Brian Whatcott wrote:
Larry wrote:
(Richard Casady) wrote in
news:49c969b9.9397015
@news.east.earthlink.net:
Eight bits, one byte, per tube.
I'd love to know the physics behind how they did that. Dual triodes,
such as 12AX7, 12AT7, 12AU7, or even earlier 6SN7 were used as
latching flip flops, but they only stored one bit...0 or 1. To get 8
reliable levels would be magic. They did use a neon counter tube that
had multiple cathodes. Perhaps that is the "tube" that did a byte.
[long nostalgic bit follows...]
Well, this was a slightly mixed memory, I'd say. the idea of 8 bit
collections called bytes came later, around the IBM 360 time frame.
The early CDC machines I worked on used 6-bit "bytes" and used 36 or 60
bit words for instructions. IIRC they even did BCD math on 4 bit
nibbles.
Previous IBM incarnations like the 1401 used eight bits plus parity:
6 data, and two marker bits for laying out data fields - if you wanted
to add two 1000 decimal place numbers, no problem. Just lay down the
field length markers and issue ADD
But computer memories were kinda weird and wonderful.
There was quite a development effort into CRT memory storage:
allocate an X and Y value for each bit and point the electron beam at it
to set and to read it out. Hundreds or thousands of bits.
Trouble was the beam position drifted, which didn't help data integrity.
I worked on a CRT that use magnetic torsion rod memory. Magnets on one
end would send torsion pulses down the rod which were picked up on the
other end. This was synchronized with a spinning disk that had the
character masks for the screen!
The serial computers used a nickel line for short term storage: you
pulsed an electromagnet at one end to set a pulse going for a '1' or no
pulse for '0' so a stream of pulses would run down the nickel wire at
the speed of sound to the sense coil at the other end, where the data
would turn round to the start coil and start over.
That was called magnetostriction.
Others used mercury delay lines. Or coil delays.
Yes, I remember the mercury delay systems. Fortunately they only gained
favor in the analog world.
One of my favorite "antique" books (next to my old Bowditch and Coast
Pilots) is "High Speed Computing Devices" published in 1950. They
describe in detail a new device called the transistor: "It seems likely
that this device will simply computer circuits considerably."
....
This was a step back in one way: you could lose power on a core
computer and when it came up again, the code and data were still there
in core. Not so with solid state memory....
My lab had one older "mini computer," a DG Nova with limited core memory
(8K?) and a teletype with paper tape reader, that we we used as a
programmable calculator. It was loaded with Basic and was ready to go
on powerup. A machine like that, but with 64 kB and floppies became my
first "personal computer" in 1979.