My prediction- Bayer sensors limited lifespan

[bfitzgerald]

With pixel densities increasing rapidly on cropped sensors, 15mp and rising..the associated problems with that, requirement for better quality lenses, AA filters, diffraction etc. Surely 20mp is not far off. Maybe even more

I just don't see long term how they can carry on with bayer sensors. Surely the way forward is a 3 layer sensor such as foveon, but done right..better at high ISO.
I bet a foveon 12mp D3 would beat a 24mp bayer sensor right off, and do so with non super demanding lenses too.
APS-C, 15mp is already punishing some lenses, it's only going to get worse as time goes on, they can only really push this so far..until it grinds to a halt.

What do you think?

[PhotoTraveler]

Bayer isn't going anywheres.

It's simple and works great and scales endlessly. It's used in sensors right now that scaled to FF would be 420MP. No problems.

Foveon Design doesn't scale at all, they haven't been able to get anywheres bumping up the resolution of their sensors. Plus it hasn't shown to be very good.

Until Someone can invents a method that is cheaper, easier to make, scales better, and or has massive image quality benefits over bayer, it's not going anyplace. People keep talking about something else. Yet 33 years later, the original solution is still the one used in over 99% of color applications.

[Dusty]

Well, they're limited by the fact that progress keeps marching on, even though it sometimes comes by accident. Someday a new and better sensor will be discovered. Someone will let one get contaminated by sulpher, and put it aside. Then an assistant, not knowing it's bad, will hook it up to an inverted power supply and it will produce triple the resolution.

Or maybe a pharmaceutical company had an experiment go wrong and discovers that the new compound is ultra light sensitive. More times than not great leaps are made by mistake, small improvements are made by lots of hard work.

Maybe it will come soon, or maybe it will be years from now, but there will eventually be a replacement. Then we can get those 420 MP FF sensors that will do 48000 ISO. There will still be someone complaining about something, though.

Dusty

[bossel]

Someone will let one get contaminated by sulpher, and put it aside. Then an assistant, not knowing it's bad, will hook it up to an inverted power supply and it will produce triple the resolution.

I think you understand how it goes

[David Kilpatrick]

Foveon and Bayer are not exclusive technologies. There is no reason why a sensor can not be made with a green-blue checkerboard filter pattern, no red filters at all, but have the entire sensor sensitive to red in a second layer. Calculating from known overall light energy (derived from adjacent GBGB square plus the underlying red RRRR square) and the known G and B values, plus the known values of red transmitted through B and through G (the high sensitivity of the silicon to red and infra-red correcting in part for the attenuated values after G and B filtration) each GR and Br (r = lower red value because of great effect of B filtration in reducing R), a calculated RGB value can be derived for each individual pixel without using a large convolution to synthesize it.

It is already possible to make a sensor which can create full colour using only two colours (typically orange and green filters) plus an exposure map - a low resolution greyscale mask which can be created by a sensor like the live view CCD in the A350, or by the metering sensor in a Nikon with 3D Matrix colour metering. It can even be done pretty well if just the overall exposure is known.\

Dr Andrew Stevens, late of Kodak, was working on ways of recovering realistic full colour using only a single colour filter and greyscale information (he did this the easy way, using b/w film - one normal exposure, one through a filter, then scan them, register them in layers in Photoshop - and the rest was his own discovery). I have not seen any recent published papers by Andrew and I do not know where he is now, but if he's working for someone like Foveon, Leaf, or whoever I'm sure alternatives to the Bayer principle will be advancing.

David

[Javelin]

I put a bunch of machines into kodak at Kodak park in Rochester when they were going through a realy rough time. in their digital camera department in fact. they were importing cameras with their name on it from ... Nikoh or Chinon? can't remember which.. 1 mp at the time maybe 2.. they were getting so many returns they had to dedicate a line to referbishing them repack and send them out again.

man I've never, even now, been to a place where everyone I spoke to was so afraid for their jobs. these were all people that worked there for their whole lives and whole departments were vanishing overnight. On a trip there once I was a little early and when security paged my contacts there was no answer and he let it slip that it was the 3rd department that day thats dissapeared... man the stories from that place were pretty grim.

I can't imagine their real talent staying there. the environment wasn't pleasant with each department fighting for it's life and everyone being scared. When their sheet film plant closed in Ontario there wasn't that sense of impending doom and people obviously liked working there right untill the end.

I had to put a machine in that plant too and the funny story is that I had to practice setting the machine up and adjusting it in the dark. then I had to train 2 people on it's operation of it in the dark, it was an interesting experience. They told me the film was for arial photography.

[youpii]

I'm not very comfortable with the Bayer filter because I feel that the picture is actually upscaled three times (each Red/Green/Blue sensor is shared among three pixels http://www.cambridgeincolour.com/tutori ... ensors.htm" onclick="window.open(this.href);return false;).
With today'ss high pixel count sensors, maybe other RGB matrix and demosaicing algorithms are more adequate.

Another road is 3CMOS like in high end video cameras. The RGB filters are replaced by a prism that eats much less light than the RGB filters. APS-C sensors are quite cheap now anyway.

I would also be interested in a B&W only digital camera, no more RGB filters to record colors that I will discard anyway.

[InTheSky]

For the 3 cmos sensors with prism , have you ever tried a Minolta RD3000 ? This was a good design ... but the size of it ! , and to support 3 CCD or CMOS today, I don't see how they can do this smaller for a FF sensor. Probably not a portable camera ... or a different look. (I have got on RD3000 for a few months, but just trying to make the camera works with my High end PC was a good reason the resold the camera ).

The other challenge, is to be able to place 24mpx perfectly align on three place. The light take off by the prism will probably be not much a problem because the Single light sensor will resist more to lost of light.

[David Kilpatrick]

The RD3000 was not a three-sensor system, it used two Bayer sensors and a prismatic lens which split the image centrally so that it joined seamlessly despite having two 'stitched' smaller RGB Bayer sensors. It was a Vectis mount system, I used one for a short period after we sold our Leaf Lumina scanning camera. The RD3000 was replaced, believe it or not, by a Minolta Dimage 7 for my use - the Dimage 7 was much better.

The RD175 was the model with three sensors.

David

[InTheSky]

My big mistake I got the RD175 the one made with a Maxxum 500SI body. With the Big PCMCIA card. The big noise going from the Harddrive when you put the power on. I think I have sold it with a Minolta 18-70mm lens.


Sorry

Frank

[youpii]

CCDs were very crappy at the time of the RD-175.
Another design where you don't have a classical mirror&prism might fit in a DLR-sized camera (maybe with a different shape).

[bossel]

Another road is 3CMOS like in high end video cameras. The RGB filters are replaced by a prism that eats much less light than the RGB filters.

If this could be done in a not too big body, why not? There are still many people out on digital medium backs, they might be thrilled by a 24Mpixel 3 sensor camera, even if'd be a tad more pricey!

I would also be interested in a B&W only digital camera, no more RGB filters to record colors that I will discard anyway.

Didn't Kodak do something like this - dunno if it was successful.

In any case, I think Bayer will be around for a long time, as will be APS-C. Maybe in 5 or 10 years the picture will be different.

[David Kilpatrick]

Didn't Kodak do something like this - dunno if it was successful.

The DCS 760M

6 megapixels, but as a monochrome camera, about as sharp as a 12 megapixel Bayer image converted to monochrome by optimum methods.

David