Luminous Landscape Forum
Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: PatrikR on August 24, 2010, 03:07:28 am
-
http://www.canon.com/news/2010/aug24e.html
Canon's new 120MP APS size chip. Pretty exciting... Maybe soon there will be cameras that record video at 4k resolutions or more.
Medium format or APS format... Who cares if it works. 120MP camera weighing at 450 grams would definetely be cool.
-
Cool, diffraction starts at f2.5 :o I hope they produce some fast lens to go with it.
-
Of course you would need a tank to mount it on for sufficent stability..
-
OK Canon, we know you know a thing or two about CMOS design but stop all this obsessing about APS-H size chips.
Use some of that know how and technology and stick it in a MFD back/camera and hopefully we might see some innovation rather than exaggeration.... >:(
-
All of 2,2 micron pixel size...
Will it be noisy?
-
Ok, that's what I call Pixel density. The question of image quality remains. But the chips is an amazing feat. Not only in terms of pixel count but also in terms of speed : 9.5 FPS. The only issue is that frame speed appears to stress image quality too, due to noise. But who nows?
-
Cool, diffraction starts at f2.5 :o I hope they produce some fast lens to go with it.
A question: How important is this? If I understand correctly the diffraction will just reduce the effective resolution, and since it is so high the effect could be easily accepted. What I mean is: If diffraction halves the resolution, it's still a 60mpixels APS sensor
-
All of 2,2 micron pixel size...
Will it be noisy?
If anything, it won't help dynamic range. Sensels of that size will hold (maybe) the charge from some 7250 photons. If (a big IF) the read noise can be limited to 5 electrons, then the maximum dynamic range (engineering definition) would be 10.5 stops. In practice that will probably mean something marginally useful at low ISO. Downsampling will help a bit, but the well depth sets the limit.
With regards to diffraction, it will mean that diffraction can be sampled quite accurately and overall sharpness will benefit (after proper sharpening), but the MP count doesn't necessarily translate directly into magnification potential.
Cheers,
Bart
-
Just a few years ago anything under 9 microns was considered the limit. Anything under 9 microns would not yield acceptable quality. But we all now those challenges are history.
Things change. :)
-
This is great news - since my current camera has 18MP, this new sensor will improve my photography 7-fold!
-
......
Things change. :)
The laws of physics stay. ;)
-
…since my current camera has 18MP, this new sensor will improve my photography 7-fold!
I'm not so sure about that but your memory cards will fill up 7 times faster!
;D
-
Of course you would need a tank to mount it on for sufficent stability..
You'll also need a computer so powerful that you run the risk of it achieving self-awareness and rising up against it's human masters...
-
At that resolution the quality of the lens becomes the limiting factor, like in the film days. Maybe at such sensor resolution Bayer interpolation is not needed anymore?
-
This would speed up shooting the gigapixel panos:)
-
The laws of physics stay. ;)
I have to agree here. I just think that the glass refraction index is too high to allow you to make lenses that could resolve the image enough for pixels that small. Plus, I believe that the best lenses in the world can only create line pairs at 5 microns with a contrast difference of 20% (if I am reading the charts rights). Maybe when we figure out a way to make diamonds big enough for lens, it will be usable. Also, I believe that a sensor that dense would produce a significant amount of heat when processing the images.
Plus, do we really need resolutions this large? Come on, it's practically useless, unless your are in the spy game.
-
If anything, it won't help dynamic range. Sensels of that size will hold (maybe) the charge from some 7250 photons. If (a big IF) the read noise can be limited to 5 electrons, then the maximum dynamic range (engineering definition) would be 10.5 stops. In practice that will probably mean something marginally useful at low ISO. Downsampling will help a bit, but the well depth sets the limit.
Canon already get below 3 electrons readnoise for their 7D, 50D, IDIV and 5DII (at medium ISOs), so beating 5 electrons shouldn't be a problem. The well depth is small enough to permit, at low ISO, the sort of gain that these other cameras employ at medium ISO.
With regards to diffraction, it will mean that diffraction can be sampled quite accurately and overall sharpness will benefit (after proper sharpening), but the MP count doesn't necessarily translate directly into magnification potential.
Cheers,
Bart
You are correct.
But the real problem will be finding lenses which have sufficiently low aberrations to even approach being diffraction-limited, across the field, at such fast f-stops. The way that compact digicams, with similarly tiny pixels, manage it is by being designed for sharpness on-axis, and let the off-axis fall as it may. And that's for a chip where "off-axis" stops 2 - 3 mm from the centre. Just imagine how such a lens would cope when illuminating an APS chip with the same pixel density, and its "off-axis" stopping a relatively whopping 13 - 14 mm from the centre!
L glass won't be enough...L^2 glass will be called for (looks like an L lens; but the weight, number of elements and price are all squared :D )
-
I am not too sure to understand why they bothered doing this.
I'd be the first one to applaude a breakthrough camera from Canon, but I am not an OEM looking for a sensor. Anybody can create a 120 mp sensor that is not checked for image quality and real usefulness as part of a camera.
Cheers,
Bernard
-
it seems to me there might be some usefulness to a lot of pixels with some cleverness -- improved dynamic range per Fujii, pixel binning at high-iso. granted, Canon hasn't shown particular interest in cleverness so far, but we're getting pretty near the manufacturing limits for 35mm lenses with the latest offerings from Canon, Nikon, Zeiss, and Leica. pixel density for pocket cameras have (fortunately) hit the wall. my s90 gives quite good RAW results in good light at base ISO, but the overall IQ is not much better than 2 or 4 year old pocket cameras with lower pixel density.
i'm concerned that SLRs continue a pixel race that results in no practical benefits. as much as i like the features of my 7D, high ISO is no better than the 40D and practical resolution is only incrementally better - both significantlly less than the 5D2.
-
OK Canon, we know you know a thing or two about CMOS design but stop all this obsessing about APS-H size chips.
The use of 29.2 x 20.2 mm format (which is not even in the same 16:9 shape as the actual APS-H film format of 30.2 × 16.7 mm, so I do not understand why Canon persists in that silly naming) is because this sensor is probably a "because we can" technology show piece, rather than a product intended for sale, and so Canon has combined its smallest current CMOS pixels with the largest sensor size it can make without the hassles of stitching, to get the biggest pixel count in the headlines. Canon did a 50MP CMOS sensor of the same size a few years ago, never used in any actual product.
I continue to doubt that Canon cares to invest in MF sensors for the sake of increasing its total sensor sales by the 0.1% or so that MF sensors would contribute. Better to leave MF with inferior sensor technology and so help its 35mm format camera and lens sales.
-
it seems to me there might be some usefulness to a lot of pixels with some cleverness -- improved dynamic range per Fujii, pixel binning at high-iso.
Perhaps binning could be used for something other than just noise. Perhaps a sensor that doesn't need de-moisaicing? What if the idea is to "bin" 4 pixels with appropriate filters as a single pixel and derive the exact color of the final pixel only from the color of those 4 pixels. So while the sensor has the traditional "bayer" filter arrangement, you have 4 pixels acting as a single pixel. End result you have a 30mp sensor with no moire, no anti-aliasing filter necessary and no de-moisacing required.
I guess the only reason I say this is it something I've wondered about for some time. Sort of like adding cores to a single CPU. Create a sensor where multiple pixels work as single pixels, not only for noise but for this issue as well.
I guess the best way to
-
Unfortunately you'll still get moire and aliasing doing that, probably more so than using a standard demosaic approach.
Graeme
-
120MP is basically P&S sized pixels (2.2µ) in a DSLR sized sensor. Please note that noise is not a fixed quantity, but varies with scale in the image. There are two components to noise in most images -- read noise (noise contributed by the camera electronics), and photon shot noise (quantum fluctuations in the light signal itself). At low to moderate ISO, photon noise tends to be the more visible noise in an image. At a fixed image scale, this noise depends only on how much light is collected, and is therefore independent of pixel size. High ISO might be compromised somewhat; with read noise unchanged from current performance of production DSLR's, read noise at a fixed scale will be about twice that of 7D (since the pixels are about half the linear dimension). Low ISO might not be so compromised for read noise; for instance the 40D has about 10-15% less read noise than the 1D3 at base ISO, so smaller pixels might have slight advantage unless Canon decides that cleaner low ISO shadows are more of a priority (they've been going backwards in this department recently). They could also mitigate the high ISO read noise issue by cutting down on the pattern noise that plagues their cameras at all ISO (though it's worse at low ISO).
And since diffraction is often discussed in the context of pixel size, note that diffraction is a property of the optics, not the sensor; having smaller pixels does not increase diffraction -- on the contrary, all it does is decrease the range of f-stops over which the sensor resolution is the limiting factor in system resolution, rather than the rest of the optics. A side benefit is that the AA filter's blur radius, being tied to the pixel size, will be smaller; and demosaic artifacts will be pushed off to finer image scales where they will be less noticeable.
What Canon really needs to do if they are serious about heading in this direction is to concentrate some development resources on compression technology -- clearly one doesn't want to have to deal with the huge files that result from high MP count cameras, since most of that information is redundant. sRAW is the lamest possible image compression method one can imagine. Much better would be the sort of compression RED uses -- preserves nearly full resolution while reducing file sizes substantially, and allowing continuous shooting without maxing out the image buffer. I did a very klugy exercise some time ago, taking a raw file, separating out the RGGB planes and doing JPEG2000 (wavelet based) compression by about a factor of 10, then uncompressing, reassembling the raw, and demosaic. The result was not bad for such heavy compression; I'm given to understand that RED does something along these lines, but of course they've spent more than an afternoon on it and it works much better. ;)
One could also wring a little more savings out of recognizing that 14-bit is wasteful at current output DR levels; 12 is sufficient for all Canon DSLR's ever made. Level-thinning methods such as Nikon uses with its lossy compression are another way to shave off a bit or two more per pixel. But to me, the issue is that one doesn't want to keep all the image info at high res (eg, not in skies and other smooth regions), so the question becomes what data to keep and what to discard. Low res cameras make the decision for you, by omitting all fine scale image information; but there are certainly more intelligent methods one can imagine (and some that exist already), where compression technology keeps high res information where it is desired, and discards it where it is not.
BTW, as I understand it Canon is not "obsessed" with APS-H; rather, it is the largest sensor size that they can make without stitching on their fab line.
-
If you read till the end you'll find this: In 2007, the company successfully developed an APS-H-size sensor with approximately 50 million pixels.
We are in the middle of 2010; where is my 50 megapixels Canon?
-
It's nice to see that the LL forum reputation as the place for the measurebators is safe and sound. ;D
It's surprising (well maybe not ::) ) to see so much serious discussion about a device that will never make it to market or, if it does, not likely in any of our productive lifetimes.
As someone else noted, this is a 'because we can' exercise. It's a headline grabber.
No doubt Canon will do testing with this sensor. That's what this kind of exercise is. It's a test bed. It's like Formula 1 as a test bed for technology that eventually migrates down to our street cars (not as much anymore with the restrictions on costs in F1, sadly).
-
120MP is basically P&S sized pixels (2.2µ) in a DSLR sized sensor. Please note that noise is not a fixed quantity, but varies with scale in the image. There are two components to noise in most images -- read noise (noise contributed by the camera electronics), and photon shot noise (quantum fluctuations in the light signal itself). At low to moderate ISO, photon noise tends to be the more visible noise in an image. At a fixed image scale, this noise depends only on how much light is collected, and is therefore independent of pixel size. High ISO might be compromised somewhat; with read noise unchanged from current performance of production DSLR's, read noise at a fixed scale will be about twice that of 7D (since the pixels are about half the linear dimension).
Emil's analysis assumes that photon collection is dependent on total sensor area and is not affected by pixel size. However, CMOS sensors have transistors in each pixel and a fill factor of well less than 100%. If one makes pixel size smaller but does not shrink the transistor area proportionally by using a smaller process for the electronics, fill factor will decrease. This can be compensated for by the use of microlenses, but full well will likely suffer because of increased charge density. As Catrysee and Wandell explain (CMOS Roadmap (http://www.imageval.com/public/Papers/EI%205678-01%20Peter%20Catrysse.pdf/)), small pixels require a larger f# since the stack height of the sensor does not scale with pixel size, greatly complicating micro lens design.
Furthermore, read noise occurs with each pixel readout, and doubling the lateral size of the sensor while holding pixel size constant would quadruple the pixel count and the total read noise would also quadruple; a large pixel can be read with about the same level of read noise as a small pixel (see Photometrics (http://www.photometrics.com/resources/learningzone/binning.php/)), so holding the pixel count constant while increasing sensor size decreases total read noise. Downsizing in software does not reduce read noise by the same amount as binning in hardware.
-
Pixel noise might be handled by deep trench. remember the number of electronic is counting in volumn (3D) not area (2D).
Theoretically, the diffraction problem could be handled by correlation across the pixels. Who said a good image must be clean per pixel?
-
it is soon time for me to sell all my Canon gear