View Full Version : Nikon develops full color sensor design...
Evin Grant
08-09-2007, 01:48 PM
Very interesting new patent, could be the answer to the Bayer naysayer.
http://www.dpreview.com/news/0708/07080901nikonimagesensor.asp
http://www.dpreview.com/news/0708/nikonsensortech.gif
Cail Young
08-09-2007, 05:27 PM
I suppose it is marginally better than triple-CCD... But you still have differing distances from the microlens to each color's photosite.
Evin Grant
08-09-2007, 06:56 PM
Yea, but only by nanometers, a piece of film is many hundreds of times that distance thick.
Cail Young
08-09-2007, 10:42 PM
Yea, but only by nanometers, a piece of film is many hundreds of times that distance thick.
Nice! So i guess this negates the Foveon's problems with transmission, so we're down now to only the quality of the dichroic mirrors...
Evin Grant
08-10-2007, 02:08 AM
And the fill factor of the micro lens/photo site combo. This may be the achilies heel of the design. In order to achieve the same sensitivity the photosites would have to be about 50% larger I'm guessing since each photosite now has three photosenitive elements that all have to be addressed and powered. Although they don't have color filters over them, so that helps. It's hard to say how much more resolution is gained by avoiding a debayer algorithm but it's certainly not 100% as the numbers might suggest. My guess is it's probably on the order of 20-30% of actual resolution. So you could achieve a similar resolution to a 10MP sensor using only a 6MP sensor. Becasue it's resolution in both directions X & Y.
however a 10MP sensor would resolve like a 15MP and that's something interesting, not to mention color accuracy.
Kevin Halverson
08-10-2007, 07:23 AM
This is a very interesting invention. I assume that the third dichoric mirror's direct output is virtually non existent due to a (not shown) IR filter proceeding each "microlens". In the attached image, this concept isn't illustrated. Likewise, I wonder if each "microlens" would negate the necessity of an OLPF in preventing aliasing?
The only downside (if you can really even call it that) is the increase in the number of photo sites and the speed that would be required to perform three times as many quantizations.
It will certainly be interesting to follow the commercialization of this invention. Thanks for posting this, great find!