39 lines
1.7 KiB
Markdown
39 lines
1.7 KiB
Markdown
---
|
|
tags: [computer-history, memory]
|
|
created: Wednesday, September 18, 2024
|
|
---
|
|
|
|
# Delay line memory
|
|
|
|
- First described in the _First Draft_ by #vonNeumann based on work by Eckert
|
|
and Mauchley.
|
|
|
|
- Information is stored as acoustic waves travelling through a medium, typically
|
|
mercury. As sound travels more slowly than electric charge (light), if
|
|
electric signals are converted to sound, the delay time involved can be used
|
|
as a type of store.
|
|
|
|
- It worked as follows:
|
|
- Imagine we have an 8-bit number. The number is input as a sequence of pulses
|
|
where a pulse is 1 and the absence of a pulse is 0.
|
|
- The pulses are converted into sound waves (using a transducer, as with a
|
|
speaker) and sent through a mercury-filled tube.
|
|
- The length of the tube is calculated so that it takes exactly the time of
|
|
one machine cycle for a pulse to travel from one end to the other
|
|
- At the other end of the tube, another transducer converts the sound waves
|
|
back to electrical pulses
|
|
- This operates in a loop: the pulses go back into the input end of the tube
|
|
and the cycle runs again.
|
|
|
|
To read the data, the computer would "listen" to the pulses coming out of the
|
|
receiving end. To write new data, the computer would inject new pulses at the
|
|
precise moment when it wants to change a '0' to a '1' or vice versa.
|
|
|
|
For an 8-bit number there would be 8 pulses for each bit and it would be
|
|
necessary to keep track of 8 specific positions in the delay line where each bit
|
|
would be stored.
|
|
|
|
The limitations were timing - difficult to keep track of each precise bit and
|
|
when it would arrive at the output. Also it was sequential rather than random
|
|
access. Not every bit could be accessed at once, you had to wait for its time
|
|
slot to come around to access it.
|