dcmiller wrote:
Even in the best case you claim a 14m sensor Sigma is the equivalent of a 12m sensor bayer, and the bayer camera is inefficient?
Actually, we're comparing a 4.7 megapixel Foveon to a 12 mega-site Bayer sensor.
Also, why do you think the older diagrams from foveon show all layers exposed at the surface? It's because light "passing through" doesn't work. It was an interesting idea that was made into a product before it was proven to be sufficiently sensitive to be a competitive sensor. Foveon played games with the design and marketing to stretch the truth as far as they could.
Then don't buy one, and stop complaining. Enjoy your 4 megapixel Bayer-sensor camera, whose manufacturer claims that it's a 12-megapixel camera. I'll stick to my 150 megapixel images and wait for sensors that actually record color
Alf Beharie wrote:
Actually, in the real world the SD9 and SD10 are equivalent to 10mp Bayer sensor using DSLR's and the SD14 is equivalent to 12mp Bayer sensor using DSLR's...As has been shown in several head to comparisons.
This is wishful thinking and always has been. It's the stuff of myths and legends. I have yet to see a head-to-head comparison that definitively supports or proves this to be the case, although many have suggested it.
I've owned the SD9 and SD10 alongside my Canons and Nikons. The Sigma SD9 was my first DSLR. The native resolution of the Sigmas is 3.4MP, period. The print size is quite small compared to a 10MP Bayer camera. When you download a Sigma SD9/SD10 file at "Double Size" via Sigma's own software, the file definitely gets much larger both in physical size as well as the size of the file in megabytes. But, unfortunately, it's loaded with noticeable artifacts and fine detail is quite smeared compared to the original native file.
The SD14 uses the same exact sensor as the previous models, but it squeezes in an additional and miniscule 1.2MP for a total of 4.6MP. In my own side-by-side of the SD14 vs. the Canon 5D, the differences were substantial. The 5D is the better camera by far in real world usage, particularly in the noise department beginning as low as ISO 200.
As long as we're talking about the "real world," where a camera should be able to adapt to the situation at hand and enable a good picture no matter what, none of the above actually matters very much. The Foveon sensor is a nice camera so long as you have plenty of light and nothing is moving too fast. Otherwise, it's practically hopeless in the kinds of circumstances and situations that the Bayer cameras handle with one hand tied behind their backs. If a single Canon or Nikon displayed the kind of hit-and-miss operational tendencies that the Sigma displays, it would be considered a design failure.
Perhaps in Sigma's hands the design of the Foveon sensor and associated electronics will improve and be brought to 2008 standards. But quite frankly, I doubt it. I think the problem is the Foveon 3-layer design itself.
veroman wrote:
Perhaps in Sigma's hands the design of the Foveon sensor and associated electronics will improve and be brought to 2008 standards. But quite frankly, I doubt it. I think the problem is the Foveon 3-layer design itself.
Why would it be any better in Sigma's hands? They know less about it than Foveon and have absolutely zero experience beyond their collaboration with Foveon. This is clearly not something that is in their area of expertise. And even the "associated electronics" in their cameras has never been up to current standards -- they are probably a one good technology generation behind everyone else. How can anyone think that Sigma could possibly have the knowledge and ability to improve and modernize the Foveon sensor? All they brought to the table was money to keep Foveon from disappearing. Now they have to figure out what to do with it. Sorry, but I am not sanguine about Sigma's ability to do any more with the Foveon sensor than what has already been done (still the same Foveon staff working on it, after all), other than perhaps keeping it on life-support for a couple more years and draining Sigma's resources that would be better spent developing new lenses.
dcmiller wrote:
Even in the best case you claim a 14m sensor Sigma is the equivalent of a 12m sensor bayer, and the bayer camera is inefficient?
Also, why do you think the older diagrams from foveon show all layers exposed at the surface? It's because light "passing through" doesn't work. It was an interesting idea that was made into a product before it was proven to be sufficiently sensitive to be a competitive sensor. Foveon played games with the design and marketing to stretch the truth as far as they could.
Sigma now owns Foveon because they have money sunk into it.. The company failed because their product doesn't match their marketing. ...Show more →
The Bayer mosaic sensors are inefficient because it takes four photo sites to pick up the entire RBG spectrum. So a 12 megapixel Bayer sensor only has 3 megapixels to record red, 3 megapixels to record blue and 6 megapixels to record green. A 4 megapixel Foveon sensor is actually three sensors deep so you have in effect 12 megapixels of photo sites. Granted each of these photo sites may or may not be as sensitive as each photo site on the Bayer sensor.
If light pass through doesn't work then why does my Foveon sensor camera actually works. Did they create the image with a Bayer sensor and fooled the public? I think not. The sensor is sufficiently sensitive enough to create a very pleasant image as far as I know. The pixel density is not quite measured the same way as a Bayor sensor and that is the main point of confusion here.
Economic health of an R & D company does not have any relationship with their technical concept.
Some people in this thread are far too emotionally involved in bashing Foveon. Take a deep breath, take a step back, get the christmas spirit and use rational, mature and fact-backed arguments please.. Religiously fervent attacks on some technology you dislike is not the way to go about these things.
Preface: I think that Foveon design is dumb. It's a fabrication nightmare, even if theoretically it could work better.
With all this banter back and forth about resolution x y z or whatever.
Remember that chrominance information is WAY WAY less important than luminance.
Pure and physical advantage to Bayer pattern.
That's not to say that I think that the photosites of modern cameras are getting small enough that other patterns may realize much greater advantage without losing material chrominance data.
Seems like lot of passion and lot "opinionated" views. Its good if put some links providing people's finding on this topic (with some nice comparisons)..
I have this link in my bookmarks,
too much "english text" here... lets go point by point whatever people are saying...
Daniel Heineck wrote:
Preface: I think that Foveon design is dumb.
More views please... It will help us in professional life too
Daniel Heineck wrote:
It's a fabrication nightmare,
Great in-depth... thanks..
But this team designed (fabrication) the first ever processor fabrication.. But in real life we can not go by "guilt by association", same way we can not take the intelligence of a person guaranteed in every field and forever (remember Shanon)...
The fabrication is not just nightmare but complex & needs lot of innovation. And I think thats the why the crappy companies like Canon, Nikon and Sony failed in this domain...
Dont go far, Sony is nothing but a text book failure case in new technology (how much time it took them to make a stable Cell processor)... Nikon and Canon are Silica company, but sorry... they know how to convert sand to glass, not to chips... they are just follower in fabrication process... good in customization....
That SD14 is equivelent to the 5D? Tell me when you wake up, in every side by side test (remember the one here?) the SD14 looks like what it is and it ain't even in the same league for resolution.
Interesting thread, there is one thing that Sigma has done, which Canon/Nikon has yet to get off their laurels and do - create a compact camera with a decent sensor in it.
You can keep your lousy G10 and G9, as photographers we should not do, is compromise on image quality to this level. The Sigma DP1 alone is an incredibility feat, great lens and sensor. Albeit slow with fixed focal length.
Lets hope to see if Sigma can really take on the G10, LX3 and P6000 digicams. I think this is where their focus should be on - compacts with large sensors, their DSLR is a bit pants to say the least. For now Canon and Nikon has lost the plot to the mega pixel race in the compact market.
asbalyan wrote:
Seems like lot of passion and lot "opinionated" views. Its good if put some links providing people's finding on this topic (with some nice comparisons)..
I have this link in my bookmarks,
Sorry, this is flawed and biased. I and others tried to recreate it with the raw files and could not. His 5D raw conversion looks a lot worse that mine. Also, he failed to consider DOF effects between the 2 camera (SD14 smaller sensor more DOF). Interestingly, after people challenged his results, he removed the raw files from the site. This comparison looks good, but has been debunked.
Beni wrote:
That SD14 is equivelent to the 5D? Tell me when you wake up, in every side by side test (remember the one here?) the SD14 looks like what it is and it ain't even in the same league for resolution.
Right. Every objective comparison (i.e., not those done by Sigma fanboys), shows that the SD14 is competitive with 8-10mp. That's it. The SD14 can't touch the 5D for resolution. I can't believe that there are some fan boys out there still clinging to these myths. I thought this was already put to bed. But hey, if all you need is say 8-10mp, and like the Sigma's output, buy it. But drop the fantasy that it's better than every thing else out there.
And the ball has moved further away from the SD14. Now it's not even competitive with all the prosumer cameras D90, 50D, K20D.And the "new" SD15 is basically the same sensor and camera with a few tweeks. Sigma is seriously falling behind.
Again the Foveon has potential and has a couple of advantages, but they way it's made limited the number of pixels they can put on the chip. I don't not see why they don't make a full frame version as that would be a big help, unless their foundry (or process) is incapable of doing so. It's doomed if they can't.
Again, all they have shown is that they should not be designing cameras, much less investing in sensor technology that they don't know how to manage (throwing more money after bad, IMHO). They should be doing what they know -- designing and building lenses.
Daniel Heineck wrote:
Preface: I think that Foveon design is dumb. It's a fabrication nightmare, even if theoretically it could work better.
It's a good thing that the people who try these new things out have actual brains.
Take for example bicMOS (bipolor CMOS). At first, it was also a fabrication nightmare.
Now look at the company that pioneered it -- it's called Intel -- you might have heard of them?
I have no way of knowing whether or not Foveon's approach will go anywhere, but if everyone thought the way that you do, we wouldn't even have digital sensors right now.
ulrikft wrote:
Hey! I tried to search the forums, but did not find any threads on this particular theme.
1) Why have we not seen any fullframe foveons? Cost? possible at all?
2) Why not more megapixels in the foveon sensors? Cost? Possible at all?
I just imagine that foveon (or foveon like) sensor tech on a fullframe with 15-20mpix would be the dream of many.
I speculated about all this when I owned a Sigma SD9 and had high hopes for at least a 1.3 crop as Foveon advanced the design for Sigma. But they really didn't advance it very much during the 5 or 6 years since its introduction. The SD14 has the same sensor as all of the previous models (there were only 2) with just a few more MP crammed in there.
So why no full-frame or even 1.6 or 1.3 crop? I personally have concluded that there three issues that prevent them making a Foveon sensor of larger physical size:
1) Too much heat generated by the sensor, which results in dramatically increased noise
2) Too much data being passed through, resulting in a painfully slow camera that simply could not compete with even a Canon x-series. Frame-rate would probably be something 1 every two or three seconds. One thing about the Foveon sensor: it does produce very large files due to the amount of data and detail each file contains. I think Sigma would have to invent a new kind of buffer to accommodate the data transferred by a true 20MP, full frame Foveon. In fact, I do think that this is the primary reason they've kept the sensor small to this point.
3) The cost of producing a 20MP full frame Foveon sensor may be prohibitive, perhaps resulting in a camera costing as much as a Canon 1Ds III, but without most of the 1Ds III features. Would you buy an $8,000 bare-bones camera from Sigma just because it was 20MP and full frame?
foveon is great in a point and shoot where it beats all others in image quality but in the slr world it is basically obsolete. sigma is a great company and I really liked their attentiveness in the releases of DP1 firmwares. I will be buying DP2 and if anyone at Sigma is listening and wants to sponsor my 10 day wilderness hike with only DP1 and DP2 then contact me at photo1 @ widereach dot net.
Tamerlin wrote:
It's a good thing that the people who try these new things out have actual brains.
Take for example bicMOS (bipolor CMOS). At first, it was also a fabrication nightmare.
Now look at the company that pioneered it -- it's called Intel -- you might have heard of them?
I have no way of knowing whether or not Foveon's approach will go anywhere, but if everyone thought the way that you do, we wouldn't even have digital sensors right now.
I represent lacking brains!
If you're not somewhat in the field of semicon fabrication--be warned
I need to back off my statement of "Foveon being dumb" a bit, but I do stand behind it being a fabrication nightmare -- specifically isolating the different photodiodes from electron diffusion. Another problem is the sensitivity of the sensor to incident light angle. I'm also wondering how much real estate they're losing to transistors--it seems more than an equivalent bayer design given their noise problems. Does Foveon X3 use microlenses to regain some of their real estate back and possibly improve the angle of incidence?
The Intels of the world have the $$$ and the vision to be able to beat their heads into the ground for several decades to get a superior technology working. Even then they screw up plenty. Think about how long and how much it took them to get HfO2 to work. Foveon and Sigma don't have that money nor momentum and are working with a foundry (Nat'l Semiconductor), which isn't going to spend the money to get it to work well either. Foveon needs a dedicated fab and a LOT of process development to really demonstrate its a better technology.
Best thing that could happen to Foveon is someone big decides its worth going for and buys them out. For a small company to make it in semi fab, they either need to have their IP purchased by someone who can actually make it a reality or invent something ridiculously simple, but conceptually brilliant. I just don't see Foveon in either of those two camps.
Canon, Nikon (Sony), ...etc aren't exactly building barn burner chips either. The design behind these chips is pretty simple and "relatively" easy to design. There aren't too many mask steps if I've done my research correctly. The biggest challenge they've got is getting high yields on absolutely enormous chips.
Anyhow, if anyone wants to talk more about this stuff, shoot me a PM. I'm a big nerd when it comes to fabrication and would love to talk shop.
Beni wrote:
That SD14 is equivelent to the 5D? Tell me when you wake up, in every side by side test (remember the one here?) the SD14 looks like what it is and it ain't even in the same league for resolution.
Nonsense...its been proved on numerous occasions by people who actually have both cameras.
Alf Beharie wrote:
Nonsense...its been proved on numerous occasions by people who actually have both cameras.
I shouldn't even wade into this anymore because in the end "who cares"? We'll all go out and use the system that gives us the results we want. I've owned both the 5D and the DP1 and can emphatically state that without a doubt the 5D (assuming good glass, in this case the 24mm TS-E and 24-70 L) is the far superior camera for large prints - no contest IMHO.
For web display and prints under 10x15" the DP1 is a great little "carry everywhere" camera and I certainly enjoyed using it for about 8 months this year.
Daniel Heineck wrote:
I need to back off my statement of "Foveon being dumb" a bit, but I do stand behind it being a fabrication nightmare -- specifically isolating the different photodiodes from electron diffusion.
I agree -- but also, building sensors in multiple layers is another nightmare, simply because every layer requires another pass through the process, which adds to cost... and combine that with the fact that the Foveon has its read circuitry on the back of the chip on the edges like most other sensors, and now suddenly you need MORE layers...
I only responded that way to the statement of Foveon being dumb because it's equivalent to saying we shouldn't bother climbing Mount Rainier, because it's hard.
Another problem is the sensitivity of the sensor to incident light angle. I'm also wondering how much real estate they're losing to transistors--it seems more than an equivalent bayer design given their noise problems. Does Foveon X3 use microlenses to regain some of their real estate back and possibly improve the angle of incidence?
I don't know for sure, but I'd expect so; being clever doesn't get them out of dealing with the same physics as everyone else
The Intels of the world have the $$$ and the vision to be able to beat their heads into the ground for several decades to get a superior technology working.
They do now, but Intel started off as a puny little upstart with a crappy chip. Intel wasn't a particularly large company as far as semiconductor manufacturers go when they started using bicMOS manufacturing. And AMD isn't exactly flush with cash, yet they're actually implementing immersion lithogrphy, which no one else had been able to do in a high-volume fab...
Even then they screw up plenty. Think about how long and how much it took them to get HfO2 to work. Foveon and Sigma don't have that money nor momentum and are working with a foundry (Nat'l Semiconductor), which isn't going to spend the money to get it to work well either. Foveon needs a dedicated fab and a LOT of process development to really demonstrate its a better technology.
That's probably why Sigma made the DP-1 and bought Foveon; if Sigma can make enough sales volume with point and shoot cameras using Foveon sensors, they can start making some gains from economies of scale. That's obviously a big if and still no guarantee, of course.
Best thing that could happen to Foveon is someone big decides its worth going for and buys them out. For a small company to make it in semi fab, they either need to have their IP purchased by someone who can actually make it a reality or invent something ridiculously simple, but conceptually brilliant. I just don't see Foveon in either of those two camps.
Well, the buying out part happened... let's see what happens next.
Canon, Nikon (Sony), ...etc aren't exactly building barn burner chips either. The design behind these chips is pretty simple and "relatively" easy to design. There aren't too many mask steps if I've done my research correctly. The biggest challenge they've got is getting high yields on absolutely enormous chips.
AFAIK that's true. The big chips are expensive because they're big, and Intel isn't making them
I think also that they're less tolerant of errors than most other chips, since most chips, including the massive current x86 monstrosities, are mostly cache... bad cache blocks aren't a problem because the designers can plan for that and build in redundancy, but how would you do that for a bad photosite? You can't just route the data from another one.
Anyhow, if anyone wants to talk more about this stuff, shoot me a PM. I'm a big nerd when it comes to fabrication and would love to talk shop.
Yeah, I guess this should probably be the last public post on the subject, unless others are interested
Not that I'm opposed to continuing the discussion, though.
Vern Dewit wrote:
I shouldn't even wade into this anymore because in the end "who cares"? We'll all go out and use the system that gives us the results we want. I've owned both the 5D and the DP1 and can emphatically state that without a doubt the 5D (assuming good glass, in this case the 24mm TS-E and 24-70 L) is the far superior camera for large prints - no contest IMHO.
For web display and prints under 10x15" the DP1 is a great little "carry everywhere" camera and I certainly enjoyed using it for about 8 months this year.
V.
I've casually tested my 5D and Leica 28 against the DP-1 and the results were very close. Printed text in the files was easier to read from the DP-1. And the Leica is WAY better than my copy of the 24-70 at 24mm.
We'll all go out and use the system that gives us the results we want. I've owned both the 5D and the DP1 and can emphatically state that without a doubt the 5D (assuming good glass, in this case the 24mm TS-E and 24-70 L) is the far superior camera for large prints - no contest IMHO.