Moore's Law relies on size reduction

[BJL]

Sensor Size and Moore's Law
It needs to be said yet again: one of the dominant factors driving the improvements in cost and speed summarized by Moore's Law is reduction in the size of the individual components on chips, and through that the size of the chips themselves. Chip size reduction reduces cost, and also can increase speed by reducing the distances that signals need to travel. At 3GHz, signals can only travel a couple of cm in a clock tick, so a 24x36mm processor chip would probably be too big to function at today's top clock rates.

I have not seen any evidence of a Moore's Law like rate of cost reduction for chips of a given size, including sensors in a given format. What retail price reductions there have been for 24x36mm format DSLR's (none at all for the 1DS series in over five years) seem to be explicable by reductions in the cost of other parts of the camera (5D compared to 1Ds and 1D models) and related increases in sales volume that improve economies of scale and thus allow lower markups.

Factors like this along with increased competition will likely bring prices down for 24x36mm format DSLR's somewhat, but to nothing like the extent of Moore's Law behavior.


To see Moore's Law at work in digital camera pricing, look at the digicam mainstream, where sensors and pixels keep getting smaller: first 2/3" and then 1/1.8" and 1/2" have been marginalized by the new favorites, 1/2.33" and 1/2.5".


Lens Size and Format Size
Also, Moore's Law does nothing to the size, weight and cost of lenses, so the smaller pixels rather consistently found in smaller sensors allow the use of shorter focal lengths to get equally detailed images over the same field of view, which is why smaller formats have a fairly steady advantage in lens bulk, particularly with telephoto lenses, and thus in overall bulk of kit with decent telephoto reach.

Anyone who has felt much need for focal lengths beyond about 300mm in 35mm format should be able to see the advantages of reducing the telephoto focal lengths needed, at least when traveling light and small is a priority. For this reason alone, formats like DX, EF-S and 4/3 will continue to be in demand with even fairly serious amateur and professional photographers, even if sometimes paired with a bulkier kit in a larger format for other purposes.

[danm628]

Moore's law will help a little on fixed size chips such as sensors.  It allows more support circuits (data  amplification, data transfer, etc.) to be placed on the chip.  A side effect of Moore's law also helps, yield of a given size chip will go up with time.  The other chips in the camera do benefit from Moore's law, they get faster and smaller.

You are correct in your main point though.  Die cost is mostly a function of area, smaller is cheaper.  Devices where the die size is fixed such as camera sensors don't gain very much from the reduced geometry size which Moore's law predicts.

I do wish that all silicon was subject to Moore's law.  That would help a lot with my continuing lust for large, expensive silicon (or oxides thereof) in lenses.

   -- Dan

[Ray]

Photography means 'painting with light', or in the digital context, perhaps more specifically, painting with photons.

The larger the sensor, the greater amount of paint we can use, especially significant in the shadows.

What's exciting about this continual miniaturisation and movement towards full nanotechnology, is the possibility that one day we might have a P&S camera that fits in a shirt pocket, which  is capable of a resolution and dynamic range exceeding that of an 8"x10" field camera.

This is not science fiction. You'll find on the Dpreview news section an announcement that Kodak have already designed a 5MP sensor just 4.47mmx3.8mm with 1.4 micron photosites. That's about double the size of the wave-length of red light.

Whilst the technology will initially be used in cell phones, there's no reason why we could not eventually have a 1 gigapixel FF 35mm sensor.

But this development of Kodak's is not just about smaller pixels. They've also employed their new CFA which has half the pixels being panchromatic (ie. no color filter at all). Not only that, they've reversed the process of assessing electrical charge. Instead of counting electrons, they are now counting the holes left by the departing electrons. The electrons can now buzz around producing all sorts of cross-talk. Never mind. It's the holes they left we're interested in.

This new way of doing things results in a 1 & 2/3rds stop S/N improvement. (slightly less than some claims for the D3 improvement over everything else   ).

The flow-on effect when implemented in larger sensors may not be this dramatic, apparently, but will certainly reduce noise at the pixel level in gigapixel 35mm sensors.

But what about dynamic range? It's true that DR with a single exposure of a small sensor is limited. But not if you can take rapid multiple exposures of varying length. Electronic shutters can be much faster than mirrors flipping up and down.

Ultimately, there's probably no technological obstacle to exposing a small sensor to a 100th second exposure at f2.8 in broad daylight with sunny and contrasty conditions; 1/4000th for the highlights and 1/200th for the shadows, with a few resets in between.

The future as I see it   .