2

Imagine that you have a round sensor that covers full circle that your lens can expose. Your camera should be able to shoot at any aspect ratio in any orientation. You won't have to rotate your camera for portraits; just set the aspect ratio 2:3 instead of 3:2...

Your lens collection is more expensive than your camera; the sensor in your camera is only a part of the camera's cost. Why won't then camera manufacturers produce circular sensors (that would be only marginally more expensive) and provide the above mentioned conveniences?

3

I've often wondered that myself. Some possible reasons:

  • You will either waste a lot of silicon in places that will never get light (near the edges) or you'll have a round wafer that is harder to manufacture.
  • The resulting software will either have to waste cycles processing the junk data in those unused parts or will be significantly more complex.
  • Sensor readout is challenging if the rows have variable length.
  • There may or may not be adequate space for sufficiently large reflex mirrors, requiring either more interesting (multi-flip) SLR designs or mirrorless designs.
  • Some lenses have parts that are shaped in ways that prevent getting a full circle.
  • Your average user will likely find a round sensor harder to use unless the default behavior is to show a normally cropped rectangular image, which requires additional hardware (e.g. a partial blackout LCD in the OVF).

That said, the idea is not without merit. There's more discussion on the subject here:

https://www.diyphotography.net/circular-image-sensor-best-thing-ever-heck-yeah/

  • 3
    There's also the issue of yield from a single wafer. Square/rectangular sensors can use almost the entire surface of a wafer. Cutting round sensors out of a wafer would leave a large percentage of the blank unused. – Michael C Nov 29 '18 at 5:15
  • Related: I once read a paper that proposed a phyllotaxis layout (like sunflower seeds, spiral arms are successive Fibonacci numbers) (I think their prototype sensor was 377x610 spirals). The point was to minimize average radial corner distance, minimize net interpolation errors under rotation, and was particularly suited for 180° fisheye projections. I wish I could remember where I read that paper, it was fascinating. Of course, they had obvious difficulty with scaling manufacturing up for high resolution sensors. – scottbb Nov 29 '18 at 6:07
  • A solution to the wasted waffer space for a round sensor is to make a hexagonal one, that still fills the surface – lijat Nov 29 '18 at 9:08
  • @MichaelClark Aren't the wafers usually round to begin with anyway? At least for silicon wafers they're round, apparently due to the way the silicon crystals grow. Rather than cutting round dies out of one they could just use smaller, whole wafers. Assuming that wouldn't turn out more expensive. It would forego the cutting process though. – G_H Nov 29 '18 at 13:13
  • The wafers are typically 100–400mm in diameter. I guess smaller wafers do exist, but the cost per unit area is probably higher as the wafer size gets smaller. Not sure. A hexagonal design likely would work better. It would also be easier in terms of wiring it to the package, because the contacts would be in a straight line. :-) – dgatwood Nov 29 '18 at 19:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.