Hi,

I'm wondering about building a RAM-chip camera for a robot.  Does anybody
have any links or additional info for one of these?

I'm guessing that light "charges-up" one of those little memory bits &
that the more light that hits it, the quicker it charges, and the amount
of charging time corresponds to a grey level.  How long can this take?
I'd settle for 64 grey levels in a 32x32 image.  Is that practical?

Are there any other small digital-ready camera devices out there?

        A ceramic-case chip with a metal lid is the easiest kind
        of chip to open without destroying.  Heat the lid to melt
        the solder and flip it off.  Epoxy on a glass window to
        protect the chip from dust.  Microscope slide "cover slips"
        are about the right size.

        Depending on the chip you will either set the memory to
        1's and they degrade to 0's or set them to 0's and they
        degrade to 1's.

        The order of address bits for the chip may have little
        to do with the physical order of the memory cells.  You
        may have to swap (and possibly invert) address bits to
        figure out how to scan top-to-bottom and left-to-right.

        Also the "imaging area" is seldom square so in a 2K chip
        you are likely to get two sections that are 16x64 instead
        of two sections that are 32x32.  You may have to go to an
        8K or 16K chip to get an image area that is at least 32
        bits across its smallest dimension.

:       You have to use static RAM (not dynamic or pesudo-static).

        Uh, not as I recall doing it circa 1975, with an Intel
1Kbit DRAM (1101?). Statics would keep replenishing the charge.
It's the acclerated leakage that you _want_ to use for sensing.

:       Depending on the chip you will either set the memory to
:       1's and they degrade to 0's or set them to 0's and they
:       degrade to 1's.

        Or set half the columns to zeroes and half to ones, on chips
that try to balance the sense-amp inputs and thus have conditional
inverters on the data lines :-). You can tell by "letting a chip rot"
and seeing what the default seems to be.

: > I'd settle for 64 grey levels in a 32x32 image.  Is that practical?

:       I don't know about grey levels but 32x32 is 1K-bits.  Many
:       RAM chips put addressing circuits down the middle of the
:       chip and, therefore, have their "image plane" divided into
:       two sections.  That would mean that you would want to
:       start with a 2k-bit chip.

        You get "grey levels" by taking your samples at longer
and longer intervals. The brightest spots "rot" first. Dark
spots can take several seconds to rot. This is _not_ high-speed
photography. I talked to one wacko who "dithered" his DRAM-Cam
by vibrating it mechanically, then processing the succesive images,
to get rid of the "dead zone crosshairs". Clearly a fellow with
_way_ too much time (and CPU power) on his hands :-)

        My two cents, worth no more than you paid :-)

In Germany some people developed a RAM-camera. You can find some links
using Alta-Vista with the keyword 'kuckuck.zip' (including the quotes).
But the papers I found were all in German (thats no problem for me, but
perhaps for you). The papers however do include theoretics and even
source-code.

Good Luck,