I don't see any obstacles to making this camera either, but it'll
be a couple of years yet (if not longer) before they can do it for
$3500 AND still make a profit. If they want to sell at a loss, then
that's another matter. Of course, this is just my opinion, but
since I design small high performance computers for a living, this
is a somewhat educated opinion.
Here are a few facts to consider:
- A full frame sensor is roughly 6 times the size of Intel's
current top of the line Pentium 4 chip. The P4 is generally
considered to be a pretty big chip by industry standards. A full
frame chip, therefore, is HUGE and yields will be AWFUL, even if
manufactured in a state of the art fab. The largest semiconductor
wafers used today are 300mm in diameter. Roughly 60 full frame
chips will fit on a 300mm wafer. That means the cost of processing
a single wafer must be amortized over at best 60 chips. And most of
those chips won't be functional, for a whole host of reasons, so
the cost to fab a single wafer will actually have to be amortized
over much fewer than 60 chips. Wafer fab costs for state of the art
300mm wafers are pretty high. After yield losses, I would be
surprised if anyone could fabricate a full frame chip for less than
$1K apiece. That's actual cost to build, not any marked up cost for
sale.
- Getting 12 MP off a sensor at 8 frames per second is a daunting
task. With the LBCAST sensor, Nikon had to go to 2 readout lines
per pixel row to get the required readout bandwidth. Scaling that
up suggests that you would need 6 readout lines per pixel row (i.e.
12 MP at 8 frames per sec. requires three times the readout
bandwidth of Nikon's sensor). You would probably need three times
as many amplifiers and A/D converters to handle all these read
lines. Amplifier gain would have to be adjusted so that you didn't
get banding. Adjusting amplifier gain drives up cost. High speed
A/D converters aren't cheap and consume a lot of power. So getting
12 MP off a sensor at 8 frames per sec. drives up the cost and
power consumption.
- To buffer 12 MP at 8 frames per second would take a huge buffer.
I agree with you that memory is cheap, but consider this. The
largest ram chips currently available are 512 megabits, or 64
megabytes. With 18 MB per image (assuming 12-bits per pixel), it
would take 6 of these very large ram chips to buffer 20 raw images.
Still not a lot of money, but this would require a lot of circuit
board space in the camera. Also, to handle the flood of pixels,
you'd have to operate these chips, and the ASIC that controlled
them, at frequencies of 100 Mhz or more. Six chips and an ASIC
operating at such a high speed consumes a lot of power, which along
with the higher expected sensor power consumption, will require a
beefier battery and power regulation system.
None of these things are impossible, of course. But they all drive
up costs. It just doesn't seem even remotely plausible to me that
Canon, or Nikon, or anyone, could make such a camera, with TODAY's
technology, for the quoted $3500. A couple of years from now it'll
be a different story, but the Canon thread is claiming it is
imminent, which means it would have to be designed with today's,
and to some extent, yesterday's technology.
Of course, I've been wrong before (more times than I care to
admit). If Canon were to pull this off, then I would happily eat
crow and get in line for one of these miracles. But I think that's
just what it would take - a miracle - for Canon, or anybody else,
to pull this off.
I do not see any obstacle to make this camera.
that the 1D
