The more than perfect chip
Nov 10, 2010
Not many replied to the previous post, the "perfect chip". Maybe all agreed - or not many read it Whatever, here comes the continuation.
So - whats wrong with "the perfect chip"?
Two things really:
1. I dont really think RGB is optimal
The eye is not pure RGB. Printers have lots of inks. The illumination varies. OK - you can make rather nice color images using RGB. And the mapping to your printer is fairly good. But - I assume you can do better with more spectral bands.
2. A regular grid is too perfect
They have to change the measuring of resolution because of the regular grid. You cannot any longer measure resolution with horizontal lines. You have to tilt them five degrees or so. Isnt that cute? When the sensor cannot resolve horizontal lines nicely you tilt the resolution pattern. Isnt something wrong here? Should not the image quality be independent of how you tilt your camera?
There are several solutions to this problem
2.1 Just more pixels!!
Instead of 30 MP (x3) you can have 300 MP or 3000 MP. At some number of pixels - the regular grid problem becomes a non problem. Because neither lenses nor printers can resolve it. The big amount of pixels makes the sensor hard to do though and the data becomes large.
2.2 Irregular pattern
You can have an irregular pattern of detectors. A pattern that you have designed to be independent of direction. Not easy to handle in files and in editing programs though. And quite hard to display on your monitor. But ... with the correct printer driver it should be VERY good.
2.3 The one bit detectors
Instead of having a counter in each pixel, you could have just a simple switch. If something hits the pixel - then the switch is turned on. This means of course that if one more photon hits the same detector then nothing happens. If you then have lots of such cells. Lots and lots and lots. Then you have a sensor with very high resolution but lousy bit depth. But ... that does not matter - because there are so many of them. NOTE - that this is very near to how film works. Note also that it is non linear. The more cells that are switched the smaller is the probability for detection. This increases the dynamic range. I assume the shoulder in film is made the same way. This solution will also mean lots of data.
(Sleeping - so the need to support it is even higher)
X3F tools : http://www.proxel.se/x3f.html