Univac I memory tank

[Christian Corti]

On Wed, 15 Mar 2017, dwight wrote:
> The Olivetti used a piece of wire for the delay line. I forget what the 
> Dielh Combitron used but I know it used a two delay lines. One was for 
> registers and the other was for lookup tables that loaded at turn on 
> time from a metal tape ( as I recall ).

I can tell you exactly what the Diehl Combitron does; I have a running 
bachelor thesis for a student who is developing an emulator and assembler 
for that machine, and we have also disassembled (but not yet 
understood) the firmware (contained on the metal tape) of the machine.

In fact, it uses two magnetostrictive delay lines, one is called the R 
delay line containing 219 bits plus one external in a flip-flop. The other 
is called the M delay line with a total of 10889+1 bits. The main clock of 
the machine is 1 MHz thus a bit time (called P) is 1 µs.
The R line holds four words à 55 bits, one in the I phase P bits time, one 
in the I phase /P bits time, one in the /I phase P bits time and one in 
the /I phase /P bits time. The instructions are always fetched from the I 
phase /P bits and executed in the /I phase.
The M line holds a total of 99 P words and 99 /P words. The phase between 
P and /P changes every M cycle

During the loading phase (e.g. at power on or after a 'e1' order) the R 
line is filled with the contents coming from tape and then executed. 
Usually the code just transfers the other three words of the R line 
somewhere into the M line and restarts the loading phase for the next 
block. After loading the last block a fill instruction "jumps" to the 
entry point of the firmware. The fill instruction transfers four words 
from the M line to the R line.

Christian