have a 15 MPix black & white camera with built-in 45 Mpix colour option :)

- if they were to scale the current sensor density up to a medium format back they could make a very big splash in that particular river. That market is quality driven.

[Charles]

the only way I can describe my love hate relationship with my DP1 is to imagine an unexpected intense relationship with a beautiful stranger who is also slightly bi-polar. The highs are quite extraordinary but you just know you're gonna be hurt...and keep coming back for more.

I have often wondered why Sigma didn't enter a relationship with another camera maker, like Fuji did with Nikon. Heck. The F5 with a foveon chip would be quite lovely.

[Michel]

Who would have ever imagined dreaming of an adapter from a Nikon, Canon or Zeiss lens to a Sigma body?

[Clayton]

I once held back a year from going digital because I thought that Foveon would take over the world. Did not go so far as to buy there stock though.
Will wait and see this time, but it sounds good. Sigma does make some good stuff, but almost everyone does now. Even the ones that don't can fix it with software.

[Ed Hawco]

When I read about advances like this in sensor technology, I dream about the day when sensors are standardized and can be transferred from one camera body to another. Of course there are software reasons why this can't be done now, but that's easy enough to resolve.

Can you imagine the day when camera manufacturers go back to being camera manufacturers? When Nikon and Canon and all the others only worry about things like build quality, lenses, and on-board features? When sensor manufacturers become like the old film manufacturers, offering choice and variety instead of being inextricably hitched to specific camera bodies?

Imagine making your choice of a camera (Nikon D90? Canon 5D? Low-end Pentax? High-end Pentax?) and THEN making your choice of a sensor to snap into it?

I'd like to think that day is coming. No time soon, but we can dream.

[Luke]

Why does it have two shutter release buttons? Maybe you can use two fingers to push one button at a time in seqence, simulating a high speed motor drive!

[robert e]

And to bang that drum again: this is huge news for digital black and white, too.

[Mike Johnston]

Nicholas,
No, I mean Canon, Nikon, Sony et al. are doing R&D on technologies they're not currently marketing.

Mike

[Max]

From what Jim said, I agree, it's really hard when you get improvements in resolution at this point it's really hard to press improvemets unless you use a tripod. I experience that with "hanholdable" medium format cameras, it's imposible to get all that resolution handheld. That's why I think image stabilization means are becoming more and more important at this point, and would require paralell developing to improved resolution means on small cameras.

[Benjamin R. George]

Putting aside all our suspicions about things we usually expect Sigma to get wrong, my biggest problem with the specs announced so far (at least if DPReview is to be trusted here) is that the focusing screen is fixed. If this is true, it'll make this a much less appealing camera than it would otherwise be for use with adapted manual focus lenses. That in turn makes the limitations of the Sigma AF lens lineup a much bigger problem than they'd otherwise be.

The second biggest problem with the announced specs is the size (which isn't awful by itself, but is a bit disappointing when you compare it with, say, the Pentax K-5 or even the Nikon D7000).

[extrasalt]

I learned long ago, from a sensor designer in Mountain View, CA, that not all manufacture's pixels are the same. So, I ignore all statements regarding pixel size (pixel site) and Bayer filtering, filter layering, etc. I prefer to test a camera in a real-world situ and evaluate the results. After all, if you take RGGB, then you divide the advertised pixel population by 3 to get the actual pixel count used to produce a file.

[John F. Opie]

I think it's correct to be skeptical: Sigma has severely disappointed in the past, far too late to market, with products that show that there was not enough money behind them to get them to market in a timely manner. The DP1 is a great example: a great camera, if it had been introduced 3 years before it was. By the time it came out, it was so damned slow in comparison that I returned mine because it was infuriating to work with.

The sensor holds enormous potential. Me, I'd like to see it at this resolution in a m4/3 camera. But if the SD1 comes out first around 2012 and has the same kind of clumsy menu interface and weak JPG production out of the camera, forcing you to work with a largely unsupported RAW format, then: meh.

More power to them. But empirically? Wait and see...

[Pärdik]

Paradoxally, It's really a closest thing to a 15Mp dedicated B&W sensor there is. Anyway It's hard for me to trust the color separation filters that are made from silicon, and silicon only (and software, yes).

[Stephen S.]

I have to admit, this SD1 announcement is all my fault. You see, just yesterday, I put my Sigma SD14 and all accessories on consignment at local camera store.

[Gavin McLelland]

Mike

I know it's photokina but can we have some photographs, some Random Excellence?

Please

Gavin

[Mark]

OK I understand the concept of being a fan of a quirky technology and agree that the Walsh comparison is a perfect one for the Sigma camera.

With that said, and as the owner of a pair of Ohm Walsh 2's for over 20 years I love quirky speakers but I just can't get behind the Foveon sensor. The Wlash 2's still work great even to this day. The Foveon never seems to get it right to me. Maybe this new one will change things but I am not holding my breath.

I will however be replacing my Ohm Walsh 2's with a new set of Walsh 2000's in the near future.

[James]

Seems like a lot of people want one camera that can do everything. I guess that's one way to approach the problem, but it seems like a pretty good way of always being dissatisfied with your tools. If the camera turned out to have poor high ISO, no video and lousy jpegs there would still be plenty of uses for it. Like taking really high quality photographs for example. I find myself wondering which lens in the sigma lineup would be best for art repro...

James.

[C. Cunningham]

But to be fair about it, how many companies aside from Sigma showed intense interest (and backed it up with cash) in Foveon technology as an alternative to Bayer?

Anyone?

So the manufacturers inattention or complacency led to the technology being marginalized. Sure, it's too bad there isn't more competition between Bayer and Foveon, but maybe if more customers had demanded it, Sigma might have made more and better camera gear, and maybe more sensor manufacturers would have licensed and been making Foveon sensors. Shouldn't capitalism and the law of supply and demand made Foveon popular?

But sometimes it seems that no matter how good a design is or how well a manufacturer makes their mousetrap, if there are enough nattering nabobs on the internet, it will never make it.

And no matter how poor a product is, if the marketroids do their job and convince the voices of the internet that their product's pixie dust is just right, people will buy it anyway, no matter if it seems that other people are having problems.

Whoa! Sorry that slipped onto a side road AND turned into a rant. Not sure how it happened.

Oh well, I stand by it.

[improbable]

There is something very beautiful about files from the Foveon sensor. I think they pointed out to me how much of the "digital look" of other files is about treating colours differently: the red rose isn't only red, it stands out from the green leaves in other ways. Sometimes it almost looks mis-focused. But the Foveon images aren't like this, they are very democratic about colour.

Still, wow, it's a big leap to commit to sigma lenses (in sigma mount too, how will you ever sell them?) and also to one funny camera (which let's face it is not going to be quite 2010 material in other regards) not to mention looking after 45MP worth of file size...

[Patrick Perez]

@Nicholas Condon: At least in the past, the actual manufacture of the Foveon sensors was done by Texas Instruments (and IIRC, they even owned Foveon and sold it to Sigma, though that memory is not as trustworthy ).

Patrick

[deejjjaaaa]

> forcing you to work with a largely unsupported RAW format,

.X3F raw files are supported by Adobe (Lightroom and Camera Raw), nowadays that means "largely supported", add here SilkyPix from ISL (which is a raw converter/raw engine used by Pentax, Panasonic and Samsung)... both options are available on Windows and Mac platforms.

[deejjjaaaa]

> I reviewed the DP1 a couple of years ago and could not get a red Porsche to render in a color aside from magenta with the Foveon sensor. I found a third-party application that ran an algorithm to compensate when shooting RAW images, but JPEGs were stuck with poorly captured red tones.

---

between your review and today there was/is a whole generation of Sigma cameras w/ "True II" engine starting from DP2 (DP1, SD14 are old pre True II designs) plus new generation of raw converters (ACR 5.x/6.x, SilkyPix) to support it

[Peter Cameron]

I'll add my voice to those who'd love to see this sensor in a DP3, preferably with an e35mm f/2 lens.

[Mike Johnston]

"as the owner of a pair of Ohm Walsh 2's for over 20 years I love quirky speakers...I will however be replacing my Ohm Walsh 2's with a new set of Walsh 2000's in the near future."

Mark,
You probably know this, but John Strohbeen will rehab your old 2's with new 2000 drivers if you want. The resulting hybrid is called the "2.2000."

Mike

[Pete Myers]

I was in touch with senior management at Foveon for a few years, trying to convince them to allow me to do a story on their technology. Despite best efforts, I could not get them to agree. Usually companies are very happy about getting press on their technology when its new, novel and has to win market share. I was not surprised when the entire technology was sold to Sigma. I think they sensed the limits of the technology, and likely a signal to noise issue on layered sensing---just a guess.

What has surprised me greatly is that none of the camera companies have tried to make a three-chip, prism-based image sensor. Ultimately this is where broadcast television cameras made their mark. It was very surprising to me that the RED camera went back to Bayer given the issues with matrix-based color sensing.

Having shot a digital monochrome for 18 months, and now with a Bayer based camera, my wife knows only too well my complaints about matrix-based color sensing.

Some one, somehow, somewhere will get'er done!

Pete

[Elwood (Woody) Spedden]

Hi Mike
You are of course correct when you say that Carver Mead coined the term "Moore's Law." However this was coined in the year 1970 (or thereabouts) and not 2000!

I have been in the semiconductor industry for the past forty years and I can hardly remember a time when Moore's law was not constantly uttered.

Gordon Moore was and is a genius.

Woody

[James]

"You probably know this, but John Strohbeen will rehab your old 2's with new 2000 drivers if you want. The resulting hybrid is called the "2.2000.""

That right there is exactly the sort of thing that could persuade me to drop silly money on an esoteric product, if only I had said silly money available. Commitment to a product over 20 years old - it practically brings a tear to my eye.

[Janne]

Three times is just empty exaggeration. I could see a justification for claiming double spatial resolution (half the sites on Bayer are green, our main component for luminance information). Realistically I'd say the difference is somewhere between 1.4-2 times, depending on the demosaicing algorithm you employ for the Bayer sensor (and disregarding the corresponding deconvolution you need for the Foveon sensor as that should affect mostly the color rendition).

But three times? Nope.

[Abdul Rahman]

How would these mega pixels contribute to real world image making? Photography and photographers would be better served with real cameras with only necessities to enable photography not the current bells & whistles which are hardly used.

Can't wait for a full review of the Fuji X100 - seems the right kind of tool for me.

[Zalman Stern]

Foveon had a partnership with National Semiconductor and National had a stake in the company early on. The first two or three Foveon sensors were manufactured by National. I don't think there was ever any involvement with Texas Instruments.

-Z-

[Will Frostmill]

Max wrote:
From what Jim said, I agree, it's really hard when you get improvements in resolution at this point it's really hard to press improvements unless you use a tripod. I experience that with "hanholdable" medium format cameras, it's impossible to get all that resolution handheld.

Hmm. You woudn't happen to have a rule of thumb for how many megapixels you need before hand-held shake makes additional pixels moot? (Or for film, what enlargement factor makes the increased area useless? E.g. 8x for MF?)

Will

[Mani Sitaraman]

I want them all! A Sigma SD1! A Fuji X100! A Pentax K-5! All these fabulously innovative darlings.

Darn you, TOP...

[Mike Johnston]

"That right there is exactly the sort of thing that could persuade me to drop silly money on an esoteric product, if only I had said silly money available. Commitment to a product over 20 years old - it practically brings a tear to my eye."

James,
"Unlike any other company I can think of, Ohm still fully supports every product they've ever made. Two years ago, I bought newly reconditioned drivers for those 20-year old Walsh 4s. They also take trade-ins no matter the vintage. Periodically, they refinish those trade-in cabinets, install the latest/greatest drivers and offer them for sale at attractive prices. Without a doubt, these guys go the extra mile to serve their customers past and present and are very serious about what they are doing."

--From my late friend John Potis's review of the MicroWalsh Talls.

Mike

[Arthur]

It's very good news. Unfortunately you can only use Sigma lenses. The one sigma I have, a 20mm, is the weakest prime I own by far. All the Minolta or Sony primes are far superior.

If they open up this camera to other mounts they can make money. If they keep it locked to the Sigma lenses the patent will wither as it runs out. Their development money will be wasted.

I expect their pride will win and they will lose just like the horn manufacturer in your article.

[anurag agnihotri]

All of us know how sharp and detailed DP1 and DP2 images are, actually that being the reason many of us jumped on to buy these crippled cameras. Before these cameras, i didn't even know my camera churned out mushy images which needed to be sharpened a lot to remove that incredible softness.

A bigger F-sensor is great news. Its good in a way that the world is not so excited about it. This will definitely take off, may be in the hands of some other manufacturer (olympus pen?, in some other form who knows...and it will go a long way.

[Graham Wood]

Mike,

"Moores Law" was coined by Gordon Moore founder of Intel 40+ years ago.

[Jim Kofron]

Pete: You probably are aware that Foveon's first studio camera was exactly what you describe: a three sensor beamsplitter camera. It ran tethered tho, so not the best for landscape...

Arthur, your 20 mm Sigma is a very old model with meh performance. The newer primes (50 f/1.4, 85 f/1.4) are very nice, as is the 70 mm macro. And they make a lot of interesting zooms (I'm a big fan of the 8-16 mm).

[Mike Johnston]

"Mike, "Moores Law" was coined by Gordon Moore founder of Intel 40+ years ago"

Gordon Moore developed the theory, Carver Mead coined the phrase.

Mike

[lith]

For the gaming geeks, I suppose it's kinda like Nvidia buying out PhysX...although Nvidia is actually capable of making decent cards, they just haven't for a while and PhysX has stagnated.

And Mike, I think a better comparison is like Kodachrome to Autochromes - Kodachrome used layers to produce continuous tone, while Autochrome used individual grains in the same layer, dithering the colours like a comic book.

There are actually tri-colours out there, or there used to be: plenty of video cameras used three CCDs, with the light split through a prism, have one CCD for red, green and blue.

Sigma, if you love something, set it free...

[David Dyer-Bennet]

Well, when Foveon was available in cameras at first, lots of people were interested. But the test results from those early cameras caused most people to lose interest. What, exactly, the issues were can be argued endlessly -- problems inherent with the sensor concept, issues with the camera built around it, or something else -- but most of the potential customer base said "oh well" and moved on.

I keep hearing people talk about subtle differences in rendering that they can't precisely characterize, associated with a sensor or a lens. Generally speaking I believe them; but I'd be a lot happier if they'd display many pairs of photos, taken at the same time and place with different cameras, really showing the difference. Of course you can only get such photos with considerable effort, starting with owning two sets of equipment and hauling it around.

Hmmm; I have three old primes which overlap my 24-70/2.8, plus I have an old 28-70/2.6-2.8. Plus the 70-210 overlaps the two other zooms just barely at the long end. Which means I could set up some scenes of different types and photograph them from a tripod with the same camera but different lenses. I wonder if I'd find the results illuminating?

[Lawrence Kaye]

I'm still waiting for Contax to get the bugs out of the Contax N. Am I wrong to wait?

[Stephen S.]

Pete Myers: "What has surprised me greatly is that none of the camera companies have tried to make a three-chip, prism-based image sensor. Ultimately this is where broadcast television cameras made their mark. It was very surprising to me that the RED camera went back to Bayer given the issues with matrix-based color sensing."

There are two main problems with that - first, video cameras often use relatively small sensors, and building a camera with three comparably gigantic APS-C sized sensors and a correspondingly gigantic prism would make for a camera that was gigantic, heavy, and awfully expensive.

The second problem is - lenses for prism cameras need to be designed differently, to focus different wavelengths of light at slightly different distances. Putting a lens designed for a single sensor (or film) on a prism camera, or vice-versa, leads to very poor image quality.

That's why the RED camera (and Arri and Aaton and Sony's equivalent cameras,) which is designed to use lenses from the film days, went with a single bayer pattern sensor.
It is able to use broadcast video camera lenses designed for cameras with prisms, but it requires a special adapter that includes optics to correct for the different colours focusing at slightly different distances.

Going the other way, a single sensor/film lens on a prism camera, is far more difficult. That's why Canon or Nikon et al is unlikely to go that route - their existing lenses would all produce very poor results, and an entirely new lens system would be required.

[Mike Johnston]

"I'm still waiting for Contax to get the bugs out of the Contax N. Am I wrong to wait?"

Lawrence,
Not considering I'm still waiting for the Spotmatic III.

Mike

[Jeremy]

I really like your Ohm/Walsh analogy. When Foveon/Sigma was first promoted, I was intrigued, but also immediately thought of the Wankel/Mazda rotary engine of the 1970's. A 'revolutionary' technology with tons of potential, that got a lot of press, but was only really developed by one company, and never made any significant impact in the market. Like the rotary engine, the Foveon sensor could easily outperform Bayer technologies of similar scale. But Sigma/Foveon was never able to achieve the scale (size of chip or camera features) needed to leapfrog other mfgs to make much of a difference for consumers. It is nice to see they're finally catching up to the size & number of photosites of other sensors.

Unfortunately, it will be very difficult for Sigma to hit a sweet spot for consumers with a single DSLR product and a single compact camera product. They're likely to continue to lag behind other camera makers for features (evf, HD video, usability/ergonomics, etc.) I certainly wish them luck though!

[ctein]

Dear Will,

You are asking a different question than the concern that Max has. There is a subset of digital photographers whose goal is to make photographs that exploit the maximum resolution that a camera is capable of. Today this is of practical consequence to very few of us. It made sense when digital cameras had only a few megapixels to try to squeeze out every last bit of sharpness. Today, it makes no more sense than being a film photographer and insisting that your photographs reach the resolution limits of the films. I can assure you that only a small fraction of photographers ever made photographs in the real world that did that, and nobody ever did it on more than rare occasion.

Your question is the sensible one: When does adding more megapixels stop contributing a visible improvement. I've analyzed that in previous columns. You can arrive at different answers depending on what assumptions you make, but as a ballpark you should not even be concerned with this question until we are nearing the 100 megapixel mark. In other words, check back with me around 2020 and I'll give you a more precise calculation.

With the kind of megapixel counts we're dealing with in the present real-world, adding more megapixels improves image quality much the same way that using a sharper film did in your 35mm camera. There's a visible improvement in the photograph even though few, if any, of the photographs approach theoretical film/sensor limits.

That having been said, you should take Thom's message to heart. Being obsessed with sharpness is not only a case of not seeing the forest for the trees, it's not even noticing the trees… Except for how tall they are.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

[titus]

Sigma, is Sigma, is Sigma, but if you want a camera that stimulates your photography passion, buy the DPx, the X100, or whatever. It's all about creativity, whether Foveon or Bayer. But i love to think differently. It always seems impossible until it is done.

[ctein]

Dear Jeremy,

"Like the rotary engine, the Foveon sensor could easily outperform Bayer technologies of similar scale"

Ummm, no. Only if you're talking about spatial detail. By any other measure of image quality-- speed, noise, exposure range, color accuracy, the Foveon sensor has been an ordinary performer. It is in no way outstanding, except for spatial detail.

As pixel counts get higher and that becomes a less important matter, the import of the Foveon design becomes less obvious.

Of course, the new sensor may very well have improved on all aspects of image quality. We will just have to wait and see.

pax / Ctein

[Mark]

"Mark,
You probably know this, but John Strohbeen will rehab your old 2's with new 2000 drivers if you want. The resulting hybrid is called the "2.2000."

Mike"

Yep, I know and that is what I am planning on doing. He will send me the kit and I will replace my old driver and crossover with the new one for a whole lot less money than a completely new set.

[Mike Plews]

Ohm Speakers, Foveon sensors, superior technology held back by poor marketing. I'd say more but this Amiga is so slow...

[Arthur]

Arthur, your 20 mm Sigma is a very old model with meh performance. The newer primes (50 f/1.4, 85 f/1.4) are very nice, as is the 70 mm macro. And they make a lot of interesting zooms (I'm a big fan of the 8-16 mm).

Posted by: Jim Kofron

No its not. Its DG so it cant be a very old design.

http://www.sigmaphoto.com/shop/20mm-f18-ex-dg-asp-rf-sigma

"meh performance" is right. If they cant make a prime well they have no hope on a zoom with far more elements.

[ObiJohn]

Wow! I was just resigning myself to the idea of selling off all of my Sigma gear and picking up an A900.

Re resolution, I've found that a Foveon's sensor compares nicely to a Bayer sensor with twice the photosite count, e.g., a 3.4 MP SD10 will outresolve a 6 MP Bayer sensor and be very close to an 8 MP Bayer sensor, and the 4.3 MP SD14 is approximately equivalent to a 10 MP Bayer sensor. That would mean the 15.3 MP SD1 would easily outresolve any '35mm'-sized full-frame dSLR camera on the market. More important, the full-frame SD1 has about the same photosite size as the APS-sized SD9/10 sensor, meaning it won't be diffraction-limited until f/16 or thereabouts. Sigma will win the '35mm' full-frame resolution war because no Bayer sensor will be able to outresolve it; upping the resolution to around 40 MP on a similar-size Bayer sensor and you'll be diffraction-limited below f/8.

If Sigma can get the SD1 performance up to the level of the A850 (and hopefully throw in in-camera image stabilization), and price this camera in the same range as the 5DMkII or thereabouts, they will sell a ton of them. I'll certainly buy one.

Oh, and put the current APS sensor in a new Sigma micro-4/3 body!

[Tom Kwas]

I was always interested in the Foveon chip as well, as seeming to be a somewhat closer "look" to film than the Bayer chips (and somewhat sharper)...the advertising images always seem to have that little bit "more" and many times looked like high quality print jobs from 120 or larger film. Having spent years going down dead-end roads for various technologies, even in the film days, I certainly decided to wait and see what the next step would be, and now this is it, and pretty slow from the last step...maybe I'm waiting for the third...

FYI for all those talking about "reproduction" and the like, it's important to understand that the dawn of digital photography was never, ever related to photographic prints, and always to printed reproduction. As a manger of a large photo department at a catalog company, I was constantly bombarded by sales people and their technologists all through the early 90's, Leaf, Phase One, Megavision, and even Sony (with some god-awful color digital camera the size of a old video studio cam!) and it became apparent early on, that the whole goal of digital reproduction for imaging was to match the look of 35mm to 120 film reproduced at a 150 line screen to a full bleed 8.5 X 11 page. It was stated as much by everyone involved in the process. Every time I would ask about actual physical output for display, I always got the run-around and mystified looks; finally I was told they never really looked at the technology that way. They sort of expected that wedding guys and portraitists would still be using film! Interesting how far it's come from then...