William Donzelli <aw288_at_osfn.org> wrote:
> It seems to me that maybe, just maybe, if you shine a bright enough image
> (with very good contrast) at a stock CRT you could detect the variations
> in the beam current. I think this was tried a while back in a CRT
> specially made for the application in an early memory device.
I imagine you're thinking of Williams Tube memory, which was used on
various computers before core memory was available, including the IBM 701
and 702 (their first mass-produced electronic computers for scientific
and business applications, respectively).
Williams tube memory worked by sensing the potential at the face of the
tube, not by sensing the beam current.
IBM apparently was never able to make Williams Tube memory work reliably
at the density that they originally quoted to customers (1K bits per tube),
so they scaled it back to 512 bits per tube with twice as many tubes.
When core memory became practical, IBM retrofitted the 701s and 702s they had
already sold. Meanwhile, Thomas Watson Jr. heard that the 705 team was
planning to use Williams Tube memory instead of core because it was cheaper.
He apparently said something about the stupidity of using something that
doesn't work instead of something that does on the basis of the economy of the
former. The 705 shipped with core memory. :-) As did allmost all IBM
computers from then until the introduction of semiconductor main memory on the
360/85 and thin-film memory on the 360/95.
Does anyone know how RCA's Selectron memory tube worked? I've seen pictures
but no explanation.
Memory wasn't the only part of early computers that used special and somewhat
bizarre electron tubes. One of the earliest CRT display devices was the
Charactron. This worked by defocusing the electron beam, using one set of
deflection coils to steer the electrons through a character stencil (like a
shadow mask), and a second set of deflection coils to steer the electrons
to the desired character location on the screen. If memory serves, this
type of display was used on Whirlwind. Presumably it was more economical
than using a standard XY oscilliscope with either a hardware vector character
generator or software.
Eric
Received on Sun Aug 15 1999 - 02:45:02 BST