They probably are cooled.  Even with cooling and near 100% efficiency
there will be an uncertainty in how many photons arrive at each well
from a uniform source in any given time frame.

See
<http://www.astro.psu.edu/xray/cubic/papers/cubic.cal.spie95/cubic.pcs.ca
libration.html#qe> Note carefully the unrealistic assumption that every
photon gives rise to a pixel detection.

Take a look at
<http://www.palomar.caltech.edu:8000/maintenance/scope/status.tcl> for a
real CCD telescope.

<http://www.ccas.ws/Newsletter.htm>

Yeah, would that it were that simple.  Electronic detectors
are obviously more practical than film for space use, although
before CCDs, there were other electronic detectors (Reticons,
video tubes, intensified phototubes...)  It's still a lot more
complicated than bolting together a camera, solar panel,
and radio.

The chip is presumably destined for some telescope that
the US Naval Observatory has access to or is building a
CCD camera for. I don't know which one.  They did have a
proposal for an astrometric satellite in one of the NASA
Explorer programs, but I think it was cancelled even before
the current return-to-Mars budget crunch put Explorers
on hold.

CCDs used for astronomical imaging are usually cooled
with either a thermoelectric cooler or liquid nitrogen - if
LN2 is used, there is also usually a electric heater
(or heaters).  One wants to regulate temperature, and keep
humidity from condensing on the dewar window.  It is
important to keep the CCD not only cold, but at a consistent
temperature.  Cold reduces dark current and read noise,
generally.  Consistency means that when you take an
image at one time, and a flat field a few hours later, they
have the same properties (like pixel-to-pixel variations in
sensitivity).  If you can't rely on that, the data will be useless.

Contrary to popular belief, science grade CCDs at the average
ground-based observatory (HST can pick and choose a little
more) are not usually free of cosmetic defects - hot pixels, bad
columns etc).  It's more important to push the envelope of detector
size than to have an absolutely perfect CCD.  Often the users
can combine multiple images to reduce the effects of cosmetic
defects.  If not you usually just suffer the loss of a little bit of
the data.

According to Jeff Rife  <wevsr@nabs.net>:

    Hmm ... as long as there is nothing moving in the shot.

    Well ... actually, the wafers *are* broken, after break lines
are either scribed or etched in the wafer.  Quite often, the technique
for breaking is to inflate a rubber sheet pressing it into another
rubber sheet, which combine to try to bend it into a dome shaped curve,
and thus causes breaking along the pre-made scribe lines.

    But just breaking a large sensor may be pretty useless, as
normally the outer border of the sensor has the amplifiers for driving
the signals as they go off the edge of the wafer for processing in the
camera body.  And unless it is very carefully designed, you will lose
functionality in some parts as a result of the break.  Or -- if you
partition it for the break, you will have zones in the sensor which do
nothing -- along the proposed break lines, so the wafer must be designed
either for partitioning into multiple smaller sensors, or to be a single
large one.

    Enjoy,
        DoN.

The snap is ten or more minutes, the move is ten seconds; might work for
whole sky imaging, but not so good for deep space objects.

The chip is designed to deliver the data from the individual elements to
(presumably) the edges of the chip where the interface electronics is
located, if you "cut bits out" they would not have interface electronics
on the edges of them. Much more importantly you would not have 10,000
square millimetres of contiguous sensor.

Producing a wafer that is completely usable (errors below a specific
threshold) is a nightmare, but it was produced by a fab-lab specializing
in one off devices. Presumably the cost of ensuring that the whole wafer
is good is justifiable, whereas when you are making thousands of wafers
with ten or more devices on each trying to get 100% yeald is not
justifiable for the market. I suspect that alternative of splitting the
image from the telescope across ten sensors is not cheap. I did not
notice any mention that it was to be used in space.