Jeff Keller LOVES the Sigma SD10
Full Review including gallery here
tis shame that Canon, Nikon and others will most likely ignore this sensor. Matsu, Eugene, Powerdoc get in here and tell us what you think!
Do we know if any other companies besides Sigma and Polaroid are using the Foveon sensor in the future?
Quote:
While the Sigma SD10 would be a nice D-SLR if it used a traditional CCD or CMOS sensor, it's the SD10's Foveon X3 image sensor which really makes it stand out from the crowd. Image quality is, in a word, stunning. The resolution and detail makes your typical 3 Megapixel camera look like a toy
While the Sigma SD10 would be a nice D-SLR if it used a traditional CCD or CMOS sensor, it's the SD10's Foveon X3 image sensor which really makes it stand out from the crowd. Image quality is, in a word, stunning. The resolution and detail makes your typical 3 Megapixel camera look like a toy
tis shame that Canon, Nikon and others will most likely ignore this sensor. Matsu, Eugene, Powerdoc get in here and tell us what you think!
Do we know if any other companies besides Sigma and Polaroid are using the Foveon sensor in the future?
Comments
Originally posted by hmurchison
Full Review including gallery here
tis shame that Canon, Nikon and others will most likely ignore this sensor. Matsu, Eugene get in here and tell us what you think!
Do we know if any other companies besides Sigma and Polaroid are using the Foveon sensor in the future?
This thread belong to digital hub.
Foveon sensor have their cons also. They have less sensitivity, less resolution, the SD 10 is sligty below a 6 MP.
Canon only use CMOS technology produced and designed by them
Nikon use Sony CCD sensor.
Both are happy with their technologic choice, and give pretty good results.
You have also Fuji with it's new 12 MP sensor.
Imagine a Foveon processor with Canon or Nikon lenses. Mmmmmmmm
With a 1.7X multiplier and "only" 3.4MP the foveon isn't scaring anyone. Look at that ISO 800 performance: Most of the products from Nikon, Canon and Fuji can do considerably better. And processing is getting better all the time too, small sensor digicams are figuring out how to kill CA, and noise. And they ain't dealing with the best raw data by any means, garbage in... so to speak. If you feed some of the new processing technologies from Canon and TI (just for instance) with higher quality (larger) imagers, you will get good results, they're getting good results already and will only get better. Is that Foveon better than a D2H with only 4.5MP?
This all forstalls the (perhaps) inevitable move to full color pixels, as does a certain lack of processing power, though the Mark II demonstrates that a big data rate is within the grasp of current tech.
Kodak has cleaned up the DCS 14n, and SLR/n with seemingly little change in process tech. With new tech that allows less dark current noise and/or similar/better dynamic range down to slightly smaller pixel sizes, we probably won't see a need to move to foveon type technologies untill manufacturers slap into 20-22 MP limits at FF sizes, and 12-14 at APS sizes. Foveon themselves have commented that they believe they can get the same DR, and noise characteristics as today from pixels as small as 4um.
But that data rate will kill you. Mark II sounds awesome, 8MP@8fps. Or 16MP @ 4fps or about 30MP @ 2fps, making a few too many assumptions about processing power, but just to give a ballpark.
Now give that number X3. Is there an engine that can process 24@8 or 48@4 or 60+@2fps on battery power and keep cool/small enough to handhold without asbestos mitts and a crane?
Not yet, but probably sooner than later.
It's going to take 5-10MP X3 product before people really look at foveon. The returns will be better when we hit the physical limits of pixel pitch. For now, I think the budget entries from Canon and Nikon are probably better cams, and except for moire, give slightly better res, noise, and low light performance.
For now... it will change.
That's kind of the idea that i'm reading which is the 3.4Mp X3 is = to a 6Mp Bayer. So the next thing to watch for is just how high Foveon can take this and will it remain affordable?
The one thing to think about with full frames is light fall off.
Time ago, everyone was worried about this, particularly with wide angle lenses, the high angle of incedence at the edges could cause a few problems that don't happen on film. Olympus' 4/3rds system is built around this problem, but as it turns out, it hasn't been much of a problem at all for bayer type sensors, with only one layer. I think the DCS14n had some issues, and the CMOS chip had to be recalibrated in the SLR/n, but generally there isn't really a problem. Foveon sensors, with their deeper wells, stacked on top of each other to boot, may have a harder time of it, and actually require lenses that keep light more perpendicular, like those of the E1, nee 4/3rds system.
Then again, it might be something that can be cured with micro-lenses. Light fall off might even be dealt with electronically in the future. HP's cams can do a digital flash trick that seems to work pretty well, maybe a little intelligent dodge and burn or other computational tweak can compensate for vignetting in the same way?
Also, a 3.4MP foveon and a 6MP bayer look to be pretty much equal.
Down sampling a 6MP image to 3.4 produced a most X3 like image, a tad bit of color artifact, but not much at all for a 1.8-1 downsample.
Upsampling the foveon to 6MP provided pretty much the same image but a tad fuzzier.
Give the (slight) edge to bayer for bigger prints, give the edge to foveon for color fidelity.
I'd love to use a foveon to do some close up work with a nice macro. Looks like it'd make real nice images there.
I think it comes down to price, ease of use, and lenses.
I'd really really like to se a 5MP+ foveon though, 6-10MP would be truly stunning. I think a 6MP (18M photo-detectors) X3 FF would win a lot of converts for Sigma.
Originally posted by Matsu
[B]Down sampling a 6MP image to 3.4 produced a most X3 like image, a tad bit of color artifact, but not much at all for a 1.8-1 downsample.
I saw a ton of sharpening artifacts in the test images, in the watch-face and bottle label.
Upsampling the foveon to 6MP provided pretty much the same image but a tad fuzzier.
I wonder if upsampling the image in 10% steps in Photoshop instead of doubling the size in Photo Pro and then downsampling again would be better.
I'd really really like to se a 5MP+ foveon though, 6-10MP would be truly stunning. I think a 6MP (18M photo-detectors) X3 FF would win a lot of converts for Sigma.
I don't think it would, though the Canon mount Kodak SLR/c in a Sigma body raises some eyebrows...
These are the two cams SD 10 and EOS 10D, both at the 3.4MP size, can you tell which is which without editing. I think you can, but be honest, it isn't that easy...***
I'm not really seeing the massive artifacting from the Canon down-sample ?
As for the upsampled SD10 (not shown), upon second look, while a bit fuzzier, the SD10 up-sample appears to hold the detail, and something like the incremental approach in PS or genuine fractals would preserve that last bit of sharpness.
BTW, the above test crops came from www.DPReview.com and everyone reading this should visit the site, it's very thorough.
*** though the color saturation does give it away immediately to those who've used at least one of the two cams.
The SD10 image also looks more three-dimensional. This is especially apparent in the close-up of the measuring spoons. Look at the metal beads. And on the face of the playing card, you can see the lines are much crisper on the SD10 sample.