"swnw" <(EMail Removed)> writes:
>Why CCD Imagers are usually smaller than 24 x 36 area?
Because of the way the cost of a chip scales with its size.
Many good consumer cameras use a chip that is about 1/5 the dimensions
of the 35 mm frame, about 5 x 7 mm. This is 1/25 the area of a
fullframe chip.
If you had a manufacturing process that turned out 100% usable chips,
using a square sheet of silicon, you would get 25 times as many of the
small chips as you would of the fullframe chips, given the same amount
of raw materials. So the large chip has to cost at least 25 times as
much as the cost of the small ones.
In addition, the raw wafer is actually round, and you can pack small
CCDs closer to the edges of the wafer than for the larger ones, so the
factor is slightly larger than 25:1.
But the think that really makes it more expensive is called "yield".
Chip manufacturing isn't perfect, and every wafer has some flaws in it
that are large enough to make the chip containing the flaw into junk.
Now, suppose that the flaws are relatively few, and that when making
the smaller CCDs 99% of them are good, and only 1% of them are bad.
In other words, there is a 99% chance of no "fatal" flaws in any one
CCD. When you make the larger fullframe CCDs, the odds of having no
fatal flaws in an area 25 times as large is (roughly) 0.99^25 = 0.78.
So 78% of the larger CCDs are good, and 22% of them are junk. The
smaller proportion of good large chips means that the good ones have to
cost more by a factor of 1/0.78 just to pay for the raw materials.
As the rate of flaws goes up, it messes up the economics of the large
chips much faster than the smaller ones. A 2% defect rate with the
smallarea chips means that 98% are good. With the larger CCDs, only
0.98^25 = 0.60 or 60% work. If 97% of the small chips work (3%
defects), only 47% of the large ones do. If 95% of the small chips
work, only 28% of the large ones do. And if 90% of the small chips
work, only 7.2% of the large ones work. If 80% of the small chips work,
only 0.4% of the large ones work  essentially none at all.
So, in the case where 90% of the small chips work, the large chips have
to cost 25 times more because of area, and 90/7.2 = 12.5 times as much
because of the difference in defect rate. So the *manufacturer's cost*
for the large chip is 312 times the cost of the small one, not 25
times.
And then there are economies of production scale that make the
difference larger yet.
Apparently semiconductor fab houses will not tell you what their defect
rate is  whether it's 1% or 5% or 10%. But the way the mathematics
works, you can pretty much guarantee that the cost of a large chip goes
up much faster than the relative area of the chip.
Dave
