D3x: hemispherical sensor

[Oz_Beachside]

Has anyone thought that maybe the flat sensor creats redundancy in the lens?

Assuming the human eye is superior in design to a dSLR, it only has a single lens, but does not have a flat sensor (retina).... so, would it not make sense (not commercial sense of course) to have a hemispherical sensor, and simplify the lens design?

[phillipb]

Makes sense to me.

[MATT]

From mad man to genius or wa it he other way around.

MAtt

[Killakoala]

I don't doubt for a minute that the boffins at Nikon have thought of that. The issue would be 'how would we manufacture a CCD curved?'

Not impossible though!!!

[Biggzie]

Sounds good, but present lens design has a flat focus plane. To use a curved sensor with a flat focus lens would mean you create focus effects arounf the edge of frame like a tilt lens does across the frame.
So my guess is that a curved design would lead to a complete new range of lenses designed for the curve.

[wider]

probably been contemplated many times but would be largely an issue of the cost of developing such a sensor. also means you can cram less pixels per sq cm due to the shape of the sensor...

the first maker to come up with this design will (with lenses to suit) probably claim a large part of the market share

[Murray Foote]

I would have thought that the retina's the rear element and the sensor is the brain, which doubles as Photoshop as well. If you have a flat field lense (macro, enlarging or reversed lense) you presumably wil have a flat image at the sensor. Otherwise, I would think the curvature of the projected image plane would vary with lense, focal length, probably aperture and maybe focus. I don't think a curved sensor is going to happen any tme soon....

[Reschsmooth]

If the photosites are effectively cylindrical silos, you would have a situation where either the silos are more conical or there are large gaps between the photosites. Either way, I would imaging this would either cause signal/noise issues or data collection problems. As I often say, I know nothing about this.

[gstark]

So my guess is that a curved design would lead to a complete new range of lenses designed for the curve.

Which is exactly what Bruce was saying.

The issue would be 'how would we manufacture a CCD curved?'

Actually, I would have thought that the issue might have been "how do we manufacture a curved shutter?"

From a practical PoV, there are just two places that you can place a shutter, either at the focal plane, or within the lens.

Placing one within the lens introduces some serious complications into the lens design - where to place it, how to keep it open so focus and metering may be achieved, how to make it perform so that rapid shutter actuations - 8 or 11fps, for instance, can be achieved. While some of these issues are addressed with leaf shutter technology, there are good reasons why the maximum shutter speed is only 1/500.

And of course you now have the added expense and weight of putting a shutter into each lens.

So ... how to make a curved fp shutter? I haven't a clue.

But getting back to Bruce's original question ... the lens in the human eye is a very simple device, and only provides for just one focal length, roughly equivalent, in 35mm terms, to an angle of view presented by a 50mm lens. A Planar lens is probably the closest we see, in general terms, to this sort of design, but that still has a flat trajectory, caused, of course, by the need to focus on a flat object, that being (originally) film.

To a large extent, view camera lenses still take a very similar - and simple - design philosophy, and that is largely why, when using a 300mm lens on a view camera, you need extension of ... 300mm.

That's not even considering the physical size that any given focal length/aperture combo dictates, and yes, we all want faster glass.

[moz]

Hmm. Astro gear does not have the problems we have with pricing and labour availability. But they choose to use flat sensors. If someone willing to spend over a million bucks just for one sensor does it the boring traditional way maybe that's a hint?

Given the way they make sensors, its effectively impossible to build a curved one. At the very least it would be quite expensive. The cost of a semi-spherical sensor requiring that all the existing chip manufacturing processes be replaced would presumably start at about the cost of a complete new chip fab or around $5 billion. If you bought a second hand, low-res fab you could possibly get it for a couple of billion, but remember that the high end fabs don't get retired they get refitted. So you'd be back in the 100nm or bigger territory. And you'd be refitting right from the very start, where they slice a cylinder of silicon into wafers. Or in your case, semi-spherical shells. Using some as-yet undiscovered technology. Or you could waste dozens of wafers worth of silicon by cutting big chunks off and machining them down to the shape you want. That would probably be cheaper.

You could instead hand build the sensors one at a time, and for the number of units that high end digital cameras sell that's probably the cheapest way. So say we tile the sensors onto a semi-spherical back plate using micromanipulators. The back plate is easy, just use steel or titanium and machine it down, spherical surfaces are pretty easy to make. Then get a box of mm-square sensor tiles and start laying them out and soldering them together.

Now you need to completely rework your lens line, and because this is high end stuff you're going to need to supply a full set of primes and a couple of zooms to start with, and that's a few years work for your main lens design team. So, no decent lenses for anyone else for a few years. But that's ok, you'll make up the loss in market share when you launch your new toy.

[Oz_Beachside]

The issue would be 'how would we manufacture a CCD curved?'

Actually, I would have thought that the issue might have been "how do we manufacture a curved shutter?"
...
So ... how to make a curved fp shutter? I haven't a clue.
...

I was thinking, that if my design steps outside the film "box" on the shape of the image capture area, that the same thinking might apply to the shutter.

Since semiconductor switching rates are always on the increase, I would imagine that we will one day be able to remove the mechanical shutter, and simply switch the sensor... so, we have removed all the crappy lens elements that flatten the image plane for us, and removed the limits of a mechnical shutter...

[gstark]

The issue would be 'how would we manufacture a CCD curved?'

Actually, I would have thought that the issue might have been "how do we manufacture a curved shutter?"
...
So ... how to make a curved fp shutter? I haven't a clue.
...

I was thinking, that if my design steps outside the film "box" on the shape of the image capture area, that the same thinking might apply to the shutter.

Which is why I mentioned the leaf shutter, which, typically, is mounted within the lens. As I said, there's no other (practical) place to put it.

Since semiconductor switching rates are always on the increase, I would imagine that we will one day be able to remove the mechanical shutter, and simply switch the sensor... so, we have removed all the crappy lens elements that flatten the image plane for us, and removed the limits of a mechnical shutter...

Yep. The D70 has that; you can sync at whatever the fastest shutter speed it has. Shows how long it's been since I picked mine up.

[digitor]

Has anyone thought that maybe the flat sensor creats redundancy in the lens?

Yes, they have! I was reading (in Photonics magazine, I think) recently that this technology will be incorporated in the next generation of mobile phone cameras, which will make the lens design much easier (and thinner).

Cheers