Why on earth doesn't Foveon...

[Joseph S Wisniewski]

what would you say about R1 type camera with Foven sensor by Kodak
or Polarod?
Just this is a right moment for some to punch in this gap.

I don't think Kodak would touch the Foveon sensor, unless they
partner like FillFactory did.

What does the R1 offer compared to a DSLR with live preview ?

A DSLR with live preview still has to use DSLR lenses. They need a lot of "back focus", a big gap between the rear element of the lens and the sensor to accommodate the moving SLR mirror. You can increase optical quality by having lens designs where the rear element comes very close to the sensor. Many point and shoots have lens designs like this.

Also, cameras with permanent lenses like R1 tend to stay cleaner inside than interchangeable lens cameras.

--
Normally, a signature this small can't open its own jumpgate.

Ciao! Joe

http://www.swissarmyfork.com

 

[Kendall Helmstetter Gelner]

Some shutters "leak" too much light to allow the mirror to be
locked up for extended periods of time. To get blades to move very
fast (for high x sync time) or very smoothly (for high max shutter
speeds) on low cost shutters, often the blades have minimum
overlap, and light can bounce between the blades a few times and
make it through to the film or sensor.

Thanks, that does explain why some cameras might skip that feature.

One thing I think is really inexcusable though is when a camera can support mirror-lockup but the camera maker leaves it out on purpose so as not to undercut a higher end model... anyone remember the 350D/10D? I seem to remember you could enable mirror lockup with a bootleg firmware update.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmasd9/user_home

 

[Joseph S Wisniewski]

Some shutters "leak" too much light to allow the mirror to be
locked up for extended periods of time. To get blades to move very
fast (for high x sync time) or very smoothly (for high max shutter
speeds) on low cost shutters, often the blades have minimum
overlap, and light can bounce between the blades a few times and
make it through to the film or sensor.

Thanks, that does explain why some cameras might skip that feature.

One thing I think is really inexcusable though is when a camera can
support mirror-lockup but the camera maker leaves it out on purpose
so as not to undercut a higher end model... anyone remember the
350D/10D? I seem to remember you could enable mirror lockup with a
bootleg firmware update.

Yes, that one is a case where the hardware was capable, and the feature was left out for "product differentiation".

But Canon learned their lesson, mainly because Nikon went "no holds barred" on their competing models. Removing features from 300D may have kept Canon from losing some 10D sales to 300D, but it made them lose a lot more 300D sales to Nikon D70. So now, if the hardware can do it, Nikon, Canon, Pentax, etc. seem to make sure the software supports it.

--
Normally, a signature this small can't open its own jumpgate.

Ciao! Joe

http://www.swissarmyfork.com

 

[dnjake]

Major camera manufacturers have not chosen Foveon.

Only a modest number of customer's have chosen the Foveon implementations that have appeared.

The conclusion seems fairly simple. The advantages of Foveon technology if they exist at all are small enough to be lost in the noise of all the other factors that determine the cameras that people buy. The truth is that the newcomer is always going to be at some disadvantage. So far Foveon does not seem anywhere close to the threshold needed to displace more established technologies that increasingly do a very effective job.
--
David Jacobson
http://www.pbase.com/dnjake

 

[Roland Karlsson]

Really, from a bussiness point of view I don't understand how
Foveon is in the bussiness.

but now we have here another Foveon Sensor in a Sigma camera.

The way I have got it explained to me is that the owners of Foveon do not have the primary plan to get rich fast. They are already rich and they think this project is fun. Of course - they want it to succeed - but they have lots of patience if it takes time.

This might not really be 100% true as it sounds too good to be true - but thats how I have understood it.

When it comes to Sigma - I have got it explained that Sigma is a privately held company and not a share holder company. It is also going very well and the owners think it is "fun" to have a camera line besides the lens manufacturing.

Also sounds somewhat weird - but it is not unreasonable.

Those two lucky circumstance makes the combination Foveon/Sigma almost invincible even if they do not make profit now.

--
Roland
http://klotjohan.mine.nu/~roland/

 

[Joseph S Wisniewski]

1. Canon manufacturers their own chips. Although one could make a
great case for image improvement in Canon cameras if they
contracted Foveon to do a 12MP x3 chip (a "SD36"), you'd have to
ask the question:

"Does Canon have a problem selling cameras?"

Well, 5D tends to sit on the shelves. And oddly enough, for much the same reason that SD10 did, although the sensor is pretty cool, the 5D body was very "anemic" for its price class.

But that's not the real issue the camera you describe. A 12x3mp Foveon produces a lot of data that has to be stored and processed. Canon's intermediate processing platform (shared by 30D and 5D) can manage to put away 40 million samples/sec of data (5 frames/sec at 8mp for 30D, 3 frames/sec at 12mp for 5D). With a 12x3 Foveon, you'd be shooting an incredible 1 frame/sec, with a 4 shot raw buffer, and 15 seconds per image to write a raw to flash.

2. Nikon uses Sony chips in both their dSLR line and their
point-n-shoot cameras. Might there be an issue with Nikon dropping
Sony's chip and going with a Foveon??

I doubt it. Doesn't seem to affect Canon or Olympus (Kodak and Panasonic chips in the DSLRs, Sony in the point and shoot). Nor did it stop Nikon from using their own LBCAST sensor in D2H and D2Hs, or from supplying bodies to Fuji.

3. Kodak would have been a very logical partner with Foveon to
build a dSLR. When in the last 20 years has Kodak shown foresight
and logic and make a good strategic business decision and stuck it
through?? (disclaimer---ex Kodaker here...)

Not much of a business model for Foveon, they've already got Sigma for low quantity sales.

4. Sony. No. Pentax. I don't know enought about them. Leica? Would
have been nice, but they got a nice Kodak chip.

Leica needs low light, it's the nature of their market. ANd Foveon would have to develop a new technology (offset microlenses) to accommodate rangefinder lenses. Poor match to Foveon.

5. Sigma---a family run business that wants to be a complete
imaging system. A great choice. Now---why don't they sell Canon and
Nikon mounts??? Where do you think Nikon, Canon, and Sigma make
their money---with the cameras, or with the lenses?

Nikon and Canon take plenty of profit from the cameras. Then again, their quantities are 100x higher than Sigma. I used to believe that Sigma could afford to run the camera as a loss leader, but Thom brought me around to the point of view that the numbers don't work. They would make a lot more money in sales of Nikon and Canon mount lenses by pumping X dollars into new lens designs or increased advertising than they would be spending it on camera R&D and maintaining a non profitable item in the distribution channel.

The SD14 is reportedly a very nice camera with a great viewfinder,
a 100,000x better shutter, and improvements in almost all major
areas where users were looking for them. The only current issue
that I see was the lead trial balloon that they sent up regarding
MSRP.

Remember, MSRP on Sigma, Tamron, and Tokina lenses is insane, too. That lets dealers offer wonderful "discounts" like 50% off MSRP.

Unfortunately, for DSLRs, Nikon, Canon, Pentax, etc. don't talk about MSRP, they talk about expected street price.

If they divide that by the crop factor, they'll be in a great
position to sell a bunch of these.

--
Normally, a signature this small can't open its own jumpgate.

Ciao! Joe

http://www.swissarmyfork.com

 

[Joseph S Wisniewski]

--partner with a major camera manufacturer (ala - Fuji and Nikon)?

To reiterate a few things.
Fuji, Canon, Sony all make their own sensors. You can forget them
right away.

Can you? Although Canon used their own sensors in their first camera, the 30D, when they launched the higher performance flagship 1D, they went to Panasonic for the sensor. CCD performance was more important that CMOS economics. Companies aren't as tied to internal sources as you might suspect.

Basically you have to sell to who is left, where it is mainly a
foveon vs Sony game. And there are many issues and obstacles.

1: Nikon, Pentax are set on 1.5 crop. 1.7 would not be acceptable.

The big question in my mind, is why build yet another 1.7 crop
sensor? Why disadvantage themselves like this?

Even if they built a 1.5 crop sensor, there is no guarantee anyone
would bite. I don't think they have a compelling case.

Not everything is built "on spec". If Nikon or Canon said "we're interested in a 1.5x crop version", Foveon would make one. And it wouldn't cost that much, if they didn't change the cell geometry. Same size pixels, just lay out more of them. Computer aided design systems are wonderful at stuff like that.

2: I surmise sigma sensors are more expensive than Sonys for a few
reasons. Economy of Scale, Penny pinching on 1.7 crop, and being
fabless (profit must be taken at two stages for Foveon and for the
Fab).

Sony is very well isolated, their sensor people take profit, even from internal customers like their own camera division.

3: Sony has demonstrated ability to deliver large volumes and ramp
new designs fairly quickly, and deliver chips that have faster
datapaths etc. Foveon has not and represents a risk.

The data paths on the Foveon chips are very fast. Remember, the Foveon is inherently a 3 channel device. The sensor used in SD10 can make 4 frames/second in an industrial camera, and it only has to clock at 13.6 MHz to do that. It's just Sigma's use of a relatively slow processor that wouldn't let it put data away faster than 1.8 frames/sec at full resolution.

4: Foveon sensor may fail to meet all design targets. High ISO
performance may still not be up to par. Off angle light may be more
problematic (hence the 1.7 crop).

And here is where I think you really nailed it. High ISO is the single biggest problem with the Foveon design, and I;m not sure it's addressable.

Off angle light isn't a problem. The fill factor for the sensor under the chip is pretty high, over 50%, so the microlenses used to increase sensitivity aren't any more restrictive on angle than those used by Sony or Canon.

5: Possibly no competetive advantage is seen in the design.

Agreed. It's a "different" sensor, better in some ways compared to Bayer filter arrays, worse in others. It has some uniques strengths, and I think the really big news is the new fixed lens camera, which exploits the strong live preview capability of the Foveon.

While it generates Buzz for Sigma, Nikon and Pentax don't need that
buzz so much and it the design is behind in any area, it could be
more of a curiousity than a serious design.

Yup.

Though a vocal minority see special quality the vast majority just
don't care and see just a different set of tradeoffs.

Exactly.

Summary:

Basically I think that is it. I think they are not getting into any
of the big guys unless they demonstrate massive success in the
market elsewhere first. Like the Sigma SD14 outselling D200s. I
think we all know that is not going to happen.

This only needs to happen proportionately. SD14 doesn't have to outsell D200. D200 sells as well as a camera in its class could be expected to do from a company like Nikon. Based on the type of company Sigma is, market analysts will expect them to cell a certain number of cameras in SD14s class. If SD14 radically outsells this expectation, it will prove that the Foveon is a success, and that other camera companies might be able to use it to sell beyond their own expectations.

I think they failed to capitalize on their second chance, just as
they did with their first. By delivering just 4.6 MP X3 in a 1.7
crop they will again be aiming for parity with entry level 10MP
bayer cameras in image quality ( the way most people perceive it).
Parity is not the way to win with a new design.

[snip, a lot of stuff I agreed with]

At minimum they should have delievered 1.5 crop 6MP X3 for this
iteration to at least start ahead of the competition and to have
the crop other potential customers might use.

Possibly not. I thought it should be up at a higher level, too. If the new pixel pitch works out for them, a boost to 1.5x crop would have got them to 5.9x3 (might as well scrunch the pixels 1% smaller and hit an even 6.0). That would give it a fighting chance against 12-15mp Bayer cameras, and would have guaranteed at least a year of life for the new sensor.

--
Normally, a signature this small can't open its own jumpgate.

Ciao! Joe

http://www.swissarmyfork.com

 

[Pete Stavrakoglou]

The Sigma isn't for you. Don't buy one.

--Richard,

Like I said, I'm sorry to sound so rough. I am amazed at the
images that I see taken from the SD-10 users in your forum. Don't
you feel that the body is lacking?

Some of use like the ergonomics of the SD9. Any setting can be changed with a dial, no need to dig into the menu system. I can't say the same for my D50 (I do realize that other Nikons have two dials while the D50 has only one). The build quality is not at all cheap, it's a well-built camera. Although I haven't handled or even seen the SD14,I would take exception with your characterizing the body being like a Hyundai. Here's a quote from a KM - Sony DSLR user (a professional) who has handled the S14:

"The Sigma SD14 is not a small, light camera - it's still got a glass prism, it's really solid, a little smaller than the SD10 due to better battery system, much more solid in feel, and has a shutter mirror sound and action to die for. Honestly, it's the best sounding and 'feeling when fired' camera around. And the new shutter in the SD14 is good for 100,000+ exposures."

the thread can be read here http://forums.dpreview.com/forums/read.asp?forum=1037&message=20202331
--
My humble photo gallery: http://ntotrr.smugmug.com

 

[Pete Stavrakoglou]

Somebody needs to introduce Foveon to Canon...

Sigma already has a Canon-mount body. It's the body they provided
to Kodak for use as the Kodak SLR-C (Canon-mount). Just like Kodak
put a Kodak sensor into it, Sigma could have just as easily put a
Foveon sensor into it.

No they couldn't have done that. Sigma was not licensed to use the mount. Sigma was a fabricator for Kodak, it was Kodak's license in use, not Sigma's.
--
My humble photo gallery: http://ntotrr.smugmug.com

 

[Paul Roberts]

JustMosing wrote:>
And by the way, according to Canon's Chuck Westfall, Canon doesn't
license their lens mount to any third parties. All the third-party
companies offering Canon mount lenses are the product of
un-licensed reverse engineering.

Tamron lenses are licensed, or so I've been told in camera shops.

--
Paul A Roberts
http://www.paul-roberts.com

 

[richard stone]

In a few words: No, I do not feel that the body is lacking, really.

It is substantial.

It has all the controls I need, and they are accessible quickly.

The autofocus and other basic elements work well.

I do not believe that the camera was intended for wedding use, sports, or even for pictures of fast moving small children.

As noted elsewhere on a thread rergarding the new Sigma cameras, it is a camera intended for artists. That is why the "focus" is on image quality, rather on modes and features of minimal benefit to the user. Instead, the camera user makes all the decisions.

Personally, as a medium format user, I never really considered the other brands after I saw the pbase pictures from the Sigma and the pictures from Stephen (Steven?) Johnson from the SD9. They were exquisite.

My present take on all of this is to question how photographers (as opposed to consumers) were ever persuaded that it is reasonable and correct, and the best way, to use a system of digital imaging that requires the image to be "blurred" to avoid artifacts, then the colors and edge details to be guessed at by mathematical formulas, and which thereafter necessitates the use of sharpening to even come close to the original image. As part of that process colors are corrected and "enhanced" to provide the sense of real life, if not the actual reality. That it works at all is miraculous and a feat of electrical engineering and computation. However, somewhere in the process the truth is lost and something else is substituted for it.

That photographers have been so persuaded is perhaps itself a miracle of marketing.

 

[dkloi]

But that's not the real issue the camera you describe. A 12x3mp
Foveon produces a lot of data that has to be stored and processed.
Canon's intermediate processing platform (shared by 30D and 5D) can
manage to put away 40 million samples/sec of data (5 frames/sec at
8mp for 30D, 3 frames/sec at 12mp for 5D). With a 12x3 Foveon,
you'd be shooting an incredible 1 frame/sec, with a 4 shot raw
buffer, and 15 seconds per image to write a raw to flash.

I'm pushing for a sensor which has 90% of the advantages of a Foveon-type sensor, but with 67% of the storage. Instead of a three-layer stack, have a two layers; one layer would be all green, and the other layer would be a red/blue checkerboard. Subsampling the red and blue image components would not substantially decrease visually significant resolution. The green layer would provide the majority of the resolution of a fully sampled sensor. It'd only cut down the frame rate by a factor of 2, not 3. Engineering such a device is a bit harder, may have to use the Fujifilm organic CMOS technology to really make it work. May help with colorimetric accuracy as well.

Anti-aliasing could be interesting, not an expert on de-mosaicing algorithms for such data. May need AA filters which don't affect green wavelengths to avoid colour moiré but not reduce resolution (much).

Actually, you could keep the subsampling of red/blue as in a conventional RGBG Bayer array and simply fill in the holes in the green array. Up the resolution significantly but only increase the storage requirements by 50% compared to Bayer (or half of Foveon x3).

My 2c,
Daniel.

 

[Peter G]

Can you? Although Canon used their own sensors in their first
camera, the 30D, when they launched the higher performance flagship
1D, they went to Panasonic for the sensor. CCD performance was more
important that CMOS economics. Companies aren't as tied to internal
sources as you might suspect.

Canon had a specific need that they couldn't meet in that case. As a long time tech geek, I remember the 1D using third party sensors. But today for a host of other reasons, I don't think they see that need. I think the internal fab gives them a competetive advantage that there would have to be a significant design advantage that they needed before they would source an external DSLR sensor again.

Not everything is built "on spec". If Nikon or Canon said "we're
interested in a 1.5x crop version", Foveon would make one. And it
wouldn't cost that much, if they didn't change the cell geometry.
Same size pixels, just lay out more of them. Computer aided design
systems are wonderful at stuff like that.

New layouts, new mask, new test runs. These are up front costs to absorb. Doing smaller custom runs means amortizing those up front costs over a smaller number of chips. It would make more sense to do 1.5 right from the start that has wider appeal unless Sigma demanded 1.7 crop. This is the one decision I really don't get. 1.7 crop make sigmas own digitial lenses off. They design for 1.5 but build 1.7??

Sony is very well isolated, their sensor people take profit, even
from internal customers like their own camera division.

Even if totally isolated, that is still better than a fabless firm, where the foundry has to take its own independent profit, before the fabless IP house takes theirs, then finally the camera maker has to take theirs. An integrated house like Canons gives them a competetive advantage IMO.

The data paths on the Foveon chips are very fast. Remember, the

It was more a discussion on being on sony being a stable supplier that you can count on for consistent supply and relatively rapid updates. Speed varitions were just an example.

Off angle light isn't a problem. The fill factor for the sensor
under the chip is pretty high, over 50%, so the microlenses used to
increase sensitivity aren't any more restrictive on angle than
those used by Sony or Canon.

I remember a presentation with one of the Foveon guys where they as much as admitted they had more angular issues than bayer designs. And I have seen some bizzare artifact on the edges with specular highlights. I still keep wondering why 1.7 crop again.

strengths, and I think the really big news is the new fixed lens
camera, which exploits the strong live preview capability of the
Foveon.

This was the most significant news for me as well. But mainly because it was a pocket camera with an APS sensors . Sony has shown similar capability with the R1. I may get a DP1 once the price drops enough and no one else enters the market before it gets cheap enough. If the prototype shown is any indication, it is quite a primitive camera in terms of handling/controls. I want someone to build this, but I won't jump until the data is in and the price is right.

certain number of cameras in SD14s class. If SD14 radically
outsells this expectation, it will prove that the Foveon is a
success, and that other camera companies might be able to use it to
sell beyond their own expectations.

I suspect it won't happen for reasons highlighted. I have read some non digicam technology boards and the camera was met with huge yawns. Those who don't understand, don't see the point at all. Of those who do undertand most are in the camp that use a 2x factor for equivalence and realize they just about where they were at previous launch. Aiming for parity.

Possibly not. I thought it should be up at a higher level, too. If
the new pixel pitch works out for them, a boost to 1.5x crop would
have got them to 5.9x3 (might as well scrunch the pixels 1%
smaller and hit an even 6.0). That would give it a fighting chance
against 12-15mp Bayer cameras, and would have guaranteed at least a
year of life for the new sensor.

This is pretty much my thinking as well. But the inital review will be against 10MP bayer. It would give them a better footing if they had 6mp chip that clearly had an edge, rather than a parity chip again.

Imagine if SD14 was the first camera released, but with a 1.5crop 6MP x3 sensor. (compared to 6MP bayer at the time) It would have demonstrated such an overwhelming advantage in resolution that there would probably be other Foveons on the market today. And they could still be using that sensor for a year or two more.

If you are trying to break into the market, you can't aim to simply equal what is there. You must surpass it. Makiing that mistake once is understandably, twice is probably lethal IMO.

As we move to greater spatial pixel counts, the advantages of a layered sensor become more and more diminished. As the true advantages are only in a 1 pixel wide locality (the blur and interpolation effects of Bayer+AA filter). The greater the pixel density the smaller the relative area of issues.

 

[Joseph S Wisniewski]

Somebody needs to introduce Foveon to Canon...

Sigma already has a Canon-mount body. It's the body they provided
to Kodak for use as the Kodak SLR-C (Canon-mount). Just like Kodak
put a Kodak sensor into it, Sigma could have just as easily put a
Foveon sensor into it.

No they couldn't have done that.

Of course they could.

Sigma was not licensed to use the mount.

True, because no license is necessary. First, the patents on the Canon mount expired several years ago. Second, even if the patents were still in force, it wouldn't matter, because courts around the world have supported the legality of properly done reverse engineering time after time, in field after field, for the last couple of decades. Sigma can't hire corporate spies to smuggle documentation out of Canon, but they can use the well supported "double clean room" methodology. One team dissects the Canon cameras and lenses and works up as thorough a specification for the interface as they can. A second, independent team then builds lenses, teleconverters, and yes, even cameras, to this specification.

--
Normally, a signature this small can't open its own jumpgate.

Ciao! Joe

http://www.swissarmyfork.com

 

[Joseph S Wisniewski]

But that's not the real issue the camera you describe. A 12x3mp
Foveon produces a lot of data that has to be stored and processed.
Canon's intermediate processing platform (shared by 30D and 5D) can
manage to put away 40 million samples/sec of data (5 frames/sec at
8mp for 30D, 3 frames/sec at 12mp for 5D). With a 12x3 Foveon,
you'd be shooting an incredible 1 frame/sec, with a 4 shot raw
buffer, and 15 seconds per image to write a raw to flash.

I'm pushing for a sensor which has 90% of the advantages of a
Foveon-type sensor, but with 67% of the storage. Instead of a
three-layer stack, have a two layers; one layer would be all green,
and the other layer would be a red/blue checkerboard. Subsampling
the red and blue image components would not substantially decrease
visually significant resolution. The green layer would provide the
majority of the resolution of a fully sampled sensor. It'd only cut
down the frame rate by a factor of 2, not 3. Engineering such a
device is a bit harder, may have to use the Fujifilm organic CMOS
technology to really make it work. May help with colorimetric
accuracy as well.

Anti-aliasing could be interesting, not an expert on de-mosaicing
algorithms for such data. May need AA filters which don't affect
green wavelengths to avoid colour moiré but not reduce resolution
(much).

Actually, you could keep the subsampling of red/blue as in a
conventional RGBG Bayer array and simply fill in the holes in the
green array. Up the resolution significantly but only increase the
storage requirements by 50% compared to Bayer (or half of Foveon
x3).

Foveon has an interesting patent where the green layer twice the resolution (4x the pixels) of the red or blue layers. Similar in principle, you get pretty much the same resolution, with 1/2 the data storage required.

--
Normally, a signature this small can't open its own jumpgate.

Ciao! Joe

http://www.swissarmyfork.com

 

[pahountisg]

--Im sure the photos are razor sharp but spare me the sigma body and mount is way too limited for me...
Make one for canon or nikon mount and it will sell like crazy...
Too bad for foveon..........

Remember VHS in video? It wasnt the best format but it was the one that succeded the most.

Sooner or later Canon or Nikon or maybe sony or pentax will make something equivalet to foveon and here goes a great company...
http://www.pahountis.gr

 

[Morven]

True, because no license is necessary. First, the patents on the
Canon mount expired several years ago. Second, even if the patents
were still in force, it wouldn't matter, because courts around the
world have supported the legality of properly done reverse
engineering time after time, in field after field, for the last
couple of decades. Sigma can't hire corporate spies to smuggle
documentation out of Canon, but they can use the well supported
"double clean room" methodology. One team dissects the Canon
cameras and lenses and works up as thorough a specification for the
interface as they can. A second, independent team then builds
lenses, teleconverters, and yes, even cameras, to this
specification.

All true. However, it's possible Sigma still does have an agreement with Canon. How quickly did Sigma come out with EF-mount lenses after the EOS system came out? If before the standard patent expiration period, it's quite likely an agreement was signed.

-Matt

 

[DavidMaven]

--
[email protected]

 

[Leo]

No Text

 

[Chuck A Wulla]

"In-camera JPEG support"

Don't laugh, that's one of the things that put me of The Sigma, which is funny because I only shoot RAW now anyway.