[GreenKeys] [External] Re: Interesting modified teletype

[Jones, Douglas W]

From: John Spigel [w1an.dxusa at gmail.com] -- Saturday, March 30, 2024 9:27 AM

> I definitely was surprised that arrays of small CRTs were used as memory storage.

It's called a Williams Tube memory.  That was the first really successful vacuum tube RAM technology.
Technically, to use modern terminology, it was vacuum tube DRAM.

The phosphor on the CRT screen was irrelevant (but useful for debugging).

What mattered was that when you throw a packet of electrons at a spot on the screen, they charge that spot (leaking away slowly because the glass screen isn't a perfect insulator).

To read a spot, you need a conducting plate on the other side of the screen.  Copper screen was usually expoxied to the face of the Williams tube because that let you see the screen for debugging.

Throw a small pulse at a spot on the screen and you get pulse on the plate, coupled capacatively through the glass of the screen.  If that spot on the face of the tube was already charged up, the read pulse will be repelled and delayed.  If the spot was not charged, you get a prompt pulse.

To make it work, you need leakage fast enough that read pulses at any spot don't accumulate much even if you focus all your memory cycles on one spot.

To make it useful, you need to refresh the memory.  Typically, half of the memory cycles were used by refresh logic that simply cycled through memory, reading and refreshing each address continuously.  The monitor CRT lets you watch the traffic to any bit in memory.

You could typically double the performance of a Williams tube computer if you turned off the refresh logic and instead did refreshes on every memory cycle.  So long as the software read every location in memory on which it depended frequently enough, the program would continue to do useful work.

In the late 1940s and early 1950s, no other memory technology could compete with Williams Tubes for speed.

The competing technology used mercury acoustic delay lines or (even slower) magnetic drums.  Both of those stored data sequentially, typically using 40 delay lines or 40 drum tracks for a 40-bit word.  To read or change a particular word, you had to wait for that word to come by in the memory cycle.

Magnetic core memory was a huge improvement!

        Doug Jones