Foveon again

I think what is really absurd is that Canon, Nikon and the rest have been allowed to market their sensors with as high a sensor count as they have. The Canon 10D doesn't really have a 6Megapixels resolution. Rather, it has 3 million green, 1.5 million red, and 1.5 million blue pixels. The pixels are spatially separated so that the original image that the lens focussed upon the sensor will forever remain a mystery. The computer algorithms then interpolate this data to'create' an image which is then touted as having a 6 million pixel 'resolution'. HOGWASH. The aesthetic of the final image is NOTHING like film. It is rather artifically created...very much like a computerized painting. Complex details, textures, of grass and leaves in the distance are simply 'colored' in.

 

The Sigma SD9 originally claimed to have 3.4 million pixel resolution. BUT that was a TRUE FULL COLOR resolution. That was 3.4Million red, 3.4 million blue, and 3.4 million green...with the three colors on top of each other to create one 'true' pixel. No need for interpolation, no need for artificial recreation...a better, more natural aesthetic. Anyone who has seen the images at PBase.com for the SD9 can attest to this.

 

Sigma realized they may have made a marketing mistake in being so 'HONEST' when all the other companies(Canon,Nikon) were so dishonest in their resolution specifications. And so they decided to join in and call each 'color pixel' a 'pixel'. Sigma can RIGHTFULLY claim to having 10.2 million color pixels...just like all the other companies misleadingly do.

 

It is true that the effective resolution of the 10.2 Megapixel SD9 or SD10 is not quite as high as the Canon 1Ds which has 11.4Million pixels. But the Sigma 10D body will sell for $1350 as opposed to over$7000 for the Canon. The SD9 and the SD10 both have considerably HIGHER resolutions than the Canon 10D or the Nikon D100. So there is no exaggeration whatsoever in Sigma's claim for resolution superiority.

 

However, resolution is not the only criteria for a camera. The main problem with the SD9 was that it was limited to iso100. Higher iso's of 200/400 were ridiculously noisy and almost worthless. THe Sigma SD10 has alleviated this problem and may have useable iso400..but I suspect it is still not as noisefree as a CMOS or CCD Bayer sensor. In any event, let me repeat that image NOISE was the main problem that had to be overcome. There was no shortage of resolution!

 

In summary, if you can live with the iso/noise limitations, and with the limitation of only using Sigma lenses, then the Foveon sensor offers images that are of a considerably higher order/aesthetic than a 6Mpixel Bayer.

Just wanted to expand on Bob's statement:

 

"It's [the Foveon] neat technology, but I've yet to see any convincing evidence it's actually better than existing sensors when it comes to performance in the real world. Maybe that will happen one day. It just hasn't happened yet."

 

Consider the attached sketch. This is a hypothetical plot of the performance of a device vs. it's time of introduction. The blue line represents the current lateral filter sensor (e.g. the Bayer scheme) and the dashed lines represents the demands of the low end and high end market (consumer vs professional). My assumptions is that the current generation of Bayer scheme CCD/CMOS sensors meet, if not exceed the requirements of professional photographers today. The green dot represents the introduction of the Foveon sensor. I've assumed that the Bayer sensor had quite a head start. Primarily based on the Sigma SD9 review on www.dpreview.com, I placed it just below the performance of the current CCD/CMOS sensors - I've assumed that it's good enough to be compared to the D60 and certainly exceeds the capabilities of the D30 (older technology).

 

Now here's the interesting part - do you believe the Foveon performance curve will follow the red line or the yellow line? If it's red, then you believe the Foveon will never exceed the capabilities of Bayer CCD/CMOS. If it's yellow, then if you're a sensor manufacturer (Canon, Fuji, Sony, etc), you better be preparing plans today to implement this technology in your products in the future. If you wait for the yellow and blue lines to intersect, it's too late - your competitors are already on the second/third generation while you're still need to figure out how to flip your factories over.

 

For a real-life example unrelated to photography, consider the 3.5" form-factor hard drive. In 1984, there were about 15 hard drive companies making 5.25" form-factor drives. Then one company introduced the smaller 3.5 FF drive. At the time, it didn't perform any better or have more capacity than the larger 5.25" drive; additionally, computer makers designed their cases for the larger drive so there wasn't much market for the smaller drive. 4 years later, only a third of the hard drive companies were able to manufacture the 3.5" FF in volume. What happened to the other companies? Out of business or bought out - they didn't adapt to the emerging technology as quickly as their competitors, thus they perished.

 

My point is, be careful of dismissing Foveon just because emerging technology may not exceed current technology (as an aside - I feel that the fair comparison is the SD9 vs D30, which I think the SD9 wins). Whether or not Sigma/Foveon is better than Canon right now is irrelevent - if Canon et. al. feel that Bayer is going to be reaching a hard physical limit that the Foveon (or Fuji SuperCCD) can exceed (and that's a HUGE if), then they will use the Foveon or have their own implementation of the Foveon.

 

The tech business is brutally competetive - companies can't afford not to invest in new technology.

 

Regards,

I-Liang

Sorry - here's the <a href="http://www.photo.net/bboard/big-image?bboard_upload_id=15756784" >figure</a>

L-Liang, Based on Phil's review? Phil concluded that the SD9's resolution was "at least equal" to the current crop of 6Mpixel cameras. So your chart is misleading by placing Foveon resolution below the current Bayer sensors. I think a fairer comparison would be the Sigma SD10(10.2 mpixels) versus the Canon 1DS (11.4 Mpixel). Admittedly the Canon 1DS might have a slight edge...but at nearly 6 times the price...I'd say the SD10 has already blown away the competition. It'll just take awahile for professionals to part with the lens collections...and their brand loyalty. Lets wait for Phil's review of the SD10.

Hi Anthony -</p>

 

I may have been a little conservative in my assessment of where the SD9 stood in comparison with the 6MP Bayer. The <a href="http://www.photo.net/bboard/big-image?bboard_upload_id=15769484" >resolution chart</a> on dpreview.com does indeed show the SD9 capable of resolving up to 1550, but there also seems to be some fuzziness of the lines around 1200 ~ 1500 (I've circled this in red in the attached clip). I'm not sure if this is an inherent hardware problem or something that can be corrected by software/firmware, but that's why I hesitate to say the SD9 matches/exceeds the Bayer resolution - can the SD9 resolve 1550 and everything lower, or is 1550 just a sweet spot?

 

At any rate, I eagerly await for Phil Askey's fully detailed review of the SD10. </p>

 

Regards,

I-Liang

<p>

Hi,

</p>

<p>

I don't own a Sigma SD9 or SD10 (yet), but I must say I have been impressed by the pictures I saw from these camera's, both on screen and in print (I printed up to A4 size, and I was impressed).

</p>

<p>

For an example of what is apparently possible in print, take a look at this article about printing to A0 format:

<a href="http://forums.dpreview.com/forums/read.asp?forum=1027&message=7182672">A0 prints just came!!! [pic] HOT STORY!!</a>

<p>

One of the printed images can be found here:

<a href="http://www.pbase.com/image/21254066">portrait of musician</a>

</p><p>

Be sure to look at the picture in original size, really very sharp and detailed.

Many more nice pictures (mainly portraits) of the same guy and camera at:

<br>

<a href="http://www.pbase.com/rytterfalk">Pbase page of Rytterfalk</a>

<br>

<a href="http://www.pbase.com/sigmasd9/carl_rytterfalk">PBase page of Rytterfalk in SD9 owners section</a>

<br>

</p>

Cheers, Luc.,

I just read this thread and chose to join photonet in order to add to the discussion. A few questions were raised in the last few months. One was about where the bayer pattern came from. It was patented by a Kodak employee named Bryce Bayer. You can see the patent on the uspto.gov site by searching for patent #3,971,065. In the patent is a good explaination for the logic behind the pixel layout.

 

One of the respondents criticized the bayer pattern as clearly inferior because it measured only a fraction of what foveon did. This is not wholely true. Real images show high degrees of correlation between adjacent pixels even where there is much detail. The sophisticated interpolation algorithms used for bayer pattern reconstruction rely on this fact. This is why there is not such a clear cut advantage (or viewable difference) for Foveon over bayer.

 

The real difference comes when considering the aliasing question. As one person pointed out both sensor types suffer from luminance (variations of light and dark) aliasing. Only the bayer pattern suffers from color aliasing. Both types of aliasing can be corrected in both types of sensor through the use of an anti-aliasing filter (bluring type). However the color aliasing is also correctable in the bayer sensor, without bluring, through the use of another class of filter (which I advocate so that you know my bias). This evens the score once again.

 

The final factor is in how the two sensors separate the light into three colored channels. Foveon uses the physics of the photoelectric material (silicon) which attenuates the light as it passes through. The blue gets absorbed first then the green then the red. By this mechanism there is significant blending of the three channels. This is claimed by Foveon as an advantage. For bayer pattern sensors, many choices of filter dyes can be applied. Those that show limited mixing of the channels are most often used. If more blending of the channels were desired, the sensor makers could choose to use broader filter dyes instead.

 

Hope this helps,

 

Bruce

i was just looking online tonight to see what people had to say about the foveon

sensor. no offense to you guys, but have you ever actually seen prints made by a

good photographer with a good lens and the sd10? i got to shoot with it for the last

3 days and 8x10 prints are astonishingly good. to my eye they are on par or better in

some cases than an eos 10d. the foveon sensor handles moire really really well. i

find that it seems to resolve slightly more detail than the 10d. BUT... canon lenses

are better, period. AND...compared to a 1ds, its no competition. i love that camera. i

want one so bad. everyone time i use it i take soooo many photos it would make you

sick.