crazy memory ideas?

Nice code:

int a[1817];main(z,p,q,r){for(p=80;q+p-80;p-=2*a[p])for(z=9;z--;)q=3&(r=time(0)

Sounds like the Atansoff-Berry computer memory, IIRC? But with frikkin' lazers instead of electrodes.

If you swap laser for electron beam, doing it in a vacuum, you've got something very much like the storage tube (CRT). That helps because secondary emission is pretty easy to deal with, and helps keep the charge in place. The photoelectric effect would do just as well of course.

Jeri Ellsworth made her own MOSFETs... nothing's impossible. Rather than wafers, it might be easier to start with something dumb, like evaporated CdS or something. Food for thought:

IIRC, it has to do with the buried charge layer in the substrate, beneath the MOSFET. As it turns off, the charge gets 'squeezed' into the next stage. There should still be attenuation though, as any normal (constant) capacitance should act to 'dilute' the effect of the variable capacitance, attenuating the signal. I think.

Tim

Cool, I'm going to have to try this. Instead of water you could use window defrost heater repair paint.

This guy made his own DRAM for his relay computer:

I don't see any refresh circuits in his design.. I guess programs have to be aware of memory decay time :-).

Oh, another one I want to try is to make an acoustic delay line: but I want to run it at audio frequency and use the air as the delay line.. (instead of wire or crystal), so that you can listen to it..

Another is analog TV + camera in a feedback loop, you can get bits to walk across the screen with the right camera angle.

Heh, relay contacts have less leakage than DRAM cells (in absolute terms, probably not, but adjusted by scale, absolutely). I'd be surprised if a bit decays (becomes unable to trip the relay coil) by the time he gets bored with running it. :)

Beware of reflection and dispersion -- keeping bits in place (the step response) may be challenging.

Heh. Which goes back to your CCD comment before. :) (Probably over half the data ends up stored in the CCD -- or Vidicon if you're nostalgic -- and the rest in phosphors and wiring.) Seems to me they always faded to blue. Wonder if that's something with NTSC, or just a coincidence?

Tim

We could distribute software on vinyl disks, 45s or LPs. We'd have to agree on a coding format.

You could melt pits into the surface of wax or ice with a laser, and read it back with a lower power laser. Bulk erase.

There was actually an early computer whose memory was a drum decorated with discrete capacitors. It rotated against a row of contact brushes. That could be done on a pcb, surface-mount. Possibly the FR4 could be the capacitors.

That was the Atanasoff-Berry Computer mentioned upthread. And software _was_ distributed on audio cassettes!

Cheers

Phil Hobbs

Kansas City?

A disk version should be easy. I calculate 62 pF each for 16 sectors of a

Records too!

There is another theory, which states that this has already happened.

Why did my read error rate go up when I moved from Toronto to Phoenix? :)

Matt Roberds

Some early memory designs had a rotating loop of magnetic wire. If you wanted to access address X, you waited X*bits clock cycles after the sync marker then read or wrote bits. Memory density wasn't too bad for that era but latency was awful.

Scientists have tried similar tricks sending light pulses around loops of fiber optic cable. It works with all the same limitations as the loop of magnetic wire. There's also short-term storage using piezo elements that bounces mechanical waves through a little plate of quartz. Longer term storage can be performed by sending mechanical waves through long coils of spring wire, like old guitar amps that have a reverb effect. Ultra-short storage uses bits of coax cable.

I don't know semiconductor physics, but I believe the bucket-brigade chips slide a clump of electrons down a continuous FET having hundreds of gates. It's essentially creating a moving pocket of conduction that the electrons must follow. It's not a chain of FET amps like some equivalent schematics. CCD sensors use the same tech to slide images off the optical area to amplifiers at an edge.

JL > We could distribute software on vinyl disks, JL > 45s or LPs. We'd have to agree on a coding format. Shades of Tarbell interface! A Minneapolis Radio computer talk show in the 1980's actually broadcast a lot of software for various computers and the interface they sold was not terribly expensive. Computer magazines also published binary software that could be scanned from the page by some of the scanner gadgets at that time. Univac sold HUGE magnetic drum storage units that used ferrous oxide coating and magnetic heads very much like common hard disk drives today. Apparently this type of storage was fast enough that it could be used in ways more traditionally thought of as a job for core memory or RAM memory. I heard that people working around them didn't want to think about what would happen if the bearings went out or the huge spinning drum were to ever break loose at speed.

Mercury filled tubes were used in early (1940's) computers to send acoustic waves through the tube and after regeneration sent back to form a shift register, with the same limitations as a rotating drum or disk storage.

The most common application for this was in PAL colour television (hundreds of millions of sets) in Europe, which required a one horizontal line (64 us) delay with approximately 1 MHz bandwidth.

You could use a TV satellite transponder with perhaps 200 Mbit/s throughput as a shift register memory. With 250 ms round trip delay, this ring could store about 6 Mbytes. Of course, there are cheaper means to store that amount of data :-).

However, 14" single platter disk drives with similar capacity were used in the 1980's.

The main advantage of drums vs. disks was that you more or less had to use a separate head for each track, while in a disk drive you could move the read head from track to track, causing the seek time.

With a rotating device, you must be very careful how you store the data. If you put data in every block, reading and processing a block may take too long, so that the next block is lost and you have to wait a full rotation until getting the next block.

However, if you put the data in every other block thus the first part in odd blocks and the rest in even numbered block, after reading the first block, you have the full memory rotation block time to process the data, before processing the next block. This assumes sequential program execution.

In floppy disks a similar method was used.

Cheers

Phil Hobbs

I tried this: here's a video:

snipped-for-privacy@downunder.com schrieb:

Hello,

they did that not only with mercury filled tubes, but also with solid metal wires wound to a spiral and carefully suspended in slots cut into sheet metal stripes. The suspension of the wire did not disturb the ultrasonic waves travelling through the wire.

Bye

The British Elliott 803 used nickel wire spiral delay lines for registers, and ferrite memory rings for logic (and for memory as well: 8192 words of 39 bits). It was not blindingly fast: 288 us per machine cycle.

Lower local literacy rate?

?-)

Why would any sane person move to Phoenix?

Because the weather in Toronto is even worse, though in a very different way.

Cheers

Phil Hobbs