discrete, historic memory, was Re: Neon bulb logic elements link?

[Brent Hilpert]

Tom Jennings wrote:
> Slightly changing subject (!), but remaining on discretely
> constructed memory, one truly practical memory that was tried and
> abandoned about 1952 (one of the Standards computers) was a
> switched capacitor memory. It was umm revived some years later and
> is now of course universal.
> 
> A discrete-capacitor, diode-switched, dynamically-refreshed memory
> was built and tested and performed apparently well was built for
> SWAC? SEAC?  or something in 51, 52 [ref in Huskey's COMPUTER
> HANDBOOK] but abandoned as solid-state diodes were prohibitively
> expensive.
> 
> It would clearly be most straight-forward to build one today, even
> with el cheapo ceramic disks. Probably could do .25 sq in per cell
> on a PC board. And since one was actually made, it would be
> historically correct.

Don't forget discrete-capacitor dynamically-refreshed memory was pioneered by Atanasoff for the ABC in 1939-41, albeit mechanically accessed (rotating drum, 1500 bits, max access time 1 second).

One of my languishing projects is a recreation of the ABC prototype (2 registers of 25 bits (capacitors) each on rotating disc, together with serial RVL[1] adder/subtractor. (Although such a recreation was already done for the Honeywell vs. Sperry-Rand trial in the 70s).

 [1] RVL: Resistor-Vacuum-tube-Logic - the original electronic logic family :) 


... recently received a counter using Philips E1T[2] tubes. Not overly useful as a general-purpose memory but it is a state-holding device (3.3 equivalent bits), and another example of the bizarre-by-today's-standards techniques from the pre-IC days.

 [2] E1T provides similar functionality to dekatrons - one tube providing combined functionality of decade counter and display - but entirely different technology. The E1T is based on an electon beam held in 1 of 10 places (states) by a combination of electrostatic and magnetic fields. The beam can be shifted from place to place with pulses to accomplish the counting function. Beam then hits a flourescent screen in accordingly different spots for the 1-of-10 display.