KIDERAL wrote:
I think it is safe to say that canon has some spare pixel capacity in its quiver and is holding out on us...or as I have always thought the bottleneck is that they can't move the data fast enough...
That G7 sensor has about 38 mm² for about 40 million photo sites. (Bayer pattern, one red subpixel, one blue and two green.) Take the root of the area per subpixel to get the subpixel pitch. We get a spacing of 0.97 µm (microns).
Now, if you want to know what CAN be done, consider this: CCD and CMOS sensors are often compared to memory chips, because both consist of very large grids with millions of 'cells'. Memory manufacturers have arrived at the 50 and 45 nanometre node with their process technology. Let that sink in... A spacing of 970 nm and the latest litho processing is at 45 nm. Err on the safe side, assume that a subpixel pitch of 90 nm is possible. That's 0.09 µm. So the bayer pattern pitch can be reduced 10 fold, which increases pixel density a 100 fold!
But I don't want to know what kind of p*ss poor ISO performance I'd get out of subpixels that are 1/100th the size of the current smallest subpixels
Philippe Arnez wrote:
Ehm, I don't want to dissapoint you, but I don't think you can actually buy stuff at the photokina... If I got you right, that's your plan. But it's a tradeshow for the big guys of the business...
The companies don't sell any gear, it's just for show
Nah mate you got me wrong. im waiting till after photokina before i buy a new body. otherwise knowing my luck, ill buy a 1dmk2 then find out that canon are updating it only to see the price of the now outdated 1dmk2 drop. id rather save my pennies for a few months before i spend it all. in the mean time my friend will let me use his D2X (he might even tempt me to the dark side!!!!)
Tentacle wrote:
But I don't want to know what kind of p*ss poor ISO performance I'd get out of subpixels that are 1/100th the size of the current smallest subpixels
Hmmmm, I wonder... It might be a wonderful experiment in noise. Sort of like a pointillist painting with different colored noise gloubules placed in the "approximate position" of your subject. At best....
Taff Reardon wrote:
Nah mate you got me wrong. im waiting till after photokina before i buy a new body. otherwise knowing my luck, ill buy a 1dmk2 then find out that canon are updating it only to see the price of the now outdated 1dmk2 drop. id rather save my pennies for a few months before i spend it all.
Ok then...
in the mean time my friend will let me use his D2X (he might even tempt me to the dark side!!!!)
Rgds Taff
Uh-oh.... watch out. And don't stand to close to the black hole....
Tentacle wrote:
That G7 sensor has about 38 mm² for about 40 million photo sites. (Bayer pattern, one red subpixel, one blue and two green.) Take the root of the area per subpixel to get the subpixel pitch. We get a spacing of 0.97 µm (microns).
gfiksel wrote:
No, it's 10 Mpixels total, not 40 Mp
Please read more carefully, I'm talking about photo sites. One pixel consists of 4 subpixels, also called photo sites. Those groups make up a socalled bayer pattern. Except for the Foveon X3 sensor that is. See http://en.wikipedia.org/wiki/Bayer_filter
Huh? The previous Canon sensors have used only one color per pixel. No subpixels, only a bayer pattern. 40M photosites would be 40 megapixels )and megabucks as well).
Tentacle wrote:
Please read more carefully, I'm talking about photo sites. One pixel consists of 4 subpixels, also called photo sites. Those groups make up a socalled bayer pattern. Except for the Foveon X3 sensor that is. See http://en.wikipedia.org/wiki/Bayer_filter
There is no subpixels. One photosite is, for this purpose, equivalent to one pixel. Each pixel corresponds to R, G, or B according to Bayers filter. That is for 10PM and R/2G/B filter there are 2.5MP (red), 2.5Mp(Blue), and 5Mp(Green) pixels
Tentacle wrote:
Please read more carefully, I'm talking about photo sites. One pixel consists of 4 subpixels, also called photo sites. Those groups make up a socalled bayer pattern. Except for the Foveon X3 sensor that is. See http://en.wikipedia.org/wiki/Bayer_filter
That wiki contradicts everything you posted in the second paragraph:
"The raw output of Bayer-filter cameras is referred to as a Bayer pattern image. Since each pixel is filtered to record only one of three colors, two-thirds of the color data is missing from each. To obtain a full-color image, various demosaicing algorithms can be used to interpolate a set of complete red, green, and blue values for each point.
gfiksel wrote:
There is no subpixels. One photosite is, for this purpose, equivalent to one pixel. Each pixel corresponds to R, G, or B according to Bayers filter. That is for 10PM and R/2G/B filter there are 2.5MP (red), 2.5Mp(Blue), and 5Mp(Green) pixels
Ok, my bad, I'll condede. As a mitigating circumstance, bayer pattern to pixel conversion can be done in various ways and we have no info on whatever proprietary algorithms are used.
The rest remains. In fact, fhe factor 4 that goes missing makes it more dramatic: The chip is 38.1976 mm², divide by (3648 x 2736) pixels, take the root, and it's 1.96 µm, or 1956 nm. The smallest litho process is at 45 nm. A factor 20 decrease in pitch, give or take some, and a factor 400 increase in pixel density. Give or take a lot more.
Well, two things come to mind when comparing the Powershots to EOS.
We can look forward to significant improvements in resolution and we are no where near the resolution limitation on glass.
(If a powershot lens can resolve 10mp with a CCD which if it were the size of 35mm sensor would be in excess of 100mp then a 1 series lens will easliy be able to resolve a full frame sensor when we get to 100 mp images.)
Except that the glass in the powershot may or may not exceed the recommended daily amount of fuzzy images. Translation: don't look at 50% or higher crops of the powershot series.
timbop wrote:
Except that the glass in the powershot may or may not exceed the recommended daily amount of fuzzy images. Translation: don't look at 50% or higher crops of the powershot series.
Yeah, a huge difference...sure. Sorry, but all the people that think the glass is the limiting factor on IQ are just dead wrong. There are at least a dozen other things that limit IQ before you run into glass limitations with current tech.
The point I'm making here is that if you can get 95% of the IQ from a SD700 as you can from a 30D (given the MUCH smaller sensor, the more aggressive noise reduction, the more compromised lens design, etc) then the 1D's are very likely nowhere near the actual limitation of L glass.
Edited by Jeff on Sep 16, 2006 at 12:37 PM GMT (Reason: un-embedded link to copyrighted material)
Edited by jwil on Sep 16, 2006 at 02:42 PM GMT (Reason: deleted 'Edited by' posted by moderator)
Edited by Jeff on Sep 19, 2006 at 08:18 AM GMT (Reason: Restored edits to reflect reality)
timbop wrote:
Except that the glass in the powershot may or may not exceed the recommended daily amount of fuzzy images. Translation: don't look at 50% or higher crops of the powershot series.
That's about exactly right, 'timbop'.
My S70's files look far worse at 50% than my 1Ds does at 100%. Does that mean the effective resolution (compared to a dSLR) would be reduced from 7.0 actual to <3.5 'effective'? I'd think that's an apt comparison.
My S70's files look far worse at 50% than my 1Ds does at 100%. Does that mean the effective resolution (compared to a dSLR) would be reduced from 7.0 actual to <3.5 'effective'? I'd think that's an apt comparison.
Take a look at what I linked I think if you make a fair comparison of roughly apples to apples (even with a 2MPx advantage to the DSLR) the difference isn't nearly as pronounced, even at 100%.
Again, my point is that there are a dozen other things you have to fix before glass becomes a problem. But then what do I know.
Isn't this much more of a parallel problem where glass quality plays its role alongside things such as resolving power of the sensor, AA filter, noise reduction, anti-reflection coatings and all the rest?
