Why are sensors flat?

I posted this over in Leica talk but this one can be a bit more lively.

I'm not an engineer or optics designer but I've always wondered why sensors are flat? Film I get, but if the sensor has a slight concave structure as the eye has and as the lens itself delivers the light to the sensor, then couldn't post processing render the picture flat? After all that's what the brain does with a curved image. This would perhaps solve the microlens issue (Leica) and allow for a larger sensor with a bigger sweet spot and more resolution/less noise. Put paid to vignetting too. (I'm also confused by what a shutter is for exactly - another day).

Maybe I'm a bit nuts but the whole design of digital cameras seems to be a bit like motorized horse carriages. Maybe a radical new design paradigm is needed?

Opinions welcome.

--
http://www.pbase.com/edward_in_chicago/root

Can you imagine the cost that it would add?

Edward in Chicago said:

I posted this over in Leica talk but this one can be a bit more lively.

I'm not an engineer or optics designer but I've always wondered why
sensors are flat? Film I get, but if the sensor has a slight concave
structure as the eye has and as the lens itself delivers the light to
the sensor, then couldn't post processing render the picture flat?
After all that's what the brain does with a curved image. This would
perhaps solve the microlens issue (Leica) and allow for a larger
sensor with a bigger sweet spot and more resolution/less noise. Put
paid to vignetting too. (I'm also confused by what a shutter is for
exactly - another day).

Maybe I'm a bit nuts but the whole design of digital cameras seems to
be a bit like motorized horse carriages. Maybe a radical new design
paradigm is needed?

Opinions welcome.

--
http://www.pbase.com/edward_in_chicago/root

Click to expand...

It is much easier and cheaper to make lenses that focus on a flat plane than to make a sensor that is curved like a telescope mirror. Besides, most Vignetting on your 5D is not a result of micro-lens shading. In fact, that is a very minor cause.

Steven

--
---
Winter 2008:
http://www.pbase.com/snoyes/images_winter_2008

2007 Paria Plateau
http://www.pbase.com/snoyes/images_spring_2007

I think its the other way around. Lenses that focus on a curved surface should be far, far easier to make. Achieving focus across a flat plane is an essential compromise and focus precision edge-to-edge is always a problem. Any optical engineers out there?
--
http://www.pbase.com/edward_in_chicago/root

that would cost a fortune. Then you have curved AA filters. Curved CFA. Curved hot mirrors.

So it is MUCH cheaper to make a lens with a flat focus compared to a sensor witha curved focus.

Steven

Edward in Chicago said:

I think its the other way around. Lenses that focus on a curved
surface should be far, far easier to make. Achieving focus across a
flat plane is an essential compromise and focus precision
edge-to-edge is always a problem. Any optical engineers out there?
--
http://www.pbase.com/edward_in_chicago/root

Click to expand...

--
---
Winter 2008:
http://www.pbase.com/snoyes/images_winter_2008

2007 Paria Plateau
http://www.pbase.com/snoyes/images_spring_2007

Edward in Chicago said:

I think its the other way around. Lenses that focus on a curved
surface should be far, far easier to make. Achieving focus across a
flat plane is an essential compromise and focus precision
edge-to-edge is always a problem. Any optical engineers out there?

Click to expand...

...sensors are flat because you want to maintain compatibility with existing lenses.

• Anders @ http://www.aobild.com *

Your subject is typically in a plane. If you focus on a wall - always good for hours of Testing Fun - you wouldn't want a lens whose focus zone was curved, would you?

Dan

Edward in Chicago said:

I think its the other way around. Lenses that focus on a curved
surface should be far, far easier to make. Achieving focus across a
flat plane is an essential compromise and focus precision
edge-to-edge is always a problem. Any optical engineers out there?
--
http://www.pbase.com/edward_in_chicago/root

Click to expand...

--
---
G Dan Mitchell
SF Bay Area
http://www.gdanmitchell.com/

Way too expensive to make (I'd go as far as to say impossible with modern technology that is based around flat plats)

Way too expensive to keep clean (I grab the term "DustBowl" before anyone else)

Would leave no room for a conventional mirror, but then maybe you'd also need a distorting mirror to bring the image into focus on the focusing screen, which brings me too....

Would need a curved focussing screen or some other wacky way of working ( I think, but could be way off here )

Again not an optical, CCD or CMOS designer but curved imaging planes are one of those "I wonder...." ideas that when you start to unpack are just not going work just yet (but never say never) and mainly because the mfr plants only do flat

If we're going with mad ideas, can we have a layerd imaging "substance" that is more than one dimensional (IE not a plane or surface) and captures all light and allows us to post-focus in RAW. (EG I'd like to push this photo forward one step)

I know through clever processing this has been done before but at very low res.

Edward in Chicago said:

I posted this over in Leica talk but this one can be a bit more lively.

I'm not an engineer or optics designer but I've always wondered why
sensors are flat?

Click to expand...

To understand why sensors are flat you need to understand a little of the silicon IC process.

Sensors are not made individually, many of them are made at the same time on a single wafer of silicon, typically 8 or 12" in diameter, depending on the manufacturer.

Sensors are flat because the silicon wafers are flat.

Silicon wafers are not made individually, they are cut from a single crystal of silicon grown with the corresponding diameter and several feet long - essentially a long cylinder that is sliced up into individual wafers.

Silicon wafers are flat because that is the only way to cut the silicon crystals.

It might be possible to cut the wafers into cap shaped domes, but that would be extremely wasteful and only a few such wafers could be cut from a single crystal, instead of many flat wafers. More significantly though, the processing of the wafer in the IC foundry usually depends critically on the crystal orientation at the surface. With a flat wafer, this orientation can be defined precisely across the entire surface of the wafer, resulting in uniform deposition of the oxides, metals and dopants necessary to create the individual sensors on the wafer. With a curved wafer surface, in addition to requiring a completely unique (and therefore ultra-expensive, on the scale of billions of dollars investment) wafer processing plant, the crystal orientation would change across the surface of the wafer, resulting in highly variable deposition and growth rates. Remember growing salt or copper sulphate crystals in junior school? The end result is something with flat surfaces - all crystals are like that, so a curved wafer must have a variable crystal orientation and/or have very high crystalline defects. Such defects cause growth and deposition non-uniformity. Even if this problem could be solved with limited trillions of dollars, which is unlikely, it would have reduced yields compared to conventional flat wafer processing.

So there you have it:
Sensors are flat because silicon wafers are flat.
Silicon wafers are flat because they are cut from single silicon crystals.
Single crystals have flat facets, so sensors are flat.

Any deviation from this basic rule requires the redevelopment of the entire semiconductor industry and I don't think that the market for photographic sensors could justify that let alone a single company, such as Canon or Sony, afford it.

Edward in Chicago said:

I think its the other way around. Lenses that focus on a curved
surface should be far, far easier to make. Achieving focus across a
flat plane is an essential compromise and focus precision
edge-to-edge is always a problem. Any optical engineers out there?
--

Click to expand...

When it comes to very short focal lengths, it may be easier to produce a flat focal plane if the image plane were curved (concave). However, keep in mind that there are lenses which have negative field curvature, i.e., they would prefer an image plane that is actually a bit convex, not concave.

For medium to long focal lengths, lenses do actually naturally work with a flat image plane. Making such lenses work with a concave image surface could be quite a challenge.

Besides all of this, producing a curved imager is simply outside of current manufacturing technology, and if one were to be introduced, it would require an entire new family of lenses to work with it.

Apparently no billions involved. I did a google: Here's one:

An electronic imaging system using a curved image sensor can use a faster lens, and cover a greater field of view, than an imaging system using a planar sensor. The simpler lens systems also weight less, a decisive advantage in portable applications. This paper describes a method to fabricate a curved silicon substrate from a flat wafer containing appropriate circuits. To curve the substrate, the processed wafer is diced, by dry-etching from the backside, into 1x1cm tiles. The tiles are separated by 0.5mm gaps, which are bridged, in turn, by a dense array of 45x100um gold leads formed by electroplating using lithographically defined leads as seeds. Two methods were used to curve the wafer. In the first one, the wafer was bonded with epoxy to a PMMA disk, and then curved by heating the sandwich, under a load of ~ 230gr, for 1.5 hrs at 130C in a concave metal mold with a radius of curvature of 7.8cm. In the second method, the wafer was put into a curved metal mold, radius 14cm, loaded with 230gr, and heated to 290C for 2 hrs. The normal and shear strains accommodated by the flexible interconnects were measured by analyzing their deformation. The experimentally measured strains are compared with a model that calculates the deformation required to deform a flat sheet into a spherical surface.

http://www.mrs.org/s_mrs/sec_subscribe.asp?CID=2597&DID=110053&action=detail
--
http://www.pbase.com/edward_in_chicago/root

One could imagine some future type of sensor which could be virtually printed or deposited on a flexible substrate, a bit like some flexible displays are now being made. This could be mounted on a curved former, though I think it would only be practical to curve in either the vertical or horizontal direction, not both.

You would need a different optical design for the lenses, so of course it won't happen - at least not in the forseeable future due the vested interests in the current lens mounts and designs.

Maybe a future EVF based camera with no mirror and a new lens mount could use this concept.

Fred

Edward in Chicago said:

Apparently no billions involved. I did a google: Here's one:

Click to expand...

Sounds more like a tessalation of 1cm square segments than a curved wafer. Also lacks any info on damage caused by such processing and is hardly a mass producible part. So replace the billions of $ for mass production with millions of $ for single samples - you reach the same conclusion: it isn't gonna happen.

Fred Briggs said:

One could imagine some future type of sensor which could be virtually
printed or deposited on a flexible substrate, a bit like some
flexible displays are now being made. This could be mounted on a
curved former, though I think it would only be practical to curve in
either the vertical or horizontal direction, not both.

Click to expand...

Of course one can imagine anything, however a completely new semiconductor process that has the same capabilities as silicon today isn't going to arrive overnight or cheaply. Yes, there are some flexible electronic circuits, such as those used on touch panel displays, but these are trivial in complexity compared to silicon circuits.

Ain't gonna happen.

Insanely Ridiculously Expensive

To do it has little advantage, since lenses can make up for it just fine. But more to the point you are asking for a bowl shaped sensor. No silicon fab method exist to create such a sensor, and no such fab will likely ever exist since it would be insane to create with virtually no use. The only such path that would see any effort there would be for artificial eyes. But even then they would probably just use a flat sensor and make it up with optics.

Everything involving the creation of sensors are flat, and as is they are very expensive.

Next you would have to re-invent a whole new lens system, which would be very expensive. Which would lead to a whole new mount.

Items like a shutter would have a big gap between the farthest back part of the shutter and the time of the light would be thrown off unless you made a bowl shaped shutter, though these differences might not matter. Or possibly remove the shutter.

A SLR design would get very weird with a bowl shaped mirror and a penta prism that would be beyond weird and expensive.

Your AF sensor would have to be either bowl shaped (more expense) or have lenses on it to flatten things out.

Such a design would have been more practical with film, but it didn't happen, there is no way it will happen with digital. Unless you completely re-invent how integrated circuits are created, it's a no go.

Also look how we don't have round sensors either, they would open up all sorts of aspect ratio options, but the cost of such a sensor is ridiculously high for the gain. But again, if you come up with a all new fab process that is cheaper than what we have, it might happen.

Something along the lines you propose was done, but as far as I can tell only with film:
http://en.wikipedia.org/wiki/Schmidt_camera

Obviously, film is very easy to shape because it's thin and flexible. Too easy, in fact: some cameras had vacuum systems designed to keep the film flat against a back plate during the exposure (Contax, for instance).

So, as you can see, even with film, the efforts were directed at keeping the "sensor" flat rather than curving it to match a curved focal plane from the lens.

Cheers
My wit gauge shows “half”. Uh-oh...

I did not read all the posts but here is my take.

Can you imagin how hard it would be to clean that puppy

...the ideal curvature of the sensor have to be different for every focal length?
--
http://www.tomasjacko.com

Only those who cannot see beyond their viewfinder consider cropping a sin.

...has been haunting me for some time now, too.
--
http://www.tomasjacko.com

Only those who cannot see beyond their viewfinder consider cropping a sin.

The ability to change the focus depth in PP has been demonstrated with a multi-lens multi-imager camera system that looks a bit like a fly's eye. When you combine the individual images (which individually have deep DOF) the parallax shift between them depends on distance so you choose the focus depth when you choose the parallax correction.