The Rumor Mill is firing up again

[rexbobcat]

Canon 75MP DSLR to Use a Non-Bayer Multilayer Sensor?

Behold! Speculation and dubious source comments!

I don't fully understand how the Foveon sensor works, but either way it's still pretty cool...if it's true.

 

[Overread]

Foveon Sensor is what Sigma has been playing with for ages with their own DSLRs, though Sigma has never really pushed for it to go mainstream.

That said it wouldn't be a 75MP; instead its 3 layers of images overlain, one for each colour channel and because they are lain over the top of each other the resolution isn't the same as a single flat 75mp sensor. It would be interesting to see if Canon does go this way, however these days just because a company has a patent doesn't mean that they'll ever use it.

 

[Big Mike]

Yes, the Foveon sensor uses separate 'layers' or 'sensors'....so when they quote the megapixel number, it's much higher than we're used to hearing from cameras that use CCD or CMOS sensors. So it's not really comparing apple to apples.

Sure makes for eye catching headlines though.

 

[Derrel]

It very well could be that Canon has some mules out in the field. Here's a Luminous Landscape review of the Sigma SD1, a Foveon-sensor dslr. It shot pretty well. Lovely color, lots of detail. The actual, real-world resolution of the SD-1 was comparable to that of the Sony A900, which was a 24-megapixel, full-frame sensor model, so the three-color Foveon style sensor does seem to "punch above its weight class". A 3x 25-MP or "seventy-five megapixel" Foveon-style sensor would probably produce exquisite files with good lenses.

Sigma SD1 Review

 

[Overread]

It would be neat to see new sensor tech coming out, I know Nikon has been pushing their own cameras with increased sharpness because they removed the AA filter and let software deal with fixing moire so to see Canon go for 3 layered sensors is an interesting move (and I'm sure the marketing people would love to utterly beat everyone with HUGE MP numbers )

 

[rexbobcat]

I don't really understand the advantages. I mean, yes, Foveon sensors are able to get very accurate and rich color depths, but in terms of DR and noise, what's the big hub-bub?

I also don't get how they say it's 75 MP when the resolution of the final image is much less. The signal from the three layers might add up to 75 MP when measured seperately, but if that does not translate to real-world resolution then how is it logical to say it's at such a high resolution?

 

[Gavjenks]

I don't really understand the advantages. I mean, yes, Foveon sensors are able to get very accurate and rich color depths, but in terms of DR and noise, what's the big hub-bub?

I also don't get how they say it's 75 MP when the resolution of the final image is much less. The signal from the three layers might add up to 75 MP when measured seperately, but if that does not translate to real-world resolution then how is it logical to say it's at such a high resolution?

No no, resolution will be way higher.

Every pixel in a standard sensor is extrapolated over 3-4 actual pixels (which are each only one color). Other pixels also share this information for their own extrapolation. Here, though, they won't share. They will actually collect all their own information. That leads to much higher actual resolution.

The reason you are confused is because current megapixels numbers are very misleading. An 18 megapixel Bayer sensor has the resolution ability of what is actually more like a 6 megapixel foveon sensor. Yes, it ends up with 25 million pixels, but they are all redundantly sharing information with one another and thus cannot actually resolve lines with as few pixels as you would expect.

This new sensor would actually be a 25 megapixel sensor, but a REAL 25 megapixel sensor, not the misleading "25 supposedly but actually acts like 8.3" that you currently get when you buy a "24 megapixel" sensor.



Long story short: Don't worry about the number confusingness. Resolution will indeed be that of an effectively 3-4x higher megapixel count. Or a linear resolution improvement (like lines per mm in a single direction) of 2x

The 75 number is just what it would be if you perpetuated the dishonest / lying sort of marketing terminology that the companies all currently use.

http://www.ephotozine.com/articles/...w-20388/images/sigma-foveon-sensor-larger.jpg


Edit: it might be possible to get the same resolution in monochrome with current sensors, with fancy software, since in Black and White, you essentially have a useful data site at every position (I'm talking about shooting in color than converting to B&W with advanced custom software). But nothing like this happens normally or by default with mdoern cameras.

 

[amolitor]

The number I've seen thrown around for a Bayer sensor is 2:1. So, an 18 megapixel Bayer array "usually" produces an effective resolution of 9 megapixels of RGB image. It depends on the image. The 75 megapixel Foveon will always, always, every single time, produce a 25 megapixel RGB image.

 

[Derrel]

The Sigma DP-1 was a 15.4 Megapixel sensor crop-sensor dslr, which Sigma referenced as "46 megapixel" in terms of equivalence.

The newer Sigma compact style camera, their DP-2 Merrill camera uses that same sensor with a fixed, f/2.8 Sigma lens, and the camera has what Michael Recihman of The Luminous Landscape refers to as "medium-format digital image quality".From a pocketable compact camera. And that comparison is coming from a guy who has owned and shot most of the best medium format digital cameras made over the last decade...so...

Take a look at how AMAZING the sheer image clarity, richness, and quality of the Foveon 15.4 megapixel, APS-C sized sensor is compared to other high-end Leica, Sony, and Fuji cameras--all with Bayer-style sensors of higher MP count.

Sigma DP2M Review

Just take a look. You'll understand what the fuss is all about. The 15.4 MP APS-C Foveon sensor is amazingly good. The pictures look clearer, sharper, more-realistic, and more three-dimensional.

 

[Scatterbrained]

Personally I hope this is true. I'd be right in line to pre-order it.

 

[amolitor]

I do not, by the way, think you can get substantially more resolution out of a Bayer Array if you elect to go b&w, no matter how clever the software.

Except, there is one special case. When re-photographing b&w content. If you happen to know the color characteristics of your light source (ideally, you have equal amounts of what your camera sees as R, G, and B) then you can custom-convert and get the full sensor resolution out. This is theoretically pretty handy for re-photographing b&w negatives as a cheap route to digitization.

 

[gsgary]

Will it take 35mm or 120

 

[Gavjenks]

I do not, by the way, think you can get substantially more resolution out of a Bayer Array if you elect to go b&w, no matter how clever the software.

Except, there is one special case. When re-photographing b&w content. If you happen to know the color characteristics of your light source (ideally, you have equal amounts of what your camera sees as R, G, and B) then you can custom-convert and get the full sensor resolution out. This is theoretically pretty handy for re-photographing b&w negatives as a cheap route to digitization.

Yeah I guess the micro filters would make it so that you can't just get intensity information usefully out of a color Bayer filter.

However, it would be possible to manufacture a single layer black and white digital sensor for the cost of current ones (actually probably much less, since there are fewer passes in manufacturing), and the double or triple or whatever it is extra resolution of a Bayer. Also 1 2/3-ish more stops of dynamic range, I think? Since each photosite can collect 3x more photons without any more noise. And/or ISO performance.

Edit: you could probably also do without an AA filter anymore, which would give a huge resolution boost.

You might get weird absorption patterns though. Might not look like normal B&W. Howeve,r this would be most efficiently dealt with by the user choosing various LENS colored filters, not by building in any standard filter on the sensor. Then, if you wanted a more standard look with only modest gains over color (still get the resolution, though), you slap on a greenish filter on your lens. If you don't mind the slightly odd tones in exchange for high ISO / DR, don't put on any filter.

Come to think of it, I wonder why nobody has done this. Or have they?

 

[amolitor]

Yeah, zapping the color filters increases resolution. There's any number of places that will pull the bayer filters off your sensor for you, and/or the AA, etc.

 

[Gavjenks]

oh sweet! I'll definitely look into that (have a spare crop sensor body sitting around I don't use much now...). I would have assumed it was too permanently integrated on the sensor to do that.