Canon develops world's largest CMOS sensor

http://www.dpzen.com/news/1008/10083101canonlargestsensor.asp Wow. Talk about large format, this sucker is is almost 80x80 inches! View: original size (Canon's ultra-large-scale CMOS sensor (left) alongside a 35 mm full-frame CMOS sensor ) What this thing for anyway? - using huge telescopes to make movies or something?
"Its advanced circuitry allows video recording at 60 frames per second with 0.3 lux illumination that according to the company is roughly one-half the brightness of a moonlit night."
  • C

Still pretty big! Ansel Adams has one on preorder. KP -- Use the new site features to hide my homely face and banjo!! View: original size http://www.ahomls.com/photo.htm http://www.phillipsphotographer.com 'The urge to save humanity is always a false front for the urge to rule it.', H. L. Mencken

Ken Phillips wrote: Still pretty big! Ansel Adams has one on preorder.
Yeah I misread cm for mm and didn't put my brain in gear :X Still its huge and high iso. Canon's reply to the D3s ?
  • C

CFynn wrote: Wow. Talk about large format, this sucker is is almost 80x80 inches!
Sorry. That should have been 8x8 inches. I read cm instead of mm
  • C

What?! It's not even 8x10 FF?? ?) 4x5 FF should be good enough for the LF shooter. This may be some military application sensor!
CFynn wrote:
CFynn wrote: Wow. Talk about large format, this sucker is is almost 80x80 inches!
Sorry. That should have been 8x8 inches. I read cm instead of mm
  • C

Well, guess we'll have to redefine 'Full-Frame' now

Steen Bay wrote: Well, guess we'll have to redefine 'Full-Frame' now
Yes. if there were a digital back made using this, you'd need an 8x10 view camera to put it on. What is that it is mounted in anyway? View: original size
  • C

... Canon has certain large lenses (that make the1200/5.6 look tiny! ) that this would fit nicely behind! KP -- Use the new site features to hide my homely face and banjo!! View: original size http://www.ahomls.com/photo.htm http://www.phillipsphotographer.com 'The urge to save humanity is always a false front for the urge to rule it.', H. L. Mencken

Ken Phillips wrote: ... Canon has certain large lenses (that make the1200/5.6 look tiny! ) that this would fit nicely behind!
But they probably aren't made for a 288mm image circle.

It doesn't matter how large Canon's lenses are, if they designed to cover only 36x24mm at infinity (43mm diagonal), they will not cover 8x10. And 8x10 lenses aren't that huge: http://www.largeformatphotography.info/lenses/LF8x10in.html
Ken Phillips wrote: ... Canon has certain large lenses (that make the1200/5.6 look tiny! ) that this would fit nicely behind! KP -- Use the new site features to hide my homely face and banjo!! View: original size http://www.ahomls.com/photo.htm http://www.phillipsphotographer.com 'The urge to save humanity is always a false front for the urge to rule it.', H. L. Mencken

I'd love to fit that to the back of an old large format camera View: original size

Andy Crowe wrote: I'd love to fit that to the back of an old large format camera View: original size
In 10 or 15 years it might turn up on e-bay
  • C

But it's a shame the old 8x10 lenses are not wonderful on a super sensor like this one!
Andy Crowe wrote: I'd love to fit that to the back of an old large format camera View: original size

Assuming the same pixel pitch as on Canon's 120 MP APS-H sensor, then this one should be about 8424 MP. Guess that should leave a little room for cropping.

Steen Bay wrote: Assuming the same pixel pitch as on Canon's 120 MP APS-H sensor, then this one should be about 8424 MP. Guess that should leave a little room for cropping.
What? I was thinking to stitch a few images to get a decent quality instead

Too bad it isn't exactly square. I hate cropping. -- Doug Dolde http://douglasdolde.com

The crop factor is about 0.15x, meaning that a 300mm f/2.8 (35mm, FoV/exposure) equivalent lens should be about 1994mm with a 712mm aperture diameter. Hopefully it'll have IS.

Wouldn't that be less magnification rather than more? My APS-C sensor has a crop factor of 1.6, which makes a 300 mm lens an effective 480. If the crop factor is less than 1, I think that decreases the effective lens length. So .15x would result in 45 mm.

yes, but you would need a 1900 mm lens with a 700mm front lens element to have the same field of view as a 300 mm f/2.8 lens on a canon 5dmk2/nikon d700/sony 900

You do need longer lenses as you move larger on the format, however, unless you are shooting for space images you won't see a 300 f2.8 eq. easily. DOF eq aside, 8x10 camera lenses are relatively "small", and you shoot f22 and on under normal conditions.
Henrik Melander wrote: yes, but you would need a 1900 mm lens with a 700mm front lens element to have the same field of view as a 300 mm f/2.8 lens on a canon 5dmk2/nikon d700/sony 900

So - we'll see this trickle down to larger format cameras humans can buy. If I can still hold a camera in 10 yrs. when I'm 70 I can now look forward so something affordable and used with a lot bigger and better sensor that I have now. I remember the wait from the 1st digital cameras until something usable came along. For me it was the 20D, and I think it was about 8 yrs. from the time I saw digital photography on the consumer horizon until I had the 20D, and a few more years until I got satisfied with the 5D II. I wonder if it's too early to expect any medium format announcement at PhotoKEENa ?

maybe they are developing something for the space or military programs -- beam me up scotty do you consider yourself lucky?

Sence 12mp is very popular in low light photography, lets say this baby is 3500x3500 pixels. 57 uM pixelsize, that equals a pixel area of 3250. The D700 has a pixelsize of 8.5 um and a pixelarea of 80. Thats 40 times the light gathered per pixel. Clean iso 52000 pictures? Tho, I would only imagine one of these beeing used in a telescope of similar

[*] What are the pixel dimensions? [*] How long until I get get one of these for my 8x10 Linhof?

It's the obvious question. Does anyone know how many megapixels this thing has? Maybe it's in the Japanese press release or something?

davidgro wrote: It's the obvious question. Does anyone know how many megapixels this thing has? Maybe it's in the Japanese press release or something?
Since its for video, it does not need many pixels. Potential applications for the new high-sensitivity CMOS sensor include the video recording of stars in the night sky and nocturnal animal behavior.

It's possible that it's only video resolution, but even in that case there is a huge difference between say 640x640 and 1920x1920 - specifically a factor of 9 in that case of course. (The wording on the press release is terribly ambiguous about whether it's any good for stills but also "video-compatible" or if it really is a huge camcorder sensor)

Made for extreme light situations.

One giant pixel. That's how it can gather data so quickly.

"The sensor makes possible the image capture in one one-hundredth the amount of light required by a 35 mm full-frame CMOS sensor, facilitating the shooting of 60 frame-per-second video with a mere 0.3 lux of illumination. 3" Yes, if you use a lens with an aperture 10 times larger. So, 100 is a pretty irrelevant number here. But imagine the dynamic range on that thing.

Cax wrote: "The sensor makes possible the image capture in one one-hundredth the amount of light required by a 35 mm full-frame CMOS sensor, facilitating the shooting of 60 frame-per-second video with a mere 0.3 lux of illumination. 3" Yes, if you use a lens with an aperture 10 times larger. So, 100 is a pretty irrelevant number here.
How about an 8.2 meter aperture. As soon as I read about this thing, I thought of Mauna Kae. The National Astronomical Observatory of Japan built a really big scope there, the "Subaru". It has an 8.2m active main mirror. It has an enormous prime focus, I bet considerably larger than this sensor, and is one of the world's largest widefield instruments.
But imagine the dynamic range on that thing.
Yeah. Cooled, with exposure averaging...

Joseph S Wisniewski wrote: As soon as I read about this thing, I thought of Mauna Kae. The National Astronomical Observatory of Japan built a really big scope there, the "Subaru". It has an 8.2m active main mirror. It has an enormous prime focus, I bet considerably larger than this sensor, and is one of the world's largest widefield instruments.
But imagine the dynamic range on that thing.
Yeah. Cooled, with exposure averaging...
Named 'Subaru' after the motor needed to drive the dust buster?

Joseph S Wisniewski wrote: As soon as I read about this thing, I thought of Mauna Kae. The National Astronomical Observatory of Japan built a really big scope there, the "Subaru". It has an 8.2m active main mirror. It has an enormous prime focus, I bet considerably larger than this sensor, and is one of the world's largest widefield instruments.
I think you meant Mauna Kea.

You explanation makes a lot sense in conjunction with the fast readout of this giant sensor. The sensor will probably be used for "lucky imaging". This technique reduces the negative effects of the atmosphere on the image. The astronomers take very many images, each with a very short exposure (a video). They keep only the images with the least atmospheric distortion, and they shift them, because the atmosphere does not only blur the image but also slightly moves them. Then the selected images are added to average out the noise, and make faint objects visible. With this technique ground based telescopes can get better resolution than the Hubble telescope. But there always have to be some well known bright stars in the image, as references to decide which images are good, and for lining the images up. Therefore this technique is mainly useful for wide field telescopes. Unknown dark parts of the sky, like the Hubble Deep Field, can not be photographed with this technique. http://en.wikipedia.org/wiki/Lucky_imaging

Eike Welk wrote: With this technique ground based telescopes ...
Adaptive optics can beat Hubble: http://en.wikipedia.org/wiki/Adaptive_optics

By son recently completed a master's project doing the image processing for lucky imaging, and gave me some interesting info on it. There's a very informative site on it here: http://www.ast.cam.ac.uk/~optics/Lucky_Web_Site/index.htm this group is using sensors much more esoteric than this Canon: http://www.ast.cam.ac.uk/~optics/Lucky_Web_Site/guide_to_l3ccds.htm these provide effectively zero read noise and the ability to count individual photons! -- Bob

bobn2 wrote: By son recently completed a master's project doing the image processing for lucky imaging, and gave me some interesting info on it. There's a very informative site on it here: http://www.ast.cam.ac.uk/~optics/Lucky_Web_Site/index.htm this group is using sensors much more esoteric than this Canon: http://www.ast.cam.ac.uk/~optics/Lucky_Web_Site/guide_to_l3ccds.htm these provide effectively zero read noise and the ability to count individual photons! -- Bob
Cool stuff and informative websites! I'll definitely read more on it. e2v is pretty innovative. Have you heard of Andor (they're nearby)?, they make some pretty nice cameras with those and other sensors. Chris

cpw wrote:
bobn2 wrote: By son recently completed a master's project doing the image processing for lucky imaging, and gave me some interesting info on it. There's a very informative site on it here: http://www.ast.cam.ac.uk/~optics/Lucky_Web_Site/index.htm this group is using sensors much more esoteric than this Canon: http://www.ast.cam.ac.uk/~optics/Lucky_Web_Site/guide_to_l3ccds.htm these provide effectively zero read noise and the ability to count individual photons! -- Bob
Cool stuff and informative websites! I'll definitely read more on it. e2v is pretty innovative. Have you heard of Andor (they're nearby)?, they make some pretty nice cameras with those and other sensors. Chris
seen their stuff at shows and the like, never had need to use anything that sophisticated. For the kind of stuff I'm involved in the camphone sensor is an incredibly useful tool and dirt cheap, too (amazing what types of sensor you can make with a camphone module looking at some other device and some image processing). -- Bob

It has a diagonal 6.7x larger than 35mm or 36x24 full-frame. A 50/1.2 lens on 35mm would be equivalent to a 335/8.04 on this camera in terms of light-gathering capability, depth of field, and angle of view (when you consider the size of the sensor it is designed to cover). There are plenty of lenses that fit that bill. (There are several 360/6.5 or 360/6.8 lenses). These would get about 1.33 times as much light (total quantity) as would a 50/1.2 on 35mm. While you couldn't expose it the same -- you would need very high ISO film -- the quality would come out as good as 50/1.2 on 35mm, actually better. There is nothing in existence that would actually get proportionately more light as is the ratio of the areas of 8x8in vs. 36x24mm (46x as much area). You would need a 335/1.2 lens capable of covering 8x8 to get an equivalent to 50/1.2 in terms of angle of view and exposure. This would result in much much better image quality, presuming the lens was good, of course. Or you could increase the ISO enormously and get equivalent quality but a shorter shutter time. Mind you, the lens then would be enormous (335/1.2 = 279mm = 11in diameter). You'd need a camera body designed for 11x14 inch, just due to the lens alone. A more practical use of this will be for us large-format photographers: really high resolution photography with tilts and shifts, at longer exposure times. Theoretically, assuming arbitrarily small ISOs, arbitrarily small photosite sizes, and optics limited only by diffraction, you can get the same resolution out of 35mm as from 4x5 or 8x10. While Cannon has produced photosites small enough to maybe equal the resolution of a 4x5 (not really, theoretically...assuming film gets 100 lp/mm, that's 20320 x 25400 "pixels" or 516 megapixels...Canon only has a 120 MP APS-C they developed), we don't have arbitrarily small ISOs. And more importantly, if you look at actual data, 4x5 optics are a lot closer to the diffraction limit at f/32 (e.g., > 80% of the diffraction limit) than are full-frame lenses at f/8 or f/9 (the full-frame equivalent f-stop in terms of DOF and angle of view as f/32 on large format). The sharpest lenses I've seen are my Olympus around 120 lp/mm max at f/4 or f/5.6 (equivalent to f/32 on 4x5); that's only 32-45% of the diffraction limit. At f/8, the Oly 50/2 is getting 105 lp/mm, or 56% of the diffraction limit.

CFynn wrote: http://www.dpzen.com/news/1008/10083101canonlargestsensor.asp Wow. Talk about large format, this sucker is is almost 80x80 inches! View: original size (Canon's ultra-large-scale CMOS sensor (left) alongside a 35 mm full-frame CMOS sensor ) What this thing for anyway? - using huge telescopes to make movies or something?
I'm guessing part "proof of concept" for their stitching process, and part a useful sensor for a big telescope. As soon as I read about this thing, I thought of Mauna Kae. The National Astronomical Observatory of Japan built a really big scope there, the "Subaru". It has an 8.2m active main mirror. It has an enormous prime focus, I bet considerably larger than this sensor, and is one of the world's largest widefield instruments.

A technology demonstrator for the future LSST was the first use that entered my mind- based partly on Canon's emphasis on data rate. Your guess may be better though: putting it on the Subaru would be a good way to show its potential before the LSST is constructed. http://www.lsst.org/lsst

Tne LSST is planned to use an array of CCDs: http://www.lsst.org/files/docs/aas/2010/215-RC-1259-AAS_Rasmussen.pdf Still I think this public announcement pretty much rules out military applicacion and therefore public space observation is the prime candidate (Joseph might be onto something here). -- http://www.flickr.com/photos/guillaumeserandour/

Travel_G wrote: Tne LSST is planned to use an array of CCDs:
fwiw this is the 3.2 Gpix array that the LSST will be using, 64cm/25" diameter, 30TB of data per night. View: original size

CFynn wrote: http://www.dpzen.com/news/1008/10083101canonlargestsensor.asp What this thing for anyway? - using huge telescopes to make movies or something?
Yes, movies of variable astronomical subjects (like pulsars?) ... and maybe wildlife in very low light through terrestrial telescopes: the article indicates that, with the words "Potential applications for the new high-sensitivity CMOS sensor include the video recording of stars in the night sky and nocturnal animal behavior." and "Its advanced circuitry allows video recording at 60 frames per second with 0.3 lux illumination that according to the company is roughly one-half the brightness of a moonlit night." The combination of the large image circle needed of low f-stop needed for low light video pretty much rules out everything except substantial telescopes as the lenses; I doubt that large format (view camera) lenses would be much use.

That high megapixel sensor didn't impress me at all. This thing does. So a digital 4x5 view camera would be real cool. -- “I’m not in this world to live up to your expectations and you’re not in this world to live up to mine.” “Showing off is the fool’s idea of glory.” Chris, Broussard, LA

What memory card will it use? The data transfer rate must be awesome to get all these pixels copied and saved... but it looks to me as a low pixel count, where the emphasis is on 'low light capability' vs high resolution. Still an awesome achievement -- I can see Phase One drooling on this one...
bikinchris wrote: That high megapixel sensor didn't impress me at all. This thing does. So a digital 4x5 view camera would be real cool. -- “I’m not in this world to live up to your expectations and you’re not in this world to live up to mine.” “Showing off is the fool’s idea of glory.” Chris, Broussard, LA

jboyer wrote: What memory card will it use?
Fusion-IO ioDrive Duo

jboyer wrote: What memory card will it use?
Probably a hard drive, since it will mostly be used behind a large telescope. But anyway, the pixel count is not that high, maybe 10MP or less. I say that because Canon claims 100 times the sensitivity of it 22MP 35mm format sensors, suggesting 100 times the photosite area, while having "only" 47 times the sensor area, which would mean 47/100 the pixel count of those 35mm format sensors: about 10MP. And after all, it is described as a "video sensor"!

BJL wrote: But anyway, the pixel count is not that high, maybe 10MP or less. I say that because Canon claims 100 times the sensitivity of it 22MP 35mm format sensors, suggesting 100 times the photosite area, while having "only" 47 times the sensor area, which would mean 47/100 the pixel count of those 35mm format sensors: about 10MP.
If it is intended for use on a telescope, it might not have a Bayer-array color filter, since wavelength selection on telescopes is normally done with external filters. So if they have a natively black-and-white sensor, that may give them a factor of 2x or more sensitivity (although obviously that's an apples-and-oranges comparison...) Also, if there is no IR-cut filter (also common on telescopes and animal behavior video cameras) that gives you even more sensitivity, depending on the color temperature of your illumination.

John Beale wrote: ... if they have a natively black-and-white sensor, that may give them a factor of 2x or more sensitivity
Good point; you are probably right about having no filters, either CFA or IR: any filters would almost certanly be interchangeable ones in front, to allow a choice of what wavelengths to look at. So I will up my estimate to about 20MP. By the way, the pixel size is then about 45nm, which might allow fabrication with larger format, lower resolution LCD steppers instead of the usual IC steppers, which could reduce costs substantially.

Add new comment

Image
More information
  • Files must be less than 2 MB.
  • Allowed file types: png gif jpg jpeg.
Attachment
More information
  • Files must be less than 2 MB.
  • Allowed file types: zip rar.
By submitting this form, you accept the Mollom privacy policy.