>> You know that there are people who do scientific experiments *about*
>> computer programs, right? ;-)
> 
> Yeah :)

OK, good. :-)

> Actually, in the vein of computer science: For the fun of it I'm 
> designing CPU from scratch. I plan to build it with TTL gates after I 
> have the design complete and have vetted it through simulation.

...this sounds so much like my life approximately 15 years ago...

> Eventually, I want to take a simplified version of that design and 
> create a CPU entirely out of relays. 
> Something about a bit stack of clacking relays with blinking 
> lights really appeals to me.

Hmm, interesting.

Perhaps you would have enjoyed a visit to Stoney Stratford telephone 
exchange. It's a sprawling building, filled with rack after rack after 
rack of cases, each case filled with dozens of relays. And they're not 
in cases or anything, just "naked", so you can see (and if you desire, 
move) the working components. There is also several tonnes of wire 
overhead feeding this equipment. And all of it is as silent as death. 
It's like a museum or something.

(Just down the corridor there's a small broom cupboard. This is the 
*real* telephone exchange, alive and operational. It's really just a few 
lasers and some circuit boards. But mess with it and you'll have just 
taken 15,000 customers offline!)

Myself, I visualised a computer made of pressurised water fed through a 
series of simple mechanical valves. Sadly, I fear that for reliable 
operation, you would need inside water pressure. And if you wanted a 
clock speed of more than about 0.02 Hz, you would have to use steel 
piping and pyrotechnics to sustain the necessary pressure! It would be a 
very "kinetic" experience though.

> It's fun stuff, and occupies a good deal of time. Also a great way to 
> learn HOW a computer like the one you're sitting at actually works. e.g. 
> how each instruction of machine code does what it does, etc...

Yeah. I've written about it before now. (As I say, a computer is a layer 
cake though. Try to explain it and laymen quickly get lost in the layers 
of abstraction upon abstraction needed to perform even the most trivial 
task.)

Also... when I tried to build my own machine out of 7400s, I quickly 
discovered that the gates don't appear to function as their truth table 
indicates that they should. (!) I also looked into playing with FPGAs, 
but the cost is prohibitive. (And, knowing me, I'd just make something 
that doesn't even work, and then spend months trying to find out why!) 
Plus, Xilinix (?) have a free simulation tool available, and it's just 
painful to use. I dred to think what actually synthesizing with it would 
be like...

> The TTL version should be fun, too. I plan to have lots of LEDs to show 
> what's happening inside the machine.

In solomn truth, it's probably simpler and easier to code a small peice 
of JavaScript that controls a little Flash animation on a computer 
screen. But there's something impressive about being able to pick up a 
physical object in your hands and see that there really are no tricks...

Post a reply to this message