dcmiller wrote:
With bayer, CCD will soon kick CMOSes butt. CMOS is about size and functionality. WIth CMOS, the ability to construct 3 color sensors is a place CCD can't follow for a long time.
So are you saying the 40D will have 10MP true-colour- IOW Bayer equivalent of about 30MP?
If so, is this based on any information, and how do they get that much data on to CF/SD cards?
Regards,
DaveMart
Yakim Peled wrote:
Undoubtedly but I - personally - tend to get much more excited by new lenses. Something like a 20-50/2.8 IS will suit me just fine....
I prefer women myself. Well, unless we are talking about something like 200 1.8 IS.
Just so we are on the same page on this whole Full Color / True Color / Foveon talk...
A 10 Mpixel Foveon sensor is a 3-layer sensor, with in each layer one third of the total. This was apparent already with the Sigma SD10, 10 Mpixel, 2268 x 1512 x 3, which would, by and large, get you the same image as a 6 Mpixel traditional Bayer sensor, like on the Canon10D.
So, if we're talking about a 40D with a layered full-color-per-pixel sensor, we should be talking about 14 or 16 mpixel, to arrive at a comparable 10 mpixel bayer resolution.
dcmiller wrote:
20mp full color has the same file size as 20 mp bayer.
Your assuming the bayer raw does not have the two missing color channels filled in. It does.
Here is the file size from the Nikon D80 (10MP, bayer) http://www.dpreview.com/reviews/nikond80/page12.asp
8.5MB for RAW, with a rez given at 3872 by 2592
Here is the Sigma S10: http://www.dpreview.com/reviews/sigmasd10/page9.asp
7.7MB 2268 x 1512
So presumably if you are indicating that the file sizes are around what you would expect from a 10MP Bayer, then the nomenclature that is being chosen for the Canon truecolour is the same as Sigma uses, and the 40D would be 2268 by 1512
Regards,
DaveMart
i read here and there that Canon has a patent on a foveon-like sensor, but does anyone know more about it? maybe some recent quotes from Westfall et al?
So presumably if you are indicating that the file sizes are around what you would expect from a 10MP Bayer, then the nomenclature that is being chosen for the Canon truecolour is the same as Sigma uses, and the 40D would be 2268 by 1512.
And that would put it in D40 territory, resolution wise. I doubt that will happen. To get a 3-layer sensor with the Bayer equivalent of 10 mpixel would have about 5 mpixel per layer, hence the 14 to 16 mpixel number I mentioned on top of this page.
If Canon introduces a new Eos 1 with a 25.5x36mm sensor (for 3/2+4/3 ratio) and if it works with most L lenses, it is a brilliant idea. I dream of such a Canon Camera.
I'm not so excited by the Foveon-like hypothesis. I prefer a classical Bayer (or why not a 4-color Bayer, see Sony f828) with excellent color gradation, wide color gamut and wide dynamic range. Foveon is not the ultimate solution : as far as I know, MF digital back are all Bayer-type. And they have excellent color gradation and gamut.
pmnx wrote:
If Canon introduces a new Eos 1 with a 25.5x36mm sensor (for 3/2+4/3 ratio) and if it works with most L lenses, it is a brilliant idea. I dream of such a Canon Camera.
The taller mirror that would require would kill the use of a lot of Leica and Zeiss glass on the new 1 series. So I hope they don't do that. In other words the mirror would hit the back of the Zeiss and Leica glass during an exposure as it flips up.
Don't get me wrong. I love Canon L glass. But at 22MP you could really get some killer landscape shots with amazing resolution using Leica or Zeiss glass. That would really compete with medium format digital.
I find myself in the minority. I do not want more MP. I think we have more than enough. I'd welcome improvements in other areas such as DR, brighter viewfinders, programmable buttons etc.
Alistair101 wrote:
Perhaps on the Digic II platform yes. The next series of new releases should well be Digic III based and therefore take 'the next step' in noise reduction following on from the amazing 5D sensor. I would hope that we will be able to see low noise ISO 1600 and a fully usable ISO 3200. Combine this capability with the wide aperture supertelephotos and wildlife shots in lower light levels should produce much higher image quality results!
phibes wrote:
i dont like the idea of a number cruncher taking care of the noise, i'd rather like to see a sensor that doesnt need to rely on noise-reduction. The 5D sensor is nothing special in terms of noise, it has big pixels, thats all ..
Thats exactly what I meant! As a 1Dxx user I would hope DIGIC III gave the capability of lose noise but as a design feature of the chip rather than via an aggressive AA filter (which 1Ds don't normally have anyway!) Damn, you have got to really be specifically semantic on this thread!
I've just taken a long and good look at the US patent 6,388,709. Conclusion: This is intended for TV or video, despite the claims that it also can be used for still camera.
The patent speaks of refresh rate and of feedback of luminance data. That will only work if you record a continuous stream of images where data analysis of each image determines the adjustments for recording of the next image.
In other words, this will not work for a traditional curtain shuttered dSLR system.
I don't think the only problem of this hybrid size 36x25.5mm sensor is the compatibility with Zeiss and Leca lenses. I think there are many other technical (and cultural) barriers that prevent Canon engineers to built such a camera.
If they make it, this camera would be revolutionnary and if the sensor image quality is high, it would be the ultimate tool for many pros.
And Canon would be very pleased to sell you razorsharp lenses: that's their business.
Tentacle wrote:
And that would put it in D40 territory, resolution wise. I doubt that will happen. To get a 3-layer sensor with the Bayer equivalent of 10 mpixel would have about 5 mpixel per layer, hence the 14 to 16 mpixel number I mentioned on top of this page.
I'm not following all this logic, but
40D - 3872 x 2592
There's a chance it remains at 3504 x 2336
Remember on both cameras that Canon hasn't told it outsiders WHY the files look so much better.
Don't confuse pixels with sensors. Sigma SD9 is a 3.54 mp camera with about 9 million photodetectors. The information contained in each pixel is worth about twice what a good bayer camera delivers.
We are talking about a qualitative improvement per pixel.
Alistair101 wrote:
Thats exactly what I meant! As a 1Dxx user I would hope DIGIC III gave the capability of lose noise but as a design feature of the chip rather than via an aggressive AA filter (which 1Ds don't normally have anyway!) Damn, you have got to really be specifically semantic on this thread!
Does 'how' matter? Looks like the new cameras go up 1EV is sensitivity. But even now tight-lipped Canon seems publicly confident there's a lot more sensitivity to come.
"I find myself in the minority. I do not want more MP. I think we have more than enough. I'd welcome improvements in other areas such as DR, brighter viewfinders, programmable buttons etc."
I'm not following all this logic, but
40D - 3872 x 2592
There's a chance it remains at 3504 x 2336
Remember on both cameras that Canon hasn't told it outsiders WHY the files look so much better.
Don't confuse pixels with sensors. Sigma SD9 is a 3.54 mp camera with about 9 million photodetectors. The information contained in each pixel is worth about twice what a good bayer camera delivers.
We are talking about a qualitative improvement per pixel.
I was wandering along the Path of Speculation on a 40D with a 10 Mpixel TrueColor aka Foveon sensor. That brought me into Sigma SD10 territory, because that has a sensor with exactly those specs, a 3-layered sensor with a total of 10,287,648 pixels. (Just like the older SD9.)
The conclusion of that particular review is that a 6.3 mpixel 10D gives you about the same detail and resolution as a 10 mpixel 3-layer sensor. Either downsample 6 Mpixel to 3.5 Mpixel, or upscale the SD10 output to 3072x2048 and you get comparable results.
So, if 10 Mpixel foveon gets you the equivalent of 6 Mpixel Bayer, then a 40D with foveon-like 3-layer sensor would need to have 16 Mpixels in 3 layers.
