RSS
It's the '60s, and you don't have access to a semiconductor fab to make piles of cheap memory for you, so how could you store data on your computer?
Well, MAKE subscriber Steve points us to one possible solution, courtesy of vintagecalcuators.com: delay line memories. Rather than having a bunch of individual units that store a bit each, these memory devices work by storing data in sound (compression) waves. The device consists of a long length of wire, with an actuator on one end to vibrate the wire, and a reader on the other end to pick up vibrations. Because the vibrations don't travel very fast along the wire, you can make a whole bunch of them before the first one reaches the end of the wire, and that becomes the 'size' of the memory. Data can be read back by looking for a vibration at a particular time- if there is one, that corresponds to a '1', and if there isn't, it would be a '0'.
It sounds a bit weird, so I like to think of it like this. If you had a hard time remembering things for very long, and happened to live in a cave, you could just shout out what you didn't want to forget, and a few seconds later you would hear an echo to remind you. Of course, the problem with this is that an echo doesn't stick around for long, so you would have to shout again every time that you heard the echo, so that you could remember again in a few seconds. Assuming you could keep this up, you would never forget your idea. Of course, that would get really tiring after a while, so you would be much better off just writing it down.
The memory shown above is from a Monroe Epic 3000 calculator, which was apparently the first programmable calculator with a printer built in.
The physics behind it is called magnetostriction. It was well known in the 60's, see a RCA semiconductor handbook of the time and you'll find the drive amplifier and sense amplifier circuits.
Currently it is used in lifts as "linear position feedback", search "Ultrasonic positioning system lifts".
Reply to this comment
Similar phenomenon but with a liquid metal, two piezos on the top and bottom of a mercury (quicksilver) tube. vibration travels through the metal and is captured on the other side. The system could be made to mimic a endless loop tape.
Reply to this comment
Of course, it wasn't a standalone concept. You could buy reverb units for your stereo, to add echo, that was basically a spring with a something to feed it at one and a transducer at the other end to turn it back into sound. A variant was in "Elementary Electronics" about 1974, where they used a garden hose as the delay, a speaker at one and and a microphone at the other.
Long coils of wire were used as delay lines in ocilliscopes and in color tv sets. Such things were even used for interfacing CPUs with dynamic RAM, they needing some delay.
What is interesting is that once semiconductor memory came along, the earliest but highest density memory was shift registers, the semiconductor equivalent of the wire delay line. Don Lancaster's earliest TV Typewriter used long shift registers for memory, as did a lot of early slow scan tv converters. They were what wsa "cheap" in the early seventies. I don't know if they made the first inroads because nobody could imagine some other format, or they were seen as easier to interface, or even if there was a specific need for long shift registers that were then used elsewhere.
Surely "RAM", ie Random Access Memory, is named because it really is random access, unlike the shift registers and the delay lines, where you had to cycle through to get to the same address location again. Being able to address a specific address by loading that into the IC was in some ways novel at the time. It explains why "Random Access" is seen as the important descriptor, when after a certain point virtually all memory, whether it was read only or read and write, was random access.
Michael
Reply to this comment
Shift registers would be useful for display memory, where you need to sequence through the memory as the electron beam on a CRT scans across a screen.
That brings to mind Williams Tube memory:
http://en.wikipedia.org/wiki/Williams_tube
As for storing bits in a long length of something, how about storing bits as optical pulses in a long fiber optic? Probably not very dense, but might be cheap to make. :-)
Dave
Reply to this comment
Over the past few years, similar concepts have been successfully tried using Metro-Area-Networks (MANs). Basically using the processing delay time between routers to store quite a bit in routing "rings". The data basically resides in the buffers of each router as it makes it's way around the ring.
I'm not entirely sure what the point was, but kinda neat anyway.
Reply to this comment
Finally, something fun to do with all that vibration/acoustical education!
This reminds me of a story a colleague once told me about communications satellites. In the old days, occasional gamma ray events would knock out the electronic clocks in military communications satellites. They used an acoustical delay line memory, impervious to gamma ray events. It stored the "last known-good timestamp" long enough that if the clock got wiped, it could query the delay loop to know what time to start counting from for encryption. http://www.va-consult.com/
Reply to this comment